This is an archive of the discontinued LLVM Phabricator instance.

Differential D112194

[OpenMP] Remove restriction on the thread count for parallel regions
ClosedPublic

Authored by jdoerfert on Oct 20 2021, 6:45 PM.

Details

Reviewers
tianshilei1992
jhuber6
Commits
rG0153ab6dbc33: [OpenMP] Remove restriction on the thread count for parallel regions
Summary

This was taken from the old runtime but it is not, or should not, be
necessary in the new one.

Diff Detail

Event Timeline

jdoerfert created this revision.Oct 20 2021, 6:45 PM
Herald added subscribers: guansong, bollu, yaxunl. · View Herald TranscriptOct 20 2021, 6:45 PM
jdoerfert requested review of this revision.Oct 20 2021, 6:45 PM
Herald added a project: Restricted Project. · View Herald TranscriptOct 20 2021, 6:45 PM
Herald added a subscriber: sstefan1. · View Herald Transcript
Harbormaster completed remote builds in B129860: Diff 381127.Oct 20 2021, 6:51 PM
jhuber6 accepted this revision.Oct 20 2021, 6:59 PM

LG

This revision is now accepted and ready to land.Oct 20 2021, 6:59 PM
Closed by commit rG0153ab6dbc33: [OpenMP] Remove restriction on the thread count for parallel regions (authored by jdoerfert). · Explain WhyMar 21 2023, 7:17 PM
This revision was automatically updated to reflect the committed changes.
jdoerfert added a commit: rG0153ab6dbc33: [OpenMP] Remove restriction on the thread count for parallel regions.
Herald added a project: Restricted Project. · View Herald TranscriptMar 21 2023, 7:17 PM
Herald added subscribers: jplehr, sunshaoce. · View Herald Transcript

Revision Contents

PathSize
openmp/
libomptarget/
DeviceRTL/
src/
6 lines

Diff 381127

openmp/libomptarget/DeviceRTL/src/Parallelism.cpp

Show First 20 Lines • Show All 48 Lines • ▼ Show 20 Lines
uint32_t determineNumberOfThreads(int32_t NumThreadsClause) { uint32_t determineNumberOfThreads(int32_t NumThreadsClause) {
uint32_t NThreadsICV = uint32_t NThreadsICV =
NumThreadsClause != -1 ? NumThreadsClause : icv::NThreads; NumThreadsClause != -1 ? NumThreadsClause : icv::NThreads;
uint32_t NumThreads = mapping::getBlockSize(); uint32_t NumThreads = mapping::getBlockSize();
if (NThreadsICV != 0 && NThreadsICV < NumThreads) if (NThreadsICV != 0 && NThreadsICV < NumThreads)
NumThreads = NThreadsICV; NumThreads = NThreadsICV;
// Round down to a multiple of WARPSIZE since it is legal to do so in OpenMP.
if (NumThreads < mapping::getWarpSize())
NumThreads = 1;
else
NumThreads = (NumThreads & ~((uint32_t)mapping::getWarpSize() - 1));
return NumThreads; return NumThreads;
} }
// Invoke an outlined parallel function unwrapping arguments (up to 32). // Invoke an outlined parallel function unwrapping arguments (up to 32).
void invokeMicrotask(int32_t global_tid, int32_t bound_tid, void *fn, void invokeMicrotask(int32_t global_tid, int32_t bound_tid, void *fn,
void **args, int64_t nargs) { void **args, int64_t nargs) {
switch (nargs) { switch (nargs) {
#include "generated_microtask_cases.gen" #include "generated_microtask_cases.gen"
▲ Show 20 Lines • Show All 140 Lines • Show Last 20 Lines