On Sat, Jan 13, 2018 at 11:11:35AM -0800, Jason Ekstrand wrote: > Sorry for all the list spam, but I'm sort of thinking out-loud and writing > it on the list for all to read. > > I'm thinking that what we want this list to return is not a bool but an enum > > /* The ordering of this enum is important */ > enum anv_fast_clear_support { > ANV_FAST_CLEAR_NONE = 0, > ANV_FAST_CLEAR_ZERO_ONLY = 1, > ANV_FAST_CLEAR_ANY = 2, > }; > > And then the predicate for whether or not to do the resolve becomes > (has_compression && !supports_compression) || (image_fast_clear > > supported_fast_clear). I still haven't quite figured out what to do with > MI_PREDICATE to make this work out. I'm sure it's possible with MI math > but maybe I can come up with something clever that doesn't require that. > In the worst case, we only have to deal with this complexity on gen9+ where > we have MI_MATH so it'll be ok. > > Does this seem like a reasonable plan?
Sounds good to me at least. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev