projects / platform / upstream / pytorch.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 4c74cf7) | patch
use acc16 only when n>128 and k>128 in Skylake (#18672)
authorJongsoo Park <jongsoo@fb.com>
Mon, 1 Apr 2019 15:49:37 +0000 (08:49 -0700)
committerFacebook Github Bot <facebook-github-bot@users.noreply.github.com>
Mon, 1 Apr 2019 15:52:28 +0000 (08:52 -0700)
commit822c8ee143a6f8ca78219daa71fe4ad0a9223e21
treeb2ebe49b03406140eb7492cf92bd2aaebf11c19btree | snapshot
parent4c74cf74897d0fb05fe0060af6825eb779bd6f5ccommit | diff
use acc16 only when n>128 and k>128 in Skylake (#18672)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18672

In Skylake, when n < 128 or k < 128, acc16 is slower.

Reviewed By: jianyuh

Differential Revision: D14700576

fbshipit-source-id: 80ca9f1af4626637eed9c5ca49f95ae744811189
caffe2/quantization/server/conv_dnnlowp_acc16_op.cc diff | blob | history
Domain: Machine Learning / ML Framework;