Skip to content

Draft: freedreno: inline drm

Rob Clark requested to merge robclark/mesa:wip/inline-drm into main

Something I've been meaning to experiment with for a while.. but according to piglit drawoverhead, it doesn't seem terribly helpful. On 845, pinned to big cores, avg 0.3% faster, pinned to little cores 1.9% slower (?!)

Interestingly, no-op'ing the actual submit ioctl made the 'no state change' tests ~20-25% faster (I do have CONFIG_LOCK_DEBUGGING_SUPPORT=y but do not have CONFIG_PROVE_LOCKING enabled..)

Merge request reports