Top oneDNN issues

How does int8 winograd convolution use avx512_vnni instructions?

question
Gxllii picture Gxllii·17 Sep 2020·6 Comments

Cmake build target for compat_libs does not look correct in some context

bug build
albanD picture albanD·4 Jun 2020·9 Comments

MKLDNN using TBB

build question
uu-praveeng picture uu-praveeng·11 Nov 2019·2 Comments

Low performance of LSTM training forward and backward with some input values and weights

not a bug
benjamim93 picture benjamim93·25 Sep 2019·5 Comments

non-gemm_driver extended_sgemm --> wrong results

bug
kruus picture kruus·18 Jul 2019·9 Comments

Question: Is convolution with signed 8bit source supported?

question
SafeteeWoW picture SafeteeWoW·3 Jan 2019·9 Comments

Add 'vzeroupper' at the end of primitives?

bug enhancement
CaoZhongZ picture CaoZhongZ·29 May 2018·9 Comments

[Question] calculate batchnorm backward using forward output instead of forward input?

question
eric-haibin-lin picture eric-haibin-lin·21 May 2018·6 Comments

relu_backward with y as input instead of x

question
eric-haibin-lin picture eric-haibin-lin·7 May 2018·6 Comments

support for filter dilation?

question
tonywang1990 picture tonywang1990·24 Mar 2018·5 Comments
 Previous issues
Next issues