Marek Olšák <mar...@gmail.com> writes: > The motivation is to prevent a double flush in Gallium. > > Gallium flushes in __DRI2flushExtensionRec::flush and dri2Throttle. > This will help consolidate the flushing.
It looks like you've changed a bunch of drivers to expose __DRI2_FLUSH_VERSION (4) instead of 3, without making them implement the new interface, so they'd just crash. Please continue to use explicit numbers in drivers, so that drivers don't break when the loader interface is revved without updating them. More generally, I think our loader vs driver implementation of swapbuffers is backwards. I'd rather see the driver get a swapbuffers-specific call as immediately as possible from the loader, let it handle flushing and throttling, and have it call to dri2 part of the loader for actually telling the other side to go swap the buffers. If we had that, then hopefully getting the flushes down to 1 is obvious, since the driver would do that after any other things it does (like msaa resolves) right before asking the server for the swap.
pgpc3ZEwj6Wfq.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
