On Fri, Dec 31, 2010 at 09:31:23AM +0000, Chris Wilson wrote: > I'm still leaning towards explicit associate/disassociate ioctls as I > think that will lead to a simpler userspace API (or at least integrate > conveniently with the existing libdrm).
I don't like a separate ioctl to associate/disassociate a buffer from a context: Conceptually there's no way to change that association without an execbuf call. Doing the book-keeping as a separate, non-atomic ioctl calls calls for trouble imho. > ... > "Minimal, consistent and difficult to misuse." i.e. violates the "difficult to misuse" paradigm ;) My idea of a special reloc type is probably not the cleanest (hm, slight understatement here ;) and could need some improvement. Anyway, I'll wait and see what the real use-case looks like. My arguments feel too hand-waivy already ... Cheers, Daniel -- Daniel Vetter Mail: [email protected] Mobile: +41 (0)79 365 57 48 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
