https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/Parallel-two-medium-size-GEMM/m-p/1155921#M27540 <P>Hi,&nbsp;</P> <P>i have a special use case which needs to compute two independent GEMMs.</P> <P>each one with a MNK in the range of [20~4000], on Xeon skylake 8180, only reaching 600~700 GFlops/sec.</P> <P>from the algorithm level, the two GEMMs has no denpendency, so they can be launched in parallel.</P> <P>how can i parallel these two GEMMs? say one socket for each one, perhaps. i suppose i can't use batch GEMM for this.</P> Sun, 15 Jul 2018 03:58:34 GMT Mingfei_M_Intel 2018-07-15T03:58:34Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/Parallel-two-medium-size-GEMM/m-p/1155921#M27540 <P>Hi,&nbsp;</P> <P>i have a special use case which needs to compute two independent GEMMs.</P> <P>each one with a MNK in the range of [20~4000], on Xeon skylake 8180, only reaching 600~700 GFlops/sec.</P> <P>from the algorithm level, the two GEMMs has no denpendency, so they can be launched in parallel.</P> <P>how can i parallel these two GEMMs? say one socket for each one, perhaps. i suppose i can't use batch GEMM for this.</P> Sun, 15 Jul 2018 03:58:34 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/Parallel-two-medium-size-GEMM/m-p/1155921#M27540 Mingfei_M_Intel 2018-07-15T03:58:34Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/Parallel-two-medium-size-GEMM/m-p/1155922#M27541 <P>Hi, in that case you may try to explicitly set MKL_NUM_THREADS=#of physical treads/2 and try to call gemm at the same time. You need also properly set the affinity mask to avoid threads migration:&nbsp;<SPAN style="font-size: 13.008px;">export KMP_AFFINITY=compact,1,0,granularity=fine</SPAN></P> Mon, 16 Jul 2018 02:29:34 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/Parallel-two-medium-size-GEMM/m-p/1155922#M27541 Gennady_F_Intel 2018-07-16T02:29:34Z