diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -2036,7 +2036,8 @@
         UndefValue::get(Int8Ty), F->getName() + ".ID");
 
     for (Use *U : ToBeReplacedStateMachineUses)
-      U->set(ConstantExpr::getBitCast(ID, U->get()->getType()));
+      U->set(ConstantExpr::getPointerBitCastOrAddrSpaceCast(
+          ID, U->get()->getType()));
 
     ++NumOpenMPParallelRegionsReplacedInGPUStateMachine;
 
@@ -3424,8 +3425,9 @@
 
     // Create local storage for the work function pointer.
     Type *VoidPtrTy = Type::getInt8PtrTy(Ctx);
-    AllocaInst *WorkFnAI = new AllocaInst(VoidPtrTy, 0, "worker.work_fn.addr",
-                                          &Kernel->getEntryBlock().front());
+    AllocaInst *WorkFnAI =
+        new AllocaInst(VoidPtrTy, (unsigned int)AddressSpace::Local,
+                       "worker.work_fn.addr", &Kernel->getEntryBlock().front());
     WorkFnAI->setDebugLoc(DLoc);
 
     auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
@@ -3445,14 +3447,23 @@
     CallInst::Create(BarrierFn, {Ident, GTid}, "", StateMachineBeginBB)
         ->setDebugLoc(DLoc);
 
+    Instruction *WorkFnAIGeneric = new AddrSpaceCastInst(
+        WorkFnAI,
+        PointerType::getWithSamePointeeType(
+            cast<PointerType>(WorkFnAI->getType()),
+            (unsigned int)AddressSpace::Generic),
+        WorkFnAI->getName() + ".generic", StateMachineBeginBB);
+    WorkFnAIGeneric->setDebugLoc(DLoc);
+
     FunctionCallee KernelParallelFn =
         OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
             M, OMPRTL___kmpc_kernel_parallel);
-    Instruction *IsActiveWorker = CallInst::Create(
-        KernelParallelFn, {WorkFnAI}, "worker.is_active", StateMachineBeginBB);
+    Instruction *IsActiveWorker =
+        CallInst::Create(KernelParallelFn, {WorkFnAIGeneric},
+                         "worker.is_active", StateMachineBeginBB);
     IsActiveWorker->setDebugLoc(DLoc);
-    Instruction *WorkFn = new LoadInst(VoidPtrTy, WorkFnAI, "worker.work_fn",
-                                       StateMachineBeginBB);
+    Instruction *WorkFn = new LoadInst(VoidPtrTy, WorkFnAIGeneric,
+                                       "worker.work_fn", StateMachineBeginBB);
     WorkFn->setDebugLoc(DLoc);
 
     FunctionType *ParallelRegionFnTy = FunctionType::get(
diff --git a/llvm/test/Transforms/OpenMP/custom_state_machines.ll b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
--- a/llvm/test/Transforms/OpenMP/custom_state_machines.ll
+++ b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
@@ -865,8 +865,8 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7:[0-9]+]]
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR8:[0-9]+]]
+; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8:[0-9]+]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR9:[0-9]+]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -908,7 +908,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_l22
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -916,8 +916,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -966,11 +967,11 @@
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-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)
-; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7]]
+; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-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)
 ; CHECK-NEXT:    ret void
@@ -982,7 +983,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9:[0-9]+]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10:[0-9]+]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1008,7 +1009,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1032,7 +1033,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1040,8 +1041,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1095,13 +1097,13 @@
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR7]]
-; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR8]]
+; CHECK-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-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)
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR7]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR8]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1133,7 +1135,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1179,7 +1181,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1187,8 +1189,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1242,7 +1245,7 @@
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-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)
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR9]]
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR10]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-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)
 ; CHECK-NEXT:    ret void
@@ -1254,7 +1257,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1280,7 +1283,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1304,7 +1307,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1312,8 +1315,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1365,7 +1369,7 @@
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-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)
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-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)
 ; CHECK-NEXT:    ret void
@@ -1377,7 +1381,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1403,7 +1407,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1427,7 +1431,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_pure_l77
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1435,8 +1439,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1485,7 +1490,7 @@
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-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)
@@ -1500,7 +1505,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1526,7 +1531,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1550,7 +1555,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1558,8 +1563,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1600,8 +1606,8 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    [[CALL:%.*]] = call i32 bitcast (i32 (...)* @omp_get_thread_num to i32 ()*)() #[[ATTR9]]
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[CALL]]) #[[ATTR7]]
+; CHECK-NEXT:    [[CALL:%.*]] = call i32 bitcast (i32 (...)* @omp_get_thread_num to i32 ()*)() #[[ATTR10]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[CALL]]) #[[ATTR8]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1617,8 +1623,8 @@
 ; CHECK-NEXT:    br label [[RETURN:%.*]]
 ; CHECK:       if.end:
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[A]], 1
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[SUB]]) #[[ATTR7]]
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after.internalized() #[[ATTR7]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[SUB]]) #[[ATTR8]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after.internalized() #[[ATTR8]]
 ; CHECK-NEXT:    br label [[RETURN]]
 ; CHECK:       return:
 ; CHECK-NEXT:    ret void
@@ -1638,8 +1644,8 @@
 ; CHECK:       if.end:
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 1
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after(i32 [[SUB]]) #[[ATTR9]]
-; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after() #[[ATTR9]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after(i32 [[SUB]]) #[[ATTR10]]
+; CHECK-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after() #[[ATTR10]]
 ; CHECK-NEXT:    br label [[RETURN]]
 ; CHECK:       return:
 ; CHECK-NEXT:    ret void
@@ -1649,7 +1655,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1657,8 +1663,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1693,7 +1700,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @weak_callee_empty() #[[ATTR7]]
+; CHECK-NEXT:    call void @weak_callee_empty() #[[ATTR8]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1710,7 +1717,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1736,7 +1743,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1784,7 +1791,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-NEXT:    ret void
 ;
 ;
@@ -1828,8 +1835,8 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7:[0-9]+]]
-; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR8:[0-9]+]]
+; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8:[0-9]+]]
+; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR9:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -1895,11 +1902,11 @@
 ; CHECK-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-DISABLED-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-DISABLED-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)
-; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-DISABLED-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)
 ; CHECK-DISABLED-NEXT:    ret void
@@ -1911,7 +1918,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9:[0-9]+]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -1937,7 +1944,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -1984,13 +1991,13 @@
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-DISABLED-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR7]]
-; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_before.internalized() #[[ATTR8]]
+; CHECK-DISABLED-NEXT:    call void @no_parallel_region_in_here.internalized() #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-DISABLED-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)
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_after.internalized() #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2022,7 +2029,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2095,7 +2102,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-DISABLED-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)
-; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 @unknown() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-DISABLED-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)
 ; CHECK-DISABLED-NEXT:    ret void
@@ -2107,7 +2114,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2133,7 +2140,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2184,7 +2191,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-DISABLED-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)
-; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-DISABLED-NEXT:    [[TMP2:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS1]] to i8**
 ; CHECK-DISABLED-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)
 ; CHECK-DISABLED-NEXT:    ret void
@@ -2196,7 +2203,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2222,7 +2229,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2270,7 +2277,7 @@
 ; CHECK-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS1:%.*]] = alloca [0 x i8*], align 8
 ; CHECK-DISABLED-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
-; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR8]]
+; CHECK-DISABLED-NEXT:    call void @unknown_no_openmp() #[[ATTR9]]
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; CHECK-DISABLED-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)
@@ -2285,7 +2292,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2311,7 +2318,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p1() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2355,8 +2362,8 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 bitcast (i32 (...)* @omp_get_thread_num to i32 ()*)() #[[ATTR9]]
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[CALL]]) #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    [[CALL:%.*]] = call i32 bitcast (i32 (...)* @omp_get_thread_num to i32 ()*)() #[[ATTR10]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[CALL]]) #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2372,8 +2379,8 @@
 ; CHECK-DISABLED-NEXT:    br label [[RETURN:%.*]]
 ; CHECK-DISABLED:       if.end:
 ; CHECK-DISABLED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[A]], 1
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[SUB]]) #[[ATTR7]]
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after.internalized() #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after.internalized(i32 [[SUB]]) #[[ATTR8]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after.internalized() #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    br label [[RETURN]]
 ; CHECK-DISABLED:       return:
 ; CHECK-DISABLED-NEXT:    ret void
@@ -2393,8 +2400,8 @@
 ; CHECK-DISABLED:       if.end:
 ; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
 ; CHECK-DISABLED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 1
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after(i32 [[SUB]]) #[[ATTR9]]
-; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after(i32 [[SUB]]) #[[ATTR10]]
+; CHECK-DISABLED-NEXT:    call void @simple_state_machine_interprocedural_nested_recursive_after_after() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    br label [[RETURN]]
 ; CHECK-DISABLED:       return:
 ; CHECK-DISABLED-NEXT:    ret void
@@ -2424,7 +2431,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @weak_callee_empty() #[[ATTR7]]
+; CHECK-DISABLED-NEXT:    call void @weak_callee_empty() #[[ATTR8]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2441,7 +2448,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2467,7 +2474,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
@@ -2515,7 +2522,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR9]]
+; CHECK-DISABLED-NEXT:    call void @p0() #[[ATTR10]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 ;
diff --git a/llvm/test/Transforms/OpenMP/spmdization.ll b/llvm/test/Transforms/OpenMP/spmdization.ll
--- a/llvm/test/Transforms/OpenMP/spmdization.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization.ll
@@ -124,9 +124,9 @@
 ; 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* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-NEXT:    call void @__omp_outlined__(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]]
+; CHECK-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
@@ -135,7 +135,7 @@
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_l5
 ; CHECK-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
 ; CHECK-DISABLED-NEXT:  entry:
-; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
@@ -144,8 +144,9 @@
 ; 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_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], 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:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -172,9 +173,9 @@
 ; CHECK-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK-DISABLED:       user_code.entry:
-; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
+; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__(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]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
@@ -227,7 +228,7 @@
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR5:[0-9]+]]
+; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__
@@ -255,7 +256,7 @@
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-DISABLED-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK-DISABLED:       for.end:
-; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR5:[0-9]+]]
+; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -297,7 +298,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @unknown() #[[ATTR6:[0-9]+]]
+; CHECK-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1
@@ -305,7 +306,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR6:[0-9]+]]
+; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -331,7 +332,7 @@
 ; 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:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; CHECK-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper
@@ -345,7 +346,7 @@
 ; CHECK-DISABLED-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -381,9 +382,9 @@
 ; 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* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-NEXT:    call void @__omp_outlined__2(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]]
+; CHECK-NEXT:    call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
@@ -392,7 +393,7 @@
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_stack_var_l20
 ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
 ; CHECK-DISABLED-NEXT:  entry:
-; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
@@ -401,8 +402,9 @@
 ; 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_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], 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:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -429,9 +431,9 @@
 ; CHECK-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK-DISABLED:       user_code.entry:
-; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__2(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]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
@@ -467,7 +469,7 @@
 ; CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; CHECK-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 1
 ; CHECK-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[TMP0]] to i32*
-; CHECK-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR5]]
+; CHECK-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
 ; CHECK-NEXT:    store i32 0, i32* [[I]], align 4
 ; CHECK-NEXT:    br label [[FOR_COND:%.*]]
 ; CHECK:       for.cond:
@@ -485,7 +487,7 @@
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__2
@@ -498,7 +500,7 @@
 ; CHECK-DISABLED-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 1
 ; CHECK-DISABLED-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[TMP0]] to i32*
-; CHECK-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR5]]
+; CHECK-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[I]], align 4
 ; CHECK-DISABLED-NEXT:    br label [[FOR_COND:%.*]]
 ; CHECK-DISABLED:       for.cond:
@@ -516,7 +518,7 @@
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-DISABLED-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK-DISABLED:       for.end:
-; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -566,7 +568,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3
@@ -574,7 +576,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -598,7 +600,7 @@
 ; 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:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; CHECK-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper
@@ -612,7 +614,7 @@
 ; CHECK-DISABLED-NEXT:    store i16 [[TMP0]], i16* [[DOTADDR]], align 2
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -640,9 +642,9 @@
 ; 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* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-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]]
+; CHECK-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]]) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
@@ -651,7 +653,7 @@
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_l35
 ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
 ; CHECK-DISABLED-NEXT:  entry:
-; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
@@ -660,8 +662,9 @@
 ; 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_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], 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:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -688,9 +691,9 @@
 ; CHECK-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK-DISABLED:       user_code.entry:
-; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-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]]
+; CHECK-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]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
@@ -743,7 +746,7 @@
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__4
@@ -773,7 +776,7 @@
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-DISABLED-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-DISABLED:       for.end:
-; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -826,7 +829,7 @@
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4
 ; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[X]], align 4
-; CHECK-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5
@@ -839,7 +842,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4
 ; CHECK-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4
-; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -873,7 +876,7 @@
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 ; CHECK-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 8
-; CHECK-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR3]]
+; CHECK-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper
@@ -891,7 +894,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 ; CHECK-DISABLED-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 ; CHECK-DISABLED-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 8
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -921,9 +924,9 @@
 ; 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* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-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]]
+; CHECK-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]]) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 true, i1 false)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
@@ -932,7 +935,7 @@
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_guarded_l50
 ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
 ; CHECK-DISABLED-NEXT:  entry:
-; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
@@ -941,8 +944,9 @@
 ; 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_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], 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:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -969,9 +973,9 @@
 ; CHECK-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK-DISABLED:       user_code.entry:
-; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
-; CHECK-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]]
+; CHECK-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]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
@@ -1039,7 +1043,7 @@
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__6
@@ -1071,7 +1075,7 @@
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[I]], align 4
 ; CHECK-DISABLED-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-DISABLED:       for.end:
-; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR5]]
+; CHECK-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1125,7 +1129,7 @@
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4
 ; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; CHECK-NEXT:    store i32 [[INC]], i32* [[X]], align 4
-; CHECK-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7
@@ -1138,7 +1142,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4
 ; CHECK-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; CHECK-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4
-; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1172,7 +1176,7 @@
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 ; CHECK-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 8
-; CHECK-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR3]]
+; CHECK-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper
@@ -1190,7 +1194,7 @@
 ; CHECK-DISABLED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
 ; CHECK-DISABLED-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32**
 ; CHECK-DISABLED-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[TMP4]], align 8
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR3]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP5]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1214,7 +1218,7 @@
 ; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
 ; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
@@ -1222,8 +1226,9 @@
 ; CHECK-NEXT:    br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
 ; CHECK:       worker_state_machine.begin:
 ; CHECK-NEXT:    call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
-; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
+; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
 ; CHECK-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
 ; CHECK-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
 ; CHECK-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1244,8 +1249,8 @@
 ; 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* @[[GLOB1]]) #[[ATTR3]]
-; CHECK-NEXT:    call void @__omp_outlined__8(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]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; CHECK-NEXT:    call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-NEXT:    ret void
 ; CHECK:       worker.exit:
@@ -1254,7 +1259,7 @@
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
 ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
 ; CHECK-DISABLED-NEXT:  entry:
-; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; CHECK-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 ; CHECK-DISABLED-NEXT:    store i32 0, i32* [[DOTZERO_ADDR]], align 4
@@ -1263,8 +1268,9 @@
 ; 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_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
+; CHECK-DISABLED-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
+; CHECK-DISABLED-NEXT:    [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], 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:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
@@ -1285,8 +1291,8 @@
 ; CHECK-DISABLED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
 ; CHECK-DISABLED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 ; CHECK-DISABLED:       user_code.entry:
-; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; CHECK-DISABLED-NEXT:    call void @__omp_outlined__8(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]]
+; CHECK-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; CHECK-DISABLED-NEXT:    call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; CHECK-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i1 false, i1 true)
 ; CHECK-DISABLED-NEXT:    ret void
 ; CHECK-DISABLED:       worker.exit:
@@ -1317,7 +1323,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-NEXT:    ret void
 ;
 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__8
@@ -1325,7 +1331,7 @@
 ; CHECK-DISABLED-NEXT:  entry:
 ; CHECK-DISABLED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 ; CHECK-DISABLED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
-; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR6]]
+; CHECK-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
 ; CHECK-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1369,19 +1375,21 @@
 ;.
 ; CHECK: attributes #[[ATTR0]] = { 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" }
 ; CHECK: attributes #[[ATTR1:[0-9]+]] = { convergent "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
-; CHECK: attributes #[[ATTR2:[0-9]+]] = { convergent "frame-pointer"="none" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
-; CHECK: attributes #[[ATTR3]] = { nounwind }
-; CHECK: attributes #[[ATTR4:[0-9]+]] = { convergent nounwind }
-; CHECK: attributes #[[ATTR5]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; CHECK: attributes #[[ATTR6]] = { convergent }
+; CHECK: attributes #[[ATTR2:[0-9]+]] = { alwaysinline }
+; CHECK: attributes #[[ATTR3:[0-9]+]] = { convergent "frame-pointer"="none" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
+; CHECK: attributes #[[ATTR4]] = { nounwind }
+; CHECK: attributes #[[ATTR5:[0-9]+]] = { convergent nounwind }
+; CHECK: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; CHECK: attributes #[[ATTR7]] = { convergent }
 ;.
 ; CHECK-DISABLED: attributes #[[ATTR0]] = { 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" }
 ; CHECK-DISABLED: attributes #[[ATTR1:[0-9]+]] = { convergent "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
-; CHECK-DISABLED: attributes #[[ATTR2:[0-9]+]] = { convergent "frame-pointer"="none" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
-; CHECK-DISABLED: attributes #[[ATTR3]] = { nounwind }
-; CHECK-DISABLED: attributes #[[ATTR4:[0-9]+]] = { convergent nounwind }
-; CHECK-DISABLED: attributes #[[ATTR5]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; CHECK-DISABLED: attributes #[[ATTR6]] = { convergent }
+; CHECK-DISABLED: attributes #[[ATTR2:[0-9]+]] = { alwaysinline }
+; CHECK-DISABLED: attributes #[[ATTR3:[0-9]+]] = { convergent "frame-pointer"="none" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ptx32,+sm_20" }
+; CHECK-DISABLED: attributes #[[ATTR4]] = { nounwind }
+; CHECK-DISABLED: attributes #[[ATTR5:[0-9]+]] = { convergent nounwind }
+; CHECK-DISABLED: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; CHECK-DISABLED: attributes #[[ATTR7]] = { convergent }
 ;.
 ; CHECK: [[META0:![0-9]+]] = !{i32 0, i32 20, i32 171231761, !"sequential_loop_to_stack_var", i32 20, i32 1}
 ; CHECK: [[META1:![0-9]+]] = !{i32 0, i32 20, i32 171231761, !"sequential_loop", i32 5, i32 0}