https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/cblas-gemm-s8s8s32-not-support/m-p/1164045#M28111 <P>Dear sir，</P><P>I hava a question：</P><P>I read the oneDNN code， that seams&nbsp;cblas_gemm_s8s8s32（） is not implemented，just cblas_gemm_s8u8s32，why？</P><P>Because ISA of intel （AVX2？）has not special instructions that can execute multiplying or adding&nbsp; operation&nbsp;when the two vectors have the same data type (either s8/s8 or u8/u8)？</P><P>&nbsp;</P><P>Thank you！</P> Thu, 16 Apr 2020 08:12:01 GMT qiang__zhang 2020-04-16T08:12:01Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/cblas-gemm-s8s8s32-not-support/m-p/1164045#M28111 <P>Dear sir，</P><P>I hava a question：</P><P>I read the oneDNN code， that seams&nbsp;cblas_gemm_s8s8s32（） is not implemented，just cblas_gemm_s8u8s32，why？</P><P>Because ISA of intel （AVX2？）has not special instructions that can execute multiplying or adding&nbsp; operation&nbsp;when the two vectors have the same data type (either s8/s8 or u8/u8)？</P><P>&nbsp;</P><P>Thank you！</P> Thu, 16 Apr 2020 08:12:01 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/cblas-gemm-s8s8s32-not-support/m-p/1164045#M28111 qiang__zhang 2020-04-16T08:12:01Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/cblas-gemm-s8s8s32-not-support/m-p/1164046#M28112 <P>Yes, mkl provides&nbsp;cblas_gemm_<STRONG>s8u8s32</STRONG> API only as there are no specific instructions to perform s8s8s32 operations. The latest CPU contains&nbsp;vpmaddubsw instruction which supports only s8u8s32.</P><P>&nbsp;</P><P>&nbsp;</P><P>&nbsp;</P> Fri, 17 Apr 2020 04:10:24 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/cblas-gemm-s8s8s32-not-support/m-p/1164046#M28112 Gennady_F_Intel 2020-04-17T04:10:24Z