dgemm_batch_ do not support multi-threads on AVX512?
HWWH123 opened this issue · comments
I use follow instructions to build BLIS
./configure -t openmp auto
make -j
And compare the result between export BLIS_NUM_THREADS=16 and export BLIS_NUM_THREADS=1.
The test results did not show any performance improvement. Is my construction method wrong or BLIS does not support multi-threading of the batch interface on AVX512?