Use common sync code for radv
!13427 (merged) adoption in radv.
Known AIs:
-
get a drm fd from libdrm_amdgpu -
Need to add sparse binding support to the submission implementation as it looks like a submit wrt semaphore/fence behavior. -
check the performance impact of not caching the temp syncobj for binary sync reset. -
check the performance impact of resetting the syncobj one by one instead of as a batch (syscall ...) -
might want to keep our faster queue wait idle, the kernel keeps our own