On Fri, Oct 16, 2015 at 05:34:55PM +0100, Tvrtko Ursulin wrote: > I have a slight unknown relating to how long would this ABI be > useful. If things are moving towards SVM, and the fact pre-gen8 > platforms can already use get_aperture, would that make it a bit > short lived? Optimist. It is certainly a better fit than the cascade of get param and hardcoded sizes based on gen (which ignore differences between kernels) which a fallback to get_aperture. Now although I can't just drop the old code (because I need to support kernels without this interface), it does make that code simpler and more robust against any future change. > And get_aperture would even be a better place for this if PPGTT size > will always be the same for all clients. Hmm, I'm regretting my earlier push for extending get_aperture. Constants like this would be better off in GET_PARAM (or CONTEXT_PER_PARAM as applicable), and only using get_aperture where we want the side-effect of scanning for available space (e.g. the extra stolen information would be more useful in get_aperture as we there do want to know the largest object we can allocate given current usage, but we may also want to stick the total amount of stolen in a param for convenience). [snip] -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
