Skip to content

Count granularity2

James Stevens requested to merge jdsteve2/loopy:count-granularity2 into master

These changes add a counting granularity parameter to MemAccess and Op that specifies whether they should be counted per thread, per warp, or per group. For example, uniform (stride-0) global memory accesses should be counted per warp.

Edited by James Stevens

Merge request reports