Magic-Mirror/llvm-capstone
1
0
Fork 0
mirror of https://github.com/capstone-engine/llvm-capstone.git synced 2024-10-07 10:54:01 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

[OpenMP][FIX] Move workaround code to avoid races

Browse Source 
The workaround code ensure we always call __kmpc_kernel_parallel, but it
did so in a racy manner as the initialization might not have been
completed yet. To avoid introducing a sync, we move the workaround into
the deinit function for now.
This commit is contained in:
Johannes Doerfert 2023-10-23 13:38:12 -07:00
parent 1f5a9d1d07
commit 0012b956f9
1 changed files with 17 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
openmp/libomptarget/DeviceRTL/src/Kernel.cpp
View File

@ -110,20 +110,8 @@ int32_t __kmpc_target_init(KernelEnvironmentTy &KernelEnvironment) {
// main thread's warp, so none of its threads can ever be active worker
// threads.
if (UseGenericStateMachine &&
mapping::getThreadIdInBlock() < mapping::getMaxTeamThreads(IsSPMD)) {
mapping::getThreadIdInBlock() < mapping::getMaxTeamThreads(IsSPMD))
genericStateMachine(KernelEnvironment.Ident);
} else {
// Retrieve the work function just to ensure we always call
// __kmpc_kernel_parallel even if a custom state machine is used.
// TODO: this is not super pretty. The problem is we create the call to
// __kmpc_kernel_parallel in the openmp-opt pass but while we optimize it is
// not there yet. Thus, we assume we never reach it from
// __kmpc_target_deinit. That allows us to remove the store in there to
// ParallelRegionFn, which leads to bad results later on.
ParallelRegionFnTy WorkFn = nullptr;
__kmpc_kernel_parallel(&WorkFn);
ASSERT(WorkFn == nullptr, nullptr);
}
return mapping::getThreadIdInBlock();
}
@ -140,8 +128,22 @@ void __kmpc_target_deinit() {
if (IsSPMD)
return;
// Signal the workers to exit the state machine and exit the kernel.
state::ParallelRegionFn = nullptr;
if (mapping::isInitialThreadInLevel0(IsSPMD)) {
// Signal the workers to exit the state machine and exit the kernel.
state::ParallelRegionFn = nullptr;
} else if (!state::getKernelEnvironment()
.Configuration.UseGenericStateMachine) {
// Retrieve the work function just to ensure we always call
// __kmpc_kernel_parallel even if a custom state machine is used.
// TODO: this is not super pretty. The problem is we create the call to
// __kmpc_kernel_parallel in the openmp-opt pass but while we optimize it
// is not there yet. Thus, we assume we never reach it from
// __kmpc_target_deinit. That allows us to remove the store in there to
// ParallelRegionFn, which leads to bad results later on.
ParallelRegionFnTy WorkFn = nullptr;
__kmpc_kernel_parallel(&WorkFn);
ASSERT(WorkFn == nullptr, nullptr);
}
}
int8_t __kmpc_is_spmd_exec_mode() { return mapping::isSPMDMode(); }