On Tue, Mar 7, 2017 at 4:28 PM, Roland Scheidegger <[email protected]> wrote: > I was thinking the same thing. It should be possible for backends to do > this on their own. > That said, for the d3d10 opcodes with shadow comparison we actually have > SAMPLE_C_LZ already since d3d10 exposes that (unlike glsl), though we > don't really make any use of it (gallivm just will put in a zero as the > lod parameter in the end...). > The TEX opcodes have a bit quirky encoding due to the "packed" > arguments, so I guess you'd have to pass the target along. > It sounds to me like this just introduces gallium opcodes to paper over > a backend deficiency. Though since IIRC most (all recent?) hw actually > supports such instructions directly, why not... > (Another question would then be, does this really warrant a cap bit (the > ever-increasing cap hell...), or should it just be mandatory.)
A cap bit is necessary due to time constraints on my end, but it can go away if other drivers adopt it. Marek _______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
