This is an archive of the discontinued LLVM Phabricator instance.

Differential D112893

[OpenMP][FIX] Introduce and use a simple generic-mode barrier
ClosedPublic

Authored by jdoerfert on Oct 31 2021, 12:52 PM.

Details

Reviewers
jhuber6
ggeorgakoudis
tianshilei1992
Commits
rG73720c8059cf: [OpenMP][FIX] Introduce and use a simple generic-mode barrier
Summary

Before we had aligned barriers the __kmpc_barrier_simple_spmd was
OK to be used in the custom state machine. Now that SPMD barriers are
assumed to be aligned we need to use a "generic" barrier in places
that are not aligned.

Diff Detail

Event Timeline

jdoerfert created this revision.Oct 31 2021, 12:52 PM
Herald added subscribers: ormris, guansong, bollu and 2 others. · View Herald TranscriptOct 31 2021, 12:52 PM
jdoerfert requested review of this revision.Oct 31 2021, 12:52 PM
Herald added projects: Restricted Project, Restricted Project. · View Herald TranscriptOct 31 2021, 12:52 PM
Herald added subscribers: llvm-commits, sstefan1. · View Herald Transcript
Harbormaster completed remote builds in B131648: Diff 383676.Oct 31 2021, 12:52 PM
jdoerfert updated this revision to Diff 383706.Oct 31 2021, 7:51 PM

Add function to old RT as well

Harbormaster completed remote builds in B131672: Diff 383706.Oct 31 2021, 8:44 PM
tianshilei1992 accepted this revision.Nov 1 2021, 11:01 AM

LG

This revision is now accepted and ready to land.Nov 1 2021, 11:01 AM
This revision was landed with ongoing or failed builds.Nov 2 2021, 9:23 PM
Closed by commit rG73720c8059cf: [OpenMP][FIX] Introduce and use a simple generic-mode barrier (authored by jdoerfert). · Explain Why
This revision was automatically updated to reflect the committed changes.
jdoerfert added a commit: rG73720c8059cf: [OpenMP][FIX] Introduce and use a simple generic-mode barrier.

Revision Contents

PathSize
llvm/
include/
llvm/
Frontend/
OpenMP/
3 lines
lib/
Transforms/
IPO/
8 lines
test/
Transforms/
OpenMP/
56 lines
48 lines
4 lines
openmp/
libomptarget/
DeviceRTL/
include/
2 lines
src/
6 lines
1 line

Diff 383676

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

Show First 20 Lines • Show All 444 Lines • ▼ Show 20 Lines
__OMP_RTL(__kmpc_alloc_shared, false, VoidPtr, SizeTy) __OMP_RTL(__kmpc_alloc_shared, false, VoidPtr, SizeTy)
__OMP_RTL(__kmpc_free_shared, false, Void, VoidPtr, SizeTy) __OMP_RTL(__kmpc_free_shared, false, Void, VoidPtr, SizeTy)
__OMP_RTL(__kmpc_begin_sharing_variables, false, Void, VoidPtrPtrPtr, SizeTy) __OMP_RTL(__kmpc_begin_sharing_variables, false, Void, VoidPtrPtrPtr, SizeTy)
__OMP_RTL(__kmpc_end_sharing_variables, false, Void, ) __OMP_RTL(__kmpc_end_sharing_variables, false, Void, )
__OMP_RTL(__kmpc_get_shared_variables, false, Void, VoidPtrPtrPtr) __OMP_RTL(__kmpc_get_shared_variables, false, Void, VoidPtrPtrPtr)
__OMP_RTL(__kmpc_parallel_level, false, Int8, ) __OMP_RTL(__kmpc_parallel_level, false, Int8, )
__OMP_RTL(__kmpc_is_spmd_exec_mode, false, Int8, ) __OMP_RTL(__kmpc_is_spmd_exec_mode, false, Int8, )
__OMP_RTL(__kmpc_barrier_simple_spmd, false, Void, IdentPtr, Int32) __OMP_RTL(__kmpc_barrier_simple_spmd, false, Void, IdentPtr, Int32)
__OMP_RTL(__kmpc_barrier_simple_generic, false, Void, IdentPtr, Int32)
__OMP_RTL(__kmpc_warp_active_thread_mask, false, Int64,) __OMP_RTL(__kmpc_warp_active_thread_mask, false, Int64,)
__OMP_RTL(__kmpc_syncwarp, false, Void, Int64) __OMP_RTL(__kmpc_syncwarp, false, Void, Int64)
__OMP_RTL(__kmpc_is_generic_main_thread_id, false, Int8, Int32) __OMP_RTL(__kmpc_is_generic_main_thread_id, false, Int8, Int32)
__OMP_RTL(__last, false, Void, ) __OMP_RTL(__last, false, Void, )
▲ Show 20 Lines • Show All 127 Lines • ▼ Show 20 Lines
#define __OMP_RTL_ATTRS(Name, FnAttrSet, RetAttrSet, ArgAttrSets) \ #define __OMP_RTL_ATTRS(Name, FnAttrSet, RetAttrSet, ArgAttrSets) \
OMP_RTL_ATTRS(OMPRTL_##Name, FnAttrSet, RetAttrSet, ArgAttrSets) OMP_RTL_ATTRS(OMPRTL_##Name, FnAttrSet, RetAttrSet, ArgAttrSets)
__OMP_RTL_ATTRS(__kmpc_barrier, BarrierAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_barrier, BarrierAttrs, AttributeSet(),
ParamAttrs(ReadOnlyPtrAttrs)) ParamAttrs(ReadOnlyPtrAttrs))
__OMP_RTL_ATTRS(__kmpc_barrier_simple_spmd, BarrierAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_barrier_simple_spmd, BarrierAttrs, AttributeSet(),
ParamAttrs(ReadOnlyPtrAttrs)) ParamAttrs(ReadOnlyPtrAttrs))
__OMP_RTL_ATTRS(__kmpc_barrier_simple_generic, BarrierAttrs, AttributeSet(),
ParamAttrs(ReadOnlyPtrAttrs))
__OMP_RTL_ATTRS(__kmpc_warp_active_thread_mask, BarrierAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_warp_active_thread_mask, BarrierAttrs, AttributeSet(),
ParamAttrs()) ParamAttrs())
__OMP_RTL_ATTRS(__kmpc_syncwarp, BarrierAttrs, AttributeSet(), ParamAttrs()) __OMP_RTL_ATTRS(__kmpc_syncwarp, BarrierAttrs, AttributeSet(), ParamAttrs())
__OMP_RTL_ATTRS(__kmpc_cancel, InaccessibleArgOnlyAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_cancel, InaccessibleArgOnlyAttrs, AttributeSet(),
ParamAttrs(ReadOnlyPtrAttrs)) ParamAttrs(ReadOnlyPtrAttrs))
__OMP_RTL_ATTRS(__kmpc_cancel_barrier, BarrierAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_cancel_barrier, BarrierAttrs, AttributeSet(),
ParamAttrs(ReadOnlyPtrAttrs)) ParamAttrs(ReadOnlyPtrAttrs))
__OMP_RTL_ATTRS(__kmpc_flush, BarrierAttrs, AttributeSet(), __OMP_RTL_ATTRS(__kmpc_flush, BarrierAttrs, AttributeSet(),
▲ Show 20 Lines • Show All 617 Lines • Show Last 20 Lines

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

Show First 20 Lines • Show All 1,871 Lines • ▼ Show 20 Lines bool runAttributor(bool IsModulePass) {
// Temporarily make these function have external linkage so the Attributor // Temporarily make these function have external linkage so the Attributor
// doesn't remove them when we try to look them up later. // doesn't remove them when we try to look them up later.
ExternalizationRAII Parallel(OMPInfoCache, OMPRTL___kmpc_kernel_parallel); ExternalizationRAII Parallel(OMPInfoCache, OMPRTL___kmpc_kernel_parallel);
ExternalizationRAII EndParallel(OMPInfoCache, ExternalizationRAII EndParallel(OMPInfoCache,
OMPRTL___kmpc_kernel_end_parallel); OMPRTL___kmpc_kernel_end_parallel);
ExternalizationRAII BarrierSPMD(OMPInfoCache, ExternalizationRAII BarrierSPMD(OMPInfoCache,
OMPRTL___kmpc_barrier_simple_spmd); OMPRTL___kmpc_barrier_simple_spmd);
ExternalizationRAII BarrierGeneric(OMPInfoCache,
OMPRTL___kmpc_barrier_simple_generic);
ExternalizationRAII ThreadId(OMPInfoCache, ExternalizationRAII ThreadId(OMPInfoCache,
OMPRTL___kmpc_get_hardware_thread_id_in_block); OMPRTL___kmpc_get_hardware_thread_id_in_block);
registerAAs(IsModulePass); registerAAs(IsModulePass);
ChangeStatus Changed = A.run(); ChangeStatus Changed = A.run();
LLVM_DEBUG(dbgs() << "[Attributor] Done with " << SCC.size() LLVM_DEBUG(dbgs() << "[Attributor] Done with " << SCC.size()
▲ Show 20 Lines • Show All 1,555 Lines • ▼ Show 20 Lines if (ReachedUnknownParallelRegions.empty()) {
} }
} }
// Create all the blocks: // Create all the blocks:
// //
// InitCB = __kmpc_target_init(...) // InitCB = __kmpc_target_init(...)
// bool IsWorker = InitCB >= 0; // bool IsWorker = InitCB >= 0;
// if (IsWorker) { // if (IsWorker) {
// SMBeginBB: __kmpc_barrier_simple_spmd(...); // SMBeginBB: __kmpc_barrier_simple_generic(...);
// void *WorkFn; // void *WorkFn;
// bool Active = __kmpc_kernel_parallel(&WorkFn); // bool Active = __kmpc_kernel_parallel(&WorkFn);
// if (!WorkFn) return; // if (!WorkFn) return;
// SMIsActiveCheckBB: if (Active) { // SMIsActiveCheckBB: if (Active) {
// SMIfCascadeCurrentBB: if (WorkFn == <ParFn0>) // SMIfCascadeCurrentBB: if (WorkFn == <ParFn0>)
// ParFn0(...); // ParFn0(...);
// SMIfCascadeCurrentBB: else if (WorkFn == <ParFn1>) // SMIfCascadeCurrentBB: else if (WorkFn == <ParFn1>)
// ParFn1(...); // ParFn1(...);
// ... // ...
// SMIfCascadeCurrentBB: else // SMIfCascadeCurrentBB: else
// ((WorkFnTy*)WorkFn)(...); // ((WorkFnTy*)WorkFn)(...);
// SMEndParallelBB: __kmpc_kernel_end_parallel(...); // SMEndParallelBB: __kmpc_kernel_end_parallel(...);
// } // }
// SMDoneBB: __kmpc_barrier_simple_spmd(...); // SMDoneBB: __kmpc_barrier_simple_generic(...);
// goto SMBeginBB; // goto SMBeginBB;
// } // }
// UserCodeEntryBB: // user code // UserCodeEntryBB: // user code
// __kmpc_target_deinit(...) // __kmpc_target_deinit(...)
// //
Function *Kernel = getAssociatedFunction(); Function *Kernel = getAssociatedFunction();
assert(Kernel && "Expected an associated function!"); assert(Kernel && "Expected an associated function!");
▲ Show 20 Lines • Show All 51 Lines • ▼ Show 20 Lines OMPInfoCache.OMPBuilder.updateToLocation(
StateMachineBeginBB->end()), StateMachineBeginBB->end()),
DLoc)); DLoc));
Value *Ident = KernelInitCB->getArgOperand(0); Value *Ident = KernelInitCB->getArgOperand(0);
Value *GTid = KernelInitCB; Value *GTid = KernelInitCB;
FunctionCallee BarrierFn = FunctionCallee BarrierFn =
OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction( OMPInfoCache.OMPBuilder.getOrCreateRuntimeFunction(
M, OMPRTL___kmpc_barrier_simple_spmd); M, OMPRTL___kmpc_barrier_simple_generic);
CallInst::Create(BarrierFn, {Ident, GTid}, "", StateMachineBeginBB) CallInst::Create(BarrierFn, {Ident, GTid}, "", StateMachineBeginBB)
->setDebugLoc(DLoc); ->setDebugLoc(DLoc);
if (WorkFnAI->getType()->getPointerAddressSpace() != if (WorkFnAI->getType()->getPointerAddressSpace() !=
(unsigned int)AddressSpace::Generic) { (unsigned int)AddressSpace::Generic) {
WorkFnAI = new AddrSpaceCastInst( WorkFnAI = new AddrSpaceCastInst(
WorkFnAI, WorkFnAI,
PointerType::getWithSamePointeeType( PointerType::getWithSamePointeeType(
▲ Show 20 Lines • Show All 1,274 Lines • Show Last 20 Lines

llvm/test/Transforms/OpenMP/custom_state_machines.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 909 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
Show All 11 Lines
; AMDGPU-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.check3: ; AMDGPU: worker_state_machine.parallel_region.check3:
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-NEXT: call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 80 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
Show All 17 Lines
; AMDGPU-NEXT: call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.check6: ; AMDGPU: worker_state_machine.parallel_region.check6:
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 97 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
Show All 13 Lines
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
Show All 11 Lines
; AMDGPU-NEXT: call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.check3: ; AMDGPU: worker_state_machine.parallel_region.check3:
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-NEXT: call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
Show All 11 Lines
; AMDGPU-NEXT: call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.check3: ; AMDGPU: worker_state_machine.parallel_region.check3:
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; AMDGPU-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-NEXT: call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
; AMDGPU: worker_state_machine.is_active.check: ; AMDGPU: worker_state_machine.is_active.check:
; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU: worker_state_machine.parallel_region.check: ; AMDGPU: worker_state_machine.parallel_region.check:
; AMDGPU-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper ; AMDGPU-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper
; AMDGPU-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; AMDGPU: worker_state_machine.parallel_region.execute: ; AMDGPU: worker_state_machine.parallel_region.execute:
; AMDGPU-NEXT: call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
; AMDGPU: worker_state_machine.is_active.check: ; AMDGPU: worker_state_machine.is_active.check:
; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; 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]]) #[[ATTR3]]
; AMDGPU-NEXT: call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; AMDGPU-NEXT: call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
▲ Show 20 Lines • Show All 188 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
Show All 10 Lines
; NVPTX-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.check3: ; NVPTX: worker_state_machine.parallel_region.check3:
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-NEXT: call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__1(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 80 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
Show All 16 Lines
; NVPTX-NEXT: call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void @__omp_outlined__18_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.check6: ; NVPTX: worker_state_machine.parallel_region.check6:
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 97 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
Show All 12 Lines
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.fallback.execute: ; NVPTX: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
Show All 10 Lines
; NVPTX-NEXT: call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void @__omp_outlined__11_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.check3: ; NVPTX: worker_state_machine.parallel_region.check3:
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-NEXT: call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__9(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
Show All 10 Lines
; NVPTX-NEXT: call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void @__omp_outlined__14_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.check3: ; NVPTX: worker_state_machine.parallel_region.check3:
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4 ; NVPTX-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-NEXT: call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__12(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
▲ Show 20 Lines • Show All 79 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX: worker_state_machine.parallel_region.check: ; NVPTX: worker_state_machine.parallel_region.check:
; NVPTX-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper ; NVPTX-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__19_wrapper
; NVPTX-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; NVPTX: worker_state_machine.parallel_region.execute: ; NVPTX: worker_state_machine.parallel_region.execute:
; NVPTX-NEXT: call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void @__omp_outlined__19_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX: worker_state_machine.parallel_region.fallback.execute: ; NVPTX: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__15(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Lines
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX: worker_state_machine.parallel_region.fallback.execute: ; NVPTX: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; 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]]) #[[ATTR3]]
; NVPTX-NEXT: call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]] ; NVPTX-NEXT: call void @__omp_outlined__16(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR3]]
; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
▲ Show 20 Lines • Show All 1,583 Lines • Show Last 20 Lines

llvm/test/Transforms/OpenMP/spmdization.ll

Show First 20 Lines • Show All 181 Lines • ▼ Show 20 Lines
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.begin: ; AMDGPU-DISABLED: worker_state_machine.begin:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.finished: ; AMDGPU-DISABLED: worker_state_machine.finished:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker_state_machine.is_active.check: ; AMDGPU-DISABLED: worker_state_machine.is_active.check:
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check:
; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__1_wrapper.ID to void (i16, i32)*) ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__1_wrapper.ID to void (i16, i32)*)
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute:
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU-DISABLED: worker_state_machine.done.barrier: ; AMDGPU-DISABLED: worker_state_machine.done.barrier:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU-DISABLED: thread.user_code.check: ; AMDGPU-DISABLED: thread.user_code.check:
; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU-DISABLED: user_code.entry: ; AMDGPU-DISABLED: user_code.entry:
; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[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 ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker.exit: ; AMDGPU-DISABLED: worker.exit:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; ;
; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_l5 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_l5
; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] { ; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] {
; NVPTX-DISABLED-NEXT: entry: ; NVPTX-DISABLED-NEXT: entry:
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.begin: ; NVPTX-DISABLED: worker_state_machine.begin:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.finished: ; NVPTX-DISABLED: worker_state_machine.finished:
; NVPTX-DISABLED-NEXT: ret void ; NVPTX-DISABLED-NEXT: ret void
; NVPTX-DISABLED: worker_state_machine.is_active.check: ; NVPTX-DISABLED: worker_state_machine.is_active.check:
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.check: ; NVPTX-DISABLED: worker_state_machine.parallel_region.check:
; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__1_wrapper.ID to void (i16, i32)*) ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__1_wrapper.ID to void (i16, i32)*)
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute:
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.end: ; NVPTX-DISABLED: worker_state_machine.parallel_region.end:
; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX-DISABLED: worker_state_machine.done.barrier: ; NVPTX-DISABLED: worker_state_machine.done.barrier:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX-DISABLED: thread.user_code.check: ; NVPTX-DISABLED: thread.user_code.check:
; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX-DISABLED: user_code.entry: ; NVPTX-DISABLED: user_code.entry:
; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4:[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 ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
▲ Show 20 Lines • Show All 342 Lines • ▼ Show 20 Lines
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.begin: ; AMDGPU-DISABLED: worker_state_machine.begin:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.finished: ; AMDGPU-DISABLED: worker_state_machine.finished:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker_state_machine.is_active.check: ; AMDGPU-DISABLED: worker_state_machine.is_active.check:
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check:
; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__3_wrapper.ID to void (i16, i32)*) ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__3_wrapper.ID to void (i16, i32)*)
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute:
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU-DISABLED: worker_state_machine.done.barrier: ; AMDGPU-DISABLED: worker_state_machine.done.barrier:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU-DISABLED: thread.user_code.check: ; AMDGPU-DISABLED: thread.user_code.check:
; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU-DISABLED: user_code.entry: ; AMDGPU-DISABLED: user_code.entry:
; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker.exit: ; AMDGPU-DISABLED: worker.exit:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; ;
; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_stack_var_l20 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_stack_var_l20
; NVPTX-DISABLED-SAME: () #[[ATTR0]] { ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
; NVPTX-DISABLED-NEXT: entry: ; NVPTX-DISABLED-NEXT: entry:
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.begin: ; NVPTX-DISABLED: worker_state_machine.begin:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.finished: ; NVPTX-DISABLED: worker_state_machine.finished:
; NVPTX-DISABLED-NEXT: ret void ; NVPTX-DISABLED-NEXT: ret void
; NVPTX-DISABLED: worker_state_machine.is_active.check: ; NVPTX-DISABLED: worker_state_machine.is_active.check:
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.check: ; NVPTX-DISABLED: worker_state_machine.parallel_region.check:
; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__3_wrapper.ID to void (i16, i32)*) ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__3_wrapper.ID to void (i16, i32)*)
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute:
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.end: ; NVPTX-DISABLED: worker_state_machine.parallel_region.end:
; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX-DISABLED: worker_state_machine.done.barrier: ; NVPTX-DISABLED: worker_state_machine.done.barrier:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX-DISABLED: thread.user_code.check: ; NVPTX-DISABLED: thread.user_code.check:
; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX-DISABLED: user_code.entry: ; NVPTX-DISABLED: user_code.entry:
; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__2(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
▲ Show 20 Lines • Show All 348 Lines • ▼ Show 20 Lines
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.begin: ; AMDGPU-DISABLED: worker_state_machine.begin:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.finished: ; AMDGPU-DISABLED: worker_state_machine.finished:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker_state_machine.is_active.check: ; AMDGPU-DISABLED: worker_state_machine.is_active.check:
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check:
; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*) ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*)
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute:
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU-DISABLED: worker_state_machine.done.barrier: ; AMDGPU-DISABLED: worker_state_machine.done.barrier:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU-DISABLED: thread.user_code.check: ; AMDGPU-DISABLED: thread.user_code.check:
; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU-DISABLED: user_code.entry: ; AMDGPU-DISABLED: user_code.entry:
; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker.exit: ; AMDGPU-DISABLED: worker.exit:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; ;
; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_l35 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_l35
; NVPTX-DISABLED-SAME: () #[[ATTR0]] { ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
; NVPTX-DISABLED-NEXT: entry: ; NVPTX-DISABLED-NEXT: entry:
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.begin: ; NVPTX-DISABLED: worker_state_machine.begin:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.finished: ; NVPTX-DISABLED: worker_state_machine.finished:
; NVPTX-DISABLED-NEXT: ret void ; NVPTX-DISABLED-NEXT: ret void
; NVPTX-DISABLED: worker_state_machine.is_active.check: ; NVPTX-DISABLED: worker_state_machine.is_active.check:
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.check: ; NVPTX-DISABLED: worker_state_machine.parallel_region.check:
; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*) ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__5_wrapper.ID to void (i16, i32)*)
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute:
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.end: ; NVPTX-DISABLED: worker_state_machine.parallel_region.end:
; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX-DISABLED: worker_state_machine.done.barrier: ; NVPTX-DISABLED: worker_state_machine.done.barrier:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX-DISABLED: thread.user_code.check: ; NVPTX-DISABLED: thread.user_code.check:
; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX-DISABLED: user_code.entry: ; NVPTX-DISABLED: user_code.entry:
; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__4(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
▲ Show 20 Lines • Show All 388 Lines • ▼ Show 20 Lines
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.begin: ; AMDGPU-DISABLED: worker_state_machine.begin:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.finished: ; AMDGPU-DISABLED: worker_state_machine.finished:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker_state_machine.is_active.check: ; AMDGPU-DISABLED: worker_state_machine.is_active.check:
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check:
; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*) ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*)
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute:
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU-DISABLED: worker_state_machine.done.barrier: ; AMDGPU-DISABLED: worker_state_machine.done.barrier:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU-DISABLED: thread.user_code.check: ; AMDGPU-DISABLED: thread.user_code.check:
; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU-DISABLED: user_code.entry: ; AMDGPU-DISABLED: user_code.entry:
; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker.exit: ; AMDGPU-DISABLED: worker.exit:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; ;
; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_guarded_l50 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_guarded_l50
; NVPTX-DISABLED-SAME: () #[[ATTR0]] { ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
; NVPTX-DISABLED-NEXT: entry: ; NVPTX-DISABLED-NEXT: entry:
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.begin: ; NVPTX-DISABLED: worker_state_machine.begin:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.finished: ; NVPTX-DISABLED: worker_state_machine.finished:
; NVPTX-DISABLED-NEXT: ret void ; NVPTX-DISABLED-NEXT: ret void
; NVPTX-DISABLED: worker_state_machine.is_active.check: ; NVPTX-DISABLED: worker_state_machine.is_active.check:
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.check: ; NVPTX-DISABLED: worker_state_machine.parallel_region.check:
; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*) ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq void (i16, i32)* [[WORKER_WORK_FN_ADDR_CAST]], bitcast (i8* @__omp_outlined__7_wrapper.ID to void (i16, i32)*)
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute:
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.end: ; NVPTX-DISABLED: worker_state_machine.parallel_region.end:
; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX-DISABLED: worker_state_machine.done.barrier: ; NVPTX-DISABLED: worker_state_machine.done.barrier:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX-DISABLED: thread.user_code.check: ; NVPTX-DISABLED: thread.user_code.check:
; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX-DISABLED: user_code.entry: ; NVPTX-DISABLED: user_code.entry:
; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; 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 ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], i32* [[DOTTHREADID_TEMP_]], align 4
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__6(i32* noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
▲ Show 20 Lines • Show All 388 Lines • ▼ Show 20 Lines
; AMDGPU-NEXT: entry: ; AMDGPU-NEXT: entry:
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU: worker_state_machine.begin: ; AMDGPU: worker_state_machine.begin:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU: worker_state_machine.finished: ; AMDGPU: worker_state_machine.finished:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
; AMDGPU: worker_state_machine.is_active.check: ; AMDGPU: worker_state_machine.is_active.check:
; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU: worker_state_machine.parallel_region.end: ; AMDGPU: worker_state_machine.parallel_region.end:
; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU: worker_state_machine.done.barrier: ; AMDGPU: worker_state_machine.done.barrier:
; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU: thread.user_code.check: ; AMDGPU: thread.user_code.check:
; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU: user_code.entry: ; AMDGPU: user_code.entry:
; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
; AMDGPU-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
; AMDGPU: worker.exit: ; AMDGPU: worker.exit:
; AMDGPU-NEXT: ret void ; AMDGPU-NEXT: ret void
; ;
; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65 ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
; NVPTX-SAME: () #[[ATTR0]] { ; NVPTX-SAME: () #[[ATTR0]] {
; NVPTX-NEXT: entry: ; NVPTX-NEXT: entry:
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX: worker_state_machine.begin: ; NVPTX: worker_state_machine.begin:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX: worker_state_machine.finished: ; NVPTX: worker_state_machine.finished:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker_state_machine.is_active.check: ; NVPTX: worker_state_machine.is_active.check:
; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX: worker_state_machine.parallel_region.fallback.execute: ; NVPTX: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX: worker_state_machine.parallel_region.end: ; NVPTX: worker_state_machine.parallel_region.end:
; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX: worker_state_machine.done.barrier: ; NVPTX: worker_state_machine.done.barrier:
; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX: thread.user_code.check: ; NVPTX: thread.user_code.check:
; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX: user_code.entry: ; NVPTX: user_code.entry:
; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
; NVPTX-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; NVPTX-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; NVPTX: worker.exit: ; NVPTX: worker.exit:
; NVPTX-NEXT: ret void ; NVPTX-NEXT: ret void
; ;
; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65 ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
; AMDGPU-DISABLED-SAME: () #[[ATTR0]] { ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] {
; AMDGPU-DISABLED-NEXT: entry: ; AMDGPU-DISABLED-NEXT: entry:
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5) ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8, addrspace(5)
; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; AMDGPU-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.begin: ; AMDGPU-DISABLED: worker_state_machine.begin:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8** ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast i8* addrspace(5)* [[WORKER_WORK_FN_ADDR]] to i8**
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]]) ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR_GENERIC]])
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; AMDGPU-DISABLED: worker_state_machine.finished: ; AMDGPU-DISABLED: worker_state_machine.finished:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker_state_machine.is_active.check: ; AMDGPU-DISABLED: worker_state_machine.is_active.check:
; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; AMDGPU-DISABLED: worker_state_machine.done.barrier: ; AMDGPU-DISABLED: worker_state_machine.done.barrier:
; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; AMDGPU-DISABLED: thread.user_code.check: ; AMDGPU-DISABLED: thread.user_code.check:
; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; AMDGPU-DISABLED: user_code.entry: ; AMDGPU-DISABLED: user_code.entry:
; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; AMDGPU-DISABLED: worker.exit: ; AMDGPU-DISABLED: worker.exit:
; AMDGPU-DISABLED-NEXT: ret void ; AMDGPU-DISABLED-NEXT: ret void
; ;
; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65 ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
; NVPTX-DISABLED-SAME: () #[[ATTR0]] { ; NVPTX-DISABLED-SAME: () #[[ATTR0]] {
; NVPTX-DISABLED-NEXT: entry: ; NVPTX-DISABLED-NEXT: entry:
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4 ; NVPTX-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true) ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i8 1, i1 false, i1 true)
; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.begin: ; NVPTX-DISABLED: worker_state_machine.begin:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; NVPTX-DISABLED: worker_state_machine.finished: ; NVPTX-DISABLED: worker_state_machine.finished:
; NVPTX-DISABLED-NEXT: ret void ; NVPTX-DISABLED-NEXT: ret void
; NVPTX-DISABLED: worker_state_machine.is_active.check: ; NVPTX-DISABLED: worker_state_machine.is_active.check:
; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute:
; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; NVPTX-DISABLED: worker_state_machine.parallel_region.end: ; NVPTX-DISABLED: worker_state_machine.parallel_region.end:
; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; NVPTX-DISABLED: worker_state_machine.done.barrier: ; NVPTX-DISABLED: worker_state_machine.done.barrier:
; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; NVPTX-DISABLED: thread.user_code.check: ; NVPTX-DISABLED: thread.user_code.check:
; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; NVPTX-DISABLED: user_code.entry: ; NVPTX-DISABLED: user_code.entry:
; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]]) #[[ATTR4]]
; NVPTX-DISABLED-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]] ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__8(i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTTHREADID_TEMP_]], i32* noalias nocapture noundef nonnull readnone align 4 dereferenceable(4) [[DOTZERO_ADDR]]) #[[ATTR4]]
; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true) ; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 1, i1 true)
▲ Show 20 Lines • Show All 213 Lines • Show Last 20 Lines

llvm/test/Transforms/OpenMP/spmdization_guarding.ll

Show First 20 Lines • Show All 174 Lines • ▼ Show 20 Lines
; CHECK-DISABLED-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { ; CHECK-DISABLED-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-DISABLED-NEXT: entry: ; CHECK-DISABLED-NEXT: entry:
; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 ; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8
; CHECK-DISABLED-NEXT: [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32 ; CHECK-DISABLED-NEXT: [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 false, i1 true) #[[ATTR4:[0-9]+]] ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 false, i1 true) #[[ATTR4:[0-9]+]]
; CHECK-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 ; CHECK-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1
; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] ; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]]
; CHECK-DISABLED: worker_state_machine.begin: ; CHECK-DISABLED: worker_state_machine.begin:
; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]])
; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 ; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8
; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* ; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)*
; CHECK-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null ; CHECK-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null
; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] ; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
; CHECK-DISABLED: worker_state_machine.finished: ; CHECK-DISABLED: worker_state_machine.finished:
; CHECK-DISABLED-NEXT: ret void ; CHECK-DISABLED-NEXT: ret void
; CHECK-DISABLED: worker_state_machine.is_active.check: ; CHECK-DISABLED: worker_state_machine.is_active.check:
; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] ; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
; CHECK-DISABLED: worker_state_machine.parallel_region.check: ; CHECK-DISABLED: worker_state_machine.parallel_region.check:
; CHECK-DISABLED-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] ; CHECK-DISABLED-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
; CHECK-DISABLED: worker_state_machine.parallel_region.execute: ; CHECK-DISABLED: worker_state_machine.parallel_region.execute:
; CHECK-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) ; CHECK-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]])
; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] ; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
; CHECK-DISABLED: worker_state_machine.parallel_region.check1: ; CHECK-DISABLED: worker_state_machine.parallel_region.check1:
; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] ; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
; CHECK-DISABLED: worker_state_machine.parallel_region.end: ; CHECK-DISABLED: worker_state_machine.parallel_region.end:
; CHECK-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() ; CHECK-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel()
; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] ; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]]
; CHECK-DISABLED: worker_state_machine.done.barrier: ; CHECK-DISABLED: worker_state_machine.done.barrier:
; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) ; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] ; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]]
; CHECK-DISABLED: thread.user_code.check: ; CHECK-DISABLED: thread.user_code.check:
; CHECK-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 ; CHECK-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
; CHECK-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] ; CHECK-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
; CHECK-DISABLED: user_code.entry: ; CHECK-DISABLED: user_code.entry:
; CHECK-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @[[GLOB1]]) #[[ATTR4]] ; CHECK-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @[[GLOB1]]) #[[ATTR4]]
; CHECK-DISABLED-NEXT: store i32 0, i32* [[X]], align 4, !noalias !8 ; CHECK-DISABLED-NEXT: store i32 0, i32* [[X]], align 4, !noalias !8
; CHECK-DISABLED-NEXT: [[ARRAYIDX1_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 1 ; CHECK-DISABLED-NEXT: [[ARRAYIDX1_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 1
▲ Show 20 Lines • Show All 208 Lines • Show Last 20 Lines

openmp/libomptarget/DeviceRTL/include/Interface.h

Show First 20 Lines • Show All 243 Lines • ▼ Show 20 Lines
void __kmpc_end_ordered(IdentTy *Loc, int32_t TId); void __kmpc_end_ordered(IdentTy *Loc, int32_t TId);
int32_t __kmpc_cancel_barrier(IdentTy *Loc_ref, int32_t TId); int32_t __kmpc_cancel_barrier(IdentTy *Loc_ref, int32_t TId);
void __kmpc_barrier(IdentTy *Loc_ref, int32_t TId); void __kmpc_barrier(IdentTy *Loc_ref, int32_t TId);
void __kmpc_barrier_simple_spmd(IdentTy *Loc_ref, int32_t TId); void __kmpc_barrier_simple_spmd(IdentTy *Loc_ref, int32_t TId);
void __kmpc_barrier_simple_generic(IdentTy *Loc_ref, int32_t TId);
int32_t __kmpc_master(IdentTy *Loc, int32_t TId); int32_t __kmpc_master(IdentTy *Loc, int32_t TId);
void __kmpc_end_master(IdentTy *Loc, int32_t TId); void __kmpc_end_master(IdentTy *Loc, int32_t TId);
int32_t __kmpc_single(IdentTy *Loc, int32_t TId); int32_t __kmpc_single(IdentTy *Loc, int32_t TId);
void __kmpc_end_single(IdentTy *Loc, int32_t TId); void __kmpc_end_single(IdentTy *Loc, int32_t TId);
▲ Show 20 Lines • Show All 92 Lines • Show Last 20 Lines

openmp/libomptarget/DeviceRTL/src/Synchronization.cpp

Show First 20 Lines • Show All 342 Lines • ▼ Show 20 Lines
} }
__attribute__((noinline)) void __kmpc_barrier_simple_spmd(IdentTy *Loc, __attribute__((noinline)) void __kmpc_barrier_simple_spmd(IdentTy *Loc,
int32_t TId) { int32_t TId) {
FunctionTracingRAII(); FunctionTracingRAII();
synchronize::threadsAligned(); synchronize::threadsAligned();
} }
__attribute__((noinline)) void __kmpc_barrier_simple_generic(IdentTy *Loc,
int32_t TId) {
FunctionTracingRAII();
synchronize::threads();
}
int32_t __kmpc_master(IdentTy *Loc, int32_t TId) { int32_t __kmpc_master(IdentTy *Loc, int32_t TId) {
FunctionTracingRAII(); FunctionTracingRAII();
return omp_get_team_num() == 0; return omp_get_team_num() == 0;
} }
void __kmpc_end_master(IdentTy *Loc, int32_t TId) { FunctionTracingRAII(); } void __kmpc_end_master(IdentTy *Loc, int32_t TId) { FunctionTracingRAII(); }
int32_t __kmpc_single(IdentTy *Loc, int32_t TId) { int32_t __kmpc_single(IdentTy *Loc, int32_t TId) {
▲ Show 20 Lines • Show All 46 Lines • Show Last 20 Lines

openmp/libomptarget/DeviceRTL/src/Utils.cpp

Show All 19 Lines
using namespace _OMP; using namespace _OMP;
namespace _OMP { namespace _OMP {
/// Helper to keep code alive without introducing a performance penalty. /// Helper to keep code alive without introducing a performance penalty.
__attribute__((used, weak, optnone)) void keepAlive() { __attribute__((used, weak, optnone)) void keepAlive() {
__kmpc_get_hardware_thread_id_in_block(); __kmpc_get_hardware_thread_id_in_block();
__kmpc_get_hardware_num_threads_in_block(); __kmpc_get_hardware_num_threads_in_block();
__kmpc_barrier_simple_spmd(nullptr, 0); __kmpc_barrier_simple_spmd(nullptr, 0);
__kmpc_barrier_simple_generic(nullptr, 0);
} }
} // namespace _OMP } // namespace _OMP
namespace impl { namespace impl {
/// AMDGCN Implementation /// AMDGCN Implementation
/// ///
///{ ///{
▲ Show 20 Lines • Show All 112 Lines • Show Last 20 Lines