Re: Why run xserver on the OpenGL API?
Keith Whitwell <keith-CdwZJljFklH+2FeEXyspIVaTQe2KTcn/@public.gmane.org>
| Newsgroups | gmane.comp.freedesktop.xserver |
|---|---|
| Message-ID | <[email protected]> |
Keith Packard wrote: > Around 17 o'clock on Nov 25, John McCutchan wrote: > > >>Why does an opengl extension need to be written to provide the frame >>buffer access? what do you need that you can't accomplish already with >>glReadPixels () and glDrawPixels ()? > > > Yes, we can synthesize direct frame buffer access with these two functions > for an unextended GL library, but performance will be dramatically impacted > as I pull huge chunks of the buffer up into the X server, draw a few pixels > and push them all right back again. > > A simple direct frame buffer mapping and locking extension will allow us to > run all of the existing dumb frame buffer code without change, dramatically > speeding the migration to an OpenGL-based X server with no negative > performance implications. Along the way, we may be able to identify a few > kinder looking GL extensions to reduce dependence on the frame buffer > mapping extension -- there are a few X pixel operations not supported by GL > that might make good candidates (plane masks, per-component alpha, conjoint/ > disjoint compositing operators). > > If you grant that it's interesting to move the X server to the OpenGL API, > please grant that it may be worth breaking some of the GL abstractions to > make the effort more likely to succeed. Keith, Just re-iterating that I don't have an objection to this as an intermediate step. I don't see it as desirable in the long run, but to get things working in the short to medium-ish term, I don't see a problem with this. It will always be a nasty hack, though... Keith