[OpenMP][FIX] Do not crash when kernels are debug wrapper functions

With debug information enabled (-g) Clang will wrap the actual target
region into a new function which is called from the "kernel". The problem
is that the "kernel" is now basically a wrapper without all the things
we expect. More importantly, if we end up asking for an AAKernelInfo
for the "target region function" we might try to turn it into SPMD mode.
That used to cause an assertion as that function doesn't have an
appropriately named `_exec_mode` global. While the global is going away
soon we still need to make sure to properly handle this case, e.g.,
perform optimizations reliably.

Differential Revision: https://reviews.llvm.org/D122043

diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index 4e05eb9..8e92362 100644 (file)
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -3168,12 +3168,6 @@ struct AAKernelInfoFunction : AAKernelInfo {
     auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
 
     Function *Fn = getAnchorScope();
-    if (!OMPInfoCache.Kernels.count(Fn))
-      return;
-
-    // Add itself to the reaching kernel and set IsKernelEntry.
-    ReachingKernelEntries.insert(Fn);
-    IsKernelEntry = true;
 
     OMPInformationCache::RuntimeFunctionInfo &InitRFI =
         OMPInfoCache.RFIs[OMPRTL___kmpc_target_init];
@@ -3207,10 +3201,12 @@ struct AAKernelInfoFunction : AAKernelInfo {
         Fn);
 
     // Ignore kernels without initializers such as global constructors.
-    if (!KernelInitCB || !KernelDeinitCB) {
-      indicateOptimisticFixpoint();
+    if (!KernelInitCB || !KernelDeinitCB)
       return;
-    }
+
+    // Add itself to the reaching kernel and set IsKernelEntry.
+    ReachingKernelEntries.insert(Fn);
+    IsKernelEntry = true;
 
     // For kernels we might need to initialize/finalize the IsSPMD state and
     // we need to register a simplification callback so that the Attributor
@@ -3376,8 +3372,17 @@ struct AAKernelInfoFunction : AAKernelInfo {
       return false;
     }
 
-    // Check if the kernel is already in SPMD mode, if so, return success.
+    // Get the actual kernel, could be the caller of the anchor scope if we have
+    // a debug wrapper.
     Function *Kernel = getAnchorScope();
+    if (Kernel->hasLocalLinkage()) {
+      assert(Kernel->hasOneUse() && "Unexpected use of debug kernel wrapper.");
+      auto *CB = cast<CallBase>(Kernel->user_back());
+      Kernel = CB->getCaller();
+    }
+    assert(OMPInfoCache.Kernels.count(Kernel) && "Expected kernel function!");
+
+    // Check if the kernel is already in SPMD mode, if so, return success.
     GlobalVariable *ExecMode = Kernel->getParent()->getGlobalVariable(
         (Kernel->getName() + "_exec_mode").str());
     assert(ExecMode && "Kernel without exec mode?");
@@ -4731,18 +4736,23 @@ void OpenMPOpt::registerFoldRuntimeCall(RuntimeFunction RF) {
 
 void OpenMPOpt::registerAAs(bool IsModulePass) {
   if (SCC.empty())
-
     return;
+
   if (IsModulePass) {
     // Ensure we create the AAKernelInfo AAs first and without triggering an
     // update. This will make sure we register all value simplification
     // callbacks before any other AA has the chance to create an AAValueSimplify
     // or similar.
-    for (Function *Kernel : OMPInfoCache.Kernels)
+    auto CreateKernelInfoCB = [&](Use &, Function &Kernel) {
       A.getOrCreateAAFor<AAKernelInfo>(
-          IRPosition::function(*Kernel), /* QueryingAA */ nullptr,
+          IRPosition::function(Kernel), /* QueryingAA */ nullptr,
           DepClassTy::NONE, /* ForceUpdate */ false,
           /* UpdateAfterInit */ false);
+      return false;
+    };
+    OMPInformationCache::RuntimeFunctionInfo &InitRFI =
+        OMPInfoCache.RFIs[OMPRTL___kmpc_target_init];
+    InitRFI.foreachUse(SCC, CreateKernelInfoCB);
 
     registerFoldRuntimeCall(OMPRTL___kmpc_is_generic_main_thread_id);
     registerFoldRuntimeCall(OMPRTL___kmpc_is_spmd_exec_mode);

diff --git a/llvm/test/Transforms/OpenMP/replace_globalization.ll b/llvm/test/Transforms/OpenMP/replace_globalization.ll
index 8d85a27..b48935a 100644 (file)
--- a/llvm/test/Transforms/OpenMP/replace_globalization.ll
+++ b/llvm/test/Transforms/OpenMP/replace_globalization.ll
@@ -18,9 +18,12 @@ target triple = "nvptx64"
 @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
+@foo_exec_mode = weak constant i8 1
+@bar_exec_mode = weak constant i8 1
+@baz_spmd_exec_mode = weak constant i8 2
 
 
-define dso_local void @foo() {
+define dso_local void @foo() "kernel" {
 entry:
   %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   %x = call align 4 i8* @__kmpc_alloc_shared(i64 4)
@@ -33,7 +36,7 @@ entry:
   ret void
 }
 
-define void @bar() {
+define void @bar() "kernel" {
   %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 1, i1 true, i1 true)
   call void @unknown_no_openmp()
   %cmp = icmp eq i32 %c, -1
@@ -61,7 +64,7 @@ exit:
   ret void
 }
 
-define void @baz_spmd() {
+define void @baz_spmd() "kernel" {
   %c = call i32 @__kmpc_target_init(%struct.ident_t* @1, i8 2, i1 true, i1 true)
   call void @unknown_no_openmp()
   %c0 = icmp eq i32 %c, -1
@@ -110,7 +113,7 @@ 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}
+!nvvm.annotations = !{!7, !8, !13}
 
 !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)
 !1 = !DIFile(filename: "replace_globalization.c", directory: "/tmp/replace_globalization.c")
@@ -130,58 +133,64 @@ declare void @unknown_no_openmp() "llvm.assume"="omp_no_openmp"
 ; CHECK: @[[S:[a-zA-Z0-9_$"\\.-]+]] = external local_unnamed_addr global i8*
 ; CHECK: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [113 x i8] c"
 ; CHECK: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([113 x i8], [113 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8
+; CHECK: @[[FOO_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
+; CHECK: @[[BAR_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1
+; CHECK: @[[BAZ_SPMD_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 2
 ; CHECK: @[[OFFSET:[a-zA-Z0-9_$"\\.-]+]] = global i32 undef
 ; CHECK: @[[STACK:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [1024 x i8] undef
 ; CHECK: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [16 x i8] undef, align 4
 ; CHECK: @[[Y_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] undef, align 4
 ;.
-; CHECK-LABEL: define {{[^@]+}}@foo() {
+; CHECK-LABEL: define {{[^@]+}}@foo
+; CHECK-SAME: () #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
-; CHECK-NEXT:    [[X:%.*]] = call align 4 i8* @__kmpc_alloc_shared(i64 4) #[[ATTR6:[0-9]+]]
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR5:[0-9]+]]
+; CHECK-NEXT:    [[X:%.*]] = call align 4 i8* @__kmpc_alloc_shared(i64 4) #[[ATTR7:[0-9]+]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR6:[0-9]+]]
 ; CHECK-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[X]] to i32*
 ; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[X_ON_STACK]] to i8*
-; CHECK-NEXT:    call void @use.internalized(i8* nofree [[TMP0]]) #[[ATTR7:[0-9]+]]
-; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[X]], i64 4) #[[ATTR8:[0-9]+]]
+; CHECK-NEXT:    call void @use.internalized(i8* nofree [[TMP0]]) #[[ATTR8:[0-9]+]]
+; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[X]], i64 4) #[[ATTR9:[0-9]+]]
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 ;
-; CHECK-LABEL: define {{[^@]+}}@bar() {
+; CHECK-LABEL: define {{[^@]+}}@bar
+; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR5]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR6]]
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[C]], -1
 ; CHECK-NEXT:    br i1 [[CMP]], label [[MASTER1:%.*]], label [[EXIT:%.*]]
 ; CHECK:       master1:
 ; CHECK-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([16 x i8], [16 x i8] addrspace(3)* @x_shared, i32 0, i32 0) to i8*) to [4 x i32]*
 ; CHECK-NEXT:    [[A0:%.*]] = bitcast [4 x i32]* [[X_ON_STACK]] to i8*
-; CHECK-NEXT:    call void @use.internalized(i8* nofree [[A0]]) #[[ATTR7]]
+; CHECK-NEXT:    call void @use.internalized(i8* nofree [[A0]]) #[[ATTR8]]
 ; CHECK-NEXT:    br label [[NEXT:%.*]]
 ; CHECK:       next:
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR5]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR6]]
 ; CHECK-NEXT:    br label [[MASTER2:%.*]]
 ; CHECK:       master2:
 ; CHECK-NEXT:    [[Y_ON_STACK:%.*]] = bitcast i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([4 x i8], [4 x i8] addrspace(3)* @y_shared, i32 0, i32 0) to i8*) to [4 x i32]*
 ; CHECK-NEXT:    [[B1:%.*]] = bitcast [4 x i32]* [[Y_ON_STACK]] to i8*
-; CHECK-NEXT:    call void @use.internalized(i8* nofree [[B1]]) #[[ATTR7]]
+; CHECK-NEXT:    call void @use.internalized(i8* nofree [[B1]]) #[[ATTR8]]
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; CHECK-NEXT:    ret void
 ;
 ;
-; CHECK-LABEL: define {{[^@]+}}@baz_spmd() {
+; CHECK-LABEL: define {{[^@]+}}@baz_spmd
+; CHECK-SAME: () #[[ATTR0]] {
 ; CHECK-NEXT:    [[C:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 2, i1 true, i1 true)
-; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR5]]
+; CHECK-NEXT:    call void @unknown_no_openmp() #[[ATTR6]]
 ; CHECK-NEXT:    [[C0:%.*]] = icmp eq i32 [[C]], -1
 ; CHECK-NEXT:    br i1 [[C0]], label [[MASTER3:%.*]], label [[EXIT:%.*]]
 ; CHECK:       master3:
-; CHECK-NEXT:    [[Z:%.*]] = call align 4 i8* @__kmpc_alloc_shared(i64 24) #[[ATTR6]], !dbg [[DBG9:![0-9]+]]
+; CHECK-NEXT:    [[Z:%.*]] = call align 4 i8* @__kmpc_alloc_shared(i64 24) #[[ATTR7]], !dbg [[DBG10:![0-9]+]]
 ; CHECK-NEXT:    [[Z_ON_STACK:%.*]] = bitcast i8* [[Z]] to [6 x i32]*
 ; CHECK-NEXT:    [[C1:%.*]] = bitcast [6 x i32]* [[Z_ON_STACK]] to i8*
-; CHECK-NEXT:    call void @use.internalized(i8* nofree [[C1]]) #[[ATTR7]]
-; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[Z]], i64 24) #[[ATTR8]]
+; CHECK-NEXT:    call void @use.internalized(i8* nofree [[C1]]) #[[ATTR8]]
+; CHECK-NEXT:    call void @__kmpc_free_shared(i8* [[Z]], i64 24) #[[ATTR9]]
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
@@ -190,7 +199,7 @@ declare void @unknown_no_openmp() "llvm.assume"="omp_no_openmp"
 ;
 ; CHECK: Function Attrs: nofree nounwind writeonly
 ; CHECK-LABEL: define {{[^@]+}}@use.internalized
-; CHECK-SAME: (i8* nofree [[X:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-SAME: (i8* nofree [[X:%.*]]) #[[ATTR1:[0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    store i8* [[X]], i8** @S, align 8
 ; CHECK-NEXT:    ret void
@@ -204,21 +213,22 @@ declare void @unknown_no_openmp() "llvm.assume"="omp_no_openmp"
 ;
 ;
 ; CHECK-LABEL: define {{[^@]+}}@__kmpc_alloc_shared
-; CHECK-SAME: (i64 [[TMP0:%.*]]) #[[ATTR1:[0-9]+]] {
+; CHECK-SAME: (i64 [[TMP0:%.*]]) #[[ATTR2:[0-9]+]] {
 ; CHECK-NEXT:    [[L:%.*]] = load i32, i32* @offset, align 4
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i8, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([1024 x i8], [1024 x i8] addrspace(3)* @stack, i32 0, i32 0) to i8*), i32 [[L]]
 ; CHECK-NEXT:    ret i8* [[GEP]]
 ;
 ;.
-; CHECK: attributes #[[ATTR0]] = { nofree nounwind writeonly }
-; CHECK: attributes #[[ATTR1]] = { nosync nounwind readonly allocsize(0) }
-; CHECK: attributes #[[ATTR2:[0-9]+]] = { nosync nounwind }
-; CHECK: attributes #[[ATTR3:[0-9]+]] = { nounwind readnone speculatable }
-; CHECK: attributes #[[ATTR4:[0-9]+]] = { nofree nosync nounwind readnone speculatable willreturn }
-; CHECK: attributes #[[ATTR5]] = { "llvm.assume"="omp_no_openmp" }
-; CHECK: attributes #[[ATTR6]] = { nounwind readonly }
-; CHECK: attributes #[[ATTR7]] = { nounwind writeonly }
-; CHECK: attributes #[[ATTR8]] = { nounwind }
+; CHECK: attributes #[[ATTR0]] = { "kernel" }
+; CHECK: attributes #[[ATTR1]] = { nofree nounwind writeonly }
+; CHECK: attributes #[[ATTR2]] = { nosync nounwind readonly allocsize(0) }
+; CHECK: attributes #[[ATTR3:[0-9]+]] = { nosync nounwind }
+; CHECK: attributes #[[ATTR4:[0-9]+]] = { nounwind readnone speculatable }
+; CHECK: attributes #[[ATTR5:[0-9]+]] = { nofree nosync nounwind readnone speculatable willreturn }
+; CHECK: attributes #[[ATTR6]] = { "llvm.assume"="omp_no_openmp" }
+; CHECK: attributes #[[ATTR7]] = { nounwind readonly }
+; CHECK: attributes #[[ATTR8]] = { nounwind writeonly }
+; CHECK: attributes #[[ATTR9]] = { nounwind }
 ;.
 ; CHECK: [[META0:![0-9]+]] = 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: [[META1:![0-9]+]] = !DIFile(filename: "replace_globalization.c", directory: "/tmp/replace_globalization.c")
@@ -229,7 +239,8 @@ declare void @unknown_no_openmp() "llvm.assume"="omp_no_openmp"
 ; CHECK: [[META6:![0-9]+]] = !{i32 7, !"openmp-device", i32 50}
 ; CHECK: [[META7:![0-9]+]] = !{void ()* @foo, !"kernel", i32 1}
 ; CHECK: [[META8:![0-9]+]] = !{void ()* @bar, !"kernel", i32 1}
-; CHECK: [[DBG9]] = !DILocation(line: 5, column: 14, scope: !10)
-; CHECK: [[META10:![0-9]+]] = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 1, type: !11, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
-; CHECK: [[META11:![0-9]+]] = !DISubroutineType(types: !2)
+; CHECK: [[META9:![0-9]+]] = !{void ()* @baz_spmd, !"kernel", i32 1}
+; CHECK: [[DBG10]] = !DILocation(line: 5, column: 14, scope: !11)
+; CHECK: [[META11:![0-9]+]] = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 1, type: !12, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+; CHECK: [[META12:![0-9]+]] = !DISubroutineType(types: !2)
 ;.

diff --git a/llvm/test/Transforms/OpenMP/spmdization.ll b/llvm/test/Transforms/OpenMP/spmdization.ll
index 760cb0f..f154efb 100644 (file)
--- a/llvm/test/Transforms/OpenMP/spmdization.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization.ll
@@ -169,6 +169,31 @@
 define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
 ; AMDGPU-SAME: () #[[ATTR0:[0-9]+]] {
+; AMDGPU-NEXT:    call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
+; AMDGPU-NEXT:    ret void
+;
+; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
+; NVPTX-SAME: () #[[ATTR0:[0-9]+]] {
+; NVPTX-NEXT:    call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
+; NVPTX-NEXT:    ret void
+;
+; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
+; AMDGPU-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
+; AMDGPU-DISABLED-NEXT:    call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
+; AMDGPU-DISABLED-NEXT:    ret void
+;
+; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
+; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
+; NVPTX-DISABLED-NEXT:    call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
+; NVPTX-DISABLED-NEXT:    ret void
+;
+  call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
+  ret void
+}
+
+define internal void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() {
+; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug
+; AMDGPU-SAME: () #[[ATTR1:[0-9]+]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
@@ -178,14 +203,14 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 {
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]]
-; AMDGPU-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; AMDGPU-NEXT:    br label [[COMMON_RET]]
 ;
-; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
-; NVPTX-SAME: () #[[ATTR0:[0-9]+]] {
+; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug
+; NVPTX-SAME: () #[[ATTR1:[0-9]+]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
@@ -195,14 +220,14 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 {
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3:[0-9]+]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]]
-; NVPTX-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; NVPTX-NEXT:    br label [[COMMON_RET]]
 ;
-; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
-; AMDGPU-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
+; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug
+; AMDGPU-DISABLED-SAME: () #[[ATTR1:[0-9]+]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -249,14 +274,14 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 {
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; 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:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
-; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5
-; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
+; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug
+; NVPTX-DISABLED-SAME: () #[[ATTR1:[0-9]+]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -302,9 +327,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 {
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; 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:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[0-9]+]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -339,7 +364,7 @@ define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU:       for.cond.cleanup:
-; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
+; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR7:[0-9]+]]
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       for.body:
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -358,7 +383,7 @@ define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX:       for.cond.cleanup:
-; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
+; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR7:[0-9]+]]
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       for.body:
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -377,7 +402,7 @@ define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU-DISABLED:       for.cond.cleanup:
-; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
+; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7:[0-9]+]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       for.body:
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -396,7 +421,7 @@ define internal void @__omp_outlined__(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX-DISABLED:       for.cond.cleanup:
-; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6:[0-9]+]]
+; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7:[0-9]+]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       for.body:
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -431,25 +456,25 @@ define internal void @__omp_outlined__1(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; AMDGPU-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8:[0-9]+]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; NVPTX-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8:[0-9]+]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; AMDGPU-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8:[0-9]+]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1
 ; NVPTX-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7:[0-9]+]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8:[0-9]+]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -460,43 +485,43 @@ entry:
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__1_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper
-; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1:[0-9]+]] {
+; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper
-; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1:[0-9]+]] {
+; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper
-; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1:[0-9]+]] {
+; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper
-; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1:[0-9]+]] {
+; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__1(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -523,9 +548,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l2
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; AMDGPU-NEXT:    br label [[COMMON_RET]]
 ;
@@ -540,9 +565,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l2
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; NVPTX-NEXT:    br label [[COMMON_RET]]
 ;
@@ -594,9 +619,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l2
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       user_code.entry:
-; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -647,9 +672,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l2
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       user_code.entry:
-; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -681,14 +706,14 @@ define internal void @__omp_outlined__2(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 4, addrspace(5)
 ; AMDGPU-NEXT:    [[MALLOC_CAST:%.*]] = addrspacecast i8 addrspace(5)* [[TMP0]] to i8*
 ; AMDGPU-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[MALLOC_CAST]] to i32*
-; AMDGPU-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
+; AMDGPU-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR7]]
 ; AMDGPU-NEXT:    br label [[FOR_COND:%.*]]
 ; AMDGPU:       for.cond:
 ; AMDGPU-NEXT:    [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ]
 ; AMDGPU-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU:       for.cond.cleanup:
-; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       for.body:
 ; AMDGPU-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -703,14 +728,14 @@ define internal void @__omp_outlined__2(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 4
 ; NVPTX-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[TMP0]] to i32*
-; NVPTX-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
+; NVPTX-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR7]]
 ; NVPTX-NEXT:    br label [[FOR_COND:%.*]]
 ; NVPTX:       for.cond:
 ; NVPTX-NEXT:    [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ]
 ; NVPTX-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX:       for.cond.cleanup:
-; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       for.body:
 ; NVPTX-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -726,14 +751,14 @@ define internal void @__omp_outlined__2(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 4, addrspace(5)
 ; AMDGPU-DISABLED-NEXT:    [[MALLOC_CAST:%.*]] = addrspacecast i8 addrspace(5)* [[TMP0]] to i8*
 ; AMDGPU-DISABLED-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[MALLOC_CAST]] to i32*
-; AMDGPU-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
+; AMDGPU-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR7]]
 ; AMDGPU-DISABLED-NEXT:    br label [[FOR_COND:%.*]]
 ; AMDGPU-DISABLED:       for.cond:
 ; AMDGPU-DISABLED-NEXT:    [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ]
 ; AMDGPU-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU-DISABLED:       for.cond.cleanup:
-; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       for.body:
 ; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -748,14 +773,14 @@ define internal void @__omp_outlined__2(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x i8*], align 8
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = alloca i8, i64 4, align 4
 ; NVPTX-DISABLED-NEXT:    [[X_ON_STACK:%.*]] = bitcast i8* [[TMP0]] to i32*
-; NVPTX-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR6]]
+; NVPTX-DISABLED-NEXT:    call void @use(i32* nocapture [[X_ON_STACK]]) #[[ATTR7]]
 ; NVPTX-DISABLED-NEXT:    br label [[FOR_COND:%.*]]
 ; NVPTX-DISABLED:       for.cond:
 ; NVPTX-DISABLED-NEXT:    [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ]
 ; NVPTX-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX-DISABLED:       for.cond.cleanup:
-; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       for.body:
 ; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]]
@@ -793,25 +818,25 @@ define internal void @__omp_outlined__3(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__3
 ; AMDGPU-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__3
 ; NVPTX-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3
 ; AMDGPU-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3
 ; NVPTX-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -822,43 +847,43 @@ entry:
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__3_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper
-; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper
-; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper
-; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper
-; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__3(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -886,9 +911,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; AMDGPU-NEXT:    br label [[COMMON_RET]]
 ;
@@ -903,9 +928,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; NVPTX-NEXT:    br label [[COMMON_RET]]
 ;
@@ -957,9 +982,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       user_code.entry:
-; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1010,9 +1035,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       user_code.entry:
-; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1047,7 +1072,7 @@ define internal void @__omp_outlined__4(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU:       for.cond.cleanup:
-; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       for.body:
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1068,7 +1093,7 @@ define internal void @__omp_outlined__4(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX:       for.cond.cleanup:
-; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       for.body:
 ; NVPTX-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1089,7 +1114,7 @@ define internal void @__omp_outlined__4(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU-DISABLED:       for.cond.cleanup:
-; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       for.body:
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1110,7 +1135,7 @@ define internal void @__omp_outlined__4(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX-DISABLED:       for.cond.cleanup:
-; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       for.body:
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1155,7 +1180,7 @@ define internal void @__omp_outlined__5(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; AMDGPU-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; AMDGPU-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__5
@@ -1164,7 +1189,7 @@ define internal void @__omp_outlined__5(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; NVPTX-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; NVPTX-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5
@@ -1173,7 +1198,7 @@ define internal void @__omp_outlined__5(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; AMDGPU-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; AMDGPU-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5
@@ -1182,7 +1207,7 @@ define internal void @__omp_outlined__5(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; NVPTX-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; NVPTX-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1196,7 +1221,7 @@ entry:
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper
-; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1205,11 +1230,11 @@ define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; AMDGPU-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; AMDGPU-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; AMDGPU-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper
-; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1218,11 +1243,11 @@ define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; NVPTX-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; NVPTX-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; NVPTX-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; NVPTX-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper
-; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1231,11 +1256,11 @@ define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-DISABLED-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; AMDGPU-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; AMDGPU-DISABLED-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper
-; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1244,7 +1269,7 @@ define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; NVPTX-DISABLED-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; NVPTX-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; NVPTX-DISABLED-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__5(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1274,9 +1299,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_g
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; AMDGPU-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1291,9 +1316,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_g
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 false)
 ; NVPTX-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1345,9 +1370,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_g
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       user_code.entry:
-; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1398,9 +1423,9 @@ define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_g
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       user_code.entry:
-; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1450,7 +1475,7 @@ define internal void @__omp_outlined__6(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU:       for.cond.cleanup:
-; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       for.body:
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1486,7 +1511,7 @@ define internal void @__omp_outlined__6(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX:       for.cond.cleanup:
-; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       for.body:
 ; NVPTX-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1509,7 +1534,7 @@ define internal void @__omp_outlined__6(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; AMDGPU-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; AMDGPU-DISABLED:       for.cond.cleanup:
-; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       for.body:
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1532,7 +1557,7 @@ define internal void @__omp_outlined__6(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[CMP:%.*]] = icmp slt i32 [[I_0]], 100
 ; NVPTX-DISABLED-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]]
 ; NVPTX-DISABLED:       for.cond.cleanup:
-; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       for.body:
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
@@ -1578,7 +1603,7 @@ define internal void @__omp_outlined__7(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; AMDGPU-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; AMDGPU-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__7
@@ -1587,7 +1612,7 @@ define internal void @__omp_outlined__7(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; NVPTX-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; NVPTX-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7
@@ -1596,7 +1621,7 @@ define internal void @__omp_outlined__7(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; AMDGPU-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; AMDGPU-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7
@@ -1605,7 +1630,7 @@ define internal void @__omp_outlined__7(i32* noalias %.global_tid., i32* noalias
 ; NVPTX-DISABLED-NEXT:    [[TMP0:%.*]] = load i32, i32* [[X]], align 4, !tbaa [[TBAA18]]
 ; NVPTX-DISABLED-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP0]], 1
 ; NVPTX-DISABLED-NEXT:    store i32 [[INC]], i32* [[X]], align 4, !tbaa [[TBAA18]]
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1619,7 +1644,7 @@ entry:
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper
-; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1628,11 +1653,11 @@ define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; AMDGPU-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; AMDGPU-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; AMDGPU-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper
-; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1641,11 +1666,11 @@ define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; NVPTX-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; NVPTX-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; NVPTX-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; NVPTX-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper
-; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1654,11 +1679,11 @@ define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-DISABLED-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; AMDGPU-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; AMDGPU-DISABLED-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper
-; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
@@ -1667,7 +1692,7 @@ define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; NVPTX-DISABLED-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[GLOBAL_ARGS]], align 8
 ; NVPTX-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
 ; NVPTX-DISABLED-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8, !tbaa [[TBAA26]]
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__7(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]], i32* [[TMP4]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1728,8 +1753,8 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65() #0
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; AMDGPU-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; AMDGPU-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1774,8 +1799,8 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65() #0
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; NVPTX-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; NVPTX-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1821,8 +1846,8 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65() #0
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; 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__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; AMDGPU-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1867,8 +1892,8 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65() #0
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; 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__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
 ; NVPTX-DISABLED-NEXT:    br label [[COMMON_RET]]
 ;
@@ -1896,25 +1921,25 @@ define internal void @__omp_outlined__8(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__8
 ; AMDGPU-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__8
 ; NVPTX-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__8
 ; AMDGPU-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__8
 ; NVPTX-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -1971,10 +1996,10 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_l74() #0 {
 ; AMDGPU:       common.ret:
 ; AMDGPU-NEXT:    ret void
 ; AMDGPU:       user_code.entry:
-; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; AMDGPU-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; AMDGPU-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR4]]
 ; AMDGPU-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to %struct.kmp_task_t_with_privates*
-; AMDGPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR3]]
+; AMDGPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; AMDGPU-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__9 to i8*), i8* @__omp_outlined__9_wrapper.ID, i8** [[TMP5]], i64 0)
 ; AMDGPU-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
@@ -2026,10 +2051,10 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_l74() #0 {
 ; NVPTX:       common.ret:
 ; NVPTX-NEXT:    ret void
 ; NVPTX:       user_code.entry:
-; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; NVPTX-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; NVPTX-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR4]]
 ; NVPTX-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to %struct.kmp_task_t_with_privates*
-; NVPTX-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR3]]
+; NVPTX-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR4]]
 ; NVPTX-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; NVPTX-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__9 to i8*), i8* @__omp_outlined__9_wrapper.ID, i8** [[TMP5]], i64 0)
 ; NVPTX-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
@@ -2082,10 +2107,10 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_l74() #0 {
 ; AMDGPU-DISABLED:       common.ret:
 ; AMDGPU-DISABLED-NEXT:    ret void
 ; AMDGPU-DISABLED:       user_code.entry:
-; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; AMDGPU-DISABLED-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; AMDGPU-DISABLED-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to %struct.kmp_task_t_with_privates*
-; AMDGPU-DISABLED-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; AMDGPU-DISABLED-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__9 to i8*), i8* @__omp_outlined__9_wrapper.ID, i8** [[TMP5]], i64 0)
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
@@ -2137,10 +2162,10 @@ define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_l74() #0 {
 ; NVPTX-DISABLED:       common.ret:
 ; NVPTX-DISABLED-NEXT:    ret void
 ; NVPTX-DISABLED:       user_code.entry:
-; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR3]]
-; NVPTX-DISABLED-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
+; NVPTX-DISABLED-NEXT:    [[TMP2:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 1, i64 40, i64 0, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @"_omp_task_entry$" to i32 (i32, i8*)*)) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to %struct.kmp_task_t_with_privates*
-; NVPTX-DISABLED-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i8* [[TMP2]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    [[TMP5:%.*]] = bitcast [0 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
 ; NVPTX-DISABLED-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__9 to i8*), i8* @__omp_outlined__9_wrapper.ID, i8** [[TMP5]], i64 0)
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
@@ -2169,27 +2194,27 @@ user_code.entry:                                  ; preds = %entry
 ; Function Attrs: alwaysinline convergent nounwind
 define internal void @.omp_outlined.(i32 %.global_tid., i32* noalias %.part_id., i8* noalias %.privates., void (i8*, ...)* noalias %.copy_fn., i8* %.task_t., %struct.anon* noalias %__context) #9 {
 ; AMDGPU-LABEL: define {{[^@]+}}@.omp_outlined.
-; AMDGPU-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2:[0-9]+]] {
+; AMDGPU-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR3:[0-9]+]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@.omp_outlined.
-; NVPTX-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2:[0-9]+]] {
+; NVPTX-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR3:[0-9]+]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@.omp_outlined.
-; AMDGPU-DISABLED-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2:[0-9]+]] {
+; AMDGPU-DISABLED-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR3:[0-9]+]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
-; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; AMDGPU-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@.omp_outlined.
-; NVPTX-DISABLED-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2:[0-9]+]] {
+; NVPTX-DISABLED-SAME: (i32 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTPART_ID_:%.*]], i8* noalias [[DOTPRIVATES_:%.*]], void (i8*, ...)* noalias [[DOTCOPY_FN_:%.*]], i8* [[DOTTASK_T_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR3:[0-9]+]] {
 ; NVPTX-DISABLED-NEXT:  entry:
-; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR6]]
+; NVPTX-DISABLED-NEXT:    call void @spmd_amenable() #[[ATTR7]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -2255,25 +2280,25 @@ define internal void @__omp_outlined__9(i32* noalias %.global_tid., i32* noalias
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; AMDGPU-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-NEXT:  entry:
-; AMDGPU-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; NVPTX-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-NEXT:  entry:
-; NVPTX-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; AMDGPU-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
-; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; AMDGPU-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9
 ; NVPTX-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
 ; NVPTX-DISABLED-NEXT:  entry:
-; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR7]]
+; NVPTX-DISABLED-NEXT:    call void @unknown() #[[ATTR8]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -2284,43 +2309,43 @@ entry:
 ; Function Attrs: convergent norecurse nounwind
 define internal void @__omp_outlined__9_wrapper(i16 zeroext %0, i32 %1) #3 {
 ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__9_wrapper
-; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-NEXT:  entry:
 ; AMDGPU-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-NEXT:    ret void
 ;
 ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__9_wrapper
-; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-NEXT:  entry:
 ; NVPTX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-NEXT:    ret void
 ;
 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9_wrapper
-; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; AMDGPU-DISABLED-NEXT:  entry:
 ; AMDGPU-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; AMDGPU-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; AMDGPU-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; AMDGPU-DISABLED-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; AMDGPU-DISABLED-NEXT:    ret void
 ;
 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__9_wrapper
-; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR1]] {
+; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] {
 ; NVPTX-DISABLED-NEXT:  entry:
 ; NVPTX-DISABLED-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 ; NVPTX-DISABLED-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca i8**, align 8
 ; NVPTX-DISABLED-NEXT:    call void @__kmpc_get_shared_variables(i8*** [[GLOBAL_ARGS]])
-; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR3]]
+; NVPTX-DISABLED-NEXT:    call void @__omp_outlined__9(i32* [[DOTADDR1]], i32* [[DOTZERO_ADDR]]) #[[ATTR4]]
 ; NVPTX-DISABLED-NEXT:    ret void
 ;
 entry:
@@ -2389,52 +2414,56 @@ attributes #11 = { convergent }
 !32 = !{!"kmp_task_t", !27, i64 0, !27, i64 8, !19, i64 16, !20, i64 24, !20, i64 32}
 ;.
 ; AMDGPU: attributes #[[ATTR0]] = { alwaysinline convergent norecurse nounwind }
-; AMDGPU: attributes #[[ATTR1]] = { convergent norecurse nounwind }
-; AMDGPU: attributes #[[ATTR2]] = { alwaysinline convergent nounwind }
-; AMDGPU: attributes #[[ATTR3]] = { nounwind }
-; AMDGPU: attributes #[[ATTR4:[0-9]+]] = { nosync nounwind }
-; AMDGPU: attributes #[[ATTR5:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
-; AMDGPU: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; AMDGPU: attributes #[[ATTR7]] = { convergent }
-; AMDGPU: attributes #[[ATTR8:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
-; AMDGPU: attributes #[[ATTR9:[0-9]+]] = { alwaysinline }
-; AMDGPU: attributes #[[ATTR10:[0-9]+]] = { convergent nounwind }
+; AMDGPU: attributes #[[ATTR1]] = { norecurse }
+; AMDGPU: attributes #[[ATTR2]] = { convergent norecurse nounwind }
+; AMDGPU: attributes #[[ATTR3]] = { alwaysinline convergent nounwind }
+; AMDGPU: attributes #[[ATTR4]] = { nounwind }
+; AMDGPU: attributes #[[ATTR5:[0-9]+]] = { nosync nounwind }
+; AMDGPU: attributes #[[ATTR6:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
+; AMDGPU: attributes #[[ATTR7]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; AMDGPU: attributes #[[ATTR8]] = { convergent }
+; AMDGPU: attributes #[[ATTR9:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
+; AMDGPU: attributes #[[ATTR10:[0-9]+]] = { alwaysinline }
+; AMDGPU: attributes #[[ATTR11:[0-9]+]] = { convergent nounwind }
 ;.
 ; NVPTX: attributes #[[ATTR0]] = { alwaysinline convergent norecurse nounwind }
-; NVPTX: attributes #[[ATTR1]] = { convergent norecurse nounwind }
-; NVPTX: attributes #[[ATTR2]] = { alwaysinline convergent nounwind }
-; NVPTX: attributes #[[ATTR3]] = { nounwind }
-; NVPTX: attributes #[[ATTR4:[0-9]+]] = { nosync nounwind }
-; NVPTX: attributes #[[ATTR5:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
-; NVPTX: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; NVPTX: attributes #[[ATTR7]] = { convergent }
-; NVPTX: attributes #[[ATTR8:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
-; NVPTX: attributes #[[ATTR9:[0-9]+]] = { alwaysinline }
-; NVPTX: attributes #[[ATTR10:[0-9]+]] = { convergent nounwind }
+; NVPTX: attributes #[[ATTR1]] = { norecurse }
+; NVPTX: attributes #[[ATTR2]] = { convergent norecurse nounwind }
+; NVPTX: attributes #[[ATTR3]] = { alwaysinline convergent nounwind }
+; NVPTX: attributes #[[ATTR4]] = { nounwind }
+; NVPTX: attributes #[[ATTR5:[0-9]+]] = { nosync nounwind }
+; NVPTX: attributes #[[ATTR6:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
+; NVPTX: attributes #[[ATTR7]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; NVPTX: attributes #[[ATTR8]] = { convergent }
+; NVPTX: attributes #[[ATTR9:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
+; NVPTX: attributes #[[ATTR10:[0-9]+]] = { alwaysinline }
+; NVPTX: attributes #[[ATTR11:[0-9]+]] = { convergent nounwind }
 ;.
 ; AMDGPU-DISABLED: attributes #[[ATTR0]] = { alwaysinline convergent norecurse nounwind }
-; AMDGPU-DISABLED: attributes #[[ATTR1]] = { convergent norecurse nounwind }
-; AMDGPU-DISABLED: attributes #[[ATTR2]] = { alwaysinline convergent nounwind }
-; AMDGPU-DISABLED: attributes #[[ATTR3]] = { nounwind }
-; AMDGPU-DISABLED: attributes #[[ATTR4:[0-9]+]] = { nosync nounwind }
-; AMDGPU-DISABLED: attributes #[[ATTR5:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
-; AMDGPU-DISABLED: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; AMDGPU-DISABLED: attributes #[[ATTR7]] = { convergent }
-; AMDGPU-DISABLED: attributes #[[ATTR8:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
-; AMDGPU-DISABLED: attributes #[[ATTR9:[0-9]+]] = { alwaysinline }
-; AMDGPU-DISABLED: attributes #[[ATTR10:[0-9]+]] = { convergent nounwind }
+; AMDGPU-DISABLED: attributes #[[ATTR1]] = { norecurse }
+; AMDGPU-DISABLED: attributes #[[ATTR2]] = { convergent norecurse nounwind }
+; AMDGPU-DISABLED: attributes #[[ATTR3]] = { alwaysinline convergent nounwind }
+; AMDGPU-DISABLED: attributes #[[ATTR4]] = { nounwind }
+; AMDGPU-DISABLED: attributes #[[ATTR5:[0-9]+]] = { nosync nounwind }
+; AMDGPU-DISABLED: attributes #[[ATTR6:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
+; AMDGPU-DISABLED: attributes #[[ATTR7]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; AMDGPU-DISABLED: attributes #[[ATTR8]] = { convergent }
+; AMDGPU-DISABLED: attributes #[[ATTR9:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
+; AMDGPU-DISABLED: attributes #[[ATTR10:[0-9]+]] = { alwaysinline }
+; AMDGPU-DISABLED: attributes #[[ATTR11:[0-9]+]] = { convergent nounwind }
 ;.
 ; NVPTX-DISABLED: attributes #[[ATTR0]] = { alwaysinline convergent norecurse nounwind }
-; NVPTX-DISABLED: attributes #[[ATTR1]] = { convergent norecurse nounwind }
-; NVPTX-DISABLED: attributes #[[ATTR2]] = { alwaysinline convergent nounwind }
-; NVPTX-DISABLED: attributes #[[ATTR3]] = { nounwind }
-; NVPTX-DISABLED: attributes #[[ATTR4:[0-9]+]] = { nosync nounwind }
-; NVPTX-DISABLED: attributes #[[ATTR5:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
-; NVPTX-DISABLED: attributes #[[ATTR6]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
-; NVPTX-DISABLED: attributes #[[ATTR7]] = { convergent }
-; NVPTX-DISABLED: attributes #[[ATTR8:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
-; NVPTX-DISABLED: attributes #[[ATTR9:[0-9]+]] = { alwaysinline }
-; NVPTX-DISABLED: attributes #[[ATTR10:[0-9]+]] = { convergent nounwind }
+; NVPTX-DISABLED: attributes #[[ATTR1]] = { norecurse }
+; NVPTX-DISABLED: attributes #[[ATTR2]] = { convergent norecurse nounwind }
+; NVPTX-DISABLED: attributes #[[ATTR3]] = { alwaysinline convergent nounwind }
+; NVPTX-DISABLED: attributes #[[ATTR4]] = { nounwind }
+; NVPTX-DISABLED: attributes #[[ATTR5:[0-9]+]] = { nosync nounwind }
+; NVPTX-DISABLED: attributes #[[ATTR6:[0-9]+]] = { nofree nosync nounwind allocsize(0) }
+; NVPTX-DISABLED: attributes #[[ATTR7]] = { convergent "llvm.assume"="ompx_spmd_amenable" }
+; NVPTX-DISABLED: attributes #[[ATTR8]] = { convergent }
+; NVPTX-DISABLED: attributes #[[ATTR9:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
+; NVPTX-DISABLED: attributes #[[ATTR10:[0-9]+]] = { alwaysinline }
+; NVPTX-DISABLED: attributes #[[ATTR11:[0-9]+]] = { convergent nounwind }
 ;.
 ; AMDGPU: [[META0:![0-9]+]] = !{i32 0, i32 64770, i32 541341486, !"do_not_spmdize_task", i32 74, i32 5}
 ; AMDGPU: [[META1:![0-9]+]] = !{i32 0, i32 64770, i32 541341486, !"sequential_loop_to_stack_var", i32 20, i32 1}