This is an archive of the discontinued LLVM Phabricator instance.

Differential D107862

[OpenMP][FIX] Disabled optimizations have to be made known
ClosedPublic

Authored by jdoerfert on Aug 10 2021, 12:39 PM.

Details

Reviewers
jhuber6
Commits
rGe0c5d83a9273: [OpenMP][FIX] Disabled optimizations have to be made known
Summary

To avoid simplification with wrong constants we need to make sure we
know that we won't perform specific optimizations based on the users
request. The non-SPMDzation and non-CustomStateMachine flags did only
prevent the final transformation but allowed to value simplification
to go ahead.

Diff Detail

Event Timeline

jdoerfert created this revision.Aug 10 2021, 12:39 PM
Herald added subscribers: ormris, guansong, bollu and 2 others. · View Herald TranscriptAug 10 2021, 12:39 PM
jdoerfert requested review of this revision.Aug 10 2021, 12:39 PM
Herald added a project: Restricted Project. · View Herald TranscriptAug 10 2021, 12:39 PM
Herald added subscribers: llvm-commits, sstefan1. · View Herald Transcript
jhuber6 accepted this revision.Aug 10 2021, 12:41 PM

LGTM

This revision is now accepted and ready to land.Aug 10 2021, 12:41 PM
Harbormaster completed remote builds in B118961: Diff 365589.Aug 10 2021, 1:07 PM
Closed by commit rGe0c5d83a9273: [OpenMP][FIX] Disabled optimizations have to be made known (authored by jdoerfert). · Explain WhyAug 10 2021, 10:50 PM
This revision was automatically updated to reflect the committed changes.
jdoerfert added a commit: rGe0c5d83a9273: [OpenMP][FIX] Disabled optimizations have to be made known.

Revision Contents

PathSize
llvm/
lib/
Transforms/
IPO/
11 lines
test/
Transforms/
OpenMP/
5 lines

Diff 365659

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

Show First 20 Lines • Show All 2,828 Lines • ▼ Show 20 Linesstruct AAKernelInfoFunction : AAKernelInfo {
/// See AbstractAttribute::initialize(...). /// See AbstractAttribute::initialize(...).
void initialize(Attributor &A) override { void initialize(Attributor &A) override {
// This is a high-level transform that might change the constant arguments // This is a high-level transform that might change the constant arguments
// of the init and dinit calls. We need to tell the Attributor about this // of the init and dinit calls. We need to tell the Attributor about this
// to avoid other parts using the current constant value for simpliication. // to avoid other parts using the current constant value for simpliication.
auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache()); auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
// If we have disabled SPMD-ization, stop
if (DisableOpenMPOptSPMDization)
SPMDCompatibilityTracker.indicatePessimisticFixpoint();
Function *Fn = getAnchorScope(); Function *Fn = getAnchorScope();
if (!OMPInfoCache.Kernels.count(Fn)) if (!OMPInfoCache.Kernels.count(Fn))
return; return;
// Add itself to the reaching kernel and set IsKernelEntry. // Add itself to the reaching kernel and set IsKernelEntry.
ReachingKernelEntries.insert(Fn); ReachingKernelEntries.insert(Fn);
IsKernelEntry = true; IsKernelEntry = true;
▲ Show 20 Lines • Show All 41 Lines • ▼ Show 20 Lines Attributor::SimplifictionCallbackTy StateMachineSimplifyCB =
bool &UsedAssumedInformation) -> Optional<Value *> { bool &UsedAssumedInformation) -> Optional<Value *> {
// IRP represents the "use generic state machine" argument of an // IRP represents the "use generic state machine" argument of an
// __kmpc_target_init call. We will answer this one with the internal // __kmpc_target_init call. We will answer this one with the internal
// state. As long as we are not in an invalid state, we will create a // state. As long as we are not in an invalid state, we will create a
// custom state machine so the value should be a `i1 false`. If we are // custom state machine so the value should be a `i1 false`. If we are
// in an invalid state, we won't change the value that is in the IR. // in an invalid state, we won't change the value that is in the IR.
if (!isValidState()) if (!isValidState())
return nullptr; return nullptr;
// If we have disabled state machine rewrites, don't make a custom one.
if (DisableOpenMPOptStateMachineRewrite)
return nullptr;
if (AA) if (AA)
A.recordDependence(*this, *AA, DepClassTy::OPTIONAL); A.recordDependence(*this, *AA, DepClassTy::OPTIONAL);
UsedAssumedInformation = !isAtFixpoint(); UsedAssumedInformation = !isAtFixpoint();
auto *FalseVal = auto *FalseVal =
ConstantInt::getBool(IRP.getAnchorValue().getContext(), 0); ConstantInt::getBool(IRP.getAnchorValue().getContext(), 0);
return FalseVal; return FalseVal;
}; };
▲ Show 20 Lines • Show All 85 Lines • ▼ Show 20 Lines ChangeStatus manifest(Attributor &A) override {
// custom state machine. // custom state machine.
if (!changeToSPMDMode(A)) if (!changeToSPMDMode(A))
buildCustomStateMachine(A); buildCustomStateMachine(A);
return ChangeStatus::CHANGED; return ChangeStatus::CHANGED;
} }
bool changeToSPMDMode(Attributor &A) { bool changeToSPMDMode(Attributor &A) {
// If we have disabled SPMD-ization, stop
if (DisableOpenMPOptSPMDization)
indicatePessimisticFixpoint();
auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache()); auto &OMPInfoCache = static_cast<OMPInformationCache &>(A.getInfoCache());
if (!SPMDCompatibilityTracker.isAssumed()) { if (!SPMDCompatibilityTracker.isAssumed()) {
for (Instruction *NonCompatibleI : SPMDCompatibilityTracker) { for (Instruction *NonCompatibleI : SPMDCompatibilityTracker) {
if (!NonCompatibleI) if (!NonCompatibleI)
continue; continue;
// Skip diagnostics on calls to known OpenMP runtime functions for now. // Skip diagnostics on calls to known OpenMP runtime functions for now.
▲ Show 20 Lines • Show All 1,647 Lines • Show Last 20 Lines

llvm/test/Transforms/OpenMP/spmdization.ll

Show First 20 Lines • Show All 132 Lines • ▼ Show 20 Lines
; CHECK-NEXT: ret void ; CHECK-NEXT: ret void
; ;
; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_l5 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_l5
; CHECK-DISABLED-SAME: () #[[ATTR0:[0-9]+]] { ; CHECK-DISABLED-SAME: () #[[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: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true) ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
; 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_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_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)*
▲ Show 20 Lines • Show All 240 Lines • ▼ Show 20 Lines
; CHECK-NEXT: ret void ; CHECK-NEXT: ret void
; ;
; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_stack_var_l20 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_stack_var_l20
; CHECK-DISABLED-SAME: () #[[ATTR0]] { ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
; 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: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true) ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
; 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_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_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)*
▲ Show 20 Lines • Show All 242 Lines • ▼ Show 20 Lines
; CHECK-NEXT: ret void ; CHECK-NEXT: ret void
; ;
; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_l35 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_l35
; CHECK-DISABLED-SAME: () #[[ATTR0]] { ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
; 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: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true) ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
; 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_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_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)*
▲ Show 20 Lines • Show All 264 Lines • ▼ Show 20 Lines
; CHECK-NEXT: ret void ; CHECK-NEXT: ret void
; ;
; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_guarded_l50 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_sequential_loop_to_shared_var_guarded_l50
; CHECK-DISABLED-SAME: () #[[ATTR0]] { ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
; 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: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true) ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
; 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_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_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)*
▲ Show 20 Lines • Show All 305 Lines • ▼ Show 20 Lines
; CHECK-NEXT: ret void ; CHECK-NEXT: ret void
; ;
; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65 ; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_14_a34ca11_do_not_spmdize_target_l65
; CHECK-DISABLED-SAME: () #[[ATTR0]] { ; CHECK-DISABLED-SAME: () #[[ATTR0]] {
; 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: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 ; CHECK-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
; CHECK-DISABLED-NEXT: store i32 0, i32* [[DOTZERO_ADDR]], align 4
; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true) ; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1]], i1 false, i1 false, i1 true)
; 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_spmd(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) ; CHECK-DISABLED-NEXT: [[WORKER_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)*
▲ Show 20 Lines • Show All 155 Lines • Show Last 20 Lines