Add AVX512 s/dgemm optimizations for compute kernel (2nd try)

aaraujomrequested to merge
aaraujom/eigen:aaraujom/gemm-avx512-opts into master

This is a follow up to resolve issues that pop up after !860 (merged) was merged but reverted.

It addresses the following:

  • Build issue on NEON64
  • Rename/move of transpose for trsm
  • Rename member in data mapper to include to something other than incr to avoid shadowing

I'm keeping the commits separate for now to help with readability. I will squash all of them on final rebase.

Edited by aaraujom

Merge request reports