diff options
| author | Jongsoo Park <jongsoo@fb.com> | 2019-03-13 06:14:32 +0300 |
|---|---|---|
| committer | Facebook Github Bot <facebook-github-bot@users.noreply.github.com> | 2019-03-13 06:17:49 +0300 |
| commit | 6011ce3b0c1fccee549e85b37e475c7a734ad742 (patch) | |
| tree | 7089177b6c7da36c2582da1cf9b42eca9dfb2ea7 /include/fbgemm/QuantUtilsAvx2.h | |
| parent | 50b43162fd1742122d01f2704945c78f13e0d73e (diff) | |
optimize requantize for float out processing (#85)
Summary:
Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/85
Optimizing performance of output processing when output is dequantized right away.
Reviewed By: protonu
Differential Revision: D14433141
fbshipit-source-id: f99a8d82000c43e554461acf036462a4e8f7e300
Diffstat (limited to 'include/fbgemm/QuantUtilsAvx2.h')
| -rw-r--r-- | include/fbgemm/QuantUtilsAvx2.h | 14 |
1 files changed, 14 insertions, 0 deletions
diff --git a/include/fbgemm/QuantUtilsAvx2.h b/include/fbgemm/QuantUtilsAvx2.h index 04aeba1..47f33a8 100644 --- a/include/fbgemm/QuantUtilsAvx2.h +++ b/include/fbgemm/QuantUtilsAvx2.h @@ -95,4 +95,18 @@ FBGEMM_API void requantizeOutputProcessingGConvAvx2( int ld_in, const requantizationParams_t& r); +template < + bool A_SYMMETRIC, + bool B_SYMMETRIC, + QuantizationGranularity Q_GRAN, + bool HAS_BIAS, + bool FUSE_RELU> +FBGEMM_API void requantizeForFloatAvx2( + float* out, + const std::int32_t* inp, + const block_type_t& block, + int ld_out, + int ld_in, + const requantizationForFloatParams_t& r); + } // namespace fbgemm |