pytorch
48b6b922 - [BE] Make Vec256 header only library (#50708)

Commit

4 years ago

[BE] Make Vec256 header only library (#50708) Summary: Do it by removing extraneous header dependencies. None of the at::vec256 primitive depend on notion of Tensor, therefore none of the headers that vec256 depends on should include <ATen/Tensor.h> Implicity test it be removing c10 and tensor dependency when building `vec256_test_all_types` Split affine_quantizer into affine_quantizer_base (that contains methods operating on raw types) and affine_quantizer (which contains Tensor specific methods) Fixes https://github.com/pytorch/pytorch/issues/50567 Pull Request resolved: https://github.com/pytorch/pytorch/pull/50708 Reviewed By: walterddr Differential Revision: D25949168 Pulled By: malfet fbshipit-source-id: c3323be7252865a52c7d94026a5a39b494e44efb

Author

malfet

Committer

facebook-github-bot

Parents

186c3da0

Files11

aten/src/ATen
- cpu/vec256
  - vec256_base.h
  - vec256_qint.h
- native
  - Math.h
  - cpu
    - RangeFactoriesKernel.cpp
  - quantized
    - affine_quantizer.cpp
    - affine_quantizer.h
    - affine_quantizer_base.cpp
    - affine_quantizer_base.h
    - cpu
      - int_repr_quant.cpp
      - make_per_tensor_quantized_tensor.cpp
caffe2
- CMakeLists.txt

pytorch 48b6b922 - [BE] Make Vec256 header only library (#50708)

pytorch
48b6b922 - [BE] Make Vec256 header only library (#50708)