diff options
author | Jianyu Huang <jianyuhuang@fb.com> | 2019-08-06 19:35:42 +0300 |
---|---|---|
committer | Facebook Github Bot <facebook-github-bot@users.noreply.github.com> | 2019-08-06 19:50:51 +0300 |
commit | d8b3323668fdd15dc70e9cb43ab16e96f4846eeb (patch) | |
tree | d48a6818c14575d92e68bf1ffb621d646a6c893e /include/fbgemm/Utils.h | |
parent | 0d5d057ca941ebb511bdc6178fc26c23e6c4a953 (diff) |
Integrate VNNI into FBGEMM master branch (#113)
Summary:
Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/113
Adding the VNNI support in FBGEMM.
Reviewed By: dskhudia
Differential Revision: D16276574
fbshipit-source-id: 832ccdb27339489ebc138f3b2678e53d107c1b79
Diffstat (limited to 'include/fbgemm/Utils.h')
-rw-r--r-- | include/fbgemm/Utils.h | 7 |
1 files changed, 6 insertions, 1 deletions
diff --git a/include/fbgemm/Utils.h b/include/fbgemm/Utils.h index 107cf07..3f8522b 100644 --- a/include/fbgemm/Utils.h +++ b/include/fbgemm/Utils.h @@ -29,7 +29,7 @@ enum class matrix_op_t { NoTranspose, Transpose }; /** * @brief Typed enum for supported instruction sets. */ -enum class inst_set_t { anyarch, avx2, avx512 }; +enum class inst_set_t { anyarch, avx2, avx512, avx512_vnni }; /** * @brief Typed enum for optimized paths for convolutions @@ -100,6 +100,11 @@ FBGEMM_API bool fbgemmHasAvx512Support(); FBGEMM_API bool fbgemmHasAvx2Support(); /** + * @brief Are we running on a AVX512_VNNI supported cpu? + */ +FBGEMM_API bool fbgemmHasAvx512VnniSupport(); + +/** * @brief Helper struct to enable autotuning of FBGEMM packing and kernels. * * This structure is optional. If not used, the default values for these |