Skip to content

Refactoring for AVX512

Dominic Kempf requested to merge feature/refactore-vector-width-treatment into master

This will be a mixed bag of things to bring us closer to KNL vectorization

  • Options that are set through other options
  • No explicit mentioning of vector width anymore
  • Saner way of passing the vectorization info back after dry run
  • Splitting of Theta matrices for the 2*4 = 8 strategy

Merge request reports