I hava a question：
I read the oneDNN code， that seams cblas_gemm_s8s8s32（） is not implemented，just cblas_gemm_s8u8s32，why？
Because ISA of intel （AVX2？）has not special instructions that can execute multiplying or adding operation when the two vectors have the same data type (either s8/s8 or u8/u8)？
Yes, mkl provides cblas_gemm_s8u8s32 API only as there are no specific instructions to perform s8s8s32 operations. The latest CPU contains vpmaddubsw instruction which supports only s8u8s32.