Skip to content

Fix AVX256/512 loop_mask to handle general loops

Andre Gunther requested to merge gunther_fixAVX256+512LoopMask into master

The MR fixes a crash currently occurring in skylake_avx512+vecwid256-centos7-gcc9-opt build where in the Forward Tracking loop sizes are in general not a multiple of 8 or 16.

cc @ahennequ

Merge request reports