James Reed d4bfa96cda int8 specialization for AVX2 Quantize routine (#120) ... Summary: This adds a specialization for `int8` to the AVX2 `Quantize` routine. I tried also adding a specialization for `int32` (the final datatype we support in PyTorch quantization), but it seemed to introduce numerical issues stemming from the difference in implementations: https://github.com/pytorch/FBGEMM/blob/master/include/fbgemm/QuantUtils.h#L63 vs https://github.com/pytorch/FBGEMM/blob/master/src/QuantUtilsAvx2.cc#L82 Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/120 Reviewed By: driazati Differential Revision: D17115198 Pulled By: jamesr66a fbshipit-source-id: 119145bb99235a7545389afa61483060200cc2b7		2019-08-29 11:26:05 -07:00
..
ConvUtils.h	Unified convolution interface	2019-06-05 12:50:08 -07:00
Fbgemm.h	Per channel support in fbgemmConv (#119)	2019-08-20 16:58:08 -07:00
FbgemmBuild.h	Only export symbols that are required while building shared library	2018-11-30 11:31:33 -08:00
FbgemmFP16.h	Add unpack to PackedGemmMatrixFP16 (#112)	2019-08-08 16:17:04 -07:00
FbgemmI8DepthwiseAvx2.h	Unpack data for 3x3 (and 3x3x3) depthwise convolution	2019-07-05 18:05:46 -07:00
FbgemmI8Spmdm.h	Update with clang format (#51)	2018-12-21 11:21:05 -08:00
OutputProcessing-inl.h	optimize requantize for float out processing (#85)	2019-03-12 20:17:49 -07:00
PackingTraits-inl.h	Integrate VNNI into FBGEMM master branch (#114)	2019-08-09 11:33:13 -07:00
QuantUtils.h	Per channel and groupwise quantization (#99)	2019-06-20 12:21:51 -07:00
QuantUtilsAvx2.h	int8 specialization for AVX2 Quantize routine (#120)	2019-08-29 11:26:05 -07:00
Types.h	clang-format (#11)	2018-11-18 20:19:53 -08:00
Utils.h	fix error message (#117)	2019-08-12 10:50:35 -07:00
UtilsAvx2.h	optimize requantize for float out processing (#85)	2019-03-12 20:17:49 -07:00