diff options
| author | SiCong Li <sicong.li@arm.com> | 2019-11-05 10:43:06 +0000 |
|---|---|---|
| committer | SiCong Li <sicong.li@arm.com> | 2019-11-12 09:52:55 +0000 |
| commit | 75041a1cb81c59a5a5ddd9b708476c0142362d9e (patch) | |
| tree | 5ef625879aaa35c757ee4070f753f769f4362011 /examples/gemm_tuner/README.md | |
| parent | 77d42528b796f3b8f5033785d3bbb8d9cb3fc637 (diff) | |
| download | ComputeLibrary-75041a1cb81c59a5a5ddd9b708476c0142362d9e.tar.gz | |
COMPMID-2563 Change how the best overall GEMM configuration is selected
* Based on a specified tolerance, each GEMMParam (GEMM Shape) now can
have a set of best GEMM configurations, instead of just a single one.
This improves the robustness and completeness of the tuned results, and
is in compliance with how we define the GEMMParam archetypes (the main
goal of this story)
* The tuner then tries to find the best overall GEMMConfig, from all the
best config sets, through the same voting mechanism: the config that
gets voted the most is the best overall GEMMConfig.
Change-Id: Ief770bb6ffc04629d91f1dc778eea69274e007f0
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2228
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Diffstat (limited to 'examples/gemm_tuner/README.md')
| -rw-r--r-- | examples/gemm_tuner/README.md | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/examples/gemm_tuner/README.md b/examples/gemm_tuner/README.md index 3d5b6376e1..e8cb7e5193 100644 --- a/examples/gemm_tuner/README.md +++ b/examples/gemm_tuner/README.md @@ -25,4 +25,5 @@ python >= 3.6. The script will output the best configuration, along with some analysis statistics for each strategy, and optionally save the parsed benchmark results into csv files (one for each strategy) for further analysis. -python GemmTuner.py -b \<benchmark_results_dir\> [-o \<out_dir\>] +python GemmTuner.py -b \<benchmark_results_dir\> [-t \<tolerance\>] +[-o \<out_dir\>] |