[Openmp-commits] [llvm] [openmp] [OpenMPOpt] Ask the runtime how many of a block's threads can be workers (PR #217080)
Larry Meadows via Openmp-commits
openmp-commits at lists.llvm.org
Wed Aug 19 03:10:30 PDT 2026
https://github.com/lfmeadow updated https://github.com/llvm/llvm-project/pull/217080
>From 5d6a409a60443e0bf60eb2bea6cc3de391addcc6 Mon Sep 17 00:00:00 2001
From: Larry Meadows <Lawrence.Meadows at amd.com>
Date: Tue, 18 Aug 2026 11:56:16 -0500
Subject: [PATCH 1/2] [OpenMPOpt] Ask the runtime how many of a block's threads
can be workers
The custom state machine gates a thread on InitCB < BlockHwSize - WarpSize,
reconstructing the number of worker threads from the block size on the
assumption that the main thread occupies a whole warp above them. The DeviceRTL
already computes that number, in mapping::getMaxTeamThreads(), and its own
generic state machine gates on it in shouldEnterStateMachine(). Export it as
__kmpc_get_max_team_threads() and call that instead, so the compiler's state
machine and the runtime's agree by construction rather than by arithmetic that
has to be kept in step with the launch geometry.
This is NFC here: getMaxTeamThreads() in generic mode is BlockSize - WarpSize,
the same three instructions folded into one call. It is not NFC for a toolchain
whose launch geometry differs. In ROCm, CGOpenMPRuntimeGPU starts a single extra
thread rather than a warp -- "Only one additional thread is started, not an
entire warp" -- so thread_limit(1024) on a 64-lane target launches 961 threads
and the runtime reports 960 workers, while the state machine's own arithmetic
says 961 - 64 = 897. The threads in between are in neither group: the state
machine returns immediately for them, and the parallel region still hands them
iterations. Asking the runtime removes the assumption rather than adding a
second one.
A device runtime that predates the new entry point is handled by the existing
runtimeFnsAvailable() check, which skips the rewrite rather than emitting a call
to a function that is not there.
The two calls this replaces, __kmpc_get_hardware_num_threads_in_block and
__kmpc_get_warp_size, keep their other uses in the pass, including the constant
folding of the former.
Co-authored-by: Cursor <cursoragent at cursor.com>
---
.../include/llvm/Frontend/OpenMP/OMPKinds.def | 2 +
llvm/lib/Transforms/IPO/OpenMPOpt.cpp | 52 ++++++--------
.../OpenMP/custom_state_machines.ll | 72 +++++++------------
llvm/test/Transforms/OpenMP/spmdization.ll | 64 +++++------------
.../Transforms/OpenMP/spmdization_guarding.ll | 6 +-
.../Transforms/OpenMP/spmdization_indirect.ll | 16 ++---
...zation_no_guarding_two_reaching_kernels.ll | 18 ++---
openmp/device/include/Interface.h | 3 +
openmp/device/src/Mapping.cpp | 4 ++
9 files changed, 83 insertions(+), 154 deletions(-)
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
index ea1cef97900f5..730027f0fb477 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
@@ -228,6 +228,7 @@ __OMP_RTL(__kmpc_omp_reg_task_with_affinity, false, Int32, IdentPtr, Int32,
__OMP_RTL(__kmpc_get_hardware_num_blocks, false, Int32, )
__OMP_RTL(__kmpc_get_hardware_num_threads_in_block, false, Int32, )
__OMP_RTL(__kmpc_get_warp_size, false, Int32, )
+__OMP_RTL(__kmpc_get_max_team_threads, false, Int32, )
__OMP_RTL(omp_get_thread_num, false, Int32, )
__OMP_RTL(omp_get_num_threads, false, Int32, )
@@ -721,6 +722,7 @@ __OMP_RTL_ATTRS(__kmpc_omp_reg_task_with_affinity, DefaultAttrs, SExt,
__OMP_RTL_ATTRS(__kmpc_get_hardware_num_blocks, GetterAttrs, ZExt, ParamAttrs())
__OMP_RTL_ATTRS(__kmpc_get_hardware_num_threads_in_block, GetterAttrs, ZExt, ParamAttrs())
__OMP_RTL_ATTRS(__kmpc_get_warp_size, GetterAttrs, ZExt, ParamAttrs())
+__OMP_RTL_ATTRS(__kmpc_get_max_team_threads, GetterAttrs, ZExt, ParamAttrs())
__OMP_RTL_ATTRS(omp_get_thread_num, GetterAttrs, SExt, ParamAttrs())
__OMP_RTL_ATTRS(omp_get_num_threads, GetterAttrs, SExt, ParamAttrs())
diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index 22a5efb402504..0f8c6b991e763 100644
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -3904,8 +3904,7 @@ struct AAKernelInfoFunction : AAKernelInfo {
Attributor::VirtualUseCallbackTy CustomStateMachineUseCB =
[&](Attributor &A, const AbstractAttribute *QueryingAA) {
// Whenever we create a custom state machine we will insert calls to
- // __kmpc_get_hardware_num_threads_in_block,
- // __kmpc_get_warp_size,
+ // __kmpc_get_max_team_threads,
// __kmpc_barrier_simple_generic,
// __kmpc_kernel_parallel, and
// __kmpc_kernel_end_parallel.
@@ -3920,9 +3919,8 @@ struct AAKernelInfoFunction : AAKernelInfo {
// Not needed if we are pre-runtime merge.
if (!KernelInitCB->getCalledFunction()->isDeclaration()) {
- RegisterVirtualUse(OMPRTL___kmpc_get_hardware_num_threads_in_block,
+ RegisterVirtualUse(OMPRTL___kmpc_get_max_team_threads,
CustomStateMachineUseCB);
- RegisterVirtualUse(OMPRTL___kmpc_get_warp_size, CustomStateMachineUseCB);
RegisterVirtualUse(OMPRTL___kmpc_barrier_simple_generic,
CustomStateMachineUseCB);
RegisterVirtualUse(OMPRTL___kmpc_kernel_parallel,
@@ -4381,10 +4379,10 @@ struct AAKernelInfoFunction : AAKernelInfo {
return false;
auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
- if (!OMPInfoCache.runtimeFnsAvailable(
- {OMPRTL___kmpc_get_hardware_num_threads_in_block,
- OMPRTL___kmpc_get_warp_size, OMPRTL___kmpc_barrier_simple_generic,
- OMPRTL___kmpc_kernel_parallel, OMPRTL___kmpc_kernel_end_parallel}))
+ if (!OMPInfoCache.runtimeFnsAvailable({OMPRTL___kmpc_get_max_team_threads,
+ OMPRTL___kmpc_barrier_simple_generic,
+ OMPRTL___kmpc_kernel_parallel,
+ OMPRTL___kmpc_kernel_end_parallel}))
return false;
ConstantStruct *ExistingKernelEnvC =
@@ -4463,13 +4461,10 @@ struct AAKernelInfoFunction : AAKernelInfo {
// Create all the blocks:
//
// InitCB = __kmpc_target_init(...)
- // BlockHwSize =
- // __kmpc_get_hardware_num_threads_in_block();
- // WarpSize = __kmpc_get_warp_size();
- // BlockSize = BlockHwSize - WarpSize;
+ // MaxTeamThreads = __kmpc_get_max_team_threads();
// IsWorkerCheckBB: bool IsWorker = InitCB != -1;
// if (IsWorker) {
- // if (InitCB >= BlockSize) return;
+ // if (InitCB >= MaxTeamThreads) return;
// SMBeginBB: __kmpc_barrier_simple_generic(...);
// void *WorkFn;
// bool Active = __kmpc_kernel_parallel(&WorkFn);
@@ -4534,26 +4529,23 @@ struct AAKernelInfoFunction : AAKernelInfo {
IsWorker->setDebugLoc(DLoc);
CondBrInst::Create(IsWorker, IsWorkerCheckBB, UserCodeEntryBB, InitBB);
+ // How many of the block's threads can be worker threads is a property of
+ // the launch geometry, which the runtime knows and the block size alone
+ // does not determine: a target may host the main thread in a whole warp
+ // above the workers, or in a single thread above them. Ask the runtime
+ // rather than subtracting a warp here, or the workers in between are left
+ // in neither group, waiting for no parallel region while it is handed
+ // iterations.
Module &M = *Kernel->getParent();
- FunctionCallee BlockHwSizeFn =
+ FunctionCallee MaxTeamThreadsFn =
OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
- M, OMPRTL___kmpc_get_hardware_num_threads_in_block);
- FunctionCallee WarpSizeFn =
- OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
- M, OMPRTL___kmpc_get_warp_size);
- CallInst *BlockHwSize =
- CallInst::Create(BlockHwSizeFn, "block.hw_size", IsWorkerCheckBB);
- OMPInfoCache.setCallingConvention(BlockHwSizeFn, BlockHwSize);
- BlockHwSize->setDebugLoc(DLoc);
- CallInst *WarpSize =
- CallInst::Create(WarpSizeFn, "warp.size", IsWorkerCheckBB);
- OMPInfoCache.setCallingConvention(WarpSizeFn, WarpSize);
- WarpSize->setDebugLoc(DLoc);
- Instruction *BlockSize = BinaryOperator::CreateSub(
- BlockHwSize, WarpSize, "block.size", IsWorkerCheckBB);
- BlockSize->setDebugLoc(DLoc);
+ M, OMPRTL___kmpc_get_max_team_threads);
+ CallInst *MaxTeamThreads =
+ CallInst::Create(MaxTeamThreadsFn, "max_team_threads", IsWorkerCheckBB);
+ OMPInfoCache.setCallingConvention(MaxTeamThreadsFn, MaxTeamThreads);
+ MaxTeamThreads->setDebugLoc(DLoc);
Instruction *IsMainOrWorker = ICmpInst::Create(
- ICmpInst::ICmp, llvm::CmpInst::ICMP_SLT, KernelInitCB, BlockSize,
+ ICmpInst::ICmp, llvm::CmpInst::ICMP_SLT, KernelInitCB, MaxTeamThreads,
"thread.is_main_or_worker", IsWorkerCheckBB);
IsMainOrWorker->setDebugLoc(DLoc);
CondBrInst::Create(IsMainOrWorker, StateMachineBeginBB,
diff --git a/llvm/test/Transforms/OpenMP/custom_state_machines.ll b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
index a03a4b7f3fa67..ac35d7a4a9765 100644
--- a/llvm/test/Transforms/OpenMP/custom_state_machines.ll
+++ b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
@@ -1152,10 +1152,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -1273,10 +1271,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -1420,10 +1416,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -1542,10 +1536,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -1662,10 +1654,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -1854,10 +1844,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: is_worker_check:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; AMDGPU-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2078,10 +2066,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2198,10 +2184,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2344,10 +2328,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2465,10 +2447,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2584,10 +2564,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -2775,10 +2753,8 @@ attributes #9 = { convergent nounwind willreturn memory(read) }
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: is_worker_check:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; NVPTX-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
diff --git a/llvm/test/Transforms/OpenMP/spmdization.ll b/llvm/test/Transforms/OpenMP/spmdization.ll
index 7d53a808ef4d3..88a4d53ae9c82 100644
--- a/llvm/test/Transforms/OpenMP/spmdization.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization.ll
@@ -271,9 +271,7 @@ define internal void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -343,9 +341,7 @@ define internal void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -753,9 +749,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -825,9 +819,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -1253,9 +1245,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -1325,9 +1315,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -1779,9 +1767,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -1851,9 +1837,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_sequential_loop_to_s
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2300,9 +2284,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_targe
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU: [[IS_WORKER_CHECK]]:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2348,9 +2330,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_targe
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX: [[IS_WORKER_CHECK]]:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2395,9 +2375,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_targe
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2461,9 +2439,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_targe
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2589,9 +2565,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU: [[IS_WORKER_CHECK]]:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2643,9 +2617,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX: [[IS_WORKER_CHECK]]:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2696,9 +2668,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU-DISABLED1: [[IS_WORKER_CHECK]]:
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -2768,9 +2738,7 @@ define weak ptx_kernel void @__omp_offloading_fd02_2044372e_do_not_spmdize_task_
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX-DISABLED1: [[IS_WORKER_CHECK]]:
-; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX-DISABLED1: [[WORKER_STATE_MACHINE_BEGIN]]:
diff --git a/llvm/test/Transforms/OpenMP/spmdization_guarding.ll b/llvm/test/Transforms/OpenMP/spmdization_guarding.ll
index 2fb1f878cce7b..4dae6889beb9d 100644
--- a/llvm/test/Transforms/OpenMP/spmdization_guarding.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization_guarding.ll
@@ -201,10 +201,8 @@ define weak ptx_kernel void @__omp_offloading_2a_fbfa7a_sequential_loop_l6(ptr %
; CHECK-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; CHECK-DISABLED: is_worker_check:
-; CHECK-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; CHECK-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; CHECK-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; CHECK-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; CHECK-DISABLED-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; CHECK-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; CHECK-DISABLED: worker_state_machine.begin:
; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
diff --git a/llvm/test/Transforms/OpenMP/spmdization_indirect.ll b/llvm/test/Transforms/OpenMP/spmdization_indirect.ll
index 1de719e1b85a8..71847b72ae24f 100644
--- a/llvm/test/Transforms/OpenMP/spmdization_indirect.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization_indirect.ll
@@ -395,9 +395,7 @@ define weak ptx_kernel void @spmd_and_non_spmd_callee(i1 %c) #0 {
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU: [[IS_WORKER_CHECK]]:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -458,9 +456,7 @@ define weak ptx_kernel void @spmd_and_non_spmd_callee(i1 %c) #0 {
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX: [[IS_WORKER_CHECK]]:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -751,9 +747,7 @@ define weak ptx_kernel void @spmd_and_non_spmd_callees_metadata(ptr %fp) #0 {
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; AMDGPU: [[IS_WORKER_CHECK]]:
-; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; AMDGPU: [[WORKER_STATE_MACHINE_BEGIN]]:
@@ -813,9 +807,7 @@ define weak ptx_kernel void @spmd_and_non_spmd_callees_metadata(ptr %fp) #0 {
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label %[[IS_WORKER_CHECK:.*]], label %[[THREAD_USER_CODE_CHECK:.*]]
; NVPTX: [[IS_WORKER_CHECK]]:
-; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
+; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = call i32 @__kmpc_get_max_team_threads()
; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label %[[WORKER_STATE_MACHINE_BEGIN:.*]], label %[[WORKER_STATE_MACHINE_FINISHED:.*]]
; NVPTX: [[WORKER_STATE_MACHINE_BEGIN]]:
diff --git a/llvm/test/Transforms/OpenMP/spmdization_no_guarding_two_reaching_kernels.ll b/llvm/test/Transforms/OpenMP/spmdization_no_guarding_two_reaching_kernels.ll
index 4f64839deaf53..dc08df96dfd64 100644
--- a/llvm/test/Transforms/OpenMP/spmdization_no_guarding_two_reaching_kernels.ll
+++ b/llvm/test/Transforms/OpenMP/spmdization_no_guarding_two_reaching_kernels.ll
@@ -78,10 +78,8 @@ define weak ptx_kernel void @__omp_offloading_2b_10393b5_spmd_l12(ptr %dyn) #0 {
; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; CHECK-DISABLE-SPMDIZATION-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; CHECK-DISABLE-SPMDIZATION: is_worker_check:
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; CHECK-DISABLE-SPMDIZATION-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; CHECK-DISABLE-SPMDIZATION-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; CHECK-DISABLE-SPMDIZATION: worker_state_machine.begin:
; CHECK-DISABLE-SPMDIZATION-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -155,10 +153,8 @@ define weak ptx_kernel void @__omp_offloading_2b_10393b5_generic_l20(ptr %dyn) #
; CHECK-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; CHECK-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; CHECK: is_worker_check:
-; CHECK-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; CHECK-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; CHECK-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; CHECK-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; CHECK-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; CHECK-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; CHECK-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; CHECK: worker_state_machine.begin:
; CHECK-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
@@ -197,10 +193,8 @@ define weak ptx_kernel void @__omp_offloading_2b_10393b5_generic_l20(ptr %dyn) #
; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; CHECK-DISABLE-SPMDIZATION-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; CHECK-DISABLE-SPMDIZATION: is_worker_check:
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size()
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]]
-; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]]
+; CHECK-DISABLE-SPMDIZATION-NEXT: [[MAX_TEAM_THREADS:%.*]] = call i32 @__kmpc_get_max_team_threads()
+; CHECK-DISABLE-SPMDIZATION-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[MAX_TEAM_THREADS]]
; CHECK-DISABLE-SPMDIZATION-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]]
; CHECK-DISABLE-SPMDIZATION: worker_state_machine.begin:
; CHECK-DISABLE-SPMDIZATION-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
diff --git a/openmp/device/include/Interface.h b/openmp/device/include/Interface.h
index 601694871597b..07ad56e8dbef0 100644
--- a/openmp/device/include/Interface.h
+++ b/openmp/device/include/Interface.h
@@ -217,6 +217,9 @@ uint32_t __kmpc_get_hardware_num_threads_in_block();
/// External interface to get the warp size.
uint32_t __kmpc_get_warp_size();
+/// External interface to get the maximum number of threads in a team.
+uint32_t __kmpc_get_max_team_threads();
+
/// Kernel
///
///{
diff --git a/openmp/device/src/Mapping.cpp b/openmp/device/src/Mapping.cpp
index 1096e5f496590..f65ccb5203056 100644
--- a/openmp/device/src/Mapping.cpp
+++ b/openmp/device/src/Mapping.cpp
@@ -175,6 +175,10 @@ extern "C" {
[[gnu::noinline]] uint32_t __kmpc_get_warp_size() {
return mapping::getWarpSize();
}
+
+[[gnu::noinline]] uint32_t __kmpc_get_max_team_threads() {
+ return mapping::getMaxTeamThreads();
+}
}
#define _TGT_KERNEL_LANGUAGE(NAME, MAPPER_NAME) \
>From 76cf6380e0e69821eb4346abc6f3766839b4b16b Mon Sep 17 00:00:00 2001
From: Larry Meadows <Lawrence.Meadows at amd.com>
Date: Wed, 19 Aug 2026 05:10:16 -0500
Subject: [PATCH 2/2] Say what __kmpc_get_max_team_threads returns
"The maximum number of threads in a team" reads as the block size. It is not:
in generic mode the threads hosting the main thread are not available to the
team, and how many those are is the runtime's to decide, which is the reason
the compiler asks instead of subtracting a warp itself.
Co-authored-by: Cursor <cursoragent at cursor.com>
---
openmp/device/include/Interface.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/openmp/device/include/Interface.h b/openmp/device/include/Interface.h
index 07ad56e8dbef0..dabec09f94524 100644
--- a/openmp/device/include/Interface.h
+++ b/openmp/device/include/Interface.h
@@ -217,7 +217,11 @@ uint32_t __kmpc_get_hardware_num_threads_in_block();
/// External interface to get the warp size.
uint32_t __kmpc_get_warp_size();
-/// External interface to get the maximum number of threads in a team.
+/// External interface to get the number of a block's threads that can execute a
+/// team's parallel region. This is not the block size: in generic mode the
+/// threads hosting the main thread are not available to the team. How many
+/// those are is the runtime's to decide, so callers must ask rather than
+/// subtract for themselves.
uint32_t __kmpc_get_max_team_threads();
/// Kernel
More information about the Openmp-commits
mailing list