On Tue, Apr 1, 2014 at 4:45 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote: > On Tue, Apr 1, 2014 at 10:14 AM, Marek Olšák <mar...@gmail.com> wrote: >> 1) Adding functions to cso_context only makes sense for features which >> need to be disabled or changed for meta ops (e.g. clear_with_quad, >> glBitmap, glDrawPixels). I don't see you using >> cso_save_sample_shading, which clearly shows it's useless. Does sample >> shading need to be set differently for meta ops? > > I don't think so, I'll remove the CSO stuff. I was going off of sample_mask. > >> >> 2) The changes of mesa/main should be in a separate patch. > > Based on some recent comments, I might not need them at all. (i.e. > passing in the min samples per fragment instead of the float.) > >> >> 3) We shouldn't call get_param in update_sample_shading. Either just >> check if pipe->set_sample_shading is not NULL, or store the result of >> get_param in st_context and use that instead. > > How about using ctx->Extensions.ARB_sample_shading? Otherwise we'd > have to conditionally set pipe->set_sample_shading in nv50 (only nva3+ > support the functionality). I guess that'd be feasible too. > > Or perhaps get rid of the cap entirely and predicate > ARB_sample_shading on ->set_sample_shading being set? Let me know what > you think is best.
The cap should stay. Using ctx->Extensions.ARB_sample_shading here sounds okay. Marek _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
