[OpenMP][Offloading] Change `bool IsSPMD` to `int8_t Mode` in `__kmpc_target_init` and `__kmpc_target_deinit`

[tianshilei1992]

Summary

This is a follow-up of D110029, which uses bitset to indicate execution mode. This patches makes the changes in the function call.

[tianshilei1992]

fix issues

[tianshilei1992]

fix issues and add more tests

[tianshilei1992]

remove unrelated changes introduced by tools

[tianshilei1992]

rebase and make changes in new device runtime

[jdoerfert]

This should not change anything, right? Please confirm running a non-trivial example, like miniqmc or so, and then this is good to go.

[tianshilei1992]

In D110279#3016332, @jdoerfert wrote:

This should not change anything, right? Please confirm running a non-trivial example, like miniqmc or so, and then this is good to go.

miniQMC report 26/26 Passed. Will land the patch if Phabricator comes all green.

➜  OMP_NUM_THREADS=4 ctest .
Test project /home/shiltian/Documents/apps/miniqmc/build
      Start  1: unit_test_utilities
 1/26 Test  #1: unit_test_utilities ............................   Passed    0.15 sec
      Start  2: unit_test_particle
 2/26 Test  #2: unit_test_particle .............................   Passed    0.06 sec
      Start  3: unit_test_omptarget_math
 3/26 Test  #3: unit_test_omptarget_math .......................   Passed    0.20 sec
      Start  4: unit_test_omptarget_reduction
 4/26 Test  #4: unit_test_omptarget_reduction ..................   Passed    0.19 sec
      Start  5: unit_test_omptarget_target_task
 5/26 Test  #5: unit_test_omptarget_target_task ................   Passed    0.33 sec
      Start  6: unit_test_omptarget_complex
 6/26 Test  #6: unit_test_omptarget_complex ....................   Passed    0.18 sec
      Start  7: unit_test_omptarget_complex_reduction
 7/26 Test  #7: unit_test_omptarget_complex_reduction ..........   Passed    0.20 sec
      Start  8: unit_test_omptarget_math_with_complex_header
 8/26 Test  #8: unit_test_omptarget_math_with_complex_header ...   Passed    0.07 sec
      Start  9: unit_test_omptarget_memory_interop
 9/26 Test  #9: unit_test_omptarget_memory_interop .............   Passed    0.18 sec
      Start 10: unit_test_omptarget_partial_update
10/26 Test #10: unit_test_omptarget_partial_update .............   Passed    0.18 sec
      Start 11: unit_test_omptarget_partial_update_full
11/26 Test #11: unit_test_omptarget_partial_update_full ........   Passed    0.35 sec
      Start 12: unit_test_omptarget_map_always
12/26 Test #12: unit_test_omptarget_map_always .................   Passed    0.19 sec
      Start 13: unit_test_omptarget_icpx_opencl_wrong_number
13/26 Test #13: unit_test_omptarget_icpx_opencl_wrong_number ...   Passed    0.21 sec
      Start 14: unit_test_omptarget_nowait_taskwait
14/26 Test #14: unit_test_omptarget_nowait_taskwait ............   Passed    0.35 sec
      Start 15: unit_test_omptarget_taskloop
15/26 Test #15: unit_test_omptarget_taskloop ...................   Passed    0.06 sec
      Start 16: unit_test_omptarget_runtime_mem
16/26 Test #16: unit_test_omptarget_runtime_mem ................   Passed    0.18 sec
      Start 17: unit_test_omptarget_map_transfer
17/26 Test #17: unit_test_omptarget_map_transfer ...............   Passed    0.55 sec
      Start 18: unit_test_omptarget_allocator
18/26 Test #18: unit_test_omptarget_allocator ..................   Passed    0.18 sec
      Start 19: unit_test_wavefunction
19/26 Test #19: unit_test_wavefunction .........................   Passed    0.07 sec
      Start 20: unit_test_Drivers
20/26 Test #20: unit_test_Drivers ..............................   Passed    0.06 sec
      Start 21: unit_test_assignDevice
21/26 Test #21: unit_test_assignDevice .........................   Passed    0.09 sec
      Start 22: miniqmc-g111-r1-t16
22/26 Test #22: miniqmc-g111-r1-t16 ............................   Passed    8.12 sec
      Start 23: miniqmc_sync_move-g111-r1-t16
23/26 Test #23: miniqmc_sync_move-g111-r1-t16 ..................   Passed    4.95 sec
      Start 24: check_spo-g111-r1-t16
24/26 Test #24: check_spo-g111-r1-t16 ..........................   Passed    8.29 sec
      Start 25: check_spo_batched-g111-r1-t16
25/26 Test #25: check_spo_batched-g111-r1-t16 ..................   Passed    5.22 sec
      Start 26: check_wfc-g111-r1-t16
26/26 Test #26: check_wfc-g111-r1-t16 ..........................   Passed    0.67 sec

100% tests passed, 0 tests failed out of 26

Label Time Summary:
check      = 226.94 sec*proc (3 tests)
miniqmc    = 209.11 sec*proc (2 tests)
unit       =   4.05 sec*proc (21 tests)

Total Test time (real) =  31.33 sec

[tianshilei1992]

update two cases

clang/test/OpenMP/declare_target_codegen_globalization.cpp

 
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6maini1v_l16
 // CHECK1-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

clang/test/OpenMP/nvptx_SPMD_codegen.cpp

 // CHECK-DAG: [[LIGHT:@.+]] = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 3, i32 0, i8* getelementptr inbounds
 // CHECK-DAG: [[DISTR_FULL:@.+]] = private unnamed_addr constant %struct.ident_t { i32 0, i32 2050, i32 1, i32 0, i8* getelementptr inbounds
 // CHECK-DAG: [[FULL:@.+]] = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 1, i32 0, i8* getelementptr inbounds
 // CHECK-DAG: [[BAR_LIGHT:@.+]] = private unnamed_addr constant %struct.ident_t { i32 0, i32 66, i32 3, i32 0, i8* getelementptr inbounds
 // CHECK-DAG: [[BAR_FULL:@.+]] = private unnamed_addr constant %struct.ident_t { i32 0, i32 66, i32 1, i32 0, i8* getelementptr inbounds
 // CHECK-NOT: @__omp_offloading_{{.+}}_exec_mode = weak constant i8 1
 
 void foo() {
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
 #pragma omp target teams distribute parallel for simd if(a)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
 int a;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
 #pragma omp target teams distribute parallel for lastprivate(a)
   for (int i = 0; i < 10; ++i)
     a = i;
 #pragma omp target teams distribute parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK: call i32 @__kmpc_target_init(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
 #pragma omp target teams
    {
      int b;
 #pragma omp distribute parallel for simd
   for (int i = 0; i < 10; ++i)
     ;

 #pragma omp target teams
 #pragma omp distribute parallel for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
 #pragma omp target teams
 #pragma omp distribute parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for schedule(static)

 #pragma omp target teams
 #pragma omp distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[DISTR_LIGHT]]
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[DISTR_FULL]]
 // CHECK-DAG: [[FULL]]
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target

 #pragma omp distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 #pragma omp target parallel for if(a)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK-DAG: [[BAR_LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK-DAG: [[BAR_LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK-DAG: [[BAR_LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
 #pragma omp target parallel if(a)
 #pragma omp for simd
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel
 #pragma omp for simd schedule(static)

 #pragma omp target parallel
 #pragma omp for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel
 #pragma omp for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK-DAG: [[BAR_LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
 // CHECK-DAG: [[BAR_LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 // CHECK-DAG: [[BAR_FULL]]
 #pragma omp target
 #pragma omp parallel
 #pragma omp for simd ordered
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target

 #pragma omp for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp parallel
 #pragma omp for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-DAG: [[FOR_LIGHT]]
 // CHECK-DAG: [[LIGHT]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 // CHECK-DAG: [[FULL]]
 #pragma omp target
 #pragma omp parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp parallel for schedule(static)
   for (int i = 0; i < 10; ++i)

clang/test/OpenMP/nvptx_data_sharing.cpp

 // CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]], i32* [[TMP8]]) #[[ATTR3]]
 // CHECK2-NEXT:    ret void
 // CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7test_dsv_l14
 // CHECK-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK-NEXT:    [[C:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [2 x i8*], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK:       user_code.entry:
 // CHECK-NEXT:    [[A:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK-NEXT:    [[A_ON_STACK:%.*]] = bitcast i8* [[A]] to i32*
 // CHECK-NEXT:    [[B:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK-NEXT:    [[B_ON_STACK:%.*]] = bitcast i8* [[B]] to i32*
 // CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])

 // CHECK-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 8
 // CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS1]], i64 0, i64 1
 // CHECK-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A_ON_STACK]] to i8*
 // CHECK-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 8
 // CHECK-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 // CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP9]], i64 2)
 // CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[B]], i64 4)
 // CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[A]], i64 4)
-// CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-NEXT:    ret void
 // CHECK:       worker.exit:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0]] {
 // CHECK-NEXT:  entry:

clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp

 // CHECK4-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
 // CHECK4-NEXT:    store [10 x i32]* [[D]], [10 x i32]** [[D_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[C_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
 // CHECK4-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[D_ADDR]], align 8
-// CHECK4-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[CONV1]], align 4
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]], [10 x i32]* [[TMP1]], i32* [[TMP2]], i64 [[TMP7]], [10 x i32]* [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[C:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[ARGC:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[D:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK5-NEXT:    store [10 x i32]* [[C]], [10 x i32]** [[C_ADDR]], align 4
 // CHECK5-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK5-NEXT:    store [10 x i32]* [[D]], [10 x i32]** [[D_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[C_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[D_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK5-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[TMP6]], i32* [[ARGC_CASTED]], align 4
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK5-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]], [10 x i32]* [[TMP1]], i32* [[TMP2]], i32 [[TMP7]], [10 x i32]* [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[C:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[ARGC:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[D:%.*]]) #[[ATTR0]] {
 // CHECK5-NEXT:  entry:

 // CHECK6-NEXT:    store [10 x i32]* [[C]], [10 x i32]** [[C_ADDR]], align 4
 // CHECK6-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK6-NEXT:    store [10 x i32]* [[D]], [10 x i32]** [[D_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[C_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[D_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK6-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[TMP6]], i32* [[ARGC_CASTED]], align 4
 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK6-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]], [10 x i32]* [[TMP1]], i32* [[TMP2]], i32 [[TMP7]], [10 x i32]* [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[C:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[ARGC:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[D:%.*]]) #[[ATTR0]] {
 // CHECK6-NEXT:  entry:

clang/test/OpenMP/nvptx_force_full_runtime_SPMD_codegen.cpp

 // Test target codegen - host bc file has to be created first.
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - -fopenmp-cuda-force-full-runtime | FileCheck %s
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm-bc %s -o %t-x86-host.bc
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - -fopenmp-cuda-force-full-runtime | FileCheck %s
 // RUN: %clang_cc1 -verify -fopenmp -fexceptions -fcxx-exceptions -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - -fopenmp-cuda-force-full-runtime | FileCheck %s
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 
 // CHECK-NOT: @__omp_offloading_{{.+}}_exec_mode = weak constant i8 1
 
 void foo() {
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target teams distribute parallel for simd
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
 int a;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target teams distribute parallel for lastprivate(a)
   for (int i = 0; i < 10; ++i)
     a = i;
 #pragma omp target teams distribute parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target teams
 #pragma omp distribute parallel for simd
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for simd schedule(static)
   for (int i = 0; i < 10; ++i)
     ;

 #pragma omp target teams
 #pragma omp distribute parallel for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target teams
 #pragma omp distribute parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;

 #pragma omp target teams
 #pragma omp distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target teams
 #pragma omp distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for schedule(static)

 #pragma omp distribute parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(static, 1)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(auto)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(runtime)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel for schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target parallel
 #pragma omp for simd
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel
 #pragma omp for simd schedule(static)
   for (int i = 0; i < 10; ++i)
     ;

 #pragma omp target parallel
 #pragma omp for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target parallel
 #pragma omp for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target
 #pragma omp parallel
 #pragma omp for simd ordered
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp parallel
 #pragma omp for simd schedule(static)

 #pragma omp for simd schedule(dynamic)
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp parallel
 #pragma omp for simd schedule(guided)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 true)
 #pragma omp target
 #pragma omp parallel for
   for (int i = 0; i < 10; ++i)
     ;
 #pragma omp target
 #pragma omp parallel for schedule(static)
   for (int i = 0; i < 10; ++i)
     ;

clang/test/OpenMP/nvptx_multi_target_parallel_codegen.cpp

int main() {
         {  use(); }
 }
 
 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l21
 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP2]], i64 0)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP1]], i64 0)
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l23
 // CHECK1-SAME: () #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    call void @_Z3usev() #[[ATTR6]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3:[0-9]+]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l21
 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP2]], i32 0)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP1]], i32 0)
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l23
 // CHECK2-SAME: () #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    call void @_Z3usev() #[[ATTR6]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3:[0-9]+]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l21
 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK3-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP2]], i32 0)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP1]], i32 0)
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l23
 // CHECK3-SAME: () #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    call void @_Z3usev() #[[ATTR6]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_nested_parallel_codegen.cpp

 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l25
 // CHECK1-SAME: (i32* nonnull align 4 dereferenceable(4) [[C:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK1-NEXT:    store i32* [[C]], i32** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[C_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    call void @_Z3usePi(i32* [[TMP0]]) #[[ATTR5:[0-9]+]]
 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 2)
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP5]], i64 1)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z3usePi
 // CHECK1-SAME: (i32* [[C:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l25
 // CHECK2-SAME: (i32* nonnull align 4 dereferenceable(4) [[C:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i32*, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK2-NEXT:    store i32* [[C]], i32** [[C_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[C_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    call void @_Z3usePi(i32* [[TMP0]]) #[[ATTR5:[0-9]+]]
 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 2)
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP5]], i32 1)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z3usePi
 // CHECK2-SAME: (i32* [[C:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:

 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l25
 // CHECK3-SAME: (i32* nonnull align 4 dereferenceable(4) [[C:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i32*, align 4
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK3-NEXT:    store i32* [[C]], i32** [[C_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[C_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @_Z3usePi(i32* [[TMP0]]) #[[ATTR5:[0-9]+]]
 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 2)
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP5]], i32 1)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z3usePi
 // CHECK3-SAME: (i32* [[C:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_parallel_codegen.cpp

 // CHECK1-SAME: (i64 [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS2:%.*]] = alloca [0 x i8*], align 8
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP2]], i64 0)
 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 0, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP3]], i64 0)
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS2]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** [[TMP4]], i64 0)
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1000
 // CHECK1-NEXT:    [[TMP4:%.*]] = zext i1 [[CMP]] to i32
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 [[TMP4]], i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__3_wrapper to i8*), i8** [[TMP5]], i64 0)
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV2]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP7]] to i32
 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
 // CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[CONV2]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP8]], 1
 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__3
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55
 // CHECK1-SAME: (i64 [[A:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[A1:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK1-NEXT:    [[A_ON_STACK:%.*]] = bitcast i8* [[A1]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[A_ON_STACK]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i32* [[A_ON_STACK]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__4 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__4_wrapper to i8*), i8** [[TMP5]], i64 1)
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A_ON_STACK]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[A_ON_STACK]], align 4
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[A1]], i64 4)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l26
 // CHECK2-SAME: (i32 [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS2:%.*]] = alloca [0 x i8*], align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP2]], i32 0)
 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 0, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** [[TMP3]], i32 0)
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS2]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** [[TMP4]], i32 0)
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 4
 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1000
 // CHECK2-NEXT:    [[TMP4:%.*]] = zext i1 [[CMP]] to i32
 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 [[TMP4]], i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__3_wrapper to i8*), i8** [[TMP5]], i32 0)
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP7]] to i32
 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
 // CHECK2-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__3
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55
 // CHECK2-SAME: (i32 [[A:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[A1:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK2-NEXT:    [[A_ON_STACK:%.*]] = bitcast i8* [[A1]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[A_ON_STACK]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i32* [[A_ON_STACK]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*)* @__omp_outlined__4 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__4_wrapper to i8*), i8** [[TMP5]], i32 1)
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A_ON_STACK]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[A_ON_STACK]], align 4
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[A1]], i32 4)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

clang/test/OpenMP/nvptx_parallel_for_codegen.cpp

 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 8
 // CHECK-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK:       user_code.entry:
 // CHECK-NEXT:    [[D:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK-NEXT:    [[D_ON_STACK:%.*]] = bitcast i8* [[D]] to i32*
 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK-NEXT:    store i32 [[TMP3]], i32* [[D_ON_STACK]], align 4
 // CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast [10 x i32]* [[TMP0]] to i8*
 // CHECK-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 8
 // CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[D_ON_STACK]] to i8*
 // CHECK-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
 // CHECK-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x i32]*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP8]], i64 2)
 // CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 3
 // CHECK-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK-NEXT:    store i32 [[ADD]], i32* [[ARRAYIDX]], align 4
 // CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[D]], i64 4)
-// CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-NEXT:    ret void
 // CHECK:       worker.exit:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
 // CHECK-NEXT:  entry:

 // CHECK-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_UB]], align 4
 // CHECK-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK:       omp.dispatch.end:
 // CHECK-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]])
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK-LABEL: define {{[^@]+}}@__omp_outlined___wrapper
-// CHECK-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[DOTADDR:%.*]] = alloca i16, align 2
 // CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 // CHECK-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 // CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 // CHECK-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
 // CHECK-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 // CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to [10 x i32]**
 // CHECK-NEXT:    [[TMP5:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP4]], align 8
 // CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 1
 // CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32**
 // CHECK-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[TMP7]], align 8
-// CHECK-NEXT:    call void @__omp_outlined__(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP5]], i32* [[TMP8]]) #[[ATTR1:[0-9]+]]
+// CHECK-NEXT:    call void @__omp_outlined__(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP5]], i32* [[TMP8]]) #[[ATTR2:[0-9]+]]
 // CHECK-NEXT:    ret void
 //

clang/test/OpenMP/nvptx_target_codegen.cpp

 // CHECK1-SAME: (i32* [[PTR1:%.*]], i32** nonnull align 8 dereferenceable(8) [[PTR2:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[PTR1_ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[PTR2_ADDR:%.*]] = alloca i32**, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 8
 // CHECK1-NEXT:    store i32* [[PTR1]], i32** [[PTR1_ADDR]], align 8
 // CHECK1-NEXT:    store i32** [[PTR2]], i32*** [[PTR2_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32**, i32*** [[PTR2_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i32** [[PTR1_ADDR]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i32** [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32**, i32**)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP7]], i64 2)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32** nonnull align 8 dereferenceable(8) [[PTR1:%.*]], i32** nonnull align 8 dereferenceable(8) [[PTR2:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP0]], align 8
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l39
 // CHECK1-SAME: () #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l47
 // CHECK1-SAME: (i64 [[AA:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 2
 // CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 8
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l53
 // CHECK1-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP8]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4

 // CHECK1-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP16]] to i32
 // CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
 // CHECK1-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
 // CHECK1-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
 // CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 8 dereferenceable(8) i64* @_ZN2TTIxcEixEi(%struct.TT* nonnull align 8 dereferenceable(16) [[TMP7]], i32 0) #[[ATTR6:[0-9]+]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[CALL]], align 8
 // CHECK1-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP17]], 1
 // CHECK1-NEXT:    store i64 [[ADD22]], i64* [[CALL]], align 8
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN2TTIxcEixEi
 // CHECK1-SAME: (%struct.TT* nonnull align 8 dereferenceable(16) [[THIS:%.*]], i32 [[I:%.*]]) #[[ATTR3:[0-9]+]] comdat align 2 {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
 // CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV2]], align 8
 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP4]] to i32
 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
 // CHECK1-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
 // CHECK1-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l108
 // CHECK1-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP5]] to double
 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
 // CHECK1-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP6:%.*]] = load double, double* [[A4]], align 8
 // CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP6]], 1.000000e+00
 // CHECK1-NEXT:    store double [[INC]], double* [[A4]], align 8
 // CHECK1-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
 // CHECK1-NEXT:    [[TMP7:%.*]] = mul nsw i64 1, [[TMP2]]
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP7]]
 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
 // CHECK1-NEXT:    [[A7:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP8:%.*]] = load double, double* [[A7]], align 8
 // CHECK1-NEXT:    [[CONV8:%.*]] = fptosi double [[TMP8]] to i32
 // CHECK1-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_Z3baziRd(i32 [[CONV8]], double* nonnull align 8 dereferenceable(8) [[A9]]) #[[ATTR6]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z3baziRd
 // CHECK1-SAME: (i32 [[F1:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR3]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[F_ON_STACK]], align 4
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[F]], i64 4)
 // CHECK1-NEXT:    ret i32 [[TMP7]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16unreachable_callv_l142
 // CHECK1-SAME: () #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    call void @_Z6asserti(i32 0) #[[ATTR7:[0-9]+]]
 // CHECK1-NEXT:    unreachable
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 // CHECK1:       1:
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l74
 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
 // CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP4]], 1
 // CHECK1-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[F:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-SAME: (i32* [[PTR1:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR2:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[PTR1_ADDR:%.*]] = alloca i32*, align 4
 // CHECK2-NEXT:    [[PTR2_ADDR:%.*]] = alloca i32**, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
 // CHECK2-NEXT:    store i32* [[PTR1]], i32** [[PTR1_ADDR]], align 4
 // CHECK2-NEXT:    store i32** [[PTR2]], i32*** [[PTR2_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32**, i32*** [[PTR2_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i32** [[PTR1_ADDR]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i32** [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32**, i32**)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP7]], i32 2)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR1:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR2:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP0]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l39
 // CHECK2-SAME: () #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l47
 // CHECK2-SAME: (i32 [[AA:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 2
 // CHECK2-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
 // CHECK2-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 4
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l53
 // CHECK2-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP8]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4

 // CHECK2-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP16]] to i32
 // CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
 // CHECK2-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
 // CHECK2-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 8
 // CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 8 dereferenceable(8) i64* @_ZN2TTIxcEixEi(%struct.TT* nonnull align 8 dereferenceable(16) [[TMP7]], i32 0) #[[ATTR6:[0-9]+]]
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[CALL]], align 8
 // CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i64 [[TMP17]], 1
 // CHECK2-NEXT:    store i64 [[ADD21]], i64* [[CALL]], align 8
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_ZN2TTIxcEixEi
 // CHECK2-SAME: (%struct.TT* nonnull align 8 dereferenceable(16) [[THIS:%.*]], i32 [[I:%.*]]) #[[ATTR3:[0-9]+]] comdat align 2 {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
 // CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV1]], align 4
 // CHECK2-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP4]] to i32
 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
 // CHECK2-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
 // CHECK2-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l108
 // CHECK2-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
 // CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    store double [[ADD]], double* [[A]], align 8
 // CHECK2-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP6:%.*]] = load double, double* [[A3]], align 8
 // CHECK2-NEXT:    [[INC:%.*]] = fadd double [[TMP6]], 1.000000e+00
 // CHECK2-NEXT:    store double [[INC]], double* [[A3]], align 8
 // CHECK2-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
 // CHECK2-NEXT:    [[TMP7:%.*]] = mul nsw i32 1, [[TMP2]]
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP7]]
 // CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
 // CHECK2-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
 // CHECK2-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP8:%.*]] = load double, double* [[A6]], align 8
 // CHECK2-NEXT:    [[CONV7:%.*]] = fptosi double [[TMP8]] to i32
 // CHECK2-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[CALL:%.*]] = call i32 @_Z3baziRd(i32 [[CONV7]], double* nonnull align 8 dereferenceable(8) [[A8]]) #[[ATTR6]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z3baziRd
 // CHECK2-SAME: (i32 [[F1:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR3]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[F_ON_STACK]], align 4
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[F]], i32 4)
 // CHECK2-NEXT:    ret i32 [[TMP7]]
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16unreachable_callv_l142
 // CHECK2-SAME: () #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    call void @_Z6asserti(i32 0) #[[ATTR7:[0-9]+]]
 // CHECK2-NEXT:    unreachable
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 // CHECK2:       1:
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l74
 // CHECK2-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
 // CHECK2-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP4]], 1
 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[F:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-SAME: (i32* [[PTR1:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR2:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[PTR1_ADDR:%.*]] = alloca i32*, align 4
 // CHECK3-NEXT:    [[PTR2_ADDR:%.*]] = alloca i32**, align 4
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
 // CHECK3-NEXT:    store i32* [[PTR1]], i32** [[PTR1_ADDR]], align 4
 // CHECK3-NEXT:    store i32** [[PTR2]], i32*** [[PTR2_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32**, i32*** [[PTR2_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i32** [[PTR1_ADDR]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i32** [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32**, i32**)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP7]], i32 2)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR1:%.*]], i32** nonnull align 4 dereferenceable(4) [[PTR2:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP0]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l39
 // CHECK3-SAME: () #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l47
 // CHECK3-SAME: (i32 [[AA:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 2
 // CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
 // CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 4
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l53
 // CHECK3-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP8]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4

 // CHECK3-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP16]] to i32
 // CHECK3-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
 // CHECK3-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
 // CHECK3-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 8
 // CHECK3-NEXT:    [[CALL:%.*]] = call nonnull align 8 dereferenceable(8) i64* @_ZN2TTIxcEixEi(%struct.TT* nonnull align 8 dereferenceable(16) [[TMP7]], i32 0) #[[ATTR6:[0-9]+]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[CALL]], align 8
 // CHECK3-NEXT:    [[ADD21:%.*]] = add nsw i64 [[TMP17]], 1
 // CHECK3-NEXT:    store i64 [[ADD21]], i64* [[CALL]], align 8
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN2TTIxcEixEi
 // CHECK3-SAME: (%struct.TT* nonnull align 8 dereferenceable(16) [[THIS:%.*]], i32 [[I:%.*]]) #[[ATTR3:[0-9]+]] comdat align 2 {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
 // CHECK3-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
 // CHECK3-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV1]], align 4
 // CHECK3-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP4]] to i32
 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
 // CHECK3-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
 // CHECK3-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l108
 // CHECK3-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 8
 // CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP6:%.*]] = load double, double* [[A3]], align 8
 // CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP6]], 1.000000e+00
 // CHECK3-NEXT:    store double [[INC]], double* [[A3]], align 8
 // CHECK3-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
 // CHECK3-NEXT:    [[TMP7:%.*]] = mul nsw i32 1, [[TMP2]]
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP7]]
 // CHECK3-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
 // CHECK3-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
 // CHECK3-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP8:%.*]] = load double, double* [[A6]], align 8
 // CHECK3-NEXT:    [[CONV7:%.*]] = fptosi double [[TMP8]] to i32
 // CHECK3-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_Z3baziRd(i32 [[CONV7]], double* nonnull align 8 dereferenceable(8) [[A8]]) #[[ATTR6]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z3baziRd
 // CHECK3-SAME: (i32 [[F1:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR3]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[F_ON_STACK]], align 4
 // CHECK3-NEXT:    call void @__kmpc_free_shared(i8* [[F]], i32 4)
 // CHECK3-NEXT:    ret i32 [[TMP7]]
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16unreachable_callv_l142
 // CHECK3-SAME: () #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    call void @_Z6asserti(i32 0) #[[ATTR7:[0-9]+]]
 // CHECK3-NEXT:    unreachable
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 // CHECK3:       1:
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l74
 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 1
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP3]] to i32
 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
 // CHECK3-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
 // CHECK3-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP4]], 1
 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[F:%.*]], double* nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_target_parallel_codegen.cpp

 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK1-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK1-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 8
 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i64 3)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:

 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK2-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK2-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i32 3)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:

 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK3-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK3-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK3-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i32 3)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:

 //
 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK4-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 8
 // CHECK4-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK4-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK4-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK4-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK4-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK4-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 8
 // CHECK4-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 8
 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 8
 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 8
 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK4-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i64 3)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK4-NEXT:  entry:

 //
 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK5-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK5-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK5-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK5-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK5-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK5-NEXT:  entry:

 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK5-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK5-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK5-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK5-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK5-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i32 3)
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK5-NEXT:  entry:

 //
 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l30
 // CHECK6-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK6-NEXT:  entry:
 // CHECK6-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK6-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK6-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK6-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK6-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK6-NEXT:  entry:

 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK6-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK6-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK6-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK6-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
 // CHECK6-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK6-NEXT:    [[TMP10:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
 // CHECK6-NEXT:    [[TMP11:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK6-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP11]], i32 3)
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK6-NEXT:  entry:

clang/test/OpenMP/nvptx_target_parallel_num_threads_codegen.cpp

 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK1-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 8
 // CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK1-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i64 3)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:

 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK2-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK2-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 4
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i32 3)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:

 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK3-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK3-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK3-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i32 3)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:

 //
 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK4-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 8
 // CHECK4-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
 // CHECK4-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK4-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
 // CHECK4-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK4-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK4-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
-// CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK4-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 8
 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK4-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 8
 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK4-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i64 3)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK4-NEXT:  entry:

 //
 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK5-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK5-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK5-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK5-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK5-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK5-NEXT:  entry:

 // CHECK5-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK5-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK5-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 4
 // CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK5-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK5-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 4
 // CHECK5-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK5-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i32 3)
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK5-NEXT:  entry:

 //
 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l25
 // CHECK6-SAME: (i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK6-NEXT:  entry:
 // CHECK6-NEXT:    [[AA_ADDR:%.*]] = alloca i16*, align 4
 // CHECK6-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
 // CHECK6-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP0]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
 // CHECK6-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK6-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i16*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK6-NEXT:  entry:

 // CHECK6-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x i8*], align 4
 // CHECK6-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    store i16* [[AA]], i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[AA_ADDR]], align 4
 // CHECK6-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK6-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP0]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK6-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK6-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP1]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP9]], i8** [[TMP8]], align 4
 // CHECK6-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK6-NEXT:    [[TMP11:%.*]] = bitcast [10 x i32]* [[TMP2]] to i8*
 // CHECK6-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 4
 // CHECK6-NEXT:    [[TMP12:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 // CHECK6-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*, [10 x i32]*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP12]], i32 3)
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
 // CHECK6-NEXT:  entry:

clang/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
 // Test target codegen - host bc file has to be created first.
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple i386-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm-bc %s -o %t-x86-host.bc
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -fexceptions -fcxx-exceptions -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 
 // Check for the data transfer medium in shared memory to transfer the reduction list to the first warp.
-// CHECK-DAG: [[TRANSFER_STORAGE:@.+]] = weak addrspace([[SHARED_ADDRSPACE:[0-9]+]]) global [32 x i32]
 
 // Check that the execution mode of all 3 target regions is set to Spmd Mode.
-// CHECK-DAG: {{@__omp_offloading_.+l27}}_exec_mode = weak constant i8 2
-// CHECK-DAG: {{@__omp_offloading_.+l32}}_exec_mode = weak constant i8 2
-// CHECK-DAG: {{@__omp_offloading_.+l38}}_exec_mode = weak constant i8 2
 
 template<typename tx>
 tx ftemplate(int n) {
   int a;
   short b;
   tx c;
   float d;
   double e;

 int bar(int n){
   int a = 0;
 
   a += ftemplate<char>(n);
 
   return a;
 }
 
-// CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l27}}(
-//
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 true)
-//
-//
 // define internal void [[PFN]](
-// CHECK: store double {{[0\.e\+]+}}, double* [[E:%.+]], align
-// CHECK: [[EV:%.+]] = load double, double* [[E]], align
-// CHECK: [[ADD:%.+]] = fadd double [[EV]], 5
-// CHECK: store double [[ADD]], double* [[E]], align
-// CHECK: [[PTR1:%.+]] = getelementptr inbounds [[RLT:.+]], [1 x i8*]* [[RL:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[E_CAST:%.+]] = bitcast double* [[E]] to i8*
-// CHECK: store i8* [[E_CAST]], i8** [[PTR1]], align
-// CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8*
-// CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 1, i{{32|64}} {{4|8}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]])
-// CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1
-// CHECK: br i1 [[CMP]], label
-
-// CHECK: [[E_INV:%.+]] = load double, double* [[E_IN:%.+]], align
-// CHECK: [[EV:%.+]] = load double, double* [[E]], align
-// CHECK: [[ADD:%.+]] = fadd double [[E_INV]], [[EV]]
-// CHECK: store double [[ADD]], double* [[E_IN]], align
-// CHECK: call void @__kmpc_nvptx_end_reduce_nowait(
-// CHECK: br label
-//
-// CHECK: ret
 
-//
+
 // Reduction function
-// CHECK: define internal void [[REDUCTION_FUNC:@.+]](i8* %0, i8* %1)
-// CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_RHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
-// CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to double*
-//
-// CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_LHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
-// CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to double*
-//
-// CHECK: [[VAR_LHS_VAL:%.+]] = load double, double* [[VAR_LHS]],
-// CHECK: [[VAR_RHS_VAL:%.+]] = load double, double* [[VAR_RHS]],
-// CHECK: [[RES:%.+]] = fadd double [[VAR_LHS_VAL]], [[VAR_RHS_VAL]]
-// CHECK: store double [[RES]], double* [[VAR_LHS]],
-// CHECK: ret void
 
-//
 // Shuffle and reduce function
-// CHECK: define internal void [[SHUFFLE_REDUCE_FN]](i8* %0, i16 {{.*}}, i16 {{.*}}, i16 {{.*}})
-// CHECK: [[REMOTE_RED_LIST:%.+]] = alloca [[RLT]], align
-// CHECK: [[REMOTE_ELT:%.+]] = alloca double
-//
-// CHECK: [[LANEID:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[LANEOFFSET:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[ALGVER:%.+]] = load i16, i16* {{.+}}, align
-//
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to double*
-//
-// CHECK: [[ELT_CAST:%.+]] = bitcast double* [[ELT]] to i64*
-// CHECK: [[REMOTE_ELT_CAST:%.+]] = bitcast double* [[REMOTE_ELT]] to i64*
-// CHECK: [[ELT_VAL:%.+]] = load i64, i64* [[ELT_CAST]], align
-// CHECK: [[WS32:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-// CHECK: [[WS:%.+]] = trunc i32 [[WS32]] to i16
-// CHECK: [[REMOTE_ELT_VAL64:%.+]] = call i64 @__kmpc_shuffle_int64(i64 [[ELT_VAL]], i16 [[LANEOFFSET]], i16 [[WS]])
-//
-// CHECK: store i64 [[REMOTE_ELT_VAL64]], i64* [[REMOTE_ELT_CAST]], align
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = bitcast double* [[REMOTE_ELT]] to i8*
-// CHECK: store i8* [[REMOTE_ELT_VOID]], i8** [[REMOTE_ELT_REF]], align
-//
 // Condition to reduce
-// CHECK: [[CONDALG0:%.+]] = icmp eq i16 [[ALGVER]], 0
-//
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp ult i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[CONDALG1:%.+]] = and i1 [[COND1]], [[COND2]]
-//
-// CHECK: [[COND3:%.+]] = icmp eq i16 [[ALGVER]], 2
-// CHECK: [[COND4:%.+]] = and i16 [[LANEID]], 1
-// CHECK: [[COND5:%.+]] = icmp eq i16 [[COND4]], 0
-// CHECK: [[COND6:%.+]] = and i1 [[COND3]], [[COND5]]
-// CHECK: [[COND7:%.+]] = icmp sgt i16 [[LANEOFFSET]], 0
-// CHECK: [[CONDALG2:%.+]] = and i1 [[COND6]], [[COND7]]
-//
-// CHECK: [[COND8:%.+]] = or i1 [[CONDALG0]], [[CONDALG1]]
-// CHECK: [[SHOULD_REDUCE:%.+]] = or i1 [[COND8]], [[CONDALG2]]
-// CHECK: br i1 [[SHOULD_REDUCE]], label {{%?}}[[DO_REDUCE:.+]], label {{%?}}[[REDUCE_ELSE:.+]]
-//
-// CHECK: [[DO_REDUCE]]
-// CHECK: [[RED_LIST1_VOID:%.+]] = bitcast [[RLT]]* [[RED_LIST]] to i8*
-// CHECK: [[RED_LIST2_VOID:%.+]] = bitcast [[RLT]]* [[REMOTE_RED_LIST]] to i8*
-// CHECK: call void [[REDUCTION_FUNC]](i8* [[RED_LIST1_VOID]], i8* [[RED_LIST2_VOID]])
-// CHECK: br label {{%?}}[[REDUCE_CONT:.+]]
-//
-// CHECK: [[REDUCE_ELSE]]
-// CHECK: br label {{%?}}[[REDUCE_CONT]]
-//
-// CHECK: [[REDUCE_CONT]]
 // Now check if we should just copy over the remote reduction list
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp uge i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[SHOULD_COPY:%.+]] = and i1 [[COND1]], [[COND2]]
-// CHECK: br i1 [[SHOULD_COPY]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
-// CHECK: [[DO_COPY]]
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = load i8*, i8** [[REMOTE_ELT_REF]],
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT:%.+]] = bitcast i8* [[REMOTE_ELT_VOID]] to double*
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to double*
-// CHECK: [[REMOTE_ELT_VAL:%.+]] = load double, double* [[REMOTE_ELT]], align
-// CHECK: store double [[REMOTE_ELT_VAL]], double* [[ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
-// CHECK: [[COPY_CONT]]
-// CHECK: void
 
-//
 // Inter warp copy function
-// CHECK: define internal void [[WARP_COPY_FN]](i8* %0, i32 %1)
-// CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31
-// CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5
-// CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]*
-// CHECK: store i32 0, i32* [[CNT_ADDR:%.+]],
-// CHECK: br label
-// CHECK: [[CNT:%.+]] = load i32, i32* [[CNT_ADDR]],
-// CHECK: [[DONE_COPY:%.+]] = icmp ult i32 [[CNT]], 2
-// CHECK: br i1 [[DONE_COPY]], label
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
-// CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
 // [[DO_COPY]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[BASE_ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[ELT:%.+]] = getelementptr i32, i32* [[BASE_ELT]], i32 [[CNT]]
-//
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]]
-// CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]],
-// CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]],
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
 // Barrier after copy to shared memory storage medium.
-// CHECK: [[COPY_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
-//
 // Read into warp 0.
-// CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
-// CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]]
-//
-// CHECK: [[DO_READ]]
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT_BASE:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[ELT:%.+]] = getelementptr i32, i32* [[ELT_BASE]], i32 [[CNT]]
-// CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]],
-// CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]],
-// CHECK: br label {{%?}}[[READ_CONT:.+]]
-//
-// CHECK: [[READ_ELSE]]
-// CHECK: br label {{%?}}[[READ_CONT]]
-//
-// CHECK: [[READ_CONT]]
-// CHECK: [[NEXT:%.+]] = add nsw i32 [[CNT]], 1
-// CHECK: store i32 [[NEXT]], i32* [[CNT_ADDR]],
-// CHECK: br label
-// CHECK: ret
 
-// CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l32}}(
-//
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 true)
-//
-//
 // define internal void [[PFN1]](
-// CHECK: store float {{1\.[0e\+]+}}, float* [[D:%.+]], align
-// CHECK: [[C_VAL:%.+]] = load i8, i8* [[C:%.+]], align
-// CHECK: [[CONV:%.+]] = sext i8 [[C_VAL]] to i32
-// CHECK: [[XOR:%.+]] = xor i32 [[CONV]], 2
-// CHECK: [[TRUNC:%.+]] = trunc i32 [[XOR]] to i8
-// CHECK: store i8 [[TRUNC]], i8* [[C]], align
-// CHECK: [[DV:%.+]] = load float, float* [[D]], align
-// CHECK: [[MUL:%.+]] = fmul float [[DV]], {{[0-9e\.\+]+}}
-// CHECK: store float [[MUL]], float* [[D]], align
-// CHECK: [[PTR1:%.+]] = getelementptr inbounds [[RLT:.+]], [2 x i8*]* [[RL:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: store i8* [[C]], i8** [[PTR1]], align
-// CHECK: [[PTR2:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RL]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[D_CAST:%.+]] = bitcast float* [[D]] to i8*
-// CHECK: store i8* [[D_CAST]], i8** [[PTR2]], align
-// CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8*
-// CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]])
-// CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1
-// CHECK: br i1 [[CMP]], label
-// CHECK: [[C_INV8:%.+]] = load i8, i8* [[C_IN:%.+]], align
-// CHECK: [[C_INV:%.+]] = sext i8 [[C_INV8]] to i32
-// CHECK: [[CV8:%.+]] = load i8, i8* [[C]], align
-// CHECK: [[CV:%.+]] = sext i8 [[CV8]] to i32
-// CHECK: [[XOR:%.+]] = xor i32 [[C_INV]], [[CV]]
-// CHECK: [[TRUNC:%.+]] = trunc i32 [[XOR]] to i8
-// CHECK: store i8 [[TRUNC]], i8* [[C_IN]], align
-// CHECK: [[D_INV:%.+]] = load float, float* [[D_IN:%.+]], align
-// CHECK: [[DV:%.+]] = load float, float* [[D]], align
-// CHECK: [[MUL:%.+]] = fmul float [[D_INV]], [[DV]]
-// CHECK: store float [[MUL]], float* [[D_IN]], align
-// CHECK: call void @__kmpc_nvptx_end_reduce_nowait(
-// CHECK: br label
-//
-// CHECK: ret
 
-//
 // Reduction function
-// CHECK: define internal void [[REDUCTION_FUNC:@.+]](i8* %0, i8* %1)
-// CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_RHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR1_RHS:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
-//
-// CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_LHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR1_LHS:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
-//
-// CHECK: [[VAR2_RHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_RHS]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[VAR2_RHS_VOID:%.+]] = load i8*, i8** [[VAR2_RHS_REF]],
-// CHECK: [[VAR2_RHS:%.+]] = bitcast i8* [[VAR2_RHS_VOID]] to float*
-//
-// CHECK: [[VAR2_LHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_LHS]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[VAR2_LHS_VOID:%.+]] = load i8*, i8** [[VAR2_LHS_REF]],
-// CHECK: [[VAR2_LHS:%.+]] = bitcast i8* [[VAR2_LHS_VOID]] to float*
-//
-// CHECK: [[VAR1_LHS_VAL8:%.+]] = load i8, i8* [[VAR1_LHS]],
-// CHECK: [[VAR1_LHS_VAL:%.+]] = sext i8 [[VAR1_LHS_VAL8]] to i32
-// CHECK: [[VAR1_RHS_VAL8:%.+]] = load i8, i8* [[VAR1_RHS]],
-// CHECK: [[VAR1_RHS_VAL:%.+]] = sext i8 [[VAR1_RHS_VAL8]] to i32
-// CHECK: [[XOR:%.+]] = xor i32 [[VAR1_LHS_VAL]], [[VAR1_RHS_VAL]]
-// CHECK: [[RES:%.+]] = trunc i32 [[XOR]] to i8
-// CHECK: store i8 [[RES]], i8* [[VAR1_LHS]],
-//
-// CHECK: [[VAR2_LHS_VAL:%.+]] = load float, float* [[VAR2_LHS]],
-// CHECK: [[VAR2_RHS_VAL:%.+]] = load float, float* [[VAR2_RHS]],
-// CHECK: [[RES:%.+]] = fmul float [[VAR2_LHS_VAL]], [[VAR2_RHS_VAL]]
-// CHECK: store float [[RES]], float* [[VAR2_LHS]],
-// CHECK: ret void
 
-//
 // Shuffle and reduce function
-// CHECK: define internal void [[SHUFFLE_REDUCE_FN]](i8* %0, i16 {{.*}}, i16 {{.*}}, i16 {{.*}})
-// CHECK: [[REMOTE_RED_LIST:%.+]] = alloca [[RLT]], align
-// CHECK: [[REMOTE_ELT1:%.+]] = alloca i8
-// CHECK: [[REMOTE_ELT2:%.+]] = alloca float
-//
-// CHECK: [[LANEID:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[LANEOFFSET:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[ALGVER:%.+]] = load i16, i16* {{.+}}, align
-//
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VAL:%.+]] = load i8, i8* [[ELT_VOID]], align
-//
-// CHECK: [[ELT_CAST:%.+]] = sext i8 [[ELT_VAL]] to i32
-// CHECK: [[WS32:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-// CHECK: [[WS:%.+]] = trunc i32 [[WS32]] to i16
-// CHECK: [[REMOTE_ELT1_VAL32:%.+]] = call i32 @__kmpc_shuffle_int32(i32 [[ELT_CAST]], i16 [[LANEOFFSET]], i16 [[WS]])
-// CHECK: [[REMOTE_ELT1_VAL:%.+]] = trunc i32 [[REMOTE_ELT1_VAL32]] to i8
-//
-// CHECK: store i8 [[REMOTE_ELT1_VAL]], i8* [[REMOTE_ELT1]], align
-// CHECK: store i8* [[REMOTE_ELT1]], i8** [[REMOTE_ELT_REF]], align
-//
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to float*
-//
-// CHECK: [[ELT_CAST:%.+]] = bitcast float* [[ELT]] to i32*
-// CHECK: [[REMOTE_ELT2_CAST:%.+]] = bitcast float* [[REMOTE_ELT2]] to i32*
-// CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT_CAST]], align
-// CHECK: [[WS32:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-// CHECK: [[WS:%.+]] = trunc i32 [[WS32]] to i16
-// CHECK: [[REMOTE_ELT2_VAL32:%.+]] = call i32 @__kmpc_shuffle_int32(i32 [[ELT_VAL]], i16 [[LANEOFFSET]], i16 [[WS]])
-//
-// CHECK: store i32 [[REMOTE_ELT2_VAL32]], i32* [[REMOTE_ELT2_CAST]], align
-// CHECK: [[REMOTE_ELT2C:%.+]] = bitcast float* [[REMOTE_ELT2]] to i8*
-// CHECK: store i8* [[REMOTE_ELT2C]], i8** [[REMOTE_ELT_REF]], align
-//
 // Condition to reduce
-// CHECK: [[CONDALG0:%.+]] = icmp eq i16 [[ALGVER]], 0
-//
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp ult i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[CONDALG1:%.+]] = and i1 [[COND1]], [[COND2]]
-//
-// CHECK: [[COND3:%.+]] = icmp eq i16 [[ALGVER]], 2
-// CHECK: [[COND4:%.+]] = and i16 [[LANEID]], 1
-// CHECK: [[COND5:%.+]] = icmp eq i16 [[COND4]], 0
-// CHECK: [[COND6:%.+]] = and i1 [[COND3]], [[COND5]]
-// CHECK: [[COND7:%.+]] = icmp sgt i16 [[LANEOFFSET]], 0
-// CHECK: [[CONDALG2:%.+]] = and i1 [[COND6]], [[COND7]]
-//
-// CHECK: [[COND8:%.+]] = or i1 [[CONDALG0]], [[CONDALG1]]
-// CHECK: [[SHOULD_REDUCE:%.+]] = or i1 [[COND8]], [[CONDALG2]]
-// CHECK: br i1 [[SHOULD_REDUCE]], label {{%?}}[[DO_REDUCE:.+]], label {{%?}}[[REDUCE_ELSE:.+]]
-//
-// CHECK: [[DO_REDUCE]]
-// CHECK: [[RED_LIST1_VOID:%.+]] = bitcast [[RLT]]* [[RED_LIST]] to i8*
-// CHECK: [[RED_LIST2_VOID:%.+]] = bitcast [[RLT]]* [[REMOTE_RED_LIST]] to i8*
-// CHECK: call void [[REDUCTION_FUNC]](i8* [[RED_LIST1_VOID]], i8* [[RED_LIST2_VOID]])
-// CHECK: br label {{%?}}[[REDUCE_CONT:.+]]
-//
-// CHECK: [[REDUCE_ELSE]]
-// CHECK: br label {{%?}}[[REDUCE_CONT]]
-//
-// CHECK: [[REDUCE_CONT]]
 // Now check if we should just copy over the remote reduction list
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp uge i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[SHOULD_COPY:%.+]] = and i1 [[COND1]], [[COND2]]
-// CHECK: br i1 [[SHOULD_COPY]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
-// CHECK: [[DO_COPY]]
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = load i8*, i8** [[REMOTE_ELT_REF]],
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_VAL:%.+]] = load i8, i8* [[REMOTE_ELT_VOID]], align
-// CHECK: store i8 [[REMOTE_ELT_VAL]], i8* [[ELT_VOID]], align
-//
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = load i8*, i8** [[REMOTE_ELT_REF]],
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT:%.+]] = bitcast i8* [[REMOTE_ELT_VOID]] to float*
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to float*
-// CHECK: [[REMOTE_ELT_VAL:%.+]] = load float, float* [[REMOTE_ELT]], align
-// CHECK: store float [[REMOTE_ELT_VAL]], float* [[ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
-// CHECK: [[COPY_CONT]]
-// CHECK: void
 
-//
 // Inter warp copy function
-// CHECK: define internal void [[WARP_COPY_FN]](i8* %0, i32 %1)
-// CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31
-// CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5
-// CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]*
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
-// CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
 // [[DO_COPY]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-//
-// CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]]
-// CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i8 addrspace([[SHARED_ADDRSPACE]])*
-// CHECK: [[ELT_VAL:%.+]] = load i8, i8* [[ELT_VOID]], align
-// CHECK: store volatile i8 [[ELT_VAL]], i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
 // Barrier after copy to shared memory storage medium.
-// CHECK: [[COPY_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
-//
 // Read into warp 0.
-// CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
-// CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]]
-//
-// CHECK: [[DO_READ]]
-// CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]]
-// CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i8 addrspace([[SHARED_ADDRSPACE]])*
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i8, i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: store i8 [[MEDIUM_ELT_VAL]], i8* [[ELT_VOID]], align
-// CHECK: br label {{%?}}[[READ_CONT:.+]]
-//
-// CHECK: [[READ_ELSE]]
-// CHECK: br label {{%?}}[[READ_CONT]]
-//
-// CHECK: [[READ_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
-// CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
 // [[DO_COPY]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-//
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]]
-// CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], align
-// CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
 // Barrier after copy to shared memory storage medium.
-// CHECK: [[COPY_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
-//
 // Read into warp 0.
-// CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
-// CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]]
-//
-// CHECK: [[DO_READ]]
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]], align
-// CHECK: br label {{%?}}[[READ_CONT:.+]]
-//
-// CHECK: [[READ_ELSE]]
-// CHECK: br label {{%?}}[[READ_CONT]]
-//
-// CHECK: [[READ_CONT]]
-// CHECK: ret
 
-// CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l38}}(
-//
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 true)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 true)
-//
-//
 // define internal void [[PFN2]](
-// CHECK: store i32 0, i32* [[A:%.+]], align
-// CHECK: store i16 -32768, i16* [[B:%.+]], align
-// CHECK: [[A_VAL:%.+]] = load i32, i32* [[A:%.+]], align
-// CHECK: [[OR:%.+]] = or i32 [[A_VAL]], 1
-// CHECK: store i32 [[OR]], i32* [[A]], align
-// CHECK: [[BV16:%.+]] = load i16, i16* [[B]], align
-// CHECK: [[BV:%.+]] = sext i16 [[BV16]] to i32
-// CHECK: [[CMP:%.+]] = icmp sgt i32 99, [[BV]]
-// CHECK: br i1 [[CMP]], label {{%?}}[[DO_MAX:.+]], label {{%?}}[[MAX_ELSE:.+]]
-//
-// CHECK: [[DO_MAX]]
-// CHECK: br label {{%?}}[[MAX_CONT:.+]]
-//
-// CHECK: [[MAX_ELSE]]
-// CHECK: [[BV:%.+]] = load i16, i16* [[B]], align
-// CHECK: [[MAX:%.+]] = sext i16 [[BV]] to i32
-// CHECK: br label {{%?}}[[MAX_CONT]]
-//
-// CHECK: [[MAX_CONT]]
-// CHECK: [[B_LVALUE:%.+]] = phi i32 [ 99, %[[DO_MAX]] ], [ [[MAX]], %[[MAX_ELSE]] ]
-// CHECK: [[TRUNC:%.+]] = trunc i32 [[B_LVALUE]] to i16
-// CHECK: store i16 [[TRUNC]], i16* [[B]], align
-// CHECK: [[PTR1:%.+]] = getelementptr inbounds [[RLT:.+]], [2 x i8*]* [[RL:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[A_CAST:%.+]] = bitcast i32* [[A]] to i8*
-// CHECK: store i8* [[A_CAST]], i8** [[PTR1]], align
-// CHECK: [[PTR2:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RL]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[B_CAST:%.+]] = bitcast i16* [[B]] to i8*
-// CHECK: store i8* [[B_CAST]], i8** [[PTR2]], align
-// CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8*
-// CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]])
-// CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1
-// CHECK: br i1 [[CMP]], label
-
-// CHECK: [[A_INV:%.+]] = load i32, i32* [[A_IN:%.+]], align
-// CHECK: [[AV:%.+]] = load i32, i32* [[A]], align
-// CHECK: [[OR:%.+]] = or i32 [[A_INV]], [[AV]]
-// CHECK: store i32 [[OR]], i32* [[A_IN]], align
-// CHECK: [[B_INV16:%.+]] = load i16, i16* [[B_IN:%.+]], align
-// CHECK: [[B_INV:%.+]] = sext i16 [[B_INV16]] to i32
-// CHECK: [[BV16:%.+]] = load i16, i16* [[B]], align
-// CHECK: [[BV:%.+]] = sext i16 [[BV16]] to i32
-// CHECK: [[CMP:%.+]] = icmp sgt i32 [[B_INV]], [[BV]]
-// CHECK: br i1 [[CMP]], label {{%?}}[[DO_MAX:.+]], label {{%?}}[[MAX_ELSE:.+]]
-//
-// CHECK: [[DO_MAX]]
-// CHECK: [[MAX1:%.+]] = load i16, i16* [[B_IN]], align
-// CHECK: br label {{%?}}[[MAX_CONT:.+]]
-//
-// CHECK: [[MAX_ELSE]]
-// CHECK: [[MAX2:%.+]] = load i16, i16* [[B]], align
-// CHECK: br label {{%?}}[[MAX_CONT]]
-//
-// CHECK: [[MAX_CONT]]
-// CHECK: [[B_MAX:%.+]] = phi i16 [ [[MAX1]], %[[DO_MAX]] ], [ [[MAX2]], %[[MAX_ELSE]] ]
-// CHECK: store i16 [[B_MAX]], i16* [[B_IN]], align
-// CHECK: call void @__kmpc_nvptx_end_reduce_nowait(
-// CHECK: br label
-//
-// CHECK: ret
 
-//
+
 // Reduction function
-// CHECK: define internal void [[REDUCTION_FUNC:@.+]](i8* %0, i8* %1)
-// CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_RHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
-// CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to i32*
-//
-// CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_LHS:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
-// CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to i32*
-//
-// CHECK: [[VAR2_RHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_RHS]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[VAR2_RHS_VOID:%.+]] = load i8*, i8** [[VAR2_RHS_REF]],
-// CHECK: [[VAR2_RHS:%.+]] = bitcast i8* [[VAR2_RHS_VOID]] to i16*
-//
-// CHECK: [[VAR2_LHS_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST_LHS]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[VAR2_LHS_VOID:%.+]] = load i8*, i8** [[VAR2_LHS_REF]],
-// CHECK: [[VAR2_LHS:%.+]] = bitcast i8* [[VAR2_LHS_VOID]] to i16*
-//
-// CHECK: [[VAR1_LHS_VAL:%.+]] = load i32, i32* [[VAR1_LHS]],
-// CHECK: [[VAR1_RHS_VAL:%.+]] = load i32, i32* [[VAR1_RHS]],
-// CHECK: [[OR:%.+]] = or i32 [[VAR1_LHS_VAL]], [[VAR1_RHS_VAL]]
-// CHECK: store i32 [[OR]], i32* [[VAR1_LHS]],
-//
-// CHECK: [[VAR2_LHS_VAL16:%.+]] = load i16, i16* [[VAR2_LHS]],
-// CHECK: [[VAR2_LHS_VAL:%.+]] = sext i16 [[VAR2_LHS_VAL16]] to i32
-// CHECK: [[VAR2_RHS_VAL16:%.+]] = load i16, i16* [[VAR2_RHS]],
-// CHECK: [[VAR2_RHS_VAL:%.+]] = sext i16 [[VAR2_RHS_VAL16]] to i32
-//
-// CHECK: [[CMP:%.+]] = icmp sgt i32 [[VAR2_LHS_VAL]], [[VAR2_RHS_VAL]]
-// CHECK: br i1 [[CMP]], label {{%?}}[[DO_MAX:.+]], label {{%?}}[[MAX_ELSE:.+]]
-//
-// CHECK: [[DO_MAX]]
-// CHECK: [[MAX1:%.+]] = load i16, i16* [[VAR2_LHS]], align
-// CHECK: br label {{%?}}[[MAX_CONT:.+]]
-//
-// CHECK: [[MAX_ELSE]]
-// CHECK: [[MAX2:%.+]] = load i16, i16* [[VAR2_RHS]], align
-// CHECK: br label {{%?}}[[MAX_CONT]]
-//
-// CHECK: [[MAX_CONT]]
-// CHECK: [[MAXV:%.+]] = phi i16 [ [[MAX1]], %[[DO_MAX]] ], [ [[MAX2]], %[[MAX_ELSE]] ]
-// CHECK: store i16 [[MAXV]], i16* [[VAR2_LHS]],
-// CHECK: ret void
 
-//
 // Shuffle and reduce function
-// CHECK: define internal void [[SHUFFLE_REDUCE_FN]](i8* %0, i16 {{.*}}, i16 {{.*}}, i16 {{.*}})
-// CHECK: [[REMOTE_RED_LIST:%.+]] = alloca [[RLT]], align
-// CHECK: [[REMOTE_ELT1:%.+]] = alloca i32
-// CHECK: [[REMOTE_ELT2:%.+]] = alloca i16
-//
-// CHECK: [[LANEID:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[LANEOFFSET:%.+]] = load i16, i16* {{.+}}, align
-// CHECK: [[ALGVER:%.+]] = load i16, i16* {{.+}}, align
-//
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], align
-//
-// CHECK: [[WS32:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-// CHECK: [[WS:%.+]] = trunc i32 [[WS32]] to i16
-// CHECK: [[REMOTE_ELT1_VAL:%.+]] = call i32 @__kmpc_shuffle_int32(i32 [[ELT_VAL]], i16 [[LANEOFFSET]], i16 [[WS]])
-//
-// CHECK: store i32 [[REMOTE_ELT1_VAL]], i32* [[REMOTE_ELT1]], align
-// CHECK: [[REMOTE_ELT1C:%.+]] = bitcast i32* [[REMOTE_ELT1]] to i8*
-// CHECK: store i8* [[REMOTE_ELT1C]], i8** [[REMOTE_ELT_REF]], align
-//
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16*
-// CHECK: [[ELT_VAL:%.+]] = load i16, i16* [[ELT]], align
-//
-// CHECK: [[ELT_CAST:%.+]] = sext i16 [[ELT_VAL]] to i32
-// CHECK: [[WS32:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-// CHECK: [[WS:%.+]] = trunc i32 [[WS32]] to i16
-// CHECK: [[REMOTE_ELT2_VAL32:%.+]] = call i32 @__kmpc_shuffle_int32(i32 [[ELT_CAST]], i16 [[LANEOFFSET]], i16 [[WS]])
-// CHECK: [[REMOTE_ELT2_VAL:%.+]] = trunc i32 [[REMOTE_ELT2_VAL32]] to i16
-//
-// CHECK: store i16 [[REMOTE_ELT2_VAL]], i16* [[REMOTE_ELT2]], align
-// CHECK: [[REMOTE_ELT2C:%.+]] = bitcast i16* [[REMOTE_ELT2]] to i8*
-// CHECK: store i8* [[REMOTE_ELT2C]], i8** [[REMOTE_ELT_REF]], align
-//
 // Condition to reduce
-// CHECK: [[CONDALG0:%.+]] = icmp eq i16 [[ALGVER]], 0
-//
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp ult i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[CONDALG1:%.+]] = and i1 [[COND1]], [[COND2]]
-//
-// CHECK: [[COND3:%.+]] = icmp eq i16 [[ALGVER]], 2
-// CHECK: [[COND4:%.+]] = and i16 [[LANEID]], 1
-// CHECK: [[COND5:%.+]] = icmp eq i16 [[COND4]], 0
-// CHECK: [[COND6:%.+]] = and i1 [[COND3]], [[COND5]]
-// CHECK: [[COND7:%.+]] = icmp sgt i16 [[LANEOFFSET]], 0
-// CHECK: [[CONDALG2:%.+]] = and i1 [[COND6]], [[COND7]]
-//
-// CHECK: [[COND8:%.+]] = or i1 [[CONDALG0]], [[CONDALG1]]
-// CHECK: [[SHOULD_REDUCE:%.+]] = or i1 [[COND8]], [[CONDALG2]]
-// CHECK: br i1 [[SHOULD_REDUCE]], label {{%?}}[[DO_REDUCE:.+]], label {{%?}}[[REDUCE_ELSE:.+]]
-//
-// CHECK: [[DO_REDUCE]]
-// CHECK: [[RED_LIST1_VOID:%.+]] = bitcast [[RLT]]* [[RED_LIST]] to i8*
-// CHECK: [[RED_LIST2_VOID:%.+]] = bitcast [[RLT]]* [[REMOTE_RED_LIST]] to i8*
-// CHECK: call void [[REDUCTION_FUNC]](i8* [[RED_LIST1_VOID]], i8* [[RED_LIST2_VOID]])
-// CHECK: br label {{%?}}[[REDUCE_CONT:.+]]
-//
-// CHECK: [[REDUCE_ELSE]]
-// CHECK: br label {{%?}}[[REDUCE_CONT]]
-//
-// CHECK: [[REDUCE_CONT]]
 // Now check if we should just copy over the remote reduction list
-// CHECK: [[COND1:%.+]] = icmp eq i16 [[ALGVER]], 1
-// CHECK: [[COND2:%.+]] = icmp uge i16 [[LANEID]], [[LANEOFFSET]]
-// CHECK: [[SHOULD_COPY:%.+]] = and i1 [[COND1]], [[COND2]]
-// CHECK: br i1 [[SHOULD_COPY]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
-// CHECK: [[DO_COPY]]
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = load i8*, i8** [[REMOTE_ELT_REF]],
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT:%.+]] = bitcast i8* [[REMOTE_ELT_VOID]] to i32*
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[REMOTE_ELT_VAL:%.+]] = load i32, i32* [[REMOTE_ELT]], align
-// CHECK: store i32 [[REMOTE_ELT_VAL]], i32* [[ELT]], align
-//
-// CHECK: [[REMOTE_ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[REMOTE_RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[REMOTE_ELT_VOID:%.+]] = load i8*, i8** [[REMOTE_ELT_REF]],
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[REMOTE_ELT:%.+]] = bitcast i8* [[REMOTE_ELT_VOID]] to i16*
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16*
-// CHECK: [[REMOTE_ELT_VAL:%.+]] = load i16, i16* [[REMOTE_ELT]], align
-// CHECK: store i16 [[REMOTE_ELT_VAL]], i16* [[ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
-//
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
-//
-// CHECK: [[COPY_CONT]]
-// CHECK: void
 
-//
 // Inter warp copy function
-// CHECK: define internal void [[WARP_COPY_FN]](i8* %0, i32 %1)
-// CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31
-// CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5
-// CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]*
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
-// CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
-//
 // [[DO_COPY]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-//
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]]
-// CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], align
-// CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
+// Barrier after copy to shared memory storage medium.
+// Read into warp 0.
+// [[DO_COPY]]
+// Barrier after copy to shared memory storage medium.
+// Read into warp 0.
+
+#endif
 //
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
 //
-// Barrier after copy to shared memory storage medium.
-// CHECK: [[COPY_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
 //
-// Read into warp 0.
-// CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
-// CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]]
 //
-// CHECK: [[DO_READ]]
-// CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32*
-// CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]], align
-// CHECK: br label {{%?}}[[READ_CONT:.+]]
-//
-// CHECK: [[READ_ELSE]]
-// CHECK: br label {{%?}}[[READ_CONT]]
-//
-// CHECK: [[READ_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
-// CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
 //
-// [[DO_COPY]]
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16*
-//
-// CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]]
-// CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i16 addrspace([[SHARED_ADDRSPACE]])*
-// CHECK: [[ELT_VAL:%.+]] = load i16, i16* [[ELT]], align
-// CHECK: store volatile i16 [[ELT_VAL]], i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: br label {{%?}}[[COPY_CONT:.+]]
 //
-// CHECK: [[COPY_ELSE]]
-// CHECK: br label {{%?}}[[COPY_CONT]]
 //
-// Barrier after copy to shared memory storage medium.
-// CHECK: [[COPY_CONT]]
-// CHECK: call void @__kmpc_barrier(%struct.ident_t* @
-// CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
 //
-// Read into warp 0.
-// CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
-// CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]]
 //
-// CHECK: [[DO_READ]]
-// CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]]
-// CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i16 addrspace([[SHARED_ADDRSPACE]])*
-// CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]],
-// CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16*
-// CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i16, i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align
-// CHECK: store i16 [[MEDIUM_ELT_VAL]], i16* [[ELT]], align
-// CHECK: br label {{%?}}[[READ_CONT:.+]]
 //
-// CHECK: [[READ_ELSE]]
-// CHECK: br label {{%?}}[[READ_CONT]]
 //
-// CHECK: [[READ_CONT]]
-// CHECK: ret
-
-#endif
+//
+// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l27
+// CHECK-64-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 8
+// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
+// CHECK-64-NEXT:    store double* [[E]], double** [[E_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
+// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
+// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-64:       user_code.entry:
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
+// CHECK-64-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP0]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-64-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-64-NEXT:    ret void
+// CHECK-64:       worker.exit:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@__omp_outlined__
+// CHECK-64-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 8
+// CHECK-64-NEXT:    [[E1:%.*]] = alloca double, align 8
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// CHECK-64-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store double* [[E]], double** [[E_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 8
+// CHECK-64-NEXT:    store double 0.000000e+00, double* [[E1]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = load double, double* [[E1]], align 8
+// CHECK-64-NEXT:    [[ADD:%.*]] = fadd double [[TMP1]], 5.000000e+00
+// CHECK-64-NEXT:    store double [[ADD]], double* [[E1]], align 8
+// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
+// CHECK-64-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast double* [[E1]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 8
+// CHECK-64-NEXT:    [[TMP6:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-64-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, i8* [[TMP6]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
+// CHECK-64-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 1
+// CHECK-64-NEXT:    br i1 [[TMP8]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-64:       .omp.reduction.then:
+// CHECK-64-NEXT:    [[TMP9:%.*]] = load double, double* [[TMP0]], align 8
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load double, double* [[E1]], align 8
+// CHECK-64-NEXT:    [[ADD2:%.*]] = fadd double [[TMP9]], [[TMP10]]
+// CHECK-64-NEXT:    store double [[ADD2]], double* [[TMP0]], align 8
+// CHECK-64-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP3]])
+// CHECK-64-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-64:       .omp.reduction.done:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 8
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to double*
+// CHECK-64-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP12]], i64 1
+// CHECK-64-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
+// CHECK-64-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP12]] to i64*
+// CHECK-64-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
+// CHECK-64-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
+// CHECK-64-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-64-NEXT:    [[TMP18:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-64-NEXT:    [[TMP19:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP18]])
+// CHECK-64-NEXT:    store i64 [[TMP19]], i64* [[TMP16]], align 8
+// CHECK-64-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP15]], i64 1
+// CHECK-64-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP16]], i64 1
+// CHECK-64-NEXT:    [[TMP22:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP22]], i8** [[TMP11]], align 8
+// CHECK-64-NEXT:    [[TMP23:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP25:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP26:%.*]] = and i1 [[TMP24]], [[TMP25]]
+// CHECK-64-NEXT:    [[TMP27:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-64-NEXT:    [[TMP28:%.*]] = and i16 [[TMP6]], 1
+// CHECK-64-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP28]], 0
+// CHECK-64-NEXT:    [[TMP30:%.*]] = and i1 [[TMP27]], [[TMP29]]
+// CHECK-64-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-64-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
+// CHECK-64-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
+// CHECK-64-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
+// CHECK-64-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
+// CHECK-64-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-64-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4:[0-9]+]]
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
+// CHECK-64-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-64:       then4:
+// CHECK-64-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[TMP40]], align 8
+// CHECK-64-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
+// CHECK-64-NEXT:    [[TMP44:%.*]] = bitcast i8* [[TMP41]] to double*
+// CHECK-64-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP43]] to double*
+// CHECK-64-NEXT:    [[TMP46:%.*]] = load double, double* [[TMP44]], align 8
+// CHECK-64-NEXT:    store double [[TMP46]], double* [[TMP45]], align 8
+// CHECK-64-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-64:       else5:
+// CHECK-64-NEXT:    br label [[IFCONT6]]
+// CHECK-64:       ifcont6:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-64-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [1 x i8*]*
+// CHECK-64-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-64-NEXT:    br label [[PRECOND:%.*]]
+// CHECK-64:       precond:
+// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-64-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[TMP5]], 2
+// CHECK-64-NEXT:    br i1 [[TMP6]], label [[BODY:%.*]], label [[EXIT:%.*]]
+// CHECK-64:       body:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-64-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[TMP7]], align 8
+// CHECK-64-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to i32*
+// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP9]], i32 [[TMP5]]
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP10]], align 4
+// CHECK-64-NEXT:    store volatile i32 [[TMP12]], i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP13]]
+// CHECK-64-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-64:       then4:
+// CHECK-64-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-64-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
+// CHECK-64-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
+// CHECK-64-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP17]], i32 [[TMP5]]
+// CHECK-64-NEXT:    [[TMP19:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP14]], align 4
+// CHECK-64-NEXT:    store i32 [[TMP19]], i32* [[TMP18]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-64:       else5:
+// CHECK-64-NEXT:    br label [[IFCONT6]]
+// CHECK-64:       ifcont6:
+// CHECK-64-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK-64-NEXT:    store i32 [[TMP20]], i32* [[DOTCNT_ADDR]], align 4
+// CHECK-64-NEXT:    br label [[PRECOND]]
+// CHECK-64:       exit:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l32
+// CHECK-64-SAME: (i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 8
+// CHECK-64-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-64:       user_code.entry:
+// CHECK-64-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-64-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[TMP4]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP1]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 8
+// CHECK-64-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-64-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i8*, float*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP7]], i64 2)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-64-NEXT:    ret void
+// CHECK-64:       worker.exit:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@__omp_outlined__1
+// CHECK-64-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR1]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK-64-NEXT:    [[C1:%.*]] = alloca i8, align 1
+// CHECK-64-NEXT:    [[D2:%.*]] = alloca float, align 4
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 8
+// CHECK-64-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK-64-NEXT:    store i8 0, i8* [[C1]], align 1
+// CHECK-64-NEXT:    store float 1.000000e+00, float* [[D2]], align 4
+// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-64-NEXT:    [[CONV:%.*]] = sext i8 [[TMP2]] to i32
+// CHECK-64-NEXT:    [[XOR:%.*]] = xor i32 [[CONV]], 2
+// CHECK-64-NEXT:    [[CONV3:%.*]] = trunc i32 [[XOR]] to i8
+// CHECK-64-NEXT:    store i8 [[CONV3]], i8* [[C1]], align 1
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load float, float* [[D2]], align 4
+// CHECK-64-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 3.300000e+01
+// CHECK-64-NEXT:    store float [[MUL]], float* [[D2]], align 4
+// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
+// CHECK-64-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    store i8* [[C1]], i8** [[TMP6]], align 8
+// CHECK-64-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP8:%.*]] = bitcast float* [[D2]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 8
+// CHECK-64-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-64-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 2, i64 16, i8* [[TMP9]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func3, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func4)
+// CHECK-64-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 1
+// CHECK-64-NEXT:    br i1 [[TMP11]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-64:       .omp.reduction.then:
+// CHECK-64-NEXT:    [[TMP12:%.*]] = load i8, i8* [[TMP0]], align 1
+// CHECK-64-NEXT:    [[CONV4:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK-64-NEXT:    [[TMP13:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-64-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-64-NEXT:    [[XOR6:%.*]] = xor i32 [[CONV4]], [[CONV5]]
+// CHECK-64-NEXT:    [[CONV7:%.*]] = trunc i32 [[XOR6]] to i8
+// CHECK-64-NEXT:    store i8 [[CONV7]], i8* [[TMP0]], align 1
+// CHECK-64-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP1]], align 4
+// CHECK-64-NEXT:    [[TMP15:%.*]] = load float, float* [[D2]], align 4
+// CHECK-64-NEXT:    [[MUL8:%.*]] = fmul float [[TMP14]], [[TMP15]]
+// CHECK-64-NEXT:    store float [[MUL8]], float* [[TMP1]], align 4
+// CHECK-64-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP5]])
+// CHECK-64-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-64:       .omp.reduction.done:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func3
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i8, align 1
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca float, align 4
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[TMP10]], i64 1
+// CHECK-64-NEXT:    [[TMP13:%.*]] = load i8, i8* [[TMP10]], align 1
+// CHECK-64-NEXT:    [[TMP14:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-64-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-64-NEXT:    [[TMP15:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-64-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP14]], i16 [[TMP7]], i16 [[TMP15]])
+// CHECK-64-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i8
+// CHECK-64-NEXT:    store i8 [[TMP17]], i8* [[DOTOMP_REDUCTION_ELEMENT]], align 1
+// CHECK-64-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP10]], i64 1
+// CHECK-64-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[DOTOMP_REDUCTION_ELEMENT]], i64 1
+// CHECK-64-NEXT:    store i8* [[DOTOMP_REDUCTION_ELEMENT]], i8** [[TMP11]], align 8
+// CHECK-64-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP21:%.*]] = load i8*, i8** [[TMP20]], align 8
+// CHECK-64-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP21]] to float*
+// CHECK-64-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i64 1
+// CHECK-64-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to i8*
+// CHECK-64-NEXT:    [[TMP26:%.*]] = bitcast float* [[TMP23]] to i32*
+// CHECK-64-NEXT:    [[TMP27:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i32*
+// CHECK-64-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK-64-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-64-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-64-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-64-NEXT:    store i32 [[TMP30]], i32* [[TMP27]], align 4
+// CHECK-64-NEXT:    [[TMP31:%.*]] = getelementptr i32, i32* [[TMP26]], i64 1
+// CHECK-64-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[TMP27]], i64 1
+// CHECK-64-NEXT:    [[TMP33:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP33]], i8** [[TMP22]], align 8
+// CHECK-64-NEXT:    [[TMP34:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-64-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP36:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP37:%.*]] = and i1 [[TMP35]], [[TMP36]]
+// CHECK-64-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-64-NEXT:    [[TMP39:%.*]] = and i16 [[TMP6]], 1
+// CHECK-64-NEXT:    [[TMP40:%.*]] = icmp eq i16 [[TMP39]], 0
+// CHECK-64-NEXT:    [[TMP41:%.*]] = and i1 [[TMP38]], [[TMP40]]
+// CHECK-64-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-64-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
+// CHECK-64-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
+// CHECK-64-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
+// CHECK-64-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-64-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-64-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
+// CHECK-64-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-64:       then6:
+// CHECK-64-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP52:%.*]] = load i8*, i8** [[TMP51]], align 8
+// CHECK-64-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[TMP53]], align 8
+// CHECK-64-NEXT:    [[TMP55:%.*]] = load i8, i8* [[TMP52]], align 1
+// CHECK-64-NEXT:    store i8 [[TMP55]], i8* [[TMP54]], align 1
+// CHECK-64-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[TMP56]], align 8
+// CHECK-64-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[TMP58]], align 8
+// CHECK-64-NEXT:    [[TMP60:%.*]] = bitcast i8* [[TMP57]] to float*
+// CHECK-64-NEXT:    [[TMP61:%.*]] = bitcast i8* [[TMP59]] to float*
+// CHECK-64-NEXT:    [[TMP62:%.*]] = load float, float* [[TMP60]], align 4
+// CHECK-64-NEXT:    store float [[TMP62]], float* [[TMP61]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-64:       else7:
+// CHECK-64-NEXT:    br label [[IFCONT8]]
+// CHECK-64:       ifcont8:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func4
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-64-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-64-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 8
+// CHECK-64-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-64-NEXT:    [[TMP8:%.*]] = bitcast i32 addrspace(3)* [[TMP7]] to i8 addrspace(3)*
+// CHECK-64-NEXT:    [[TMP9:%.*]] = load i8, i8* [[TMP6]], align 1
+// CHECK-64-NEXT:    store volatile i8 [[TMP9]], i8 addrspace(3)* [[TMP8]], align 1
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-64-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-64:       then4:
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-64-NEXT:    [[TMP12:%.*]] = bitcast i32 addrspace(3)* [[TMP11]] to i8 addrspace(3)*
+// CHECK-64-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[TMP13]], align 8
+// CHECK-64-NEXT:    [[TMP15:%.*]] = load volatile i8, i8 addrspace(3)* [[TMP12]], align 1
+// CHECK-64-NEXT:    store i8 [[TMP15]], i8* [[TMP14]], align 1
+// CHECK-64-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-64:       else5:
+// CHECK-64-NEXT:    br label [[IFCONT6]]
+// CHECK-64:       ifcont6:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-64-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-64:       then8:
+// CHECK-64-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
+// CHECK-64-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i32*
+// CHECK-64-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP18]], align 4
+// CHECK-64-NEXT:    store volatile i32 [[TMP20]], i32 addrspace(3)* [[TMP19]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-64:       else9:
+// CHECK-64-NEXT:    br label [[IFCONT10]]
+// CHECK-64:       ifcont10:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP21]]
+// CHECK-64-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-64:       then12:
+// CHECK-64-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-64-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[TMP23]], align 8
+// CHECK-64-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to i32*
+// CHECK-64-NEXT:    [[TMP26:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP22]], align 4
+// CHECK-64-NEXT:    store i32 [[TMP26]], i32* [[TMP25]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-64:       else13:
+// CHECK-64-NEXT:    br label [[IFCONT14]]
+// CHECK-64:       ifcont14:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l38
+// CHECK-64-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
+// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
+// CHECK-64-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-64:       user_code.entry:
+// CHECK-64-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-64-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 8
+// CHECK-64-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
+// CHECK-64-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-64-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP8]], i64 2)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-64-NEXT:    ret void
+// CHECK-64:       worker.exit:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@__omp_outlined__5
+// CHECK-64-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR1]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
+// CHECK-64-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
+// CHECK-64-NEXT:    [[A1:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[B2:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// CHECK-64-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
+// CHECK-64-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
+// CHECK-64-NEXT:    store i32 0, i32* [[A1]], align 4
+// CHECK-64-NEXT:    store i16 -32768, i16* [[B2]], align 2
+// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-64-NEXT:    [[OR:%.*]] = or i32 [[TMP2]], 1
+// CHECK-64-NEXT:    store i32 [[OR]], i32* [[A1]], align 4
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-64-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
+// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 99, [[CONV]]
+// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK-64:       cond.true:
+// CHECK-64-NEXT:    br label [[COND_END:%.*]]
+// CHECK-64:       cond.false:
+// CHECK-64-NEXT:    [[TMP4:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-64-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP4]] to i32
+// CHECK-64-NEXT:    br label [[COND_END]]
+// CHECK-64:       cond.end:
+// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
+// CHECK-64-NEXT:    [[CONV4:%.*]] = trunc i32 [[COND]] to i16
+// CHECK-64-NEXT:    store i16 [[CONV4]], i16* [[B2]], align 2
+// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
+// CHECK-64-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A1]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 8
+// CHECK-64-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP10:%.*]] = bitcast i16* [[B2]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 8
+// CHECK-64-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-64-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 2, i64 16, i8* [[TMP11]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func7, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func8)
+// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 1
+// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-64:       .omp.reduction.then:
+// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-64-NEXT:    [[OR5:%.*]] = or i32 [[TMP14]], [[TMP15]]
+// CHECK-64-NEXT:    store i32 [[OR5]], i32* [[TMP0]], align 4
+// CHECK-64-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-64-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP16]] to i32
+// CHECK-64-NEXT:    [[TMP17:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-64-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP17]] to i32
+// CHECK-64-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[CONV6]], [[CONV7]]
+// CHECK-64-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
+// CHECK-64:       cond.true9:
+// CHECK-64-NEXT:    [[TMP18:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-64-NEXT:    br label [[COND_END11:%.*]]
+// CHECK-64:       cond.false10:
+// CHECK-64-NEXT:    [[TMP19:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-64-NEXT:    br label [[COND_END11]]
+// CHECK-64:       cond.end11:
+// CHECK-64-NEXT:    [[COND12:%.*]] = phi i16 [ [[TMP18]], [[COND_TRUE9]] ], [ [[TMP19]], [[COND_FALSE10]] ]
+// CHECK-64-NEXT:    store i16 [[COND12]], i16* [[TMP1]], align 2
+// CHECK-64-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP6]])
+// CHECK-64-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-64:       .omp.reduction.done:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func7
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 8
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca i16, align 2
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-64-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-64-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-64-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+// CHECK-64-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i64 1
+// CHECK-64-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to i8*
+// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 4
+// CHECK-64-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-64-NEXT:    [[TMP16:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-64-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP15]], i16 [[TMP7]], i16 [[TMP16]])
+// CHECK-64-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_REDUCTION_ELEMENT]], align 4
+// CHECK-64-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP12]], i64 1
+// CHECK-64-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[DOTOMP_REDUCTION_ELEMENT]], i64 1
+// CHECK-64-NEXT:    [[TMP20:%.*]] = bitcast i32* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP20]], i8** [[TMP11]], align 8
+// CHECK-64-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
+// CHECK-64-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP24:%.*]] = bitcast i8* [[TMP22]] to i16*
+// CHECK-64-NEXT:    [[TMP25:%.*]] = getelementptr i16, i16* [[TMP24]], i64 1
+// CHECK-64-NEXT:    [[TMP26:%.*]] = bitcast i16* [[TMP25]] to i8*
+// CHECK-64-NEXT:    [[TMP27:%.*]] = load i16, i16* [[TMP24]], align 2
+// CHECK-64-NEXT:    [[TMP28:%.*]] = sext i16 [[TMP27]] to i32
+// CHECK-64-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-64-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-64-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-64-NEXT:    [[TMP31:%.*]] = trunc i32 [[TMP30]] to i16
+// CHECK-64-NEXT:    store i16 [[TMP31]], i16* [[DOTOMP_REDUCTION_ELEMENT4]], align 2
+// CHECK-64-NEXT:    [[TMP32:%.*]] = getelementptr i16, i16* [[TMP24]], i64 1
+// CHECK-64-NEXT:    [[TMP33:%.*]] = getelementptr i16, i16* [[DOTOMP_REDUCTION_ELEMENT4]], i64 1
+// CHECK-64-NEXT:    [[TMP34:%.*]] = bitcast i16* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-64-NEXT:    store i8* [[TMP34]], i8** [[TMP23]], align 8
+// CHECK-64-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-64-NEXT:    [[TMP36:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP37:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP38:%.*]] = and i1 [[TMP36]], [[TMP37]]
+// CHECK-64-NEXT:    [[TMP39:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-64-NEXT:    [[TMP40:%.*]] = and i16 [[TMP6]], 1
+// CHECK-64-NEXT:    [[TMP41:%.*]] = icmp eq i16 [[TMP40]], 0
+// CHECK-64-NEXT:    [[TMP42:%.*]] = and i1 [[TMP39]], [[TMP41]]
+// CHECK-64-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-64-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
+// CHECK-64-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
+// CHECK-64-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
+// CHECK-64-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-64-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-64-NEXT:    call void @"_omp$reduction$reduction_func6"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-64-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-64-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
+// CHECK-64-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-64:       then6:
+// CHECK-64-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP53:%.*]] = load i8*, i8** [[TMP52]], align 8
+// CHECK-64-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[TMP54]], align 8
+// CHECK-64-NEXT:    [[TMP56:%.*]] = bitcast i8* [[TMP53]] to i32*
+// CHECK-64-NEXT:    [[TMP57:%.*]] = bitcast i8* [[TMP55]] to i32*
+// CHECK-64-NEXT:    [[TMP58:%.*]] = load i32, i32* [[TMP56]], align 4
+// CHECK-64-NEXT:    store i32 [[TMP58]], i32* [[TMP57]], align 4
+// CHECK-64-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[TMP59]], align 8
+// CHECK-64-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[TMP61]], align 8
+// CHECK-64-NEXT:    [[TMP63:%.*]] = bitcast i8* [[TMP60]] to i16*
+// CHECK-64-NEXT:    [[TMP64:%.*]] = bitcast i8* [[TMP62]] to i16*
+// CHECK-64-NEXT:    [[TMP65:%.*]] = load i16, i16* [[TMP63]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP65]], i16* [[TMP64]], align 2
+// CHECK-64-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-64:       else7:
+// CHECK-64-NEXT:    br label [[IFCONT8]]
+// CHECK-64:       ifcont8:
+// CHECK-64-NEXT:    ret void
+//
+//
+// CHECK-64-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func8
+// CHECK-64-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-64-NEXT:  entry:
+// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-64-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-64-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-64-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-64-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// CHECK-64-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-64-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-64:       then:
+// CHECK-64-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 8
+// CHECK-64-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to i32*
+// CHECK-64-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK-64-NEXT:    store volatile i32 [[TMP9]], i32 addrspace(3)* [[TMP8]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-64:       else:
+// CHECK-64-NEXT:    br label [[IFCONT]]
+// CHECK-64:       ifcont:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-64-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-64:       then4:
+// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 0
+// CHECK-64-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
+// CHECK-64-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
+// CHECK-64-NEXT:    [[TMP15:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-64-NEXT:    store i32 [[TMP15]], i32* [[TMP14]], align 4
+// CHECK-64-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-64:       else5:
+// CHECK-64-NEXT:    br label [[IFCONT6]]
+// CHECK-64:       ifcont6:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-64-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-64:       then8:
+// CHECK-64-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
+// CHECK-64-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i16*
+// CHECK-64-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-64-NEXT:    [[TMP20:%.*]] = bitcast i32 addrspace(3)* [[TMP19]] to i16 addrspace(3)*
+// CHECK-64-NEXT:    [[TMP21:%.*]] = load i16, i16* [[TMP18]], align 2
+// CHECK-64-NEXT:    store volatile i16 [[TMP21]], i16 addrspace(3)* [[TMP20]], align 2
+// CHECK-64-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-64:       else9:
+// CHECK-64-NEXT:    br label [[IFCONT10]]
+// CHECK-64:       ifcont10:
+// CHECK-64-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-64-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP22]]
+// CHECK-64-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-64:       then12:
+// CHECK-64-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-64-NEXT:    [[TMP24:%.*]] = bitcast i32 addrspace(3)* [[TMP23]] to i16 addrspace(3)*
+// CHECK-64-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i64 0, i64 1
+// CHECK-64-NEXT:    [[TMP26:%.*]] = load i8*, i8** [[TMP25]], align 8
+// CHECK-64-NEXT:    [[TMP27:%.*]] = bitcast i8* [[TMP26]] to i16*
+// CHECK-64-NEXT:    [[TMP28:%.*]] = load volatile i16, i16 addrspace(3)* [[TMP24]], align 2
+// CHECK-64-NEXT:    store i16 [[TMP28]], i16* [[TMP27]], align 2
+// CHECK-64-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-64:       else13:
+// CHECK-64-NEXT:    br label [[IFCONT14]]
+// CHECK-64:       ifcont14:
+// CHECK-64-NEXT:    ret void
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l27
+// CHECK-32-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
+// CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[E1:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    store double 0.000000e+00, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD:%.*]] = fadd double [[TMP1]], 5.000000e+00
+// CHECK-32-NEXT:    store double [[ADD]], double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast double* [[E1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, i8* [[TMP6]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
+// CHECK-32-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 1
+// CHECK-32-NEXT:    br i1 [[TMP8]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load double, double* [[TMP0]], align 8
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD2:%.*]] = fadd double [[TMP9]], [[TMP10]]
+// CHECK-32-NEXT:    store double [[ADD2]], double* [[TMP0]], align 8
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP3]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to double*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP12]] to i64*
+// CHECK-32-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP18:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP19:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP18]])
+// CHECK-32-NEXT:    store i64 [[TMP19]], i64* [[TMP16]], align 8
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP15]], i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP16]], i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP22]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP26:%.*]] = and i1 [[TMP24]], [[TMP25]]
+// CHECK-32-NEXT:    [[TMP27:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP28]], 0
+// CHECK-32-NEXT:    [[TMP30:%.*]] = and i1 [[TMP27]], [[TMP29]]
+// CHECK-32-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
+// CHECK-32-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
+// CHECK-32-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
+// CHECK-32-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4:[0-9]+]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
+// CHECK-32-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[TMP40]], align 4
+// CHECK-32-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 4
+// CHECK-32-NEXT:    [[TMP44:%.*]] = bitcast i8* [[TMP41]] to double*
+// CHECK-32-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP43]] to double*
+// CHECK-32-NEXT:    [[TMP46:%.*]] = load double, double* [[TMP44]], align 8
+// CHECK-32-NEXT:    store double [[TMP46]], double* [[TMP45]], align 8
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [1 x i8*]*
+// CHECK-32-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND:%.*]]
+// CHECK-32:       precond:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[TMP5]], 2
+// CHECK-32-NEXT:    br i1 [[TMP6]], label [[BODY:%.*]], label [[EXIT:%.*]]
+// CHECK-32:       body:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to i32*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP9]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP10]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP12]], i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP13]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
+// CHECK-32-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP17]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP14]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP19]], i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK-32-NEXT:    store i32 [[TMP20]], i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND]]
+// CHECK-32:       exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l32
+// CHECK-32-SAME: (i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i8*, float*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP7]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__1
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[C1:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[D2:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    store i8 0, i8* [[C1]], align 1
+// CHECK-32-NEXT:    store float 1.000000e+00, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i8 [[TMP2]] to i32
+// CHECK-32-NEXT:    [[XOR:%.*]] = xor i32 [[CONV]], 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = trunc i32 [[XOR]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV3]], i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 3.300000e+01
+// CHECK-32-NEXT:    store float [[MUL]], float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[C1]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast float* [[D2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 2, i32 8, i8* [[TMP9]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func3, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func4)
+// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 1
+// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i8, i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[CONV4:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[XOR6:%.*]] = xor i32 [[CONV4]], [[CONV5]]
+// CHECK-32-NEXT:    [[CONV7:%.*]] = trunc i32 [[XOR6]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV7]], i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP1]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL8:%.*]] = fmul float [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store float [[MUL8]], float* [[TMP1]], align 4
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP5]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func3
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[TMP10]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP15:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP14]], i16 [[TMP7]], i16 [[TMP15]])
+// CHECK-32-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i8
+// CHECK-32-NEXT:    store i8 [[TMP17]], i8* [[DOTOMP_REDUCTION_ELEMENT]], align 1
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    store i8* [[DOTOMP_REDUCTION_ELEMENT]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i8*, i8** [[TMP20]], align 4
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP21]] to float*
+// CHECK-32-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i32 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to i8*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast float* [[TMP23]] to i32*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i32*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    store i32 [[TMP30]], i32* [[TMP27]], align 4
+// CHECK-32-NEXT:    [[TMP31:%.*]] = getelementptr i32, i32* [[TMP26]], i32 1
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[TMP27]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP33]], i8** [[TMP22]], align 4
+// CHECK-32-NEXT:    [[TMP34:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP37:%.*]] = and i1 [[TMP35]], [[TMP36]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP40:%.*]] = icmp eq i16 [[TMP39]], 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = and i1 [[TMP38]], [[TMP40]]
+// CHECK-32-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
+// CHECK-32-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
+// CHECK-32-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
+// CHECK-32-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP52:%.*]] = load i8*, i8** [[TMP51]], align 4
+// CHECK-32-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[TMP53]], align 4
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8, i8* [[TMP52]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP55]], i8* [[TMP54]], align 1
+// CHECK-32-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[TMP56]], align 4
+// CHECK-32-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[TMP58]], align 4
+// CHECK-32-NEXT:    [[TMP60:%.*]] = bitcast i8* [[TMP57]] to float*
+// CHECK-32-NEXT:    [[TMP61:%.*]] = bitcast i8* [[TMP59]] to float*
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load float, float* [[TMP60]], align 4
+// CHECK-32-NEXT:    store float [[TMP62]], float* [[TMP61]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func4
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32 addrspace(3)* [[TMP7]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i8, i8* [[TMP6]], align 1
+// CHECK-32-NEXT:    store volatile i8 [[TMP9]], i8 addrspace(3)* [[TMP8]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i32 addrspace(3)* [[TMP11]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[TMP13]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i8, i8 addrspace(3)* [[TMP12]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP15]], i8* [[TMP14]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i32*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP20]], i32 addrspace(3)* [[TMP19]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP21]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to i32*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP22]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP26]], i32* [[TMP25]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l38
+// CHECK-32-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP8]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__5
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[A1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[B2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    store i32 0, i32* [[A1]], align 4
+// CHECK-32-NEXT:    store i16 -32768, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR:%.*]] = or i32 [[TMP2]], 1
+// CHECK-32-NEXT:    store i32 [[OR]], i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
+// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 99, [[CONV]]
+// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK-32:       cond.true:
+// CHECK-32-NEXT:    br label [[COND_END:%.*]]
+// CHECK-32:       cond.false:
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP4]] to i32
+// CHECK-32-NEXT:    br label [[COND_END]]
+// CHECK-32:       cond.end:
+// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
+// CHECK-32-NEXT:    [[CONV4:%.*]] = trunc i32 [[COND]] to i16
+// CHECK-32-NEXT:    store i16 [[CONV4]], i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP10:%.*]] = bitcast i16* [[B2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 2, i32 8, i8* [[TMP11]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func7, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func8)
+// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 1
+// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR5:%.*]] = or i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store i32 [[OR5]], i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP16]] to i32
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP17]] to i32
+// CHECK-32-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[CONV6]], [[CONV7]]
+// CHECK-32-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
+// CHECK-32:       cond.true9:
+// CHECK-32-NEXT:    [[TMP18:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11:%.*]]
+// CHECK-32:       cond.false10:
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11]]
+// CHECK-32:       cond.end11:
+// CHECK-32-NEXT:    [[COND12:%.*]] = phi i16 [ [[TMP18]], [[COND_TRUE9]] ], [ [[TMP19]], [[COND_FALSE10]] ]
+// CHECK-32-NEXT:    store i16 [[COND12]], i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP6]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func7
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP16:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP15]], i16 [[TMP7]], i16 [[TMP16]])
+// CHECK-32-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_REDUCTION_ELEMENT]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP20]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i8* [[TMP22]] to i16*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast i16* [[TMP25]] to i8*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = load i16, i16* [[TMP24]], align 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = sext i16 [[TMP27]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    [[TMP31:%.*]] = trunc i32 [[TMP30]] to i16
+// CHECK-32-NEXT:    store i16 [[TMP31]], i16* [[DOTOMP_REDUCTION_ELEMENT4]], align 2
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = getelementptr i16, i16* [[DOTOMP_REDUCTION_ELEMENT4]], i32 1
+// CHECK-32-NEXT:    [[TMP34:%.*]] = bitcast i16* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP34]], i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = and i1 [[TMP36]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP40:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP41:%.*]] = icmp eq i16 [[TMP40]], 0
+// CHECK-32-NEXT:    [[TMP42:%.*]] = and i1 [[TMP39]], [[TMP41]]
+// CHECK-32-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
+// CHECK-32-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
+// CHECK-32-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func6"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
+// CHECK-32-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP53:%.*]] = load i8*, i8** [[TMP52]], align 4
+// CHECK-32-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[TMP54]], align 4
+// CHECK-32-NEXT:    [[TMP56:%.*]] = bitcast i8* [[TMP53]] to i32*
+// CHECK-32-NEXT:    [[TMP57:%.*]] = bitcast i8* [[TMP55]] to i32*
+// CHECK-32-NEXT:    [[TMP58:%.*]] = load i32, i32* [[TMP56]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP58]], i32* [[TMP57]], align 4
+// CHECK-32-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[TMP59]], align 4
+// CHECK-32-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[TMP61]], align 4
+// CHECK-32-NEXT:    [[TMP63:%.*]] = bitcast i8* [[TMP60]] to i16*
+// CHECK-32-NEXT:    [[TMP64:%.*]] = bitcast i8* [[TMP62]] to i16*
+// CHECK-32-NEXT:    [[TMP65:%.*]] = load i16, i16* [[TMP63]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP65]], i16* [[TMP64]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func8
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to i32*
+// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP9]], i32 addrspace(3)* [[TMP8]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP15]], i32* [[TMP14]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i16*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32 addrspace(3)* [[TMP19]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i16, i16* [[TMP18]], align 2
+// CHECK-32-NEXT:    store volatile i16 [[TMP21]], i16 addrspace(3)* [[TMP20]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP22]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i32 addrspace(3)* [[TMP23]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load i8*, i8** [[TMP25]], align 4
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[TMP26]] to i16*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load volatile i16, i16 addrspace(3)* [[TMP24]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP28]], i16* [[TMP27]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l27
+// CHECK-32-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[E1:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    store double 0.000000e+00, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD:%.*]] = fadd double [[TMP1]], 5.000000e+00
+// CHECK-32-NEXT:    store double [[ADD]], double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast double* [[E1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, i8* [[TMP6]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
+// CHECK-32-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 1
+// CHECK-32-NEXT:    br i1 [[TMP8]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load double, double* [[TMP0]], align 8
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD2:%.*]] = fadd double [[TMP9]], [[TMP10]]
+// CHECK-32-NEXT:    store double [[ADD2]], double* [[TMP0]], align 8
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP3]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to double*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP12]] to i64*
+// CHECK-32-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP18:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP19:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP18]])
+// CHECK-32-NEXT:    store i64 [[TMP19]], i64* [[TMP16]], align 8
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP15]], i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP16]], i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP22]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP26:%.*]] = and i1 [[TMP24]], [[TMP25]]
+// CHECK-32-NEXT:    [[TMP27:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP28]], 0
+// CHECK-32-NEXT:    [[TMP30:%.*]] = and i1 [[TMP27]], [[TMP29]]
+// CHECK-32-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
+// CHECK-32-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
+// CHECK-32-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
+// CHECK-32-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
+// CHECK-32-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[TMP40]], align 4
+// CHECK-32-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 4
+// CHECK-32-NEXT:    [[TMP44:%.*]] = bitcast i8* [[TMP41]] to double*
+// CHECK-32-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP43]] to double*
+// CHECK-32-NEXT:    [[TMP46:%.*]] = load double, double* [[TMP44]], align 8
+// CHECK-32-NEXT:    store double [[TMP46]], double* [[TMP45]], align 8
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [1 x i8*]*
+// CHECK-32-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND:%.*]]
+// CHECK-32:       precond:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[TMP5]], 2
+// CHECK-32-NEXT:    br i1 [[TMP6]], label [[BODY:%.*]], label [[EXIT:%.*]]
+// CHECK-32:       body:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to i32*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP9]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP10]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP12]], i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP13]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
+// CHECK-32-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP17]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP14]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP19]], i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK-32-NEXT:    store i32 [[TMP20]], i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND]]
+// CHECK-32:       exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l32
+// CHECK-32-SAME: (i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i8*, float*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP7]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__1
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[C1:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[D2:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    store i8 0, i8* [[C1]], align 1
+// CHECK-32-NEXT:    store float 1.000000e+00, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i8 [[TMP2]] to i32
+// CHECK-32-NEXT:    [[XOR:%.*]] = xor i32 [[CONV]], 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = trunc i32 [[XOR]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV3]], i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 3.300000e+01
+// CHECK-32-NEXT:    store float [[MUL]], float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[C1]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast float* [[D2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 2, i32 8, i8* [[TMP9]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func3, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func4)
+// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 1
+// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i8, i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[CONV4:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[XOR6:%.*]] = xor i32 [[CONV4]], [[CONV5]]
+// CHECK-32-NEXT:    [[CONV7:%.*]] = trunc i32 [[XOR6]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV7]], i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP1]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL8:%.*]] = fmul float [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store float [[MUL8]], float* [[TMP1]], align 4
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP5]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func3
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[TMP10]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP15:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP14]], i16 [[TMP7]], i16 [[TMP15]])
+// CHECK-32-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i8
+// CHECK-32-NEXT:    store i8 [[TMP17]], i8* [[DOTOMP_REDUCTION_ELEMENT]], align 1
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    store i8* [[DOTOMP_REDUCTION_ELEMENT]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i8*, i8** [[TMP20]], align 4
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP21]] to float*
+// CHECK-32-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i32 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to i8*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast float* [[TMP23]] to i32*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i32*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    store i32 [[TMP30]], i32* [[TMP27]], align 4
+// CHECK-32-NEXT:    [[TMP31:%.*]] = getelementptr i32, i32* [[TMP26]], i32 1
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[TMP27]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP33]], i8** [[TMP22]], align 4
+// CHECK-32-NEXT:    [[TMP34:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP37:%.*]] = and i1 [[TMP35]], [[TMP36]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP40:%.*]] = icmp eq i16 [[TMP39]], 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = and i1 [[TMP38]], [[TMP40]]
+// CHECK-32-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
+// CHECK-32-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
+// CHECK-32-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
+// CHECK-32-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP52:%.*]] = load i8*, i8** [[TMP51]], align 4
+// CHECK-32-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[TMP53]], align 4
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8, i8* [[TMP52]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP55]], i8* [[TMP54]], align 1
+// CHECK-32-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[TMP56]], align 4
+// CHECK-32-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[TMP58]], align 4
+// CHECK-32-NEXT:    [[TMP60:%.*]] = bitcast i8* [[TMP57]] to float*
+// CHECK-32-NEXT:    [[TMP61:%.*]] = bitcast i8* [[TMP59]] to float*
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load float, float* [[TMP60]], align 4
+// CHECK-32-NEXT:    store float [[TMP62]], float* [[TMP61]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func4
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32 addrspace(3)* [[TMP7]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i8, i8* [[TMP6]], align 1
+// CHECK-32-NEXT:    store volatile i8 [[TMP9]], i8 addrspace(3)* [[TMP8]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i32 addrspace(3)* [[TMP11]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[TMP13]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i8, i8 addrspace(3)* [[TMP12]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP15]], i8* [[TMP14]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i32*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP20]], i32 addrspace(3)* [[TMP19]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP21]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to i32*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP22]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP26]], i32* [[TMP25]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l38
+// CHECK-32-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP8]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__5
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[A1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[B2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    store i32 0, i32* [[A1]], align 4
+// CHECK-32-NEXT:    store i16 -32768, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR:%.*]] = or i32 [[TMP2]], 1
+// CHECK-32-NEXT:    store i32 [[OR]], i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
+// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 99, [[CONV]]
+// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK-32:       cond.true:
+// CHECK-32-NEXT:    br label [[COND_END:%.*]]
+// CHECK-32:       cond.false:
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP4]] to i32
+// CHECK-32-NEXT:    br label [[COND_END]]
+// CHECK-32:       cond.end:
+// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
+// CHECK-32-NEXT:    [[CONV4:%.*]] = trunc i32 [[COND]] to i16
+// CHECK-32-NEXT:    store i16 [[CONV4]], i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP10:%.*]] = bitcast i16* [[B2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 2, i32 8, i8* [[TMP11]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func7, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func8)
+// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 1
+// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR5:%.*]] = or i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store i32 [[OR5]], i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP16]] to i32
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP17]] to i32
+// CHECK-32-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[CONV6]], [[CONV7]]
+// CHECK-32-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
+// CHECK-32:       cond.true9:
+// CHECK-32-NEXT:    [[TMP18:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11:%.*]]
+// CHECK-32:       cond.false10:
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11]]
+// CHECK-32:       cond.end11:
+// CHECK-32-NEXT:    [[COND12:%.*]] = phi i16 [ [[TMP18]], [[COND_TRUE9]] ], [ [[TMP19]], [[COND_FALSE10]] ]
+// CHECK-32-NEXT:    store i16 [[COND12]], i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP6]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func7
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP16:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP15]], i16 [[TMP7]], i16 [[TMP16]])
+// CHECK-32-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_REDUCTION_ELEMENT]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP20]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i8* [[TMP22]] to i16*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast i16* [[TMP25]] to i8*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = load i16, i16* [[TMP24]], align 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = sext i16 [[TMP27]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    [[TMP31:%.*]] = trunc i32 [[TMP30]] to i16
+// CHECK-32-NEXT:    store i16 [[TMP31]], i16* [[DOTOMP_REDUCTION_ELEMENT4]], align 2
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = getelementptr i16, i16* [[DOTOMP_REDUCTION_ELEMENT4]], i32 1
+// CHECK-32-NEXT:    [[TMP34:%.*]] = bitcast i16* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP34]], i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = and i1 [[TMP36]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP40:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP41:%.*]] = icmp eq i16 [[TMP40]], 0
+// CHECK-32-NEXT:    [[TMP42:%.*]] = and i1 [[TMP39]], [[TMP41]]
+// CHECK-32-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
+// CHECK-32-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
+// CHECK-32-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func6"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
+// CHECK-32-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP53:%.*]] = load i8*, i8** [[TMP52]], align 4
+// CHECK-32-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[TMP54]], align 4
+// CHECK-32-NEXT:    [[TMP56:%.*]] = bitcast i8* [[TMP53]] to i32*
+// CHECK-32-NEXT:    [[TMP57:%.*]] = bitcast i8* [[TMP55]] to i32*
+// CHECK-32-NEXT:    [[TMP58:%.*]] = load i32, i32* [[TMP56]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP58]], i32* [[TMP57]], align 4
+// CHECK-32-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[TMP59]], align 4
+// CHECK-32-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[TMP61]], align 4
+// CHECK-32-NEXT:    [[TMP63:%.*]] = bitcast i8* [[TMP60]] to i16*
+// CHECK-32-NEXT:    [[TMP64:%.*]] = bitcast i8* [[TMP62]] to i16*
+// CHECK-32-NEXT:    [[TMP65:%.*]] = load i16, i16* [[TMP63]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP65]], i16* [[TMP64]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func8
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to i32*
+// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP9]], i32 addrspace(3)* [[TMP8]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP15]], i32* [[TMP14]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i16*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32 addrspace(3)* [[TMP19]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i16, i16* [[TMP18]], align 2
+// CHECK-32-NEXT:    store volatile i16 [[TMP21]], i16 addrspace(3)* [[TMP20]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP22]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i32 addrspace(3)* [[TMP23]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load i8*, i8** [[TMP25]], align 4
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[TMP26]] to i16*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load volatile i16, i16 addrspace(3)* [[TMP24]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP28]], i16* [[TMP27]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l27
+// CHECK-32-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[E1:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    store double 0.000000e+00, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD:%.*]] = fadd double [[TMP1]], 5.000000e+00
+// CHECK-32-NEXT:    store double [[ADD]], double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast double* [[E1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, i8* [[TMP6]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
+// CHECK-32-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 1
+// CHECK-32-NEXT:    br i1 [[TMP8]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load double, double* [[TMP0]], align 8
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD2:%.*]] = fadd double [[TMP9]], [[TMP10]]
+// CHECK-32-NEXT:    store double [[ADD2]], double* [[TMP0]], align 8
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP3]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to double*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP12]] to i64*
+// CHECK-32-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP18:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP19:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP18]])
+// CHECK-32-NEXT:    store i64 [[TMP19]], i64* [[TMP16]], align 8
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP15]], i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP16]], i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP22]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP26:%.*]] = and i1 [[TMP24]], [[TMP25]]
+// CHECK-32-NEXT:    [[TMP27:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP28]], 0
+// CHECK-32-NEXT:    [[TMP30:%.*]] = and i1 [[TMP27]], [[TMP29]]
+// CHECK-32-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
+// CHECK-32-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
+// CHECK-32-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
+// CHECK-32-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
+// CHECK-32-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[TMP40]], align 4
+// CHECK-32-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 4
+// CHECK-32-NEXT:    [[TMP44:%.*]] = bitcast i8* [[TMP41]] to double*
+// CHECK-32-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP43]] to double*
+// CHECK-32-NEXT:    [[TMP46:%.*]] = load double, double* [[TMP44]], align 8
+// CHECK-32-NEXT:    store double [[TMP46]], double* [[TMP45]], align 8
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [1 x i8*]*
+// CHECK-32-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND:%.*]]
+// CHECK-32:       precond:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[TMP5]], 2
+// CHECK-32-NEXT:    br i1 [[TMP6]], label [[BODY:%.*]], label [[EXIT:%.*]]
+// CHECK-32:       body:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to i32*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP9]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP10]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP12]], i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP13]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
+// CHECK-32-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP17]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP14]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP19]], i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK-32-NEXT:    store i32 [[TMP20]], i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND]]
+// CHECK-32:       exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l32
+// CHECK-32-SAME: (i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i8*, float*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP7]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__1
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[C1:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[D2:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    store i8 0, i8* [[C1]], align 1
+// CHECK-32-NEXT:    store float 1.000000e+00, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i8 [[TMP2]] to i32
+// CHECK-32-NEXT:    [[XOR:%.*]] = xor i32 [[CONV]], 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = trunc i32 [[XOR]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV3]], i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 3.300000e+01
+// CHECK-32-NEXT:    store float [[MUL]], float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[C1]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast float* [[D2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 2, i32 8, i8* [[TMP9]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func3, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func4)
+// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 1
+// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i8, i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[CONV4:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[XOR6:%.*]] = xor i32 [[CONV4]], [[CONV5]]
+// CHECK-32-NEXT:    [[CONV7:%.*]] = trunc i32 [[XOR6]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV7]], i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP1]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL8:%.*]] = fmul float [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store float [[MUL8]], float* [[TMP1]], align 4
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP5]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func3
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[TMP10]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP15:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP14]], i16 [[TMP7]], i16 [[TMP15]])
+// CHECK-32-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i8
+// CHECK-32-NEXT:    store i8 [[TMP17]], i8* [[DOTOMP_REDUCTION_ELEMENT]], align 1
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    store i8* [[DOTOMP_REDUCTION_ELEMENT]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i8*, i8** [[TMP20]], align 4
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP21]] to float*
+// CHECK-32-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i32 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to i8*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast float* [[TMP23]] to i32*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i32*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    store i32 [[TMP30]], i32* [[TMP27]], align 4
+// CHECK-32-NEXT:    [[TMP31:%.*]] = getelementptr i32, i32* [[TMP26]], i32 1
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[TMP27]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP33]], i8** [[TMP22]], align 4
+// CHECK-32-NEXT:    [[TMP34:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP37:%.*]] = and i1 [[TMP35]], [[TMP36]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP40:%.*]] = icmp eq i16 [[TMP39]], 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = and i1 [[TMP38]], [[TMP40]]
+// CHECK-32-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
+// CHECK-32-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
+// CHECK-32-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
+// CHECK-32-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP52:%.*]] = load i8*, i8** [[TMP51]], align 4
+// CHECK-32-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[TMP53]], align 4
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8, i8* [[TMP52]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP55]], i8* [[TMP54]], align 1
+// CHECK-32-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[TMP56]], align 4
+// CHECK-32-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[TMP58]], align 4
+// CHECK-32-NEXT:    [[TMP60:%.*]] = bitcast i8* [[TMP57]] to float*
+// CHECK-32-NEXT:    [[TMP61:%.*]] = bitcast i8* [[TMP59]] to float*
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load float, float* [[TMP60]], align 4
+// CHECK-32-NEXT:    store float [[TMP62]], float* [[TMP61]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func4
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32 addrspace(3)* [[TMP7]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i8, i8* [[TMP6]], align 1
+// CHECK-32-NEXT:    store volatile i8 [[TMP9]], i8 addrspace(3)* [[TMP8]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i32 addrspace(3)* [[TMP11]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[TMP13]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i8, i8 addrspace(3)* [[TMP12]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP15]], i8* [[TMP14]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i32*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP20]], i32 addrspace(3)* [[TMP19]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP21]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to i32*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP22]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP26]], i32* [[TMP25]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l38
+// CHECK-32-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP8]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__5
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[A1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[B2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    store i32 0, i32* [[A1]], align 4
+// CHECK-32-NEXT:    store i16 -32768, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR:%.*]] = or i32 [[TMP2]], 1
+// CHECK-32-NEXT:    store i32 [[OR]], i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
+// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 99, [[CONV]]
+// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK-32:       cond.true:
+// CHECK-32-NEXT:    br label [[COND_END:%.*]]
+// CHECK-32:       cond.false:
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP4]] to i32
+// CHECK-32-NEXT:    br label [[COND_END]]
+// CHECK-32:       cond.end:
+// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
+// CHECK-32-NEXT:    [[CONV4:%.*]] = trunc i32 [[COND]] to i16
+// CHECK-32-NEXT:    store i16 [[CONV4]], i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP10:%.*]] = bitcast i16* [[B2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 2, i32 8, i8* [[TMP11]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func7, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func8)
+// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 1
+// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR5:%.*]] = or i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store i32 [[OR5]], i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP16]] to i32
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP17]] to i32
+// CHECK-32-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[CONV6]], [[CONV7]]
+// CHECK-32-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
+// CHECK-32:       cond.true9:
+// CHECK-32-NEXT:    [[TMP18:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11:%.*]]
+// CHECK-32:       cond.false10:
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11]]
+// CHECK-32:       cond.end11:
+// CHECK-32-NEXT:    [[COND12:%.*]] = phi i16 [ [[TMP18]], [[COND_TRUE9]] ], [ [[TMP19]], [[COND_FALSE10]] ]
+// CHECK-32-NEXT:    store i16 [[COND12]], i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP6]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func7
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP16:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP15]], i16 [[TMP7]], i16 [[TMP16]])
+// CHECK-32-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_REDUCTION_ELEMENT]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP20]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i8* [[TMP22]] to i16*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast i16* [[TMP25]] to i8*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = load i16, i16* [[TMP24]], align 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = sext i16 [[TMP27]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    [[TMP31:%.*]] = trunc i32 [[TMP30]] to i16
+// CHECK-32-NEXT:    store i16 [[TMP31]], i16* [[DOTOMP_REDUCTION_ELEMENT4]], align 2
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = getelementptr i16, i16* [[DOTOMP_REDUCTION_ELEMENT4]], i32 1
+// CHECK-32-NEXT:    [[TMP34:%.*]] = bitcast i16* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP34]], i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = and i1 [[TMP36]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP40:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP41:%.*]] = icmp eq i16 [[TMP40]], 0
+// CHECK-32-NEXT:    [[TMP42:%.*]] = and i1 [[TMP39]], [[TMP41]]
+// CHECK-32-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
+// CHECK-32-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
+// CHECK-32-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func6"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
+// CHECK-32-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP53:%.*]] = load i8*, i8** [[TMP52]], align 4
+// CHECK-32-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[TMP54]], align 4
+// CHECK-32-NEXT:    [[TMP56:%.*]] = bitcast i8* [[TMP53]] to i32*
+// CHECK-32-NEXT:    [[TMP57:%.*]] = bitcast i8* [[TMP55]] to i32*
+// CHECK-32-NEXT:    [[TMP58:%.*]] = load i32, i32* [[TMP56]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP58]], i32* [[TMP57]], align 4
+// CHECK-32-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[TMP59]], align 4
+// CHECK-32-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[TMP61]], align 4
+// CHECK-32-NEXT:    [[TMP63:%.*]] = bitcast i8* [[TMP60]] to i16*
+// CHECK-32-NEXT:    [[TMP64:%.*]] = bitcast i8* [[TMP62]] to i16*
+// CHECK-32-NEXT:    [[TMP65:%.*]] = load i16, i16* [[TMP63]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP65]], i16* [[TMP64]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func8
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to i32*
+// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP9]], i32 addrspace(3)* [[TMP8]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP15]], i32* [[TMP14]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i16*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32 addrspace(3)* [[TMP19]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i16, i16* [[TMP18]], align 2
+// CHECK-32-NEXT:    store volatile i16 [[TMP21]], i16 addrspace(3)* [[TMP20]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP22]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i32 addrspace(3)* [[TMP23]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load i8*, i8** [[TMP25]], align 4
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[TMP26]] to i16*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load volatile i16, i16 addrspace(3)* [[TMP24]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP28]], i16* [[TMP27]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l27
+// CHECK-32-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i32 1)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
+// CHECK-32-NEXT:    [[E1:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
+// CHECK-32-NEXT:    store double 0.000000e+00, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD:%.*]] = fadd double [[TMP1]], 5.000000e+00
+// CHECK-32-NEXT:    store double [[ADD]], double* [[E1]], align 8
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast double* [[E1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, i8* [[TMP6]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
+// CHECK-32-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 1
+// CHECK-32-NEXT:    br i1 [[TMP8]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load double, double* [[TMP0]], align 8
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load double, double* [[E1]], align 8
+// CHECK-32-NEXT:    [[ADD2:%.*]] = fadd double [[TMP9]], [[TMP10]]
+// CHECK-32-NEXT:    store double [[ADD2]], double* [[TMP0]], align 8
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP3]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to double*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP12]] to i64*
+// CHECK-32-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP18:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP19:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP18]])
+// CHECK-32-NEXT:    store i64 [[TMP19]], i64* [[TMP16]], align 8
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP15]], i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP16]], i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP22]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP26:%.*]] = and i1 [[TMP24]], [[TMP25]]
+// CHECK-32-NEXT:    [[TMP27:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP28]], 0
+// CHECK-32-NEXT:    [[TMP30:%.*]] = and i1 [[TMP27]], [[TMP29]]
+// CHECK-32-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
+// CHECK-32-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
+// CHECK-32-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
+// CHECK-32-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
+// CHECK-32-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[TMP40]], align 4
+// CHECK-32-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 4
+// CHECK-32-NEXT:    [[TMP44:%.*]] = bitcast i8* [[TMP41]] to double*
+// CHECK-32-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP43]] to double*
+// CHECK-32-NEXT:    [[TMP46:%.*]] = load double, double* [[TMP44]], align 8
+// CHECK-32-NEXT:    store double [[TMP46]], double* [[TMP45]], align 8
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [1 x i8*]*
+// CHECK-32-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND:%.*]]
+// CHECK-32:       precond:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = icmp ult i32 [[TMP5]], 2
+// CHECK-32-NEXT:    br i1 [[TMP6]], label [[BODY:%.*]], label [[EXIT:%.*]]
+// CHECK-32:       body:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to i32*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP9]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP10]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP12]], i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP13]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
+// CHECK-32-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP17]], i32 [[TMP5]]
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP14]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP19]], i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP5]], 1
+// CHECK-32-NEXT:    store i32 [[TMP20]], i32* [[DOTCNT_ADDR]], align 4
+// CHECK-32-NEXT:    br label [[PRECOND]]
+// CHECK-32:       exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l32
+// CHECK-32-SAME: (i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP6]], i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i8*, float*)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP7]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__1
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[C_ADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 4
+// CHECK-32-NEXT:    [[C1:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[D2:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i8* [[C]], i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    store float* [[D]], float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i8*, i8** [[C_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load float*, float** [[D_ADDR]], align 4
+// CHECK-32-NEXT:    store i8 0, i8* [[C1]], align 1
+// CHECK-32-NEXT:    store float 1.000000e+00, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i8 [[TMP2]] to i32
+// CHECK-32-NEXT:    [[XOR:%.*]] = xor i32 [[CONV]], 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = trunc i32 [[XOR]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV3]], i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 3.300000e+01
+// CHECK-32-NEXT:    store float [[MUL]], float* [[D2]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    store i8* [[C1]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast float* [[D2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 2, i32 8, i8* [[TMP9]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func3, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func4)
+// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 1
+// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP12:%.*]] = load i8, i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[CONV4:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[C1]], align 1
+// CHECK-32-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[XOR6:%.*]] = xor i32 [[CONV4]], [[CONV5]]
+// CHECK-32-NEXT:    [[CONV7:%.*]] = trunc i32 [[XOR6]] to i8
+// CHECK-32-NEXT:    store i8 [[CONV7]], i8* [[TMP0]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP1]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load float, float* [[D2]], align 4
+// CHECK-32-NEXT:    [[MUL8:%.*]] = fmul float [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store float [[MUL8]], float* [[TMP1]], align 4
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP5]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func3
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i8, align 1
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca float, align 4
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8, i8* [[TMP10]], align 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = sext i8 [[TMP13]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP15:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP14]], i16 [[TMP7]], i16 [[TMP15]])
+// CHECK-32-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i8
+// CHECK-32-NEXT:    store i8 [[TMP17]], i8* [[DOTOMP_REDUCTION_ELEMENT]], align 1
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP10]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    store i8* [[DOTOMP_REDUCTION_ELEMENT]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i8*, i8** [[TMP20]], align 4
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP21]] to float*
+// CHECK-32-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i32 1
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to i8*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast float* [[TMP23]] to i32*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i32*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    store i32 [[TMP30]], i32* [[TMP27]], align 4
+// CHECK-32-NEXT:    [[TMP31:%.*]] = getelementptr i32, i32* [[TMP26]], i32 1
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[TMP27]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = bitcast float* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP33]], i8** [[TMP22]], align 4
+// CHECK-32-NEXT:    [[TMP34:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP37:%.*]] = and i1 [[TMP35]], [[TMP36]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP39:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP40:%.*]] = icmp eq i16 [[TMP39]], 0
+// CHECK-32-NEXT:    [[TMP41:%.*]] = and i1 [[TMP38]], [[TMP40]]
+// CHECK-32-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
+// CHECK-32-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
+// CHECK-32-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
+// CHECK-32-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP52:%.*]] = load i8*, i8** [[TMP51]], align 4
+// CHECK-32-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[TMP53]], align 4
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8, i8* [[TMP52]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP55]], i8* [[TMP54]], align 1
+// CHECK-32-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[TMP56]], align 4
+// CHECK-32-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[TMP58]], align 4
+// CHECK-32-NEXT:    [[TMP60:%.*]] = bitcast i8* [[TMP57]] to float*
+// CHECK-32-NEXT:    [[TMP61:%.*]] = bitcast i8* [[TMP59]] to float*
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load float, float* [[TMP60]], align 4
+// CHECK-32-NEXT:    store float [[TMP62]], float* [[TMP61]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func4
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32 addrspace(3)* [[TMP7]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i8, i8* [[TMP6]], align 1
+// CHECK-32-NEXT:    store volatile i8 [[TMP9]], i8 addrspace(3)* [[TMP8]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i32 addrspace(3)* [[TMP11]] to i8 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[TMP13]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i8, i8 addrspace(3)* [[TMP12]], align 1
+// CHECK-32-NEXT:    store i8 [[TMP15]], i8* [[TMP14]], align 1
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i32*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP18]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP20]], i32 addrspace(3)* [[TMP19]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP21]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to i32*
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP22]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP26]], i32* [[TMP25]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l38
+// CHECK-32-SAME: (i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
+// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP2]], -1
+// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// CHECK-32:       user_code.entry:
+// CHECK-32-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP0]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast [2 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
+// CHECK-32-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32*, i16*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP8]], i32 2)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
+// CHECK-32-NEXT:    ret void
+// CHECK-32:       worker.exit:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@__omp_outlined__5
+// CHECK-32-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR1]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
+// CHECK-32-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
+// CHECK-32-NEXT:    [[A1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[B2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
+// CHECK-32-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
+// CHECK-32-NEXT:    store i32 0, i32* [[A1]], align 4
+// CHECK-32-NEXT:    store i16 -32768, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR:%.*]] = or i32 [[TMP2]], 1
+// CHECK-32-NEXT:    store i32 [[OR]], i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
+// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 99, [[CONV]]
+// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK-32:       cond.true:
+// CHECK-32-NEXT:    br label [[COND_END:%.*]]
+// CHECK-32:       cond.false:
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP4]] to i32
+// CHECK-32-NEXT:    br label [[COND_END]]
+// CHECK-32:       cond.end:
+// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
+// CHECK-32-NEXT:    [[CONV4:%.*]] = trunc i32 [[COND]] to i16
+// CHECK-32-NEXT:    store i16 [[CONV4]], i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP8:%.*]] = bitcast i32* [[A1]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP8]], i8** [[TMP7]], align 4
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP10:%.*]] = bitcast i16* [[B2]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// CHECK-32-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 2, i32 8, i8* [[TMP11]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func7, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func8)
+// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 1
+// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
+// CHECK-32:       .omp.reduction.then:
+// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A1]], align 4
+// CHECK-32-NEXT:    [[OR5:%.*]] = or i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT:    store i32 [[OR5]], i32* [[TMP0]], align 4
+// CHECK-32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP16]] to i32
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP17]] to i32
+// CHECK-32-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[CONV6]], [[CONV7]]
+// CHECK-32-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
+// CHECK-32:       cond.true9:
+// CHECK-32-NEXT:    [[TMP18:%.*]] = load i16, i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11:%.*]]
+// CHECK-32:       cond.false10:
+// CHECK-32-NEXT:    [[TMP19:%.*]] = load i16, i16* [[B2]], align 2
+// CHECK-32-NEXT:    br label [[COND_END11]]
+// CHECK-32:       cond.end11:
+// CHECK-32-NEXT:    [[COND12:%.*]] = phi i16 [ [[TMP18]], [[COND_TRUE9]] ], [ [[TMP19]], [[COND_FALSE10]] ]
+// CHECK-32-NEXT:    store i16 [[COND12]], i16* [[TMP1]], align 2
+// CHECK-32-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP6]])
+// CHECK-32-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
+// CHECK-32:       .omp.reduction.done:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func7
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i16 signext [[TMP1:%.*]], i16 signext [[TMP2:%.*]], i16 signext [[TMP3:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [2 x i8*], align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[DOTOMP_REDUCTION_ELEMENT4:%.*]] = alloca i16, align 2
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
+// CHECK-32-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
+// CHECK-32-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
+// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to i8*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 4
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP16:%.*]] = trunc i32 [[NVPTX_WARP_SIZE]] to i16
+// CHECK-32-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP15]], i16 [[TMP7]], i16 [[TMP16]])
+// CHECK-32-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_REDUCTION_ELEMENT]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = getelementptr i32, i32* [[TMP12]], i32 1
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[DOTOMP_REDUCTION_ELEMENT]], i32 1
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP20]], i8** [[TMP11]], align 4
+// CHECK-32-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 4
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i8* [[TMP22]] to i16*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = bitcast i16* [[TMP25]] to i8*
+// CHECK-32-NEXT:    [[TMP27:%.*]] = load i16, i16* [[TMP24]], align 2
+// CHECK-32-NEXT:    [[TMP28:%.*]] = sext i16 [[TMP27]] to i32
+// CHECK-32-NEXT:    [[NVPTX_WARP_SIZE5:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
+// CHECK-32-NEXT:    [[TMP29:%.*]] = trunc i32 [[NVPTX_WARP_SIZE5]] to i16
+// CHECK-32-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP28]], i16 [[TMP7]], i16 [[TMP29]])
+// CHECK-32-NEXT:    [[TMP31:%.*]] = trunc i32 [[TMP30]] to i16
+// CHECK-32-NEXT:    store i16 [[TMP31]], i16* [[DOTOMP_REDUCTION_ELEMENT4]], align 2
+// CHECK-32-NEXT:    [[TMP32:%.*]] = getelementptr i16, i16* [[TMP24]], i32 1
+// CHECK-32-NEXT:    [[TMP33:%.*]] = getelementptr i16, i16* [[DOTOMP_REDUCTION_ELEMENT4]], i32 1
+// CHECK-32-NEXT:    [[TMP34:%.*]] = bitcast i16* [[DOTOMP_REDUCTION_ELEMENT4]] to i8*
+// CHECK-32-NEXT:    store i8* [[TMP34]], i8** [[TMP23]], align 4
+// CHECK-32-NEXT:    [[TMP35:%.*]] = icmp eq i16 [[TMP8]], 0
+// CHECK-32-NEXT:    [[TMP36:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP37:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP38:%.*]] = and i1 [[TMP36]], [[TMP37]]
+// CHECK-32-NEXT:    [[TMP39:%.*]] = icmp eq i16 [[TMP8]], 2
+// CHECK-32-NEXT:    [[TMP40:%.*]] = and i16 [[TMP6]], 1
+// CHECK-32-NEXT:    [[TMP41:%.*]] = icmp eq i16 [[TMP40]], 0
+// CHECK-32-NEXT:    [[TMP42:%.*]] = and i1 [[TMP39]], [[TMP41]]
+// CHECK-32-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
+// CHECK-32-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
+// CHECK-32-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
+// CHECK-32-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
+// CHECK-32-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
+// CHECK-32-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
+// CHECK-32-NEXT:    call void @"_omp$reduction$reduction_func6"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
+// CHECK-32-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
+// CHECK-32-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]
+// CHECK-32:       then6:
+// CHECK-32-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP53:%.*]] = load i8*, i8** [[TMP52]], align 4
+// CHECK-32-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[TMP54]], align 4
+// CHECK-32-NEXT:    [[TMP56:%.*]] = bitcast i8* [[TMP53]] to i32*
+// CHECK-32-NEXT:    [[TMP57:%.*]] = bitcast i8* [[TMP55]] to i32*
+// CHECK-32-NEXT:    [[TMP58:%.*]] = load i32, i32* [[TMP56]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP58]], i32* [[TMP57]], align 4
+// CHECK-32-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[TMP59]], align 4
+// CHECK-32-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[TMP61]], align 4
+// CHECK-32-NEXT:    [[TMP63:%.*]] = bitcast i8* [[TMP60]] to i16*
+// CHECK-32-NEXT:    [[TMP64:%.*]] = bitcast i8* [[TMP62]] to i16*
+// CHECK-32-NEXT:    [[TMP65:%.*]] = load i16, i16* [[TMP63]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP65]], i16* [[TMP64]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT8:%.*]]
+// CHECK-32:       else7:
+// CHECK-32-NEXT:    br label [[IFCONT8]]
+// CHECK-32:       ifcont8:
+// CHECK-32-NEXT:    ret void
+//
+//
+// CHECK-32-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func8
+// CHECK-32-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
+// CHECK-32-NEXT:  entry:
+// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
+// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
+// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
+// CHECK-32-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[NVPTX_TID:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_TID2:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[NVPTX_TID2]], 31
+// CHECK-32-NEXT:    [[NVPTX_TID3:%.*]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+// CHECK-32-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[NVPTX_TID3]], 5
+// CHECK-32-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 4
+// CHECK-32-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to [2 x i8*]*
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
+// CHECK-32:       then:
+// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[TMP5]], align 4
+// CHECK-32-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to i32*
+// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK-32-NEXT:    store volatile i32 [[TMP9]], i32 addrspace(3)* [[TMP8]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT:%.*]]
+// CHECK-32:       else:
+// CHECK-32-NEXT:    br label [[IFCONT]]
+// CHECK-32:       ifcont:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP10]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
+// CHECK-32:       then4:
+// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 0
+// CHECK-32-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
+// CHECK-32-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
+// CHECK-32-NEXT:    [[TMP15:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP11]], align 4
+// CHECK-32-NEXT:    store i32 [[TMP15]], i32* [[TMP14]], align 4
+// CHECK-32-NEXT:    br label [[IFCONT6:%.*]]
+// CHECK-32:       else5:
+// CHECK-32-NEXT:    br label [[IFCONT6]]
+// CHECK-32:       ifcont6:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[WARP_MASTER7:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
+// CHECK-32-NEXT:    br i1 [[WARP_MASTER7]], label [[THEN8:%.*]], label [[ELSE9:%.*]]
+// CHECK-32:       then8:
+// CHECK-32-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 4
+// CHECK-32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[TMP17]] to i16*
+// CHECK-32-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
+// CHECK-32-NEXT:    [[TMP20:%.*]] = bitcast i32 addrspace(3)* [[TMP19]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP21:%.*]] = load i16, i16* [[TMP18]], align 2
+// CHECK-32-NEXT:    store volatile i16 [[TMP21]], i16 addrspace(3)* [[TMP20]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT10:%.*]]
+// CHECK-32:       else9:
+// CHECK-32-NEXT:    br label [[IFCONT10]]
+// CHECK-32:       ifcont10:
+// CHECK-32-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
+// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTADDR1]], align 4
+// CHECK-32-NEXT:    [[IS_ACTIVE_THREAD11:%.*]] = icmp ult i32 [[NVPTX_TID]], [[TMP22]]
+// CHECK-32-NEXT:    br i1 [[IS_ACTIVE_THREAD11]], label [[THEN12:%.*]], label [[ELSE13:%.*]]
+// CHECK-32:       then12:
+// CHECK-32-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_TID]]
+// CHECK-32-NEXT:    [[TMP24:%.*]] = bitcast i32 addrspace(3)* [[TMP23]] to i16 addrspace(3)*
+// CHECK-32-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP4]], i32 0, i32 1
+// CHECK-32-NEXT:    [[TMP26:%.*]] = load i8*, i8** [[TMP25]], align 4
+// CHECK-32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[TMP26]] to i16*
+// CHECK-32-NEXT:    [[TMP28:%.*]] = load volatile i16, i16 addrspace(3)* [[TMP24]], align 2
+// CHECK-32-NEXT:    store i16 [[TMP28]], i16* [[TMP27]], align 2
+// CHECK-32-NEXT:    br label [[IFCONT14:%.*]]
+// CHECK-32:       else13:
+// CHECK-32-NEXT:    br label [[IFCONT14]]
+// CHECK-32:       ifcont14:
+// CHECK-32-NEXT:    ret void
+//

clang/test/OpenMP/nvptx_target_parallel_reduction_codegen_tbaa_PR46146.cpp

 }
 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIfEvv_l16
 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8:![0-9]+]]
 // CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIdEvv_l16
 // CHECK1-SAME: () #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8]]
 // CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIfEvv_l16
 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8:![0-9]+]]
 // CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIdEvv_l16
 // CHECK2-SAME: () #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8]]
 // CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIfEvv_l16
 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8:![0-9]+]]
 // CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z17complex_reductionIdEvv_l16
 // CHECK3-SAME: () #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA8]]
 // CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_target_printf_codegen.c

 //
 //
 //
 // CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckSimple_l13
 // CHECK-64-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK-64-NEXT:  entry:
 // CHECK-64-NEXT:    [[FMT:%.*]] = alloca i8*, align 8
 // CHECK-64-NEXT:    [[TMP:%.*]] = alloca [[PRINTF_ARGS:%.*]], align 8
-// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-64:       user_code.entry:
 // CHECK-64-NEXT:    store i8* getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0), i8** [[FMT]], align 8
 // CHECK-64-NEXT:    [[TMP1:%.*]] = load i8*, i8** [[FMT]], align 8
 // CHECK-64-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 0
 // CHECK-64-NEXT:    store i32 1, i32* [[TMP2]], align 4
 // CHECK-64-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 1
 // CHECK-64-NEXT:    store i64 2, i64* [[TMP3]], align 8
 // CHECK-64-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 2
 // CHECK-64-NEXT:    store double 3.000000e+00, double* [[TMP4]], align 8
 // CHECK-64-NEXT:    [[TMP5:%.*]] = bitcast %printf_args* [[TMP]] to i8*
 // CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @vprintf(i8* [[TMP1]], i8* [[TMP5]])
-// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-64-NEXT:    ret void
 // CHECK-64:       worker.exit:
 // CHECK-64-NEXT:    ret void
 //
 //
 // CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckNoArgs_l25
 // CHECK-64-SAME: () #[[ATTR0]] {
 // CHECK-64-NEXT:  entry:
-// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-64:       user_code.entry:
 // CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @vprintf(i8* getelementptr inbounds ([14 x i8], [14 x i8]* @.str1, i64 0, i64 0), i8* null)
-// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-64-NEXT:    ret void
 // CHECK-64:       worker.exit:
 // CHECK-64-NEXT:    ret void
 //
 //
 // CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckAllocaIsInEntryBlock_l36
 // CHECK-64-SAME: (i64 [[FOO:%.*]]) #[[ATTR0]] {
 // CHECK-64-NEXT:  entry:
 // CHECK-64-NEXT:    [[FOO_ADDR:%.*]] = alloca i64, align 8
 // CHECK-64-NEXT:    [[TMP:%.*]] = alloca [[PRINTF_ARGS_0:%.*]], align 8
 // CHECK-64-NEXT:    store i64 [[FOO]], i64* [[FOO_ADDR]], align 8
 // CHECK-64-NEXT:    [[CONV:%.*]] = bitcast i64* [[FOO_ADDR]] to i32*
-// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-64:       user_code.entry:
 // CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 // CHECK-64:       if.then:
 // CHECK-64-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PRINTF_ARGS_0]], %printf_args.0* [[TMP]], i32 0, i32 0
 // CHECK-64-NEXT:    store i32 42, i32* [[TMP2]], align 4
 // CHECK-64-NEXT:    [[TMP3:%.*]] = bitcast %printf_args.0* [[TMP]] to i8*
 // CHECK-64-NEXT:    [[TMP4:%.*]] = call i32 @vprintf(i8* getelementptr inbounds ([3 x i8], [3 x i8]* @.str2, i64 0, i64 0), i8* [[TMP3]])
 // CHECK-64-NEXT:    br label [[IF_END]]
 // CHECK-64:       worker.exit:
 // CHECK-64-NEXT:    ret void
 // CHECK-64:       if.end:
-// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-64-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-64-NEXT:    ret void
 //
 //
 //
 //
 //
 // CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckSimple_l13
 // CHECK-32-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK-32-NEXT:  entry:
 // CHECK-32-NEXT:    [[FMT:%.*]] = alloca i8*, align 4
 // CHECK-32-NEXT:    [[TMP:%.*]] = alloca [[PRINTF_ARGS:%.*]], align 8
-// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-32:       user_code.entry:
 // CHECK-32-NEXT:    store i8* getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i32 0, i32 0), i8** [[FMT]], align 4
 // CHECK-32-NEXT:    [[TMP1:%.*]] = load i8*, i8** [[FMT]], align 4
 // CHECK-32-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 0
 // CHECK-32-NEXT:    store i32 1, i32* [[TMP2]], align 4
 // CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 1
 // CHECK-32-NEXT:    store i64 2, i64* [[TMP3]], align 8
 // CHECK-32-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[PRINTF_ARGS]], %printf_args* [[TMP]], i32 0, i32 2
 // CHECK-32-NEXT:    store double 3.000000e+00, double* [[TMP4]], align 8
 // CHECK-32-NEXT:    [[TMP5:%.*]] = bitcast %printf_args* [[TMP]] to i8*
 // CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @vprintf(i8* [[TMP1]], i8* [[TMP5]])
-// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-32-NEXT:    ret void
 // CHECK-32:       worker.exit:
 // CHECK-32-NEXT:    ret void
 //
 //
 // CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckNoArgs_l25
 // CHECK-32-SAME: () #[[ATTR0]] {
 // CHECK-32-NEXT:  entry:
-// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-32:       user_code.entry:
 // CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @vprintf(i8* getelementptr inbounds ([14 x i8], [14 x i8]* @.str1, i32 0, i32 0), i8* null)
-// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-32-NEXT:    ret void
 // CHECK-32:       worker.exit:
 // CHECK-32-NEXT:    ret void
 //
 //
 // CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_CheckAllocaIsInEntryBlock_l36
 // CHECK-32-SAME: (i32 [[FOO:%.*]]) #[[ATTR0]] {
 // CHECK-32-NEXT:  entry:
 // CHECK-32-NEXT:    [[FOO_ADDR:%.*]] = alloca i32, align 4
 // CHECK-32-NEXT:    [[TMP:%.*]] = alloca [[PRINTF_ARGS_0:%.*]], align 8
 // CHECK-32-NEXT:    store i32 [[FOO]], i32* [[FOO_ADDR]], align 4
-// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK-32:       user_code.entry:
 // CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, i32* [[FOO_ADDR]], align 4
 // CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 // CHECK-32:       if.then:
 // CHECK-32-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PRINTF_ARGS_0]], %printf_args.0* [[TMP]], i32 0, i32 0
 // CHECK-32-NEXT:    store i32 42, i32* [[TMP2]], align 4
 // CHECK-32-NEXT:    [[TMP3:%.*]] = bitcast %printf_args.0* [[TMP]] to i8*
 // CHECK-32-NEXT:    [[TMP4:%.*]] = call i32 @vprintf(i8* getelementptr inbounds ([3 x i8], [3 x i8]* @.str2, i32 0, i32 0), i8* [[TMP3]])
 // CHECK-32-NEXT:    br label [[IF_END]]
 // CHECK-32:       worker.exit:
 // CHECK-32-NEXT:    ret void
 // CHECK-32:       if.end:
-// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK-32-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK-32-NEXT:    ret void
 //

clang/test/OpenMP/nvptx_target_simd_codegen.cpp

int bar(int n){
   int a = 0;
 
   a += ftemplate<int>(n);
 
   return a;
 }
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l32}}(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l37}}(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l42}}(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l47}}(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
 // CHECK-NOT: call void @__kmpc_nvptx_end_reduce_nowait(
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 // CHECK: ret void
 
 #endif

clang/test/OpenMP/nvptx_target_teams_codegen.cpp

 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i8*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i8*
 // CHECK1-NEXT:    store i8 [[TMP2]], i8* [[CONV1]], align 1
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP3]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP3]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[A_ADDR]] to i8*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i32* [[A_CASTED]] to i8*
 // CHECK2-NEXT:    store i8 [[TMP2]], i8* [[CONV1]], align 1
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[A_ADDR]] to i8*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[A_CASTED]] to i8*
 // CHECK3-NEXT:    store i8 [[TMP2]], i8* [[CONV1]], align 1
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_target_teams_distribute_codegen.cpp

 // CHECK6-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR3]]
 // CHECK6-NEXT:    ret void
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l16
 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4:[0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 // CHECK1-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 8
-// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l16
 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4:[0-9]+]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 // CHECK2-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i32 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l16
 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4:[0-9]+]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 // CHECK3-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp

 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[L]], i64* [[L_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[L_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[L_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV3]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[L_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[L:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK1:       .omp.lastprivate.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]+}}__Z9ftemplateIiET_i_l34
-// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[F_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[F]], i64* [[F_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[F_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[F_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[F_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i64 [[F:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__8
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[V]], i32** [[V_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[V_ADDR]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__10
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[V:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store i64 [[L]], i64* [[L_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[L_ADDR]] to i32*
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[L_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV3]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[L_CASTED]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[L:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK2:       .omp.lastprivate.done:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l34
-// CHECK2-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK2-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 8
 // CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK2-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 8
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[F_CASTED:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK2-NEXT:    store i64 [[F]], i64* [[F_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[F_ADDR]] to i32*
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[F_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[F_CASTED]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i64 [[F:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK2-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 8
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__8
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store i32* [[V]], i32** [[V_ADDR]], align 8
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[V_ADDR]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__10
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[V:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-NEXT:    [[L_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[L]], i32* [[L_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[L_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[L_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[L_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[L:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK3:       .omp.lastprivate.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]+}}__Z9ftemplateIiET_i_l34
-// CHECK3-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK3-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[F_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[F_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[F]], i32* [[F_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[F_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[F_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[F_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[F:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [10 x [10 x i32]]*, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__8
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32* [[V]], i32** [[V_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[V_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__10
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[V:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK4-NEXT:    [[L_CASTED:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[L]], i32* [[L_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[L_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[L_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[L_CASTED]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[L:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK4:       .omp.lastprivate.done:
 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK4:       omp.precond.end:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l34
-// CHECK4-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK4-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 4
 // CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[F_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[F_CASTED:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[F]], i32* [[F_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[F_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[F_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[F_CASTED]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR1]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR2]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[F:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [10 x [10 x i32]]*, align 4
 // CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR1]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [10 x [10 x i32]]* [[TMP0]]) #[[ATTR2]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__8
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK4-NEXT:    store i32* [[V]], i32** [[V_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[V_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR1]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__omp_outlined__10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32* [[TMP5]]) #[[ATTR2]]
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__10
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[V:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_generic_mode_codegen.cpp

 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV3]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], i32* [[TMP0]], i64 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I7]], align 4
-// CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR4:[0-9]+]]
-// CHECK1-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR5:[0-9]+]]
+// CHECK1-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CALL]], [[CALL13]]
-// CHECK1-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
+// CHECK1-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR5]]
 // CHECK1-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CALL15]]
 // CHECK1-NEXT:    store i32 [[ADD16]], i32* [[TMP0]], 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:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1

 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[ARGC_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], i32* [[TMP0]], i32 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
-// CHECK2-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR4:[0-9]+]]
-// CHECK2-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK2-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR5:[0-9]+]]
+// CHECK2-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[CALL]], [[CALL8]]
-// CHECK2-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR4]]
+// CHECK2-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR5]]
 // CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CALL10]]
 // CHECK2-NEXT:    store i32 [[ADD11]], i32* [[TMP0]], align 4
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP20]], 1

 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[ARGC_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], i32* [[TMP0]], i32 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK3-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
-// CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR4:[0-9]+]]
-// CHECK3-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR5:[0-9]+]]
+// CHECK3-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[CALL]], [[CALL8]]
-// CHECK3-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR4]]
+// CHECK3-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR5]]
 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CALL10]]
 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[TMP0]], 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:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP20]], 1

 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
 // CHECK4-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
-// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[CONV3]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], i32* [[TMP0]], i64 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK4-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I7]], align 4
-// CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR4:[0-9]+]]
-// CHECK4-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR5:[0-9]+]]
+// CHECK4-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CALL]], [[CALL13]]
-// CHECK4-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
+// CHECK4-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR5]]
 // CHECK4-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CALL15]]
 // CHECK4-NEXT:    store i32 [[ADD16]], i32* [[TMP0]], align 4
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1

 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK5-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
-// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK5-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK5-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK5:       user_code.entry:
 // CHECK5-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[TMP3]], i32* [[ARGC_CASTED]], align 4
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK5-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], i32* [[TMP0]], i32 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK5-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK5-NEXT:    ret void
 // CHECK5:       worker.exit:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK5-NEXT:  entry:

 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK5-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
-// CHECK5-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR4:[0-9]+]]
-// CHECK5-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK5-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR5:[0-9]+]]
+// CHECK5-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[CALL]], [[CALL8]]
-// CHECK5-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR4]]
+// CHECK5-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR5]]
 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CALL10]]
 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[TMP0]], 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:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP20]], 1

 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK6-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK6-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK6-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK6-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 4
-// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK6-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK6-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK6-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK6:       user_code.entry:
 // CHECK6-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[TMP3]], i32* [[ARGC_CASTED]], align 4
 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
 // CHECK6-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], i32* [[TMP0]], i32 [[TMP6]]) #[[ATTR3:[0-9]+]]
-// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK6-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK6-NEXT:    ret void
 // CHECK6:       worker.exit:
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
 // CHECK6-NEXT:  entry:

 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
 // CHECK6-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK6:       omp.inner.for.body:
 // CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
-// CHECK6-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR4:[0-9]+]]
-// CHECK6-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK6-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I4]]) #[[ATTR5:[0-9]+]]
+// CHECK6-NEXT:    [[CALL8:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR5]]
 // CHECK6-NEXT:    [[ADD9:%.*]] = add nsw i32 [[CALL]], [[CALL8]]
-// CHECK6-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR4]]
+// CHECK6-NEXT:    [[CALL10:%.*]] = call i32 @_Z3fooPi(i32* [[ARGC_ADDR]]) #[[ATTR5]]
 // CHECK6-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CALL10]]
 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[TMP0]], align 4
 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK6:       omp.body.continue:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK6:       omp.inner.for.inc:
 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP20]], 1

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp

 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[L]], i64* [[L_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[L_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[L_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV3]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[L_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i32]* [[TMP0]], i64 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[L:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK1:       .omp.lastprivate.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]+}}__Z9ftemplateIiET_i_l32
-// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
 // CHECK1-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i64 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[F_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[F]], i64* [[F_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[F_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[F_CASTED]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[F_CASTED]], align 8
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK1-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i64 [[TMP4]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i64 [[F:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-NEXT:    [[L_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[L]], i32* [[L_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[L_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[L_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[L_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[L:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK2:       .omp.lastprivate.done:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l32
-// CHECK2-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK2-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 4
 // CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[F_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[F_CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[F]], i32* [[F_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[F_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[F_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[F_CASTED]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK2-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[F:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-NEXT:    [[L_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[L]], i32* [[L_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4:[0-9]+]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[L_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[L_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[L_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i32]* [[TMP0]], i32 [[TMP6]]) #[[ATTR2:[0-9]+]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[L:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // CHECK3:       .omp.lastprivate.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]+}}__Z9ftemplateIiET_i_l32
-// CHECK3-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR2:[0-9]+]] {
+// CHECK3-SAME: (i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca [1000 x i16]*, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store [1000 x i16]* [[AA]], [1000 x i16]** [[AA_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i16]*, [1000 x i16]** [[AA_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32 [[TMP4]], [1000 x i16]* [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__2
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i16]* nonnull align 2 dereferenceable(2000) [[AA:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x i32]* [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__4
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[F_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[F_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store [10 x [10 x i32]]* [[C]], [10 x [10 x i32]]** [[C_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[F]], i32* [[F_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[C_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 false)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 false)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB4]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[F_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[F_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[F_CASTED]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
+// CHECK3-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], [10 x [10 x i32]]* [[TMP0]], i32 [[TMP4]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__6
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x i32]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[F:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

clang/test/OpenMP/nvptx_target_teams_distribute_simd_codegen.cpp

int bar(int n){
   int a = 0;
 
   a += ftemplate<int>(n);
 
   return a;
 }
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l37(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 
 // CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l43(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 
 // CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l48(
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 
 // CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
 // CHECK: ret void
 
 // CHECK: define {{.*}}void {{@__omp_offloading_.+}}_l53({{.+}}, i{{32|64}} [[F_IN:%.+]])
 // CHECK: store {{.+}} [[F_IN]], {{.+}}* {{.+}},
-// CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
-// CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
+// CHECK: call i32 @__kmpc_target_init({{.*}}, i8 2, i1 false, i1 false)
+// CHECK: call void @__kmpc_target_deinit({{.*}}, i8 2, i1 false)
 
 // CHECK: store {{.+}} 99, {{.+}}* [[COMB_UB:%.+]], align
 // CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91, {{.+}}, {{.+}}, {{.+}}* [[COMB_UB]],
 // CHECK: call void @__kmpc_for_static_fini(
 // CHECK: ret void
 
 #endif

clang/test/OpenMP/nvptx_teams_codegen.cpp

 // CHECK1-SAME: (i64 [[ARGC:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK1-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i32*
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[ARGC_ON_STACK]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR1:[0-9]+]]
+// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR2:[0-9]+]]
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i64 4)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIPPcEiT__l15
 // CHECK1-SAME: (i8** [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i8**, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i8** [[ARGC]], i8*** [[ARGC_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i64 8)
 // CHECK1-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i8***
 // CHECK1-NEXT:    store i8** [[TMP1]], i8*** [[ARGC_ON_STACK]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR2]]
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i64 8)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l23
 // CHECK2-SAME: (i32 [[ARGC:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK2-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i32*
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[ARGC_ON_STACK]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR1:[0-9]+]]
+// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR2:[0-9]+]]
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i32 4)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIPPcEiT__l15
 // CHECK2-SAME: (i8** [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i8**, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i8** [[ARGC]], i8*** [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK2-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i8***
 // CHECK2-NEXT:    store i8** [[TMP1]], i8*** [[ARGC_ON_STACK]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR2]]
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i32 4)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK3-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
 // CHECK3-NEXT:    store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i32*
 // CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV2]], align 8
 // CHECK3-NEXT:    [[ARGC3:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK3-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC3]] to i32*
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[ARGC_ON_STACK]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR1:[0-9]+]]
+// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR2:[0-9]+]]
 // CHECK3-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC3]], i64 4)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK3-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
 // CHECK3-NEXT:    store i8** [[ARGC]], i8*** [[ARGC_ADDR]], align 8
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i32*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGC_ADDR]], align 8
 // CHECK3-NEXT:    [[ARGC2:%.*]] = call i8* @__kmpc_alloc_shared(i64 8)
 // CHECK3-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC2]] to i8***
 // CHECK3-NEXT:    store i8** [[TMP1]], i8*** [[ARGC_ON_STACK]], align 8
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR2]]
 // CHECK3-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC2]], i64 8)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK4-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i32*
 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[ARGC_ON_STACK]], align 4
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR1:[0-9]+]]
+// CHECK4-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[ARGC_ON_STACK]]) #[[ATTR2:[0-9]+]]
 // CHECK4-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i32 4)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i8**, align 4
 // CHECK4-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK4-NEXT:    store i8** [[ARGC]], i8*** [[ARGC_ADDR]], align 4
-// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK4-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK4-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK4:       user_code.entry:
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    [[ARGC1:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK4-NEXT:    [[ARGC_ON_STACK:%.*]] = bitcast i8* [[ARGC1]] to i8***
 // CHECK4-NEXT:    store i8** [[TMP1]], i8*** [[ARGC_ON_STACK]], align 4
 // CHECK4-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK4-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR1]]
+// CHECK4-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8*** [[ARGC_ON_STACK]]) #[[ATTR2]]
 // CHECK4-NEXT:    call void @__kmpc_free_shared(i8* [[ARGC1]], i32 4)
-// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK4-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK4-NEXT:    ret void
 // CHECK4:       worker.exit:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:

clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp

 // CHECK1-SAME: (i64 [[E:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[E_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[E]], i64* [[E_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[E_ADDR]] to double*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load double, double* [[CONV]], align 8
 // CHECK1-NEXT:    [[E1:%.*]] = call i8* @__kmpc_alloc_shared(i64 8)
 // CHECK1-NEXT:    [[E_ON_STACK:%.*]] = bitcast i8* [[E1]] to double*
 // CHECK1-NEXT:    store double [[TMP1]], double* [[E_ON_STACK]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E_ON_STACK]]) #[[ATTR1:[0-9]+]]
+// CHECK1-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E_ON_STACK]]) #[[ATTR4:[0-9]+]]
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[E1]], i64 8)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK1-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
 // CHECK1-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
 // CHECK1-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
 // CHECK1-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK1:       then:
 // CHECK1-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
 // CHECK1-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
 // CHECK1-NEXT:    br label [[IFCONT:%.*]]
 // CHECK1:       else:
 // CHECK1-NEXT:    br label [[IFCONT]]
 // CHECK1:       ifcont:
 // CHECK1-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK1-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
 // CHECK1-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]

 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK1-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1024 x double], [1024 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
 // CHECK1-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK1-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1024 x double], [1024 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l26
 // CHECK1-SAME: (i64 [[C:%.*]], i64 [[D:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[C_ADDR]] to i8*
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[D_ADDR]] to float*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 8
 // CHECK1-NEXT:    [[C2:%.*]] = call i8* @__kmpc_alloc_shared(i64 1)
 // CHECK1-NEXT:    store i8 [[TMP1]], i8* [[C2]], align 1
 // CHECK1-NEXT:    [[TMP2:%.*]] = load float, float* [[CONV1]], align 8
 // CHECK1-NEXT:    [[D3:%.*]] = call i8* @__kmpc_alloc_shared(i64 4)
 // CHECK1-NEXT:    [[D_ON_STACK:%.*]] = bitcast i8* [[D3]] to float*
 // CHECK1-NEXT:    store float [[TMP2]], float* [[D_ON_STACK]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR4]]
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[D3]], i64 4)
 // CHECK1-NEXT:    call void @__kmpc_free_shared(i8* [[C2]], i64 1)
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK1-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
 // CHECK1-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
 // CHECK1-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
 // CHECK1-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK1:       then:
 // CHECK1-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
 // CHECK1-NEXT:    br label [[IFCONT:%.*]]
 // CHECK1:       else:
 // CHECK1-NEXT:    br label [[IFCONT]]
 // CHECK1:       ifcont:
 // CHECK1-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK1-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
 // CHECK1-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
 // CHECK1-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1024 x float], [1024 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func7
 // CHECK1-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
 // CHECK1-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1024 x float], [1024 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l33
 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[CONV]], i16* [[CONV1]]) #[[ATTR1]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK1-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[CONV]], i16* [[CONV1]]) #[[ATTR4]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined__9
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK1-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK1-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK1-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK1-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK1:       then:
 // CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK1-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK1-NEXT:    br label [[IFCONT:%.*]]
 // CHECK1:       else:
 // CHECK1-NEXT:    br label [[IFCONT]]
 // CHECK1:       ifcont:
 // CHECK1-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK1-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK1-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK1-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK1-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK1-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK1-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK1:       then:
 // CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK1-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK1-NEXT:    br label [[IFCONT:%.*]]
 // CHECK1:       else:
 // CHECK1-NEXT:    br label [[IFCONT]]
 // CHECK1:       ifcont:
 // CHECK1-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK1-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK1-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
 // CHECK1-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1024 x i16], [1024 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func19
 // CHECK1-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
 // CHECK1-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1024 x i16], [1024 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 8
-// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK1-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l20
 // CHECK2-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
 // CHECK2-NEXT:    [[E1:%.*]] = alloca double, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    [[TMP3:%.*]] = load double, double* [[TMP0]], align 8
 // CHECK2-NEXT:    store double [[TMP3]], double* [[E1]], align 8
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E1]]) #[[ATTR1:[0-9]+]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E1]]) #[[ATTR4:[0-9]+]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK2-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
 // CHECK2-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
 // CHECK2-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
 // CHECK2-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK2:       then:
 // CHECK2-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
 // CHECK2-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
 // CHECK2-NEXT:    br label [[IFCONT:%.*]]
 // CHECK2:       else:
 // CHECK2-NEXT:    br label [[IFCONT]]
 // CHECK2:       ifcont:
 // CHECK2-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK2-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
 // CHECK2-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]

 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
 // CHECK2-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1024 x double], [1024 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
 // CHECK2-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
 // CHECK2-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1024 x double], [1024 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l26
 // CHECK2-SAME: (i32 [[C:%.*]], i32 [[D:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[C_ADDR]] to i8*
 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i32* [[D_ADDR]] to float*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 4
 // CHECK2-NEXT:    [[C2:%.*]] = call i8* @__kmpc_alloc_shared(i32 1)
 // CHECK2-NEXT:    store i8 [[TMP1]], i8* [[C2]], align 1
 // CHECK2-NEXT:    [[TMP2:%.*]] = load float, float* [[CONV1]], align 4
 // CHECK2-NEXT:    [[D3:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK2-NEXT:    [[D_ON_STACK:%.*]] = bitcast i8* [[D3]] to float*
 // CHECK2-NEXT:    store float [[TMP2]], float* [[D_ON_STACK]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR4]]
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[D3]], i32 4)
 // CHECK2-NEXT:    call void @__kmpc_free_shared(i8* [[C2]], i32 1)
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK2-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
 // CHECK2-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
 // CHECK2-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
 // CHECK2-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK2:       then:
 // CHECK2-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK2-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
 // CHECK2-NEXT:    br label [[IFCONT:%.*]]
 // CHECK2:       else:
 // CHECK2-NEXT:    br label [[IFCONT]]
 // CHECK2:       ifcont:
 // CHECK2-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK2-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
 // CHECK2-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK2-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK2-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1024 x float], [1024 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func7
 // CHECK2-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK2-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK2-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1024 x float], [1024 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l33
 // CHECK2-SAME: (i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
-// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[A_ADDR]], i16* [[CONV]]) #[[ATTR1]]
-// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK2-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[A_ADDR]], i16* [[CONV]]) #[[ATTR4]]
+// CHECK2-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__omp_outlined__9
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:

 // CHECK2-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK2-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK2-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK2-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK2-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK2:       then:
 // CHECK2-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK2-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK2-NEXT:    br label [[IFCONT:%.*]]
 // CHECK2:       else:
 // CHECK2-NEXT:    br label [[IFCONT]]
 // CHECK2:       ifcont:
 // CHECK2-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK2-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK2-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK2-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK2-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK2-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK2-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK2-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK2:       then:
 // CHECK2-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK2-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK2-NEXT:    br label [[IFCONT:%.*]]
 // CHECK2:       else:
 // CHECK2-NEXT:    br label [[IFCONT]]
 // CHECK2:       ifcont:
 // CHECK2-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK2-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK2-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK2-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK2-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1024 x i16], [1024 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 4
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func19
 // CHECK2-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK2-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK2-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1024 x i16], [1024 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 4
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK2-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l20
 // CHECK3-SAME: (double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 4
 // CHECK3-NEXT:    [[E1:%.*]] = alloca double, align 8
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store double* [[E]], double** [[E_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP3:%.*]] = load double, double* [[TMP0]], align 8
 // CHECK3-NEXT:    store double [[TMP3]], double* [[E1]], align 8
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E1]]) #[[ATTR1:[0-9]+]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], double* [[E1]]) #[[ATTR4:[0-9]+]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[E:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP31:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK3-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
 // CHECK3-NEXT:    [[TMP33:%.*]] = or i1 [[TMP23]], [[TMP26]]
 // CHECK3-NEXT:    [[TMP34:%.*]] = or i1 [[TMP33]], [[TMP32]]
 // CHECK3-NEXT:    br i1 [[TMP34]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK3:       then:
 // CHECK3-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
 // CHECK3-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP35]], i8* [[TMP36]]) #[[ATTR4]]
 // CHECK3-NEXT:    br label [[IFCONT:%.*]]
 // CHECK3:       else:
 // CHECK3-NEXT:    br label [[IFCONT]]
 // CHECK3:       ifcont:
 // CHECK3-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK3-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
 // CHECK3-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]

 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
 // CHECK3-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2048 x double], [2048 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP9]], i8* [[TMP10]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
 // CHECK3-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTADDR1]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
 // CHECK3-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], %struct._globalized_locals_ty* [[TMP4]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2048 x double], [2048 x double]* [[E]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast double* [[TMP7]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP10]], i8* [[TMP9]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l26
 // CHECK3-SAME: (i32 [[C:%.*]], i32 [[D:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[C_ADDR]] to i8*
 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[D_ADDR]] to float*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 4
 // CHECK3-NEXT:    [[C2:%.*]] = call i8* @__kmpc_alloc_shared(i32 1)
 // CHECK3-NEXT:    store i8 [[TMP1]], i8* [[C2]], align 1
 // CHECK3-NEXT:    [[TMP2:%.*]] = load float, float* [[CONV1]], align 4
 // CHECK3-NEXT:    [[D3:%.*]] = call i8* @__kmpc_alloc_shared(i32 4)
 // CHECK3-NEXT:    [[D_ON_STACK:%.*]] = bitcast i8* [[D3]] to float*
 // CHECK3-NEXT:    store float [[TMP2]], float* [[D_ON_STACK]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @__omp_outlined__1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i8* [[C2]], float* [[D_ON_STACK]]) #[[ATTR4]]
 // CHECK3-NEXT:    call void @__kmpc_free_shared(i8* [[D3]], i32 4)
 // CHECK3-NEXT:    call void @__kmpc_free_shared(i8* [[C2]], i32 1)
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__1
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8* nonnull align 1 dereferenceable(1) [[C:%.*]], float* nonnull align 4 dereferenceable(4) [[D:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP42:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK3-NEXT:    [[TMP43:%.*]] = and i1 [[TMP41]], [[TMP42]]
 // CHECK3-NEXT:    [[TMP44:%.*]] = or i1 [[TMP34]], [[TMP37]]
 // CHECK3-NEXT:    [[TMP45:%.*]] = or i1 [[TMP44]], [[TMP43]]
 // CHECK3-NEXT:    br i1 [[TMP45]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK3:       then:
 // CHECK3-NEXT:    [[TMP46:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP46]], i8* [[TMP47]]) #[[ATTR4]]
 // CHECK3-NEXT:    br label [[IFCONT:%.*]]
 // CHECK3:       else:
 // CHECK3-NEXT:    br label [[IFCONT]]
 // CHECK3:       ifcont:
 // CHECK3-NEXT:    [[TMP48:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK3-NEXT:    [[TMP49:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP50:%.*]] = and i1 [[TMP48]], [[TMP49]]
 // CHECK3-NEXT:    br i1 [[TMP50]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK3-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK3-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2048 x float], [2048 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP11]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func7
 // CHECK3-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK3-NEXT:    store i8* [[TMP7]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK3-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], %struct._globalized_locals_ty.0* [[TMP4]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2048 x float], [2048 x float]* [[D]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP10]], i8** [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func2"(i8* [[TMP12]], i8* [[TMP11]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIcET_i_l33
 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 true, i1 false, i1 true)
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 false, i1 true)
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[A_ADDR]], i16* [[CONV]]) #[[ATTR1]]
-// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+// CHECK3-NEXT:    call void @__omp_outlined__9(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]], i32* [[A_ADDR]], i16* [[CONV]]) #[[ATTR4]]
+// CHECK3-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__omp_outlined__9
 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:

 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK3-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK3-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK3-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK3-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK3:       then:
 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK3-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func11"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK3-NEXT:    br label [[IFCONT:%.*]]
 // CHECK3:       else:
 // CHECK3-NEXT:    br label [[IFCONT]]
 // CHECK3:       ifcont:
 // CHECK3-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK3-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK3-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp sgt i16 [[TMP7]], 0
 // CHECK3-NEXT:    [[TMP44:%.*]] = and i1 [[TMP42]], [[TMP43]]
 // CHECK3-NEXT:    [[TMP45:%.*]] = or i1 [[TMP35]], [[TMP38]]
 // CHECK3-NEXT:    [[TMP46:%.*]] = or i1 [[TMP45]], [[TMP44]]
 // CHECK3-NEXT:    br i1 [[TMP46]], label [[THEN:%.*]], label [[ELSE:%.*]]
 // CHECK3:       then:
 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [2 x i8*]* [[TMP5]] to i8*
 // CHECK3-NEXT:    [[TMP48:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP47]], i8* [[TMP48]]) #[[ATTR4]]
 // CHECK3-NEXT:    br label [[IFCONT:%.*]]
 // CHECK3:       else:
 // CHECK3-NEXT:    br label [[IFCONT]]
 // CHECK3:       ifcont:
 // CHECK3-NEXT:    [[TMP49:%.*]] = icmp eq i16 [[TMP8]], 1
 // CHECK3-NEXT:    [[TMP50:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP51:%.*]] = and i1 [[TMP49]], [[TMP50]]
 // CHECK3-NEXT:    br i1 [[TMP51]], label [[THEN6:%.*]], label [[ELSE7:%.*]]

 // CHECK3-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK3-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2048 x i16], [2048 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP12]], i8* [[TMP13]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func19
 // CHECK3-SAME: (i8* [[TMP0:%.*]], i32 [[TMP1:%.*]], i8* [[TMP2:%.*]]) #[[ATTR2]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4

 // CHECK3-NEXT:    store i8* [[TMP8]], i8** [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
 // CHECK3-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], %struct._globalized_locals_ty.1* [[TMP4]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2048 x i16], [2048 x i16]* [[B]], i32 0, i32 [[TMP5]]
 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to i8*
 // CHECK3-NEXT:    store i8* [[TMP11]], i8** [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[DOTADDR2]], align 4
-// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR1]]
+// CHECK3-NEXT:    call void @"_omp$reduction$reduction_func14"(i8* [[TMP13]], i8* [[TMP12]]) #[[ATTR4]]
 // CHECK3-NEXT:    ret void
 //

clang/test/OpenMP/target_parallel_debug_codegen.cpp

 // CHECK1-NEXT:    [[TMP2:%.*]] = load [10 x [10 x [10 x i32]]]*, [10 x [10 x [10 x i32]]]** [[TMP]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    [[TMP3:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[B_ADDR]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP3]], [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    [[TMP4:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    [[TMP6:%.*]] = addrspacecast i8 addrspace(1)* [[TMP5]] to i8*, !dbg [[DBG47]]
 // CHECK1-NEXT:    store i8* [[TMP6]], i8** [[_TMP2]], align 8, !dbg [[DBG47]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[_TMP2]], align 8, !dbg [[DBG47]]
-// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true), !dbg [[DBG47]]
+// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true), !dbg [[DBG47]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP8]], -1, !dbg [[DBG47]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG47]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG48:![0-9]+]]
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*, !dbg [[DBG48]]
 // CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CONV]], align 4, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[A_CASTED]], align 8, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG48]]
 // CHECK1-NEXT:    store i8* [[TMP13]], i8** [[TMP12]], align 8, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP11]] to i8*, !dbg [[DBG48]]
 // CHECK1-NEXT:    store i8* [[TMP15]], i8** [[TMP14]], align 8, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast [10 x [10 x i32]]* [[TMP4]] to i8*, !dbg [[DBG48]]
 // CHECK1-NEXT:    store i8* [[TMP17]], i8** [[TMP16]], align 8, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG48]]
 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[TMP18]], align 8, !dbg [[DBG48]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG48]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i64, [10 x [10 x i32]]*, i8*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP19]], i64 4), !dbg [[DBG48]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB5:[0-9]+]], i1 true, i1 true), !dbg [[DBG49:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB5:[0-9]+]], i8 2, i1 true), !dbg [[DBG49:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG51:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG47]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 [[A:%.*]], [10 x [10 x i32]]* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG52:![0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP3:%.*]] = load [10 x [10 x i32]] addrspace(1)*, [10 x [10 x i32]] addrspace(1)** [[B_ADDR]], align 8, !dbg [[DBG137]]
 // CHECK1-NEXT:    [[TMP4:%.*]] = addrspacecast [10 x [10 x i32]] addrspace(1)* [[TMP3]] to [10 x [10 x i32]]*, !dbg [[DBG137]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP4]], [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG137]]
 // CHECK1-NEXT:    [[TMP5:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG137]]
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG137]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = addrspacecast i8 addrspace(1)* [[TMP6]] to i8*, !dbg [[DBG137]]
 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[_TMP2]], align 8, !dbg [[DBG137]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[_TMP2]], align 8, !dbg [[DBG137]]
-// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB7:[0-9]+]], i1 true, i1 false, i1 true), !dbg [[DBG137]]
+// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB7:[0-9]+]], i8 2, i1 false, i1 true), !dbg [[DBG137]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP9]], -1, !dbg [[DBG137]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG137]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB9:[0-9]+]])
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG138:![0-9]+]]
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*, !dbg [[DBG138]]
 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG138]]
 // CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = inttoptr i64 [[TMP12]] to i8*, !dbg [[DBG138]]
 // CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast [10 x [10 x i32]]* [[TMP5]] to i8*, !dbg [[DBG138]]
 // CHECK1-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 8, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG138]]
 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP19]], align 8, !dbg [[DBG138]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG138]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB9]], i32 [[TMP10]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i64, [10 x [10 x i32]]*, i8*)* @__omp_outlined__2 to i8*), i8* null, i8** [[TMP20]], i64 4), !dbg [[DBG138]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB11:[0-9]+]], i1 true, i1 true), !dbg [[DBG139:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB11:[0-9]+]], i8 2, i1 true), !dbg [[DBG139:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG141:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG137]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 [[A:%.*]], [10 x [10 x i32]] addrspace(1)* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG142:![0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP6:%.*]] = load [10 x [10 x i32]] addrspace(1)*, [10 x [10 x i32]] addrspace(1)** [[B_ADDR]], align 8, !dbg [[DBG212]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = addrspacecast [10 x [10 x i32]] addrspace(1)* [[TMP6]] to [10 x [10 x i32]]*, !dbg [[DBG212]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP7]], [10 x [10 x i32]]** [[_TMP2]], align 8, !dbg [[DBG212]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP2]], align 8, !dbg [[DBG212]]
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG212]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = addrspacecast i8 addrspace(1)* [[TMP9]] to i8*, !dbg [[DBG212]]
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[_TMP3]], align 8, !dbg [[DBG212]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[_TMP3]], align 8, !dbg [[DBG212]]
-// CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB13:[0-9]+]], i1 true, i1 false, i1 true), !dbg [[DBG212]]
+// CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB13:[0-9]+]], i8 2, i1 false, i1 true), !dbg [[DBG212]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP12]], -1, !dbg [[DBG212]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG212]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB15:[0-9]+]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG213:![0-9]+]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG213]]
 // CHECK1-NEXT:    store i8* [[TMP15]], i8** [[TMP14]], align 8, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP5]] to i8*, !dbg [[DBG213]]
 // CHECK1-NEXT:    store i8* [[TMP17]], i8** [[TMP16]], align 8, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast [10 x [10 x i32]]* [[TMP8]] to i8*, !dbg [[DBG213]]
 // CHECK1-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 8, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG213]]
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP20]], align 8, !dbg [[DBG213]]
 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG213]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB15]], i32 [[TMP13]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i32*, [10 x [10 x i32]]*, i8*)* @__omp_outlined__4 to i8*), i8* null, i8** [[TMP21]], i64 4), !dbg [[DBG213]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB17:[0-9]+]], i1 true, i1 true), !dbg [[DBG214:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB17:[0-9]+]], i8 2, i1 true), !dbg [[DBG214:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG216:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG212]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__3
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 addrspace(1)* noalias [[A:%.*]], [10 x [10 x i32]] addrspace(1)* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG217:![0-9]+]] {
 // CHECK1-NEXT:  entry:

clang/test/OpenMP/target_parallel_for_debug_codegen.cpp

 // CHECK1-NEXT:    [[TMP2:%.*]] = load [10 x [10 x [10 x i32]]]*, [10 x [10 x [10 x i32]]]** [[TMP]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    [[TMP3:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[B_ADDR]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP3]], [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    [[TMP4:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    [[TMP6:%.*]] = addrspacecast i8 addrspace(1)* [[TMP5]] to i8*, !dbg [[DBG41]]
 // CHECK1-NEXT:    store i8* [[TMP6]], i8** [[_TMP2]], align 8, !dbg [[DBG41]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[_TMP2]], align 8, !dbg [[DBG41]]
-// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false), !dbg [[DBG41]]
+// CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false), !dbg [[DBG41]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP8]], -1, !dbg [[DBG41]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG41]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG42:![0-9]+]]
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*, !dbg [[DBG42]]
 // CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CONV]], align 4, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[A_CASTED]], align 8, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG42]]
 // CHECK1-NEXT:    store i8* [[TMP13]], i8** [[TMP12]], align 8, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP11]] to i8*, !dbg [[DBG42]]
 // CHECK1-NEXT:    store i8* [[TMP15]], i8** [[TMP14]], align 8, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast [10 x [10 x i32]]* [[TMP4]] to i8*, !dbg [[DBG42]]
 // CHECK1-NEXT:    store i8* [[TMP17]], i8** [[TMP16]], align 8, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG42]]
 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[TMP18]], align 8, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR__ADDR]], align 1, !dbg [[DBG43:![0-9]+]]
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP19]] to i1, !dbg [[DBG43]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = zext i1 [[TOBOOL]] to i32, !dbg [[DBG42]]
 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG42]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB6]], i32 [[TMP9]], i32 [[TMP20]], i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i64, [10 x [10 x i32]]*, i8*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP21]], i64 4), !dbg [[DBG42]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB8:[0-9]+]], i1 true, i1 false), !dbg [[DBG45:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB8:[0-9]+]], i8 2, i1 false), !dbg [[DBG45:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG46:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG41]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 [[A:%.*]], [10 x [10 x i32]]* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG47:![0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP3:%.*]] = load [10 x [10 x i32]] addrspace(1)*, [10 x [10 x i32]] addrspace(1)** [[B_ADDR]], align 8, !dbg [[DBG146]]
 // CHECK1-NEXT:    [[TMP4:%.*]] = addrspacecast [10 x [10 x i32]] addrspace(1)* [[TMP3]] to [10 x [10 x i32]]*, !dbg [[DBG146]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP4]], [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG146]]
 // CHECK1-NEXT:    [[TMP5:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP1]], align 8, !dbg [[DBG146]]
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG146]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = addrspacecast i8 addrspace(1)* [[TMP6]] to i8*, !dbg [[DBG146]]
 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[_TMP2]], align 8, !dbg [[DBG146]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8*, i8** [[_TMP2]], align 8, !dbg [[DBG146]]
-// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB10:[0-9]+]], i1 true, i1 false, i1 false), !dbg [[DBG146]]
+// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB10:[0-9]+]], i8 2, i1 false, i1 false), !dbg [[DBG146]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP9]], -1, !dbg [[DBG146]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG146]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB15:[0-9]+]])
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG147:![0-9]+]]
 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*, !dbg [[DBG147]]
 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG147]]
 // CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = inttoptr i64 [[TMP12]] to i8*, !dbg [[DBG147]]
 // CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast [10 x [10 x i32]]* [[TMP5]] to i8*, !dbg [[DBG147]]
 // CHECK1-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 8, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG147]]
 // CHECK1-NEXT:    store i8* [[TMP8]], i8** [[TMP19]], align 8, !dbg [[DBG147]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG147]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB15]], i32 [[TMP10]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i64, [10 x [10 x i32]]*, i8*)* @__omp_outlined__2 to i8*), i8* null, i8** [[TMP20]], i64 4), !dbg [[DBG147]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB17:[0-9]+]], i1 true, i1 false), !dbg [[DBG148:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB17:[0-9]+]], i8 2, i1 false), !dbg [[DBG148:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG150:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG146]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__1
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 [[A:%.*]], [10 x [10 x i32]] addrspace(1)* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG151:![0-9]+]] {
 // CHECK1-NEXT:  entry:

 // CHECK1-NEXT:    [[TMP6:%.*]] = load [10 x [10 x i32]] addrspace(1)*, [10 x [10 x i32]] addrspace(1)** [[B_ADDR]], align 8, !dbg [[DBG236]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = addrspacecast [10 x [10 x i32]] addrspace(1)* [[TMP6]] to [10 x [10 x i32]]*, !dbg [[DBG236]]
 // CHECK1-NEXT:    store [10 x [10 x i32]]* [[TMP7]], [10 x [10 x i32]]** [[_TMP2]], align 8, !dbg [[DBG236]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load [10 x [10 x i32]]*, [10 x [10 x i32]]** [[_TMP2]], align 8, !dbg [[DBG236]]
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8 addrspace(1)*, i8 addrspace(1)** [[BB_ADDR]], align 8, !dbg [[DBG236]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = addrspacecast i8 addrspace(1)* [[TMP9]] to i8*, !dbg [[DBG236]]
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[_TMP3]], align 8, !dbg [[DBG236]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[_TMP3]], align 8, !dbg [[DBG236]]
-// CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB19:[0-9]+]], i1 true, i1 false, i1 false), !dbg [[DBG236]]
+// CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB19:[0-9]+]], i8 2, i1 false, i1 false), !dbg [[DBG236]]
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP12]], -1, !dbg [[DBG236]]
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]], !dbg [[DBG236]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB24:[0-9]+]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0, !dbg [[DBG237:![0-9]+]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast [10 x [10 x [10 x i32]]]* [[TMP2]] to i8*, !dbg [[DBG237]]
 // CHECK1-NEXT:    store i8* [[TMP15]], i8** [[TMP14]], align 8, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP5]] to i8*, !dbg [[DBG237]]
 // CHECK1-NEXT:    store i8* [[TMP17]], i8** [[TMP16]], align 8, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast [10 x [10 x i32]]* [[TMP8]] to i8*, !dbg [[DBG237]]
 // CHECK1-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 8, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3, !dbg [[DBG237]]
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP20]], align 8, !dbg [[DBG237]]
 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**, !dbg [[DBG237]]
 // CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB24]], i32 [[TMP13]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, [10 x [10 x [10 x i32]]]*, i32*, [10 x [10 x i32]]*, i8*)* @__omp_outlined__4 to i8*), i8* null, i8** [[TMP21]], i64 4), !dbg [[DBG237]]
-// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB26:[0-9]+]], i1 true, i1 false), !dbg [[DBG238:![0-9]+]]
+// CHECK1-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB26:[0-9]+]], i8 2, i1 false), !dbg [[DBG238:![0-9]+]]
 // CHECK1-NEXT:    ret void, !dbg [[DBG240:![0-9]+]]
 // CHECK1:       worker.exit:
 // CHECK1-NEXT:    ret void, !dbg [[DBG236]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__omp_outlined___debug__3
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x i32]]] addrspace(1)* noalias [[C:%.*]], i32 addrspace(1)* noalias [[A:%.*]], [10 x [10 x i32]] addrspace(1)* noalias [[B:%.*]], i8 addrspace(1)* noalias [[BB:%.*]]) #[[ATTR0]] !dbg [[DBG241:![0-9]+]] {
 // CHECK1-NEXT:  entry:

llvm/include/llvm/Frontend/OpenMP/OMPKinds.def

__OMP_RTL(__tgt_target_data_update_nowait_mapper, false, Void, IdentPtr, Int64, Int32,
           VoidPtrPtr, VoidPtrPtr, Int64Ptr, Int64Ptr, VoidPtrPtr, VoidPtrPtr)
 __OMP_RTL(__tgt_mapper_num_components, false, Int64, VoidPtr)
 __OMP_RTL(__tgt_push_mapper_component, false, Void, VoidPtr, VoidPtr, VoidPtr,
           Int64, Int64, VoidPtr)
 __OMP_RTL(__kmpc_task_allow_completion_event, false, VoidPtr, IdentPtr,
           /* Int */ Int32, /* kmp_task_t */ VoidPtr)
 
 /// OpenMP Device runtime functions
-__OMP_RTL(__kmpc_target_init, false, Int32, IdentPtr, Int1, Int1, Int1)
-__OMP_RTL(__kmpc_target_deinit, false, Void, IdentPtr, Int1, Int1)
+__OMP_RTL(__kmpc_target_init, false, Int32, IdentPtr, Int8, Int1, Int1)
+__OMP_RTL(__kmpc_target_deinit, false, Void, IdentPtr, Int8, Int1)
 __OMP_RTL(__kmpc_kernel_prepare_parallel, false, Void, VoidPtr)
 __OMP_RTL(__kmpc_parallel_51, false, Void, IdentPtr, Int32, Int32, Int32, Int32,
           VoidPtr, VoidPtr, VoidPtrPtr, SizeTy)
 __OMP_RTL(__kmpc_kernel_parallel, false, Int1, VoidPtrPtr)
 __OMP_RTL(__kmpc_kernel_end_parallel, false, Void, )
 __OMP_RTL(__kmpc_serialized_parallel, false, Void, IdentPtr, Int32)
 __OMP_RTL(__kmpc_end_serialized_parallel, false, Void, IdentPtr, Int32)
 __OMP_RTL(__kmpc_shuffle_int32, false, Int32, Int32, Int16, Int16)

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

 OpenMPIRBuilder::InsertPointTy
 OpenMPIRBuilder::createTargetInit(const LocationDescription &Loc, bool IsSPMD,
                                   bool RequiresFullRuntime) {
   if (!updateToLocation(Loc))
     return Loc.IP;
 
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc);
   Value *Ident = getOrCreateIdent(SrcLocStr);
-  ConstantInt *IsSPMDVal = ConstantInt::getBool(Int32->getContext(), IsSPMD);
+  ConstantInt *IsSPMDVal = ConstantInt::getSigned(
+      IntegerType::getInt8Ty(Int8->getContext()),
+      IsSPMD ? OMP_TGT_EXEC_MODE_SPMD : OMP_TGT_EXEC_MODE_GENERIC);
   ConstantInt *UseGenericStateMachine =
       ConstantInt::getBool(Int32->getContext(), !IsSPMD);
   ConstantInt *RequiresFullRuntimeVal =
       ConstantInt::getBool(Int32->getContext(), RequiresFullRuntime);
 
   Function *Fn = getOrCreateRuntimeFunctionPtr(
       omp::RuntimeFunction::OMPRTL___kmpc_target_init);
 

 void OpenMPIRBuilder::createTargetDeinit(const LocationDescription &Loc,
                                          bool IsSPMD,
                                          bool RequiresFullRuntime) {
   if (!updateToLocation(Loc))
     return;
 
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc);
   Value *Ident = getOrCreateIdent(SrcLocStr);
-  ConstantInt *IsSPMDVal = ConstantInt::getBool(Int32->getContext(), IsSPMD);
+  ConstantInt *IsSPMDVal = ConstantInt::getSigned(
+      IntegerType::getInt8Ty(Int8->getContext()),
+      IsSPMD ? OMP_TGT_EXEC_MODE_SPMD : OMP_TGT_EXEC_MODE_GENERIC);
   ConstantInt *RequiresFullRuntimeVal =
       ConstantInt::getBool(Int32->getContext(), RequiresFullRuntime);
 
   Function *Fn = getOrCreateRuntimeFunctionPtr(
       omp::RuntimeFunction::OMPRTL___kmpc_target_deinit);
 
   Builder.CreateCall(Fn, {Ident, IsSPMDVal, RequiresFullRuntimeVal});
 }

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

if (!C)
       return false;
 
     // Match: -1 == __kmpc_target_init (for non-SPMD kernels only!)
     if (C->isAllOnesValue()) {
       auto *CB = dyn_cast<CallBase>(Cmp->getOperand(0));
       CB = CB ? OpenMPOpt::getCallIfRegularCall(*CB, &RFI) : nullptr;
       if (!CB)
         return false;
-      const int InitIsSPMDArgNo = 1;
-      auto *IsSPMDModeCI =
-          dyn_cast<ConstantInt>(CB->getOperand(InitIsSPMDArgNo));
-      return IsSPMDModeCI && IsSPMDModeCI->isZero();
+      const int InitModeArgNo = 1;
+      auto *ModeCI = dyn_cast<ConstantInt>(CB->getOperand(InitModeArgNo));
+      return ModeCI && (ModeCI->getSExtValue() & OMP_TGT_EXEC_MODE_GENERIC);
     }
 
     if (C->isZero()) {
       // Match: 0 == llvm.nvvm.read.ptx.sreg.tid.x()
       if (auto *II = dyn_cast<IntrinsicInst>(Cmp->getOperand(0)))
         if (II->getIntrinsicID() == Intrinsic::nvvm_read_ptx_sreg_tid_x)
           return true;
 
Attributor::SimplifictionCallbackTy StateMachineSimplifyCB =
       if (AA)
         A.recordDependence(*this, *AA, DepClassTy::OPTIONAL);
       UsedAssumedInformation = !isAtFixpoint();
       auto *FalseVal =
           ConstantInt::getBool(IRP.getAnchorValue().getContext(), 0);
       return FalseVal;
     };
 
-    Attributor::SimplifictionCallbackTy IsSPMDModeSimplifyCB =
+    Attributor::SimplifictionCallbackTy ModeSimplifyCB =
         [&](const IRPosition &IRP, const AbstractAttribute *AA,
             bool &UsedAssumedInformation) -> Optional<Value *> {
       // IRP represents the "SPMDCompatibilityTracker" argument of an
       // __kmpc_target_init or
       // __kmpc_target_deinit call. We will answer this one with the internal
       // state.
       if (!SPMDCompatibilityTracker.isValidState())
         return nullptr;
       if (!SPMDCompatibilityTracker.isAtFixpoint()) {
         if (AA)
           A.recordDependence(*this, *AA, DepClassTy::OPTIONAL);
         UsedAssumedInformation = true;
       } else {
         UsedAssumedInformation = false;
       }
-      auto *Val = ConstantInt::getBool(IRP.getAnchorValue().getContext(),
-                                       SPMDCompatibilityTracker.isAssumed());
+      auto *Val = ConstantInt::getSigned(
+          IntegerType::getInt8Ty(IRP.getAnchorValue().getContext()),
+          SPMDCompatibilityTracker.isAssumed() ? OMP_TGT_EXEC_MODE_SPMD
+                                               : OMP_TGT_EXEC_MODE_GENERIC);
       return Val;
     };
 
     Attributor::SimplifictionCallbackTy IsGenericModeSimplifyCB =
         [&](const IRPosition &IRP, const AbstractAttribute *AA,
             bool &UsedAssumedInformation) -> Optional<Value *> {
       // IRP represents the "RequiresFullRuntime" argument of an
       // __kmpc_target_init or __kmpc_target_deinit call. We will answer this
       // one with the internal state of the SPMDCompatibilityTracker, so if
       // generic then true, if SPMD then false.
       if (!SPMDCompatibilityTracker.isValidState())
         return nullptr;
       if (!SPMDCompatibilityTracker.isAtFixpoint()) {
         if (AA)
           A.recordDependence(*this, *AA, DepClassTy::OPTIONAL);
         UsedAssumedInformation = true;
       } else {
         UsedAssumedInformation = false;
       }
       auto *Val = ConstantInt::getBool(IRP.getAnchorValue().getContext(),
                                        !SPMDCompatibilityTracker.isAssumed());
       return Val;
     };
 
-    constexpr const int InitIsSPMDArgNo = 1;
-    constexpr const int DeinitIsSPMDArgNo = 1;
+    constexpr const int InitModeArgNo = 1;
+    constexpr const int DeinitModeArgNo = 1;
     constexpr const int InitUseStateMachineArgNo = 2;
     constexpr const int InitRequiresFullRuntimeArgNo = 3;
     constexpr const int DeinitRequiresFullRuntimeArgNo = 2;
     A.registerSimplificationCallback(
         IRPosition::callsite_argument(*KernelInitCB, InitUseStateMachineArgNo),
         StateMachineSimplifyCB);
     A.registerSimplificationCallback(
-        IRPosition::callsite_argument(*KernelInitCB, InitIsSPMDArgNo),
-        IsSPMDModeSimplifyCB);
+        IRPosition::callsite_argument(*KernelInitCB, InitModeArgNo),
+        ModeSimplifyCB);
     A.registerSimplificationCallback(
-        IRPosition::callsite_argument(*KernelDeinitCB, DeinitIsSPMDArgNo),
-        IsSPMDModeSimplifyCB);
+        IRPosition::callsite_argument(*KernelDeinitCB, DeinitModeArgNo),
+        ModeSimplifyCB);
     A.registerSimplificationCallback(
         IRPosition::callsite_argument(*KernelInitCB,
                                       InitRequiresFullRuntimeArgNo),
         IsGenericModeSimplifyCB);
     A.registerSimplificationCallback(
         IRPosition::callsite_argument(*KernelDeinitCB,
                                       DeinitRequiresFullRuntimeArgNo),
         IsGenericModeSimplifyCB);
 
     // Check if we know we are in SPMD-mode already.
-    ConstantInt *IsSPMDArg =
-        dyn_cast<ConstantInt>(KernelInitCB->getArgOperand(InitIsSPMDArgNo));
-    if (IsSPMDArg && !IsSPMDArg->isZero())
+    ConstantInt *ModeArg =
+        dyn_cast<ConstantInt>(KernelInitCB->getArgOperand(InitModeArgNo));
+    if (ModeArg && (ModeArg->getSExtValue() & OMP_TGT_EXEC_MODE_SPMD))
       SPMDCompatibilityTracker.indicateOptimisticFixpoint();
     // This is a generic region but SPMDization is disabled so stop tracking.
     else if (DisableOpenMPOptSPMDization)
       SPMDCompatibilityTracker.indicatePessimisticFixpoint();
   }
 
   /// Modify the IR based on the KernelInfoState as the fixpoint iteration is
   /// finished now.
const int8_t ExecModeVal =
         cast<ConstantInt>(ExecMode->getInitializer())->getSExtValue();
     assert(ExecModeVal == OMP_TGT_EXEC_MODE_GENERIC &&
            "Initially non-SPMD kernel has SPMD exec mode!");
     ExecMode->setInitializer(
         ConstantInt::get(ExecMode->getInitializer()->getType(),
                          ExecModeVal | OMP_TGT_EXEC_MODE_GENERIC_SPMD));
 
     // Next rewrite the init and deinit calls to indicate we use SPMD-mode now.
-    const int InitIsSPMDArgNo = 1;
-    const int DeinitIsSPMDArgNo = 1;
+    const int InitModeArgNo = 1;
+    const int DeinitModeArgNo = 1;
     const int InitUseStateMachineArgNo = 2;
     const int InitRequiresFullRuntimeArgNo = 3;
     const int DeinitRequiresFullRuntimeArgNo = 2;
 
     auto &Ctx = getAnchorValue().getContext();
-    A.changeUseAfterManifest(KernelInitCB->getArgOperandUse(InitIsSPMDArgNo),
-                             *ConstantInt::getBool(Ctx, 1));
+    A.changeUseAfterManifest(
+        KernelInitCB->getArgOperandUse(InitModeArgNo),
+        *ConstantInt::getSigned(IntegerType::getInt8Ty(Ctx),
+                                OMP_TGT_EXEC_MODE_SPMD));
     A.changeUseAfterManifest(
         KernelInitCB->getArgOperandUse(InitUseStateMachineArgNo),
         *ConstantInt::getBool(Ctx, 0));
     A.changeUseAfterManifest(
-        KernelDeinitCB->getArgOperandUse(DeinitIsSPMDArgNo),
-        *ConstantInt::getBool(Ctx, 1));
+        KernelDeinitCB->getArgOperandUse(DeinitModeArgNo),
+        *ConstantInt::getSigned(IntegerType::getInt8Ty(Ctx),
+                                OMP_TGT_EXEC_MODE_SPMD));
     A.changeUseAfterManifest(
         KernelInitCB->getArgOperandUse(InitRequiresFullRuntimeArgNo),
         *ConstantInt::getBool(Ctx, 0));
     A.changeUseAfterManifest(
         KernelDeinitCB->getArgOperandUse(DeinitRequiresFullRuntimeArgNo),
         *ConstantInt::getBool(Ctx, 0));
 
     ++NumOpenMPTargetRegionKernelsSPMD;
 
     auto Remark = [&](OptimizationRemark OR) {
       return OR << "Transformed generic-mode kernel to SPMD-mode.";
     };
     A.emitRemark<OptimizationRemark>(KernelInitCB, "OMP120", Remark);
     return true;
   };
 
   ChangeStatus buildCustomStateMachine(Attributor &A) {
     // If we have disabled state machine rewrites, don't make a custom one
     if (DisableOpenMPOptStateMachineRewrite)
       return indicatePessimisticFixpoint();
 
     assert(ReachedKnownParallelRegions.isValidState() &&
            "Custom state machine with invalid parallel region states?");
 
-    const int InitIsSPMDArgNo = 1;
+    const int InitModeArgNo = 1;
     const int InitUseStateMachineArgNo = 2;
 
     // Check if the current configuration is non-SPMD and generic state machine.
     // If we already have SPMD mode or a custom state machine we do not need to
     // go any further. If it is anything but a constant something is weird and
     // we give up.
     ConstantInt *UseStateMachine = dyn_cast<ConstantInt>(
         KernelInitCB->getArgOperand(InitUseStateMachineArgNo));
-    ConstantInt *IsSPMD =
-        dyn_cast<ConstantInt>(KernelInitCB->getArgOperand(InitIsSPMDArgNo));
+    ConstantInt *Mode =
+        dyn_cast<ConstantInt>(KernelInitCB->getArgOperand(InitModeArgNo));
 
     // If we are stuck with generic mode, try to create a custom device (=GPU)
     // state machine which is specialized for the parallel regions that are
     // reachable by the kernel.
-    if (!UseStateMachine || UseStateMachine->isZero() || !IsSPMD ||
-        !IsSPMD->isZero())
+    if (!UseStateMachine || UseStateMachine->isZero() || !Mode ||
+        (Mode->getSExtValue() & OMP_TGT_EXEC_MODE_GENERIC))
       return ChangeStatus::UNCHANGED;
 
     // If not SPMD mode, indicate we use a custom state machine now.
     auto &Ctx = getAnchorValue().getContext();
     auto *FalseVal = ConstantInt::getBool(Ctx, 0);
     A.changeUseAfterManifest(
         KernelInitCB->getArgOperandUse(InitUseStateMachineArgNo), *FalseVal);
 

llvm/test/Transforms/OpenMP/always_inline_device.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes
 ; RUN: opt < %s -S -passes=openmp-opt -openmp-opt-inline-device | FileCheck %s
 
 %struct.ident_t = type { i32, i32, i32, i32, i8* }
 @0 = private unnamed_addr constant [23 x i8] c";unknown;unknown;0;0;;\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @0, i32 0, i32 0) }, align 8
 @__omp_offloading_fd02_c0934fc2_foo_l4_exec_mode = weak constant i8 1
 @llvm.compiler.used = appending global [1 x i8*] [i8* @__omp_offloading_fd02_c0934fc2_foo_l4_exec_mode], section "llvm.metadata"
 
 ; Function Attrs: convergent norecurse nounwind
 define weak void @__omp_offloading_fd02_c0934fc2_foo_l4() #0 {
 ; CHECK: Function Attrs: convergent norecurse nounwind
 ; CHECK-LABEL: @__omp_offloading_fd02_c0934fc2_foo_l4(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 true)
+; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK:       user_code.entry:
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
 ; CHECK-NEXT:    ret void
 ;
 entry:
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   call void @bar() #2
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 ; Function Attrs: convergent nounwind
 define hidden void @bar() #1 {
 ; CHECK: Function Attrs: alwaysinline convergent nounwind
 ; CHECK-LABEL: @bar(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    ret void
 ;

llvm/test/Transforms/OpenMP/custom_state_machines.ll

 @3 = private unnamed_addr constant %struct.ident_t { i32 0, i32 322, i32 2, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @0, i32 0, i32 0) }, align 8
 @llvm.compiler.used = appending global [8 x i8*] [i8* @__omp_offloading_14_a36502b_no_state_machine_needed_l14_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_l22_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_pure_l77_exec_mode, i8* @__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92_exec_mode, i8* @__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112_exec_mode], section "llvm.metadata"
 
 define weak void @__omp_offloading_14_a36502b_no_state_machine_needed_l14() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
 define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
   %.global_tid..addr = alloca i32*, align 8
   %.bound_tid..addr = alloca i32*, align 8
   store i32* %.global_tid., i32** %.global_tid..addr, align 8
   store i32* %.bound_tid., i32** %.bound_tid..addr, align 8
   call void @no_parallel_region_in_here() #7
omp_if.end:                                       ; preds = %omp_if.then, %entry
   call void @__kmpc_barrier(%struct.ident_t* @3, i32 %0)
   ret void
 }
 
 declare void @unknown_no_openmp() #2
 
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*) #3
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_l22() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__1(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__1(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
entry:
   ret void
 }
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__4(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__4(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
entry:
   ret void
 }
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__6(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__6(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
entry:
   ret void
 }
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__9(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__9(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
entry:
   ret void
 }
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_pure_l77() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__12(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__12(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:
entry:
   ret void
 }
 
 define weak void @__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__15(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__15(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:

 
 declare i32 @omp_get_thread_num(...) #4
 
 define weak void @__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112() #0 {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__16(i32* %.threadid_temp., i32* %.zero.addr) #3
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__16(i32* noalias %.global_tid., i32* noalias %.bound_tid.) #0 {
 entry:

 !17 = !{i32 7, !"openmp", i32 50}
 !18 = !{i32 7, !"openmp-device", i32 50}
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_needed_l14
 ; AMDGPU-SAME: () #[[ATTR0:[0-9]+]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 true)
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
 ; AMDGPU-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__
 ; AMDGPU-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ; AMDGPU-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_l22
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__2_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__2_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute2:
-; AMDGPU-NEXT:    call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.check3:
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-NEXT:    call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; AMDGPU-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; AMDGPU-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef @__omp_outlined__2_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; AMDGPU-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* noundef @__omp_outlined__3_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__3_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__2
 ; AMDGPU-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; AMDGPU-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__17_wrapper
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__17_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute2:
-; AMDGPU-NEXT:    call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.check3:
-; AMDGPU-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE5:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK6:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute5:
-; AMDGPU-NEXT:    call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.check6:
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-NEXT:    call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__4
 ; AMDGPU-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; AMDGPU-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; AMDGPU-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR8]]
 ; AMDGPU-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__5 to i8*), i8* noundef @__omp_outlined__5_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__5 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__5_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; AMDGPU-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@simple_state_machine_interprocedural_before.internalized
 ; AMDGPU-SAME: () #[[ATTR1]] {
 ; AMDGPU-NEXT:  entry:

 ; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__18 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__18_wrapper to i8*), i8** [[TMP1]], i64 0)
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__8_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute2:
-; AMDGPU-NEXT:    call void @__omp_outlined__8_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.fallback.execute:
-; AMDGPU-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-NEXT:    call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__6
 ; AMDGPU-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__7 to i8*), i8* noundef @__omp_outlined__7_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__7 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__7_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; AMDGPU-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR10]]
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__8 to i8*), i8* noundef @__omp_outlined__8_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__8 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__8_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__7
 ; AMDGPU-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; AMDGPU-NEXT:    call void @__omp_outlined__8(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__10_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__10_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute2:
-; AMDGPU-NEXT:    call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.check3:
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-NEXT:    call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; AMDGPU-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__10 to i8*), i8* noundef @__omp_outlined__10_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__10 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__10_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; AMDGPU-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__11 to i8*), i8* noundef @__omp_outlined__11_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__11 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__11_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__10
 ; AMDGPU-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; AMDGPU-NEXT:    call void @__omp_outlined__11(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_pure_l77
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__13_wrapper.ID to void (i16, i32)*)
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__13_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute2:
-; AMDGPU-NEXT:    call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.check3:
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-NEXT:    call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__12
 ; AMDGPU-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; AMDGPU-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; AMDGPU-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__13 to i8*), i8* noundef @__omp_outlined__13_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__13 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__13_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__14 to i8*), i8* noundef @__omp_outlined__14_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; AMDGPU-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__14 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__14_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__13
 ; AMDGPU-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; AMDGPU-NEXT:    call void @__omp_outlined__14(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper
-; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.execute:
-; AMDGPU-NEXT:    call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.fallback.execute:
-; AMDGPU-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__15
 ; AMDGPU-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ; AMDGPU:       return:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112
 ; AMDGPU-SAME: () #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; AMDGPU-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; AMDGPU-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.begin:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
-; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
-; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; AMDGPU:       worker_state_machine.finished:
-; AMDGPU-NEXT:    ret void
-; AMDGPU:       worker_state_machine.is_active.check:
-; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.fallback.execute:
-; AMDGPU-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; AMDGPU:       worker_state_machine.parallel_region.end:
-; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; AMDGPU:       worker_state_machine.done.barrier:
-; AMDGPU-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; AMDGPU:       thread.user_code.check:
+; AMDGPU-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU:       user_code.entry:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-NEXT:    call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       worker.exit:
 ; AMDGPU-NEXT:    ret void
 ;
 ;
 ; AMDGPU: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__16
 ; AMDGPU-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_needed_l14
 ; NVPTX-SAME: () #[[ATTR0:[0-9]+]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 true)
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
 ; NVPTX-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__
 ; NVPTX-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ; NVPTX-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_l22
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__2_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__2_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute2:
-; NVPTX-NEXT:    call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.check3:
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-NEXT:    call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; NVPTX-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; NVPTX-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; NVPTX-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef @__omp_outlined__2_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; NVPTX-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; NVPTX-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* noundef @__omp_outlined__3_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__3 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__3_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__2
 ; NVPTX-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; NVPTX-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__17_wrapper
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__17_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute2:
-; NVPTX-NEXT:    call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.check3:
-; NVPTX-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE5:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK6:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute5:
-; NVPTX-NEXT:    call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.check6:
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-NEXT:    call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__4
 ; NVPTX-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; NVPTX-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; NVPTX-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR8]]
 ; NVPTX-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; NVPTX-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__5 to i8*), i8* noundef @__omp_outlined__5_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__5 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__5_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; NVPTX-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@simple_state_machine_interprocedural_before.internalized
 ; NVPTX-SAME: () #[[ATTR1]] {
 ; NVPTX-NEXT:  entry:

 ; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__18 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__18_wrapper to i8*), i8** [[TMP1]], i64 0)
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__8_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute2:
-; NVPTX-NEXT:    call void @__omp_outlined__8_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.fallback.execute:
-; NVPTX-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-NEXT:    call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__6
 ; NVPTX-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; NVPTX-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__7 to i8*), i8* noundef @__omp_outlined__7_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__7 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__7_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; NVPTX-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR10]]
 ; NVPTX-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__8 to i8*), i8* noundef @__omp_outlined__8_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__8 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__8_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__7
 ; NVPTX-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; NVPTX-NEXT:    call void @__omp_outlined__8(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__10_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__10_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute2:
-; NVPTX-NEXT:    call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.check3:
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-NEXT:    call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; NVPTX-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; NVPTX-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__10 to i8*), i8* noundef @__omp_outlined__10_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__10 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__10_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; NVPTX-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; NVPTX-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__11 to i8*), i8* noundef @__omp_outlined__11_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__11 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__11_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__10
 ; NVPTX-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; NVPTX-NEXT:    call void @__omp_outlined__11(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_pure_l77
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__13_wrapper.ID to void (i16, i32)*)
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__13_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute2:
-; NVPTX-NEXT:    call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.check3:
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-NEXT:    call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__12
 ; NVPTX-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; NVPTX-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; NVPTX-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__13 to i8*), i8* noundef @__omp_outlined__13_wrapper.ID, i8** noundef [[TMP1]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__13 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__13_wrapper to i8*), i8** noundef [[TMP1]], i64 noundef 0)
 ; NVPTX-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
-; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__14 to i8*), i8* noundef @__omp_outlined__14_wrapper.ID, i8** noundef [[TMP2]], i64 noundef 0)
+; NVPTX-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noundef @[[GLOB1]], i32 [[TMP0]], i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__14 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__14_wrapper to i8*), i8** noundef [[TMP2]], i64 noundef 0)
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__13
 ; NVPTX-SAME: (i32* noalias nocapture nofree readnone [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree readnone [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8

 ; NVPTX-NEXT:    call void @__omp_outlined__14(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper
-; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.execute:
-; NVPTX-NEXT:    call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.fallback.execute:
-; NVPTX-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__15
 ; NVPTX-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ; NVPTX:       return:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112
 ; NVPTX-SAME: () #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
-; NVPTX-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
-; NVPTX-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.begin:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
-; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
-; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
-; NVPTX:       worker_state_machine.finished:
-; NVPTX-NEXT:    ret void
-; NVPTX:       worker_state_machine.is_active.check:
-; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.fallback.execute:
-; NVPTX-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
-; NVPTX:       worker_state_machine.parallel_region.end:
-; NVPTX-NEXT:    call void @__kmpc_kernel_end_parallel()
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
-; NVPTX:       worker_state_machine.done.barrier:
-; NVPTX-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_BEGIN]]
-; NVPTX:       thread.user_code.check:
+; NVPTX-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX:       user_code.entry:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-NEXT:    call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       worker.exit:
 ; NVPTX-NEXT:    ret void
 ;
 ;
 ; NVPTX: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__16
 ; NVPTX-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_needed_l14
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_l22
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__4
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__6
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_pure_l77
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__12
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__15
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112
 ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; AMDGPU-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; AMDGPU-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; AMDGPU-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; AMDGPU-DISABLED:       user_code.entry:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       worker.exit:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ;
 ; AMDGPU-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__16
 ; AMDGPU-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_needed_l14
 ; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_l22
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__4
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__6
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_pure_l77
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__12
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__15
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112
 ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
-; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; NVPTX-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
+; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; NVPTX-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; NVPTX-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; NVPTX-DISABLED:       user_code.entry:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
 ; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
+; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       worker.exit:
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 ;
 ; NVPTX-DISABLED: Function Attrs: convergent noinline norecurse nounwind
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__16
 ; NVPTX-DISABLED-SAME: (i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTGLOBAL_TID_:%.*]], i32* noalias nocapture nofree nonnull readnone align 4 dereferenceable(4) [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {

llvm/test/Transforms/OpenMP/custom_state_machines_remarks.ll

 ; RUN: opt -passes=openmp-opt -pass-remarks=openmp-opt -pass-remarks-missed=openmp-opt -pass-remarks-analysis=openmp-opt -disable-output < %s 2>&1 | FileCheck %s
 target triple = "nvptx64"
 
 ; CHECK: remark: llvm/test/Transforms/OpenMP/custom_state_machines_remarks.c:11:1: Generic-mode kernel is executed with a customized state machine that requires a fallback.
 ; CHECK: remark: llvm/test/Transforms/OpenMP/custom_state_machines_remarks.c:13:5: Call may contain unknown parallel regions. Use `__attribute__((assume("omp_no_parallelism")))` to override.
 ; CHECK: remark: llvm/test/Transforms/OpenMP/custom_state_machines_remarks.c:15:5: Call may contain unknown parallel regions. Use `__attribute__((assume("omp_no_parallelism")))` to override.
 ; CHECK: remark: llvm/test/Transforms/OpenMP/custom_state_machines_remarks.c:20:1: Rewriting generic-mode kernel with a customized state machine.
 
 
 ;; void unknown(void);
 ;; void known(void) {
 ;;   #pragma omp parallel
 ;;   {
 ;;     unknown();
 ;;   }
 ;; }
-;; 
+;;
 ;; void test_fallback(void) {
 ;;   #pragma omp target teams
 ;;   {
 ;;     unknown();
 ;;     known();
 ;;     unknown();
 ;;   }
 ;; }

 @13 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 2, i32 0, i8* getelementptr inbounds ([73 x i8], [73 x i8]* @12, i32 0, i32 0) }, align 8
 @G = external global i32
 @llvm.compiler.used = appending global [2 x i8*] [i8* @__omp_offloading_2a_d80d3d_test_fallback_l11_exec_mode, i8* @__omp_offloading_2a_d80d3d_test_no_fallback_l20_exec_mode], section "llvm.metadata"
 
 ; Function Attrs: convergent norecurse nounwind
 define weak void @__omp_offloading_2a_d80d3d_test_fallback_l11() local_unnamed_addr #0 !dbg !15 {
 entry:
   %captured_vars_addrs.i.i = alloca [0 x i8*], align 8
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 true, i1 true) #3, !dbg !18
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 true, i1 true) #3, !dbg !18
   %exec_user_code = icmp eq i32 %0, -1, !dbg !18
   br i1 %exec_user_code, label %user_code.entry, label %common.ret, !dbg !18
 
 common.ret:                                       ; preds = %entry, %user_code.entry
   ret void, !dbg !19
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @3) #3
   call void @unknown() #6, !dbg !20
   %2 = bitcast [0 x i8*]* %captured_vars_addrs.i.i to i8*
   call void @llvm.lifetime.start.p0i8(i64 0, i8* nonnull %2) #3
   %3 = call i32 @__kmpc_global_thread_num(%struct.ident_t* noundef nonnull @13) #3
   %4 = getelementptr inbounds [0 x i8*], [0 x i8*]* %captured_vars_addrs.i.i, i64 0, i64 0, !dbg !23
   call void @__kmpc_parallel_51(%struct.ident_t* noundef nonnull @13, i32 %3, i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef nonnull %4, i64 noundef 0) #3, !dbg !23
   call void @llvm.lifetime.end.p0i8(i64 0, i8* nonnull %2) #3, !dbg !26
   call void @unknown() #6, !dbg !27
-  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @5, i1 false, i1 true) #3, !dbg !28
+  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @5, i8 1, i1 true) #3, !dbg !28
   br label %common.ret
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1) local_unnamed_addr
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1) local_unnamed_addr
 
 ; Function Attrs: convergent
 declare void @unknown() local_unnamed_addr #1
 
 ; Function Attrs: nounwind
 define hidden void @known() local_unnamed_addr #2 !dbg !29 {
 entry:
   %captured_vars_addrs = alloca [0 x i8*], align 8
   %0 = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @13)
   %1 = getelementptr inbounds [0 x i8*], [0 x i8*]* %captured_vars_addrs, i64 0, i64 0, !dbg !30
   call void @__kmpc_parallel_51(%struct.ident_t* nonnull @13, i32 %0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** nonnull %1, i64 0) #3, !dbg !30
   ret void, !dbg !31
 }
 
 ; Function Attrs: nounwind
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*) local_unnamed_addr #3
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1) local_unnamed_addr
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1) local_unnamed_addr
 
 ; Function Attrs: norecurse nounwind
 define weak void @__omp_offloading_2a_d80d3d_test_no_fallback_l20() local_unnamed_addr #4 !dbg !32 {
 entry:
   %captured_vars_addrs.i2.i = alloca [0 x i8*], align 8
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @7, i1 false, i1 true, i1 true) #3, !dbg !33
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @7, i8 1, i1 true, i1 true) #3, !dbg !33
   %exec_user_code = icmp eq i32 %0, -1, !dbg !33
   br i1 %exec_user_code, label %user_code.entry, label %common.ret, !dbg !33
 
 common.ret:                                       ; preds = %entry, %user_code.entry
   ret void, !dbg !34
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @9) #3
   %2 = bitcast [0 x i8*]* %captured_vars_addrs.i2.i to i8*
   call void @llvm.lifetime.start.p0i8(i64 0, i8* nonnull %2) #3
   %3 = call i32 @__kmpc_global_thread_num(%struct.ident_t* noundef nonnull @13) #3
   %4 = getelementptr inbounds [0 x i8*], [0 x i8*]* %captured_vars_addrs.i2.i, i64 0, i64 0, !dbg !35
   call void @__kmpc_parallel_51(%struct.ident_t* noundef nonnull @13, i32 %3, i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef nonnull %4, i64 noundef 0) #3, !dbg !35
   call void @llvm.lifetime.end.p0i8(i64 0, i8* nonnull %2) #3, !dbg !39
   call void @llvm.lifetime.start.p0i8(i64 0, i8* nonnull %2) #3
   %5 = call i32 @__kmpc_global_thread_num(%struct.ident_t* noundef nonnull @13) #3
   call void @__kmpc_parallel_51(%struct.ident_t* noundef nonnull @13, i32 %5, i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef nonnull %4, i64 noundef 0) #3, !dbg !40
   call void @llvm.lifetime.end.p0i8(i64 0, i8* nonnull %2) #3, !dbg !42
   call void @llvm.lifetime.start.p0i8(i64 0, i8* nonnull %2) #3
   %6 = call i32 @__kmpc_global_thread_num(%struct.ident_t* noundef nonnull @13) #3
   call void @__kmpc_parallel_51(%struct.ident_t* noundef nonnull @13, i32 %6, i32 noundef 1, i32 noundef -1, i32 noundef -1, i8* noundef bitcast (void (i32*, i32*)* @__omp_outlined__2 to i8*), i8* noundef bitcast (void (i16, i32)* @__omp_outlined__2_wrapper to i8*), i8** noundef nonnull %4, i64 noundef 0) #3, !dbg !43
   call void @llvm.lifetime.end.p0i8(i64 0, i8* nonnull %2) #3, !dbg !45
   call void @no_openmp()
   call void @no_parallelism()
-  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @11, i1 false, i1 true) #3, !dbg !46
+  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @11, i8 1, i1 true) #3, !dbg !46
   br label %common.ret
 }
 
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__2(i32* noalias nocapture nofree readnone %.global_tid., i32* noalias nocapture nofree readnone %.bound_tid.) #0 !dbg !47 {
 entry:
   call void @unknown() #6, !dbg !48
   ret void, !dbg !49

llvm/test/Transforms/OpenMP/deduplication_target.ll

 
 declare void @use(i32)
 
 define weak void @__omp_offloading_50_a3e09bf8_foo_l2() #0 {
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_50_a3e09bf8_foo_l2
 ; CHECK-SAME: () #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
-; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 true)
+; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK:       user_code.entry:
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 true)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %captured_vars_addrs = alloca [0 x i8*], align 8
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 true, i1 false, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 2, i1 false, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
   %2 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 true, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 2, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*) #1
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 attributes #0 = { convergent noinline norecurse nounwind "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
 attributes #1 = { nounwind }
 
 !omp_offload.info = !{!0}
 !nvvm.annotations = !{!1}
 !llvm.module.flags = !{!2, !3, !4}
 
 !0 = !{i32 0, i32 80, i32 -1545561096, !"foo", i32 2, i32 0}
 !1 = !{void ()* @__omp_offloading_50_a3e09bf8_foo_l2, !"kernel", i32 1}
 !2 = !{i32 1, !"wchar_size", i32 4}
 !3 = !{i32 7, !"openmp", i32 50}
 !4 = !{i32 7, !"openmp-device", i32 50}

llvm/test/Transforms/OpenMP/fold_generic_main_thread.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature
 ; RUN: opt -S -passes='openmp-opt' < %s | FileCheck %s
 ; RUN: opt -openmp-opt-disable-folding -S -passes='openmp-opt' < %s | FileCheck %s --check-prefix=CHECK-DISABLED
 ; ModuleID = 'single_threaded_exeuction.c'
 
 %struct.ident_t = type { i32, i32, i32, i32, i8* }
 
 @0 = private unnamed_addr constant [1 x i8] c"\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([1 x i8], [1 x i8]* @0, i32 0, i32 0) }, align 8
 
 define void @kernel() {
 ; CHECK-LABEL: define {{[^@]+}}@kernel() {
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 false)
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i8 1, i1 false, i1 false)
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], -1
 ; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
 ; CHECK-NEXT:    call void @foo()
 ; CHECK-NEXT:    br label [[IF_END:%.*]]
 ; CHECK:       if.else:
 ; CHECK-NEXT:    call void @bar()
 ; CHECK-NEXT:    br label [[IF_END]]
 ; CHECK:       if.end:
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@kernel() {
-; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 false)
+; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i8 1, i1 false, i1 false)
 ; CHECK-DISABLED-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK-DISABLED:       if.then:
 ; CHECK-DISABLED-NEXT:    call void @foo()
 ; CHECK-DISABLED-NEXT:    br label [[IF_END:%.*]]
 ; CHECK-DISABLED:       if.else:
 ; CHECK-DISABLED-NEXT:    call void @bar()
 ; CHECK-DISABLED-NEXT:    br label [[IF_END]]
 ; CHECK-DISABLED:       if.end:
-; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ;
-  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 false, i1 false)
+  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 false, i1 false)
   %cmp = icmp eq i32 %call, -1
   br i1 %cmp, label %if.then, label %if.else
 if.then:
   call void @foo()
   br label %if.end
 if.else:
   call void @bar()
   br label %if.end
 if.end:
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
   ret void
 }
 
 ; Function Attrs: noinline
 define internal void @foo() {
 ; CHECK-LABEL: define {{[^@]+}}@foo() {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TID:%.*]] = call i32 @__kmpc_get_hardware_thread_id()

 }
 
 declare void @baz()
 
 declare i8 @__kmpc_is_generic_main_thread_id(i32)
 
 declare i32 @__kmpc_get_hardware_thread_id()
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5, !6}
 !nvvm.annotations = !{!7}
 
 !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 13.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
 !1 = !DIFile(filename: "fold_generic_main_thread.c", directory: "/tmp/fold_generic_main_thread.c")
 !2 = !{}
 !3 = !{i32 2, !"Debug Info Version", i32 3}
 !4 = !{i32 1, !"wchar_size", i32 4}
 !5 = !{i32 7, !"openmp", i32 50}
 !6 = !{i32 7, !"openmp-device", i32 50}
 !7 = !{void ()* @kernel, !"kernel", i32 1}

llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals
 ; RUN: opt -S -passes=openmp-opt < %s | FileCheck %s
 target triple = "nvptx64"
 
 %struct.ident_t = type { i32, i32, i32, i32, i8* }
 
 @kernel0_exec_mode = weak constant i8 1
 
 @G = external global i32
 ;.
 ; CHECK: @[[KERNEL0_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
 ; CHECK: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i32
 ; CHECK: @[[KERNEL1_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
-; CHECK: @[[KERNEL2_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 3
-; CHECK: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c"
-; CHECK: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8
+; CHECK: @[[KERNEL2_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
 ;.
 define weak void @kernel0() #0 {
 ; CHECK-LABEL: define {{[^@]+}}@kernel0
 ; CHECK-SAME: () #[[ATTR0:[0-9]+]] {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    call void @helper0() #[[ATTR1:[0-9]+]]
 ; CHECK-NEXT:    call void @helper1() #[[ATTR1]]
 ; CHECK-NEXT:    call void @helper2() #[[ATTR1]]
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
   call void @helper0()
   call void @helper1()
   call void @helper2()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
   ret void
 }
 
 @kernel1_exec_mode = weak constant i8 1
 
 define weak void @kernel1() #0 {
 ; CHECK-LABEL: define {{[^@]+}}@kernel1
 ; CHECK-SAME: () #[[ATTR0]] {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    call void @helper1() #[[ATTR1]]
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
   call void @helper1()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
   ret void
 }
 
 @kernel2_exec_mode = weak constant i8 1
 
 define weak void @kernel2() #0 {
 ; CHECK-LABEL: define {{[^@]+}}@kernel2
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 true, i1 true)
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[I]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]]
 ; CHECK:       common.ret:
 ; CHECK-NEXT:    ret void
 ; CHECK:       user_code.entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* null) #[[ATTR1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-NEXT:    call void @helper0() #[[ATTR1]]
 ; CHECK-NEXT:    call void @helper1() #[[ATTR1]]
 ; CHECK-NEXT:    call void @helper2() #[[ATTR1]]
 ; CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* null, i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP1]], i64 0)
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %captured_vars_addrs = alloca [0 x i8*], align 8
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 true, i1 true)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %i, -1
   br i1 %exec_user_code, label %user_code.entry, label %common.ret
 
 common.ret:
   ret void
 
 user_code.entry:
   %0 = call i32 @__kmpc_global_thread_num(%struct.ident_t* null)
   %1 = bitcast [0 x i8*]* %captured_vars_addrs to i8**
   call void @helper0()
   call void @helper1()
   call void @helper2()
   call void @__kmpc_parallel_51(%struct.ident_t* null, i32 %0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** %1, i64 0)
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
   ret void
 }
 
 define internal void @helper0() {
 ; CHECK-LABEL: define {{[^@]+}}@helper0
 ; CHECK-SAME: () #[[ATTR1]] {
-; CHECK-NEXT:    br label [[REGION_CHECK_TID:%.*]]
-; CHECK:       region.check.tid:
-; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 0
-; CHECK-NEXT:    br i1 [[TMP2]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]]
-; CHECK:       region.guarded:
 ; CHECK-NEXT:    store i32 666, i32* @G, align 4
-; CHECK-NEXT:    br label [[REGION_GUARDED_END:%.*]]
-; CHECK:       region.guarded.end:
-; CHECK-NEXT:    br label [[REGION_BARRIER]]
-; CHECK:       region.barrier:
-; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
-; CHECK-NEXT:    br label [[REGION_EXIT:%.*]]
-; CHECK:       region.exit:
 ; CHECK-NEXT:    ret void
 ;
   %threadLimit = call i32 @__kmpc_get_hardware_num_threads_in_block()
   store i32 %threadLimit, i32* @G
   ret void
 }
 
 define internal void @helper1() {
t:
   ret void
 f:
   ret void
 }
 
 define internal void @helper2() {
 ; CHECK-LABEL: define {{[^@]+}}@helper2
 ; CHECK-SAME: () #[[ATTR1]] {
-; CHECK-NEXT:    br label [[REGION_CHECK_TID:%.*]]
-; CHECK:       region.check.tid:
-; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 0
-; CHECK-NEXT:    br i1 [[TMP2]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]]
-; CHECK:       region.guarded:
 ; CHECK-NEXT:    store i32 666, i32* @G, align 4
-; CHECK-NEXT:    br label [[REGION_GUARDED_END:%.*]]
-; CHECK:       region.guarded.end:
-; CHECK-NEXT:    br label [[REGION_BARRIER]]
-; CHECK:       region.barrier:
-; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
-; CHECK-NEXT:    br label [[REGION_EXIT:%.*]]
-; CHECK:       region.exit:
 ; CHECK-NEXT:    ret void
 ;
   %threadLimit = call i32 @__kmpc_get_hardware_num_threads_in_block()
   store i32 %threadLimit, i32* @G
   ret void
 }
 
 define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias %.bound_tid.) {

 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    ret void
 ;
 entry:
   ret void
 }
 
 declare i32 @__kmpc_get_hardware_num_threads_in_block()
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1 zeroext, i1 zeroext, i1 zeroext) #1
-declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i1 zeroext, i1 zeroext) #1
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1 zeroext, i1 zeroext) #1
+declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i8, i1 zeroext) #1
 declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64)
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*)
 
 
 !llvm.module.flags = !{!0, !1}
 !nvvm.annotations = !{!2, !3, !4}
 
 attributes #0 = { "omp_target_thread_limit"="666" "omp_target_num_teams"="777"}
 
 !0 = !{i32 7, !"openmp", i32 50}
 !1 = !{i32 7, !"openmp-device", i32 50}
 !2 = !{void ()* @kernel0, !"kernel", i32 1}
 !3 = !{void ()* @kernel1, !"kernel", i32 1}
 !4 = !{void ()* @kernel2, !"kernel", i32 1}
 ;
 ;.
 ; CHECK: attributes #[[ATTR0]] = { "omp_target_num_teams"="777" "omp_target_thread_limit"="666" }
 ; CHECK: attributes #[[ATTR1]] = { nounwind }
 ; CHECK: attributes #[[ATTR2:[0-9]+]] = { alwaysinline }
-; CHECK: attributes #[[ATTR3:[0-9]+]] = { convergent nounwind }
 ;.
 ; CHECK: [[META0:![0-9]+]] = !{i32 7, !"openmp", i32 50}
 ; CHECK: [[META1:![0-9]+]] = !{i32 7, !"openmp-device", i32 50}
 ; CHECK: [[META2:![0-9]+]] = !{void ()* @kernel0, !"kernel", i32 1}
 ; CHECK: [[META3:![0-9]+]] = !{void ()* @kernel1, !"kernel", i32 1}
 ; CHECK: [[META4:![0-9]+]] = !{void ()* @kernel2, !"kernel", i32 1}
 ;.

llvm/test/Transforms/OpenMP/global_constructor.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes --include-generated-funcs
 ; RUN: opt -S -passes=openmp-opt < %s | FileCheck %s
 
 %struct.ident_t = type { i32, i32, i32, i32, i8* }
 
 @0 = private unnamed_addr constant [23 x i8] c";unknown;unknown;0;0;;\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @0, i32 0, i32 0) }, align 8
 @_ZL6Device = internal global double 0.000000e+00, align 8
 @__omp_offloading_fd02_85283c04_main_l11_exec_mode = weak constant i8 0
 
 define weak void @__omp_offloading_fd02_85283c04_main_l11(double* nonnull align 8 dereferenceable(8) %X) local_unnamed_addr {
 entry:
-  %0 = tail call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 true, i1 false, i1 false) #0
+  %0 = tail call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 2, i1 false, i1 false) #0
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %common.ret
 
 common.ret:
   ret void
 
 user_code.entry:
   %1 = load double, double* @_ZL6Device, align 8, !tbaa !11
   %2 = tail call i32 @__kmpc_get_hardware_thread_id_in_block() #0
   %3 = icmp eq i32 %2, 0
   br i1 %3, label %region.guarded, label %region.barrier
 
 region.guarded:
   store double %1, double* %X, align 8, !tbaa !11
   br label %region.barrier
 
 region.barrier:
   tail call void @__kmpc_barrier_simple_spmd(%struct.ident_t* nonnull @1, i32 %2)
-  tail call void @__kmpc_target_deinit(%struct.ident_t* nonnull @1, i1 true, i1 false) #0
+  tail call void @__kmpc_target_deinit(%struct.ident_t* nonnull @1, i8 2, i1 false) #0
   br label %common.ret
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1) local_unnamed_addr
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1) local_unnamed_addr
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1) local_unnamed_addr
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1) local_unnamed_addr
 
 define internal void @__omp_offloading__fd02_85283c04_Device_l6_ctor() {
 entry:
   %call.i = tail call double @__nv_log(double noundef 2.000000e+00) #1
   %call.i2 = tail call double @__nv_log(double noundef 2.000000e+00) #1
   %div = fdiv double %call.i, %call.i2
   store double %div, double* @_ZL6Device, align 8, !tbaa !11
   ret void

 !10 = !{!"clang version 14.0.0"}
 !11 = !{!12, !12, i64 0}
 !12 = !{!"double", !13, i64 0}
 !13 = !{!"omnipotent char", !14, i64 0}
 !14 = !{!"Simple C++ TBAA"}
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_fd02_85283c04_main_l11
 ; CHECK-SAME: (double* nonnull align 8 dereferenceable(8) [[X:%.*]]) local_unnamed_addr {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i1 true, i1 false, i1 false) #[[ATTR1:[0-9]+]]
+; CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 false) #[[ATTR1:[0-9]+]]
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]]
 ; CHECK:       common.ret:
 ; CHECK-NEXT:    ret void
 ; CHECK:       user_code.entry:
 ; CHECK-NEXT:    [[TMP1:%.*]] = load double, double* @_ZL6Device, align 8, !tbaa [[TBAA11:![0-9]+]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = tail call i32 @__kmpc_get_hardware_thread_id_in_block() #[[ATTR1]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]]
 ; CHECK:       region.guarded:
 ; CHECK-NEXT:    store double [[TMP1]], double* [[X]], align 8, !tbaa [[TBAA11]]
 ; CHECK-NEXT:    br label [[REGION_BARRIER]]
 ; CHECK:       region.barrier:
 ; CHECK-NEXT:    tail call void @__kmpc_barrier_simple_spmd(%struct.ident_t* nonnull @[[GLOB1]], i32 [[TMP2]]) #[[ATTR1]]
-; CHECK-NEXT:    tail call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i1 true, i1 false) #[[ATTR1]]
+; CHECK-NEXT:    tail call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i8 2, i1 false) #[[ATTR1]]
 ; CHECK-NEXT:    br label [[COMMON_RET]]
 ;

llvm/test/Transforms/OpenMP/gpu_state_machine_function_ptr_replacement.ll

 @2 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 2, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @0, i32 0, i32 0) }, align 8
 @llvm.compiler.used = appending global [1 x i8*] [i8* @__omp_offloading_10301_87b2c_foo_l7_exec_mode], section "llvm.metadata"
 
 define weak void @__omp_offloading_10301_87b2c_foo_l7() {
 entry:
   %.zero.addr = alloca i32, align 4
   %.threadid_temp. = alloca i32, align 4
   store i32 0, i32* %.zero.addr, align 4
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
   store i32 %1, i32* %.threadid_temp., align 4
   call void @__omp_outlined__(i32* %.threadid_temp., i32* %.zero.addr)
-  call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* @1, i8 1, i1 true)
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 declare void @unknown()
 
 define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias %.bound_tid.) {
 entry:
   %.global_tid..addr = alloca i32*, align 8
   %.bound_tid..addr = alloca i32*, align 8
   %captured_vars_addrs = alloca [0 x i8*], align 8
   %captured_vars_addrs1 = alloca [0 x i8*], align 8
entry:
   store i32 %1, i32* %.addr1, align 4
   call void @__kmpc_get_shared_variables(i8*** %global_args)
   call void @__omp_outlined__2(i32* %.addr1, i32* %.zero.addr)
   ret void
 }
 
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*)
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 define internal void @__omp_outlined__3(i32* noalias %.global_tid., i32* noalias %.bound_tid.) {
 entry:
   %.global_tid..addr = alloca i32*, align 8
   %.bound_tid..addr = alloca i32*, align 8
   store i32* %.global_tid., i32** %.global_tid..addr, align 8
   store i32* %.bound_tid., i32** %.bound_tid..addr, align 8
   ret void

llvm/test/Transforms/OpenMP/is_spmd_exec_mode_fold.ll

 ; CHECK: @[[WILL_BE_SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 3
 ; CHECK: @[[NON_SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
 ; CHECK: @[[WILL_NOT_BE_SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
 ; CHECK: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i8
 ; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [4 x i8*] [i8* @is_spmd_exec_mode, i8* @will_be_spmd_exec_mode, i8* @non_spmd_exec_mode, i8* @will_not_be_spmd_exec_mode], section "llvm.metadata"
 ;.
 define weak void @is_spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@is_spmd() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    call void @is_spmd_helper1()
 ; CHECK-NEXT:    call void @is_spmd_helper2()
 ; CHECK-NEXT:    call void @is_mixed_helper()
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
   call void @is_spmd_helper1()
   call void @is_spmd_helper2()
   call void @is_mixed_helper()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
   ret void
 }
 
 define weak void @will_be_spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@will_be_spmd() {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[I]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]]
 ; CHECK:       common.ret:
 ; CHECK-NEXT:    ret void
 ; CHECK:       user_code.entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* null) #[[ATTR2:[0-9]+]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-NEXT:    call void @is_spmd_helper2()
 ; CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* null, i32 [[TMP0]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** [[TMP1]], i64 0)
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %captured_vars_addrs = alloca [0 x i8*], align 8
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 true, i1 true)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 true, i1 true)
   %exec_user_code = icmp eq i32 %i, -1
   br i1 %exec_user_code, label %user_code.entry, label %common.ret
 
 common.ret:
   ret void
 
 user_code.entry:
   %0 = call i32 @__kmpc_global_thread_num(%struct.ident_t* null)
   %1 = bitcast [0 x i8*]* %captured_vars_addrs to i8**
   call void @is_spmd_helper2()
   call void @__kmpc_parallel_51(%struct.ident_t* null, i32 %0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__ to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined___wrapper to i8*), i8** %1, i64 0)
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
   ret void
 }
 
 define weak void @non_spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@non_spmd() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
 ; CHECK-NEXT:    call void @is_generic_helper1()
 ; CHECK-NEXT:    call void @is_generic_helper2()
 ; CHECK-NEXT:    call void @is_mixed_helper()
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
   call void @is_generic_helper1()
   call void @is_generic_helper2()
   call void @is_mixed_helper()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
   ret void
 }
 
 define weak void @will_not_be_spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@will_not_be_spmd() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
 ; CHECK-NEXT:    call void @is_generic_helper1()
 ; CHECK-NEXT:    call void @is_generic_helper2()
 ; CHECK-NEXT:    call void @is_mixed_helper()
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
   call void @is_generic_helper1()
   call void @is_generic_helper2()
   call void @is_mixed_helper()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
   ret void
 }
 
 define internal void @is_spmd_helper1() {
 ; CHECK-LABEL: define {{[^@]+}}@is_spmd_helper1() {
 ; CHECK-NEXT:    store i8 1, i8* @G, align 1
 ; CHECK-NEXT:    ret void
 ;

 ; CHECK-NEXT:    ret void
 ;
 entry:
   ret void
 }
 
 declare void @spmd_compatible() "llvm.assume"="ompx_spmd_amenable"
 declare i8 @__kmpc_is_spmd_exec_mode()
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1 zeroext, i1 zeroext, i1 zeroext)
-declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i1 zeroext, i1 zeroext)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1 zeroext, i1 zeroext)
+declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i8, i1 zeroext)
 declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64)
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*)
 declare void @foo()
 declare void @bar()
 
 !llvm.module.flags = !{!0, !1}
 !nvvm.annotations = !{!2, !3, !4, !5}
 

llvm/test/Transforms/OpenMP/parallel_level_fold.ll

 ; CHECK: @[[NO_SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
 ; CHECK: @[[SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 0
 ; CHECK: @[[PARALLEL_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 0
 ; CHECK: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i8
 ; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [3 x i8*] [i8* @no_spmd_exec_mode, i8* @spmd_exec_mode, i8* @parallel_exec_mode], section "llvm.metadata"
 ;.
 define weak void @none_spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@none_spmd() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
 ; CHECK-NEXT:    call void @none_spmd_helper()
 ; CHECK-NEXT:    call void @mixed_helper()
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 false, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 1, i1 false, i1 false)
   call void @none_spmd_helper()
   call void @mixed_helper()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 false)
   ret void
 }
 
 define weak void @spmd() {
 ; CHECK-LABEL: define {{[^@]+}}@spmd() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    call void @spmd_helper()
 ; CHECK-NEXT:    call void @mixed_helper()
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
   call void @spmd_helper()
   call void @mixed_helper()
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
   ret void
 }
 
 define weak void @parallel() {
 ; CHECK-LABEL: define {{[^@]+}}@parallel() {
-; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* align 1073741824 null, i1 true, i1 false, i1 false)
+; CHECK-NEXT:    [[I:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* align 1073741824 null, i8 2, i1 false, i1 false)
 ; CHECK-NEXT:    call void @spmd_helper()
 ; CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* noalias noundef align 1073741824 null, i32 noundef 0, i32 noundef 0, i32 noundef 0, i32 noundef 0, i8* noalias noundef align 1073741824 null, i8* noalias noundef align 1073741824 null, i8** noalias noundef align 1073741824 null, i64 noundef 0)
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
 ; CHECK-NEXT:    ret void
 ;
-  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i1 true, i1 false, i1 false)
+  %i = call i32 @__kmpc_target_init(%struct.ident_t* null, i8 2, i1 false, i1 false)
   call void @spmd_helper()
   call void @__kmpc_parallel_51(%struct.ident_t* null, i32 0, i32 0, i32 0, i32 0, i8* null, i8* null, i8** null, i64 0)
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 true, i1 false)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 2, i1 false)
   ret void
 }
 
 define internal void @mixed_helper() {
 ; CHECK-LABEL: define {{[^@]+}}@mixed_helper() {
 ; CHECK-NEXT:    [[LEVEL:%.*]] = call i8 @__kmpc_parallel_level()
 ; CHECK-NEXT:    store i8 [[LEVEL]], i8* @G, align 1
 ; CHECK-NEXT:    ret void
;
   %level = call i8 @__kmpc_parallel_level()
   store i8 %level, i8* @G
   ret void
 }
 
 declare void @foo()
 declare void @bar()
 declare i8 @__kmpc_parallel_level()
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1 zeroext, i1 zeroext, i1 zeroext) #1
-declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i1 zeroext, i1 zeroext) #1
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8 zeroext, i1 zeroext, i1 zeroext) #1
+declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i8 zeroext, i1 zeroext) #1
 
 !llvm.module.flags = !{!0, !1}
 !nvvm.annotations = !{!2, !3, !4}
 
 !0 = !{i32 7, !"openmp", i32 50}
 !1 = !{i32 7, !"openmp-device", i32 50}
 !2 = !{void ()* @none_spmd, !"kernel", i32 1}
 !3 = !{void ()* @spmd, !"kernel", i32 1}

llvm/test/Transforms/OpenMP/replace_globalization.ll

 
 @S = external local_unnamed_addr global i8*
 @0 = private unnamed_addr constant [113 x i8] c";llvm/test/Transforms/OpenMP/custom_state_machines_remarks.c;__omp_offloading_2a_d80d3d_test_fallback_l11;11;1;;\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([113 x i8], [113 x i8]* @0, i32 0, i32 0) }, align 8
 
 
 define dso_local void @foo() {
 entry:
-  %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %x = call i8* @__kmpc_alloc_shared(i64 4)
   call void @unknown_no_openmp()
   %x_on_stack = bitcast i8* %x to i32*
   %0 = bitcast i32* %x_on_stack to i8*
   call void @use(i8* %0)
   call void @__kmpc_free_shared(i8* %x, i64 4)
   call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
   ret void
 }
 
 define void @bar() {
-  %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i1 false, i1 true, i1 true)
+  %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   call void @unknown_no_openmp()
   call void @baz()
   call void @qux()
   call void @negative_qux_spmd()
   call void @__kmpc_target_deinit(%struct.ident_t* @1, i1 false, i1 true)
   ret void
 }
 
 define internal void @baz() {
 entry:
-  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 false, i1 true)
+  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 false, i1 true)
   call void @unknown_no_openmp()
   %cmp = icmp eq i32 %call, -1
   br i1 %cmp, label %master, label %exit
 master:
   %x = call i8* @__kmpc_alloc_shared(i64 16), !dbg !11
   %x_on_stack = bitcast i8* %x to [4 x i32]*
   %0 = bitcast [4 x i32]* %x_on_stack to i8*
   call void @use(i8* %0)
   call void @__kmpc_free_shared(i8* %x, i64 16)
   br label %exit
 exit:
   ret void
 }
 
 define internal void @qux() {
 entry:
-  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 true, i1 true)
+  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 true, i1 true)
   call void @unknown_no_openmp()
   %0 = icmp eq i32 %call, -1
   br i1 %0, label %master, label %exit
 master:
   %y = call i8* @__kmpc_alloc_shared(i64 4), !dbg !12
   %y_on_stack = bitcast i8* %y to [4 x i32]*
   %1 = bitcast [4 x i32]* %y_on_stack to i8*
   call void @use(i8* %1)
   call void @__kmpc_free_shared(i8* %y, i64 4)
   br label %exit
 exit:
   ret void
 }
 
 define internal void @negative_qux_spmd() {
 entry:
-  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 true, i1 true, i1 true)
+  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 2, i1 true, i1 true)
   call void @unknown_no_openmp()
   %0 = icmp eq i32 %call, -1
   br i1 %0, label %master, label %exit
 master:
   %y = call i8* @__kmpc_alloc_shared(i64 24), !dbg !12
   %y_on_stack = bitcast i8* %y to [6 x i32]*
   %1 = bitcast [6 x i32]* %y_on_stack to i8*
   call void @use(i8* %1)

 declare void @__kmpc_free_shared(i8*, i64)
 
 declare i32 @llvm.nvvm.read.ptx.sreg.tid.x()
 
 declare i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
 
 declare i32 @llvm.nvvm.read.ptx.sreg.warpsize()
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 declare void @unknown_no_openmp() "llvm.assume"="omp_no_openmp"
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5, !6}
 !nvvm.annotations = !{!7, !8}
 
 !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 12.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)

 ; CHECK: @[[S:[a-zA-Z0-9_$"\\.-]+]] = external local_unnamed_addr global i8*
 ; CHECK: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [113 x i8] c"
 ; CHECK: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([113 x i8], [113 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8
 ; CHECK: @[[X:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [16 x i8] undef, align 32
 ; CHECK: @[[Y:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] undef, align 32
 ;.
 ; CHECK-LABEL: define {{[^@]+}}@foo() {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
+; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
 ; CHECK-NEXT:    [[X:%.*]] = call i8* @__kmpc_alloc_shared(i64 4) #[[ATTR4:[0-9]+]]
 ; CHECK-NEXT:    call void @unknown_no_openmp()
 ; CHECK-NEXT:    call void @use.internalized(i8* nofree writeonly [[X]]) #[[ATTR5:[0-9]+]]
 ; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[X]], i64 4) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 ;
 ; CHECK-LABEL: define {{[^@]+}}@bar() {
-; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 true, i1 true)
+; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 true, i1 true)
 ; CHECK-NEXT:    call void @unknown_no_openmp()
 ; CHECK-NEXT:    call void @baz()
 ; CHECK-NEXT:    call void @qux()
 ; CHECK-NEXT:    call void @negative_qux_spmd()
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 ;
 ; CHECK-LABEL: define {{[^@]+}}@baz() {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i1 noundef false, i1 noundef false, i1 noundef true)
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i8 noundef 1, i1 noundef false, i1 noundef true)
 ; CHECK-NEXT:    call void @unknown_no_openmp()
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], -1
 ; CHECK-NEXT:    br i1 [[CMP]], label [[MASTER:%.*]], label [[EXIT:%.*]]
 ; CHECK:       master:
 ; CHECK-NEXT:    call void @use.internalized(i8* nofree writeonly addrspacecast (i8 addrspace(3)* getelementptr inbounds ([16 x i8], [16 x i8] addrspace(3)* @x, i32 0, i32 0) to i8*)) #[[ATTR5]]
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
 ;
 ; CHECK-LABEL: define {{[^@]+}}@qux() {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i1 noundef false, i1 noundef true, i1 noundef true)
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i8 noundef 1, i1 noundef true, i1 noundef true)
 ; CHECK-NEXT:    call void @unknown_no_openmp()
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[CALL]], -1
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MASTER:%.*]], label [[EXIT:%.*]]
 ; CHECK:       master:
 ; CHECK-NEXT:    call void @use.internalized(i8* nofree writeonly addrspacecast (i8 addrspace(3)* getelementptr inbounds ([4 x i8], [4 x i8] addrspace(3)* @y, i32 0, i32 0) to i8*)) #[[ATTR5]]
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
 ;
 ; CHECK-LABEL: define {{[^@]+}}@negative_qux_spmd() {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i1 noundef true, i1 noundef true, i1 noundef true)
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* noundef nonnull @[[GLOB1]], i8 noundef 2, i1 noundef true, i1 noundef true)
 ; CHECK-NEXT:    call void @unknown_no_openmp()
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[CALL]], -1
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MASTER:%.*]], label [[EXIT:%.*]]
 ; CHECK:       master:
 ; CHECK-NEXT:    [[Y:%.*]] = call i8* @__kmpc_alloc_shared(i64 noundef 24) #[[ATTR4]], !dbg [[DBG9:![0-9]+]]
 ; CHECK-NEXT:    call void @use.internalized(i8* nofree writeonly [[Y]]) #[[ATTR5]]
 ; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[Y]], i64 noundef 24) #[[ATTR4]]
 ; CHECK-NEXT:    br label [[EXIT]]

llvm/test/Transforms/OpenMP/single_threaded_execution.ll

 ; RUN: opt -passes=openmp-opt -debug-only=openmp-opt -disable-output < %s 2>&1 | FileCheck %s
 ; RUN: opt -passes=openmp-opt -pass-remarks-analysis=openmp-opt -disable-output < %s 2>&1 | FileCheck %s --check-prefix=REMARKS
 ; REQUIRES: asserts
 ; ModuleID = 'single_threaded_exeuction.c'
 
 %struct.ident_t = type { i32, i32, i32, i32, i8* }
 
 @0 = private unnamed_addr constant [1 x i8] c"\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([1 x i8], [1 x i8]* @0, i32 0, i32 0) }, align 8
 
 
 ; CHECK-NOT: [openmp-opt] Basic block @kernel entry is executed by a single thread.
 ; CHECK: [openmp-opt] Basic block @kernel if.then is executed by a single thread.
 ; CHECK-NOT: [openmp-opt] Basic block @kernel if.else is executed by a single thread.
 ; CHECK-NOT: [openmp-opt] Basic block @kernel if.end is executed by a single thread.
 define void @kernel() {
-  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 false, i1 false)
+  %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 false, i1 false)
   %cmp = icmp eq i32 %call, -1
   br i1 %cmp, label %if.then, label %if.else
 if.then:
   br label %if.end
 if.else:
   br label %if.end
 if.end:
-  call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+  call void @__kmpc_target_deinit(%struct.ident_t* null, i8 1, i1 true)
   ret void
 }
 
 ; REMARKS: remark: single_threaded_execution.c:1:0: Could not internalize function. Some optimizations may not be possible.
 ; REMARKS-NOT: remark: single_threaded_execution.c:1:0: Could not internalize function. Some optimizations may not be possible.
 
 ; CHECK-NOT: [openmp-opt] Basic block @nvptx entry is executed by a single thread.
 ; CHECK-DAG: [openmp-opt] Basic block @nvptx if.then is executed by a single thread.

 }
 
 declare i32 @llvm.nvvm.read.ptx.sreg.tid.x()
 
 declare i32 @llvm.amdgcn.workitem.id.x()
 
 declare void @__kmpc_kernel_prepare_parallel(i8*)
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 attributes #0 = { cold noinline }
 
 !llvm.dbg.cu = !{!0}
 !llvm.module.flags = !{!3, !4, !5, !6}
 !nvvm.annotations = !{!7}
 
 !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 12.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)

llvm/test/Transforms/OpenMP/spmdization_guarding.ll

 ; CHECK-DISABLED: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [1 x i8*] [i8* @__omp_offloading_2a_fbfa7a_sequential_loop_l6_exec_mode], section "llvm.metadata"
 ; CHECK-DISABLED: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef
 ;.
 define weak void @__omp_offloading_2a_fbfa7a_sequential_loop_l6(i32* %x, i64 %N) #0 {
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_2a_fbfa7a_sequential_loop_l6
 ; CHECK-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32
-; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i1 true, i1 false, i1 false) #[[ATTR4:[0-9]+]]
+; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 2, i1 false, i1 false) #[[ATTR4:[0-9]+]]
 ; CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK:       user_code.entry:
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-NEXT:    [[ARRAYIDX1_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 1
 ; CHECK-NEXT:    [[SEXT:%.*]] = shl i64 [[N]], 32
 ; CHECK-NEXT:    [[IDXPROM_I:%.*]] = ashr exact i64 [[SEXT]], 32
 ; CHECK-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM_I]]

 ; CHECK-NEXT:    br label [[REGION_BARRIER17]]
 ; CHECK:       region.barrier17:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
 ; CHECK-NEXT:    br label [[REGION_EXIT18:%.*]]
 ; CHECK:       region.exit18:
 ; CHECK-NEXT:    [[CALL14_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
 ; CHECK-NEXT:    [[CALL15_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
 ; CHECK-NEXT:    [[CALL16_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
-; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i1 true, i1 false) #[[ATTR4]]
+; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i8 2, i1 false) #[[ATTR4]]
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_2a_fbfa7a_sequential_loop_l6
 ; CHECK-DISABLED-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; CHECK-DISABLED-NEXT:    [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32
-; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i1 false, i1 false, i1 true) #[[ATTR4:[0-9]+]]
+; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 false, i1 true) #[[ATTR4:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK-DISABLED:       worker_state_machine.begin:
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 ; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
 ; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
 ; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-DISABLED-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null

 ; CHECK-DISABLED-NEXT:    store i32 [[CALL8_I]], i32* [[ARRAYIDX10_I]], align 4, !noalias !8
 ; CHECK-DISABLED-NEXT:    [[CALL11_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
 ; CHECK-DISABLED-NEXT:    [[IDXPROM12_I:%.*]] = sext i32 [[CALL11_I]] to i64
 ; CHECK-DISABLED-NEXT:    [[ARRAYIDX13_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM12_I]]
 ; CHECK-DISABLED-NEXT:    store i32 [[CALL11_I]], i32* [[ARRAYIDX13_I]], align 4, !noalias !8
 ; CHECK-DISABLED-NEXT:    [[CALL14_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
 ; CHECK-DISABLED-NEXT:    [[CALL15_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
 ; CHECK-DISABLED-NEXT:    [[CALL16_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8
-; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i1 false, i1 true) #[[ATTR4]]
+; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 true) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
   %N.addr.sroa.0.0.extract.trunc = trunc i64 %N to i32
-  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 true, i1 true) #3
+  %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 true, i1 true) #3
   %exec_user_code = icmp eq i32 %0, -1
   br i1 %exec_user_code, label %user_code.entry, label %worker.exit
 
 user_code.entry:                                  ; preds = %entry
   %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @1)
   store i32 0, i32* %x, align 4, !noalias !8
   %arrayidx1.i = getelementptr inbounds i32, i32* %x, i64 1
   store i32 1, i32* %arrayidx1.i, align 4, !noalias !8
__omp_outlined__.exit:                            ; preds = %for.cond.i
   store i32 %call8.i, i32* %arrayidx10.i, align 4, !noalias !8
   %call11.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8
   %idxprom12.i = sext i32 %call11.i to i64
   %arrayidx13.i = getelementptr inbounds i32, i32* %x, i64 %idxprom12.i
   store i32 %call11.i, i32* %arrayidx13.i, align 4, !noalias !8
   %call14.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8
   %call15.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8
   %call16.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8
-  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @1, i1 false, i1 true) #3
+  call void @__kmpc_target_deinit(%struct.ident_t* nonnull @1, i8 1, i1 true) #3
   ret void
 
 worker.exit:                                      ; preds = %entry
   ret void
 }
 
 define internal void @__omp_outlined__1(i32* noalias %.global_tid., i32* noalias %.bound_tid.) {
 ; CHECK-LABEL: define {{[^@]+}}@__omp_outlined__1

 ; CHECK-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) {
 ; CHECK-DISABLED-NEXT:    ret void
 ;
   ret void
 }
 
 declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64)
 
-declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1)
 
 ; Function Attrs: convergent
 declare i32 @no_openmp(i32*) #1
 
 ; Function Attrs: convergent nounwind readonly willreturn
 declare void @pure() #2
 
 ; Function Attrs: nounwind
 declare i32 @__kmpc_global_thread_num(%struct.ident_t*) #3
 
-declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1)
 
 ; Function Attrs: inaccessiblememonly nofree nosync nounwind willreturn
 declare void @llvm.experimental.noalias.scope.decl(metadata) #4
 
 attributes #0 = { convergent norecurse nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" }
 attributes #1 = { convergent "frame-pointer"="all" "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" }
 attributes #2 = { convergent nounwind readonly willreturn "frame-pointer"="all" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" }
 attributes #3 = { nounwind }

openmp/libomptarget/deviceRTLs/common/include/target.h

 /// well. The runtime calls are therefore also desiged to carry sufficient
 /// information necessary for optimizations.
 ///
 ///
 /// Intended usage:
 ///
 /// \code
 /// void kernel(...) {
-///   ThreadKind = __kmpc_target_init(Ident, /* IsSPMD */ false,
+///   ThreadKind = __kmpc_target_init(Ident, /* Mode */ 1,
 ///                                   /* UseGenericStateMachine */ true,
 ///                                   /* RequiresFullRuntime */ ... );
 ///   if (ThreadKind == -1) {
 ///     // User defined kernel code.
 ///   }
 ///   __kmpc_target_deinit(...);
 /// }
 /// \endcode
 ///
 /// Which can be transformed to:
 ///
 /// \code
 /// void kernel(...) {
-///   ThreadKind = __kmpc_target_init(Ident, /* IsSPMD */ false,
+///   ThreadKind = __kmpc_target_init(Ident, /* Mode */ 1,
 ///                                   /* UseGenericStateMachine */ false,
 ///                                   /* RequiresFullRuntime */ ... );
 ///   if (ThreadKind == -1) {
 ///     // User defined kernel code.
 ///   } else {
 ///     assume(ThreadKind == ThreadId);
 ///     // Custom, kernel-specific state machine code.
 ///   }
 ///   __kmpc_target_deinit(...);
 /// }
 /// \endcode
 ///
 ///
 ///{
 
 /// Initialization
 ///
 /// Must be called by all threads.
 ///
 /// \param Ident               Source location identification, can be NULL.
 ///
-int32_t __kmpc_target_init(ident_t *Ident, bool IsSPMD,
+int32_t __kmpc_target_init(ident_t *Ident, int8_t Mode,
                            bool UseGenericStateMachine,
                            bool RequiresFullRuntime);
 
 /// De-Initialization
 ///
 /// Must be called by the main thread in generic mode, can be called by all
 /// threads. Must be called by all threads in SPMD mode.
 ///
 /// In non-SPMD, this function releases the workers trapped in a state machine
 /// and also any memory dynamically allocated by the runtime.
 ///
 /// \param Ident Source location identification, can be NULL.
 ///
-void __kmpc_target_deinit(ident_t *Ident, bool IsSPMD,
+void __kmpc_target_deinit(ident_t *Ident, int8_t Mode,
                           bool RequiresFullRuntime);
 
 ///}
 }
 #endif

openmp/libomptarget/deviceRTLs/common/omptarget.h

 //===---- omptarget.h - OpenMP GPU initialization ---------------- CUDA -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file contains the declarations of all library macros, types,
 // and functions.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef OMPTARGET_H
 #define OMPTARGET_H
 
 #include "common/allocator.h"
 #include "common/debug.h" // debug
 #include "common/state-queue.h"
 #include "common/support.h"
 #include "interface.h" // interfaces with omp, compiler, and user
 #include "target_impl.h"
 
 #define OMPTARGET_NVPTX_VERSION 1.1
 

 extern uint8_t EXTERN_SHARED(parallelLevel)[MAX_THREADS_PER_TEAM / WARPSIZE];
 #endif
 extern uint16_t EXTERN_SHARED(threadLimit);
 extern uint16_t EXTERN_SHARED(threadsInTeam);
 extern uint16_t EXTERN_SHARED(nThreads);
 extern omptarget_nvptx_ThreadPrivateContext *
     EXTERN_SHARED(omptarget_nvptx_threadPrivateContext);
 
-extern uint32_t EXTERN_SHARED(execution_param);
+extern int8_t EXTERN_SHARED(execution_param);
 extern void *EXTERN_SHARED(ReductionScratchpadPtr);
 
 ////////////////////////////////////////////////////////////////////////////////
 // work function (outlined parallel/simd functions) and arguments.
 // needed for L1 parallelism only.
 ////////////////////////////////////////////////////////////////////////////////
 
 typedef void *omptarget_nvptx_WorkFn;

openmp/libomptarget/deviceRTLs/common/src/omp_data.cu

 // communicate with the workers.  Since it is in shared memory, there is one
 // copy of these variables for each kernel, instance, and team.
 ////////////////////////////////////////////////////////////////////////////////
 omptarget_nvptx_WorkFn SHARED(omptarget_nvptx_workFn);
 
 ////////////////////////////////////////////////////////////////////////////////
 // OpenMP kernel execution parameters
 ////////////////////////////////////////////////////////////////////////////////
-uint32_t SHARED(execution_param);
+int8_t SHARED(execution_param);
 
 ////////////////////////////////////////////////////////////////////////////////
 // Scratchpad for teams reduction.
 ////////////////////////////////////////////////////////////////////////////////
 void *SHARED(ReductionScratchpadPtr);
 
 #pragma omp end declare target

openmp/libomptarget/deviceRTLs/common/src/omptarget.cu

static void __kmpc_generic_kernel_init() {
 
   if (GetLaneId() == 0)
     parallelLevel[GetWarpId()] = 0;
 
   int threadIdInBlock = __kmpc_get_hardware_thread_id_in_block();
   if (threadIdInBlock != GetMasterThreadID())
     return;
 
-  setExecutionParameters(Generic, RuntimeInitialized);
+  setExecutionParameters(OMP_TGT_EXEC_MODE_GENERIC, OMP_TGT_RUNTIME_INITIALIZED);
   ASSERT0(LT_FUSSY, threadIdInBlock == GetMasterThreadID(),
           "__kmpc_kernel_init() must be called by team master warp only!");
   PRINT0(LD_IO, "call to __kmpc_kernel_init for master\n");
 
   // Get a state object from the queue.
   int slot = __kmpc_impl_smid() % MAX_SM;
   usedSlotIdx = slot;
   omptarget_nvptx_threadPrivateContext =
omptarget_nvptx_device_State[slot].Enqueue(
       omptarget_nvptx_threadPrivateContext);
   // Done with work.  Kill the workers.
   omptarget_nvptx_workFn = 0;
 }
 
 static void __kmpc_spmd_kernel_init(bool RequiresFullRuntime) {
   PRINT0(LD_IO, "call to __kmpc_spmd_kernel_init\n");
 
-  setExecutionParameters(Spmd, RequiresFullRuntime ? RuntimeInitialized
-                         : RuntimeUninitialized);
+  setExecutionParameters(OMP_TGT_EXEC_MODE_SPMD,
+                         RequiresFullRuntime ? OMP_TGT_RUNTIME_INITIALIZED
+                                             : OMP_TGT_RUNTIME_UNINITIALIZED);
   int threadId = __kmpc_get_hardware_thread_id_in_block();
   if (threadId == 0) {
     usedSlotIdx = __kmpc_impl_smid() % MAX_SM;
   }
 
   if (GetLaneId() == 0) {
     parallelLevel[GetWarpId()] =
         1 + (__kmpc_get_hardware_num_threads_in_block() > 1
if (threadId == 0) {
     int slot = usedSlotIdx;
     omptarget_nvptx_device_State[slot].Enqueue(
         omptarget_nvptx_threadPrivateContext);
   }
 }
 
 // Return true if the current target region is executed in SPMD mode.
 EXTERN int8_t __kmpc_is_spmd_exec_mode() {
-  return (execution_param & ModeMask) == Spmd;
+  return execution_param & OMP_TGT_EXEC_MODE_SPMD;
 }
 
 EXTERN int8_t __kmpc_is_generic_main_thread(kmp_int32 Tid) {
   return !__kmpc_is_spmd_exec_mode() && __kmpc_is_generic_main_thread_id(Tid);
 }
 
 NOINLINE EXTERN int8_t __kmpc_is_generic_main_thread_id(kmp_int32 Tid) {
   return GetMasterThreadID() == Tid;
do {
     }
 
     __kmpc_barrier_simple_spmd(Ident, TId);
 
   } while (true);
 }
 
 EXTERN
-int32_t __kmpc_target_init(ident_t *Ident, bool IsSPMD,
+int32_t __kmpc_target_init(ident_t *Ident, int8_t Mode,
                            bool UseGenericStateMachine,
                            bool RequiresFullRuntime) {
+  const bool IsSPMD = Mode & OMP_TGT_EXEC_MODE_SPMD;
   int TId = __kmpc_get_hardware_thread_id_in_block();
   if (IsSPMD)
     __kmpc_spmd_kernel_init(RequiresFullRuntime);
   else
     __kmpc_generic_kernel_init();
 
    if (IsSPMD) {
     __kmpc_barrier_simple_spmd(Ident, TId);
      return -1;
    }
 
    if (TId == GetMasterThreadID())
      return -1;
 
   if (UseGenericStateMachine)
     __kmpc_target_region_state_machine(Ident);
 
   return TId;
 }
 
 EXTERN
-void __kmpc_target_deinit(ident_t *Ident, bool IsSPMD,
-                           bool RequiresFullRuntime) {
+void __kmpc_target_deinit(ident_t *Ident, int8_t Mode,
+  const bool IsSPMD = Mode & OMP_TGT_EXEC_MODE_SPMD;
   if (IsSPMD)
     __kmpc_spmd_kernel_deinit(RequiresFullRuntime);
   else
     __kmpc_generic_kernel_deinit();
 }
 
-
 #pragma omp end declare target

openmp/libomptarget/deviceRTLs/common/src/support.cu

 #include "common/debug.h"
 #include "common/omptarget.h"
 #include "common/support.h"
 
 ////////////////////////////////////////////////////////////////////////////////
 // Execution Parameters
 ////////////////////////////////////////////////////////////////////////////////
 
-void setExecutionParameters(ExecutionMode EMode, RuntimeMode RMode) {
+void setExecutionParameters(OMPTgtExecModeFlags EMode,
+                            OMPTgtRuntimeModeFlags RMode) {
   execution_param = EMode;
   execution_param |= RMode;
 }
 
-bool isGenericMode() { return (execution_param & ModeMask) == Generic; }
+bool isGenericMode() { return execution_param & OMP_TGT_EXEC_MODE_GENERIC; }
 
-bool isRuntimeUninitialized() {
-  return (execution_param & RuntimeMask) == RuntimeUninitialized;
-}
+bool isRuntimeUninitialized() { return !isRuntimeInitialized(); }
 
 bool isRuntimeInitialized() {
-  return (execution_param & RuntimeMask) == RuntimeInitialized;
+  return execution_param & OMP_TGT_RUNTIME_INITIALIZED;
 }
 
 ////////////////////////////////////////////////////////////////////////////////
 // support: get info from machine
 ////////////////////////////////////////////////////////////////////////////////
 
 ////////////////////////////////////////////////////////////////////////////////
 //

openmp/libomptarget/deviceRTLs/common/support.h

 //===--------- support.h - OpenMP GPU support functions ---------- CUDA -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // Wrapper to some functions natively supported by the GPU.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef OMPTARGET_SUPPORT_H
 #define OMPTARGET_SUPPORT_H
 
 #include "interface.h"
 #include "target_impl.h"
 
 ////////////////////////////////////////////////////////////////////////////////
 // Execution Parameters
 ////////////////////////////////////////////////////////////////////////////////
-enum ExecutionMode {
-  Spmd = 0x00u,
-  Generic = 0x01u,
-  ModeMask = 0x01u,
+enum OMPTgtExecModeFlags : int8_t {
+  OMP_TGT_EXEC_MODE_GENERIC = 1 << 0,
+  OMP_TGT_EXEC_MODE_SPMD = 1 << 1
 };
 
-enum RuntimeMode {
-  RuntimeInitialized = 0x00u,
-  RuntimeUninitialized = 0x02u,
-  RuntimeMask = 0x02u,
+enum OMPTgtRuntimeModeFlags : int8_t {
+  OMP_TGT_RUNTIME_UNINITIALIZED = 0,
+  OMP_TGT_RUNTIME_INITIALIZED = 1 << 2
 };
 
-void setExecutionParameters(ExecutionMode EMode, RuntimeMode RMode);
+void setExecutionParameters(OMPTgtExecModeFlags EMode,
+                            OMPTgtRuntimeModeFlags RMode);
 bool isGenericMode();
 bool isRuntimeUninitialized();
 bool isRuntimeInitialized();
 
 ////////////////////////////////////////////////////////////////////////////////
 // get info from machine
 ////////////////////////////////////////////////////////////////////////////////
 

openmp/libomptarget/deviceRTLs/interface.h

 typedef enum omp_proc_bind_t {
   omp_proc_bind_false = 0,
   omp_proc_bind_true = 1,
   omp_proc_bind_master = 2,
   omp_proc_bind_close = 3,
   omp_proc_bind_spread = 4
 } omp_proc_bind_t;
 
 EXTERN double omp_get_wtick(void);
 EXTERN double omp_get_wtime(void);
 
 EXTERN void omp_set_num_threads(int num);
 EXTERN int omp_get_num_threads(void);
 EXTERN int omp_get_max_threads(void);
 EXTERN int omp_get_thread_limit(void);
 EXTERN int omp_get_thread_num(void);
 EXTERN int omp_get_num_procs(void);
 EXTERN int omp_in_parallel(void);
 EXTERN int omp_in_final(void);
 EXTERN void omp_set_dynamic(int flag);
 EXTERN int omp_get_dynamic(void);
 EXTERN void omp_set_nested(int flag);
 EXTERN int omp_get_nested(void);
 EXTERN void omp_set_max_active_levels(int level);
 EXTERN int omp_get_max_active_levels(void);
 EXTERN int omp_get_level(void);
 EXTERN int omp_get_active_level(void);
 EXTERN int omp_get_ancestor_thread_num(int level);
 EXTERN int omp_get_team_size(int level);
 
 EXTERN void omp_init_lock(omp_lock_t *lock);
 EXTERN void omp_init_nest_lock(omp_nest_lock_t *lock);
 EXTERN void omp_destroy_lock(omp_lock_t *lock);
 EXTERN void omp_destroy_nest_lock(omp_nest_lock_t *lock);
 EXTERN void omp_set_lock(omp_lock_t *lock);
 EXTERN void omp_set_nest_lock(omp_nest_lock_t *lock);

 
 // cancel
 EXTERN int32_t __kmpc_cancellationpoint(kmp_Ident *loc, int32_t global_tid,
                                         int32_t cancelVal);
 EXTERN int32_t __kmpc_cancel(kmp_Ident *loc, int32_t global_tid,
                              int32_t cancelVal);
 
 // non standard
-EXTERN int32_t __kmpc_target_init(ident_t *Ident, bool IsSPMD,
+EXTERN int32_t __kmpc_target_init(ident_t *Ident, int8_t Mode,
                                   bool UseGenericStateMachine,
                                   bool RequiresFullRuntime);
-EXTERN void __kmpc_target_deinit(ident_t *Ident, bool IsSPMD,
+EXTERN void __kmpc_target_deinit(ident_t *Ident, int8_t Mode,
                                  bool RequiresFullRuntime);
 EXTERN void __kmpc_kernel_prepare_parallel(void *WorkFn);
 EXTERN bool __kmpc_kernel_parallel(void **WorkFn);
 EXTERN void __kmpc_kernel_end_parallel();
 
 EXTERN void __kmpc_data_sharing_init_stack();
 EXTERN void __kmpc_begin_sharing_variables(void ***GlobalArgs, size_t nArgs);
 EXTERN void __kmpc_end_sharing_variables();