diff options
| author | Omar Al Khatib <omar.alkhatib@arm.com> | 2024-05-09 16:06:23 +0100 |
|---|---|---|
| committer | Omar Al Khatib <omar.alkhatib@arm.com> | 2024-05-17 14:20:06 +0000 |
| commit | f5053f782daa942126bd61ac1bcfc0af627b7b31 (patch) | |
| tree | eb96c568d3c95021860e06f06271345f68969260 /src/runtime/OMP/OMPScheduler.cpp | |
| parent | 8710385a9feb050f1b5a422ed57df691e8ba078f (diff) | |
| download | ComputeLibrary-main.tar.gz | |
On systems with BIG/MID/LITTLE cores, we need to exclude the LITTLE cores.
This is make changes to CPUInfo to detect number of LITTLE cores and set the num_threads to TOTAL_CORES-NUM_LITTLE cores
Resolves [COMPMID-7014]
Signed-off-by: Omar Al Khatib <omar.alkhatib@arm.com>
Change-Id: I3e1772e5b64d1c45304860be43233b7e5dd8dba1
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11565
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/OMP/OMPScheduler.cpp')
| -rw-r--r-- | src/runtime/OMP/OMPScheduler.cpp | 28 |
1 files changed, 25 insertions, 3 deletions
diff --git a/src/runtime/OMP/OMPScheduler.cpp b/src/runtime/OMP/OMPScheduler.cpp index d4d6193fce..2a5abb5f7a 100644 --- a/src/runtime/OMP/OMPScheduler.cpp +++ b/src/runtime/OMP/OMPScheduler.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2023 Arm Limited. + * Copyright (c) 2017-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -32,10 +32,24 @@ namespace arm_compute { +#if !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__) OMPScheduler::OMPScheduler() // NOLINT - : _num_threads(omp_get_max_threads()) + : _num_threads(cpu_info().get_cpu_num_excluding_little()), + _has_lmb(cpu_info().cpu_has_little_mid_big()), + _nonlittle_num_cpus(cpu_info().get_cpu_num_excluding_little()) { } +#else /* !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__)*/ +OMPScheduler::OMPScheduler() // NOLINT + : _num_threads(omp_get_max_threads()), + _has_lmb(cpu_info().cpu_has_little_mid_big()), + _nonlittle_num_cpus(cpu_info().get_cpu_num_excluding_little()) +{ +} +#endif /* !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__)*/ unsigned int OMPScheduler::num_threads() const { @@ -45,7 +59,15 @@ unsigned int OMPScheduler::num_threads() const void OMPScheduler::set_num_threads(unsigned int num_threads) { const unsigned int num_cores = omp_get_max_threads(); - _num_threads = (num_threads == 0) ? num_cores : num_threads; +#if !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__) + const unsigned int adjusted_num_threads = (_has_lmb) ? _nonlittle_num_cpus : num_threads; + _num_threads = (num_threads == 0) ? num_cores : adjusted_num_threads; +#else /* !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__)*/ + _num_threads = (num_threads == 0) ? num_cores : num_threads; +#endif /* !defined(_WIN64) && !defined(BARE_METAL) && !defined(__APPLE__) && !defined(__OpenBSD__) && \ + (defined(__arm__) || defined(__aarch64__)) && defined(__ANDROID__)*/ } void OMPScheduler::schedule(ICPPKernel *kernel, const Hints &hints) |