Skip to content

frontends/va, radeonsi/vcn: Skip clear on surface creation

David Rosca requested to merge nowrep/mesa:va-clears into main

What does this MR do and why?

amdgpu now clears VRAM on allocation, so we can skip clearing surfaces on creation. Also stop clearing decode internal buffers, as FW will clear them if needed.

With this, we can do full transcode pipeline with no gfx jobs.

Merge request reports