On Tue, Feb 2, 2016 at 8:37 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote: > On Tue, Feb 2, 2016 at 2:28 PM, Marek Olšák <mar...@gmail.com> wrote: >> On Tue, Feb 2, 2016 at 8:16 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote: >>> Thanks! From discussion with Dave re r600 needs, it sounds like we >>> need to throw a "writable" property into pipe_image_view, and I need >>> to do a proper job of setting the WR flag on the image in patch 6. >>> I'll do those as add-on patches. >>> >>> Did you leave patch 6 unreviewed because you had issues with it, or >>> because it was long and touched glsl_to_tgsi? >> >> Mostly a lack of focus today and not remembering the small details of >> glsl_to_tgsi. > > OK, just making sure there were no explicit issues. > >> >> I don't think r600 needs changes as long as it can analyze TGSI for >> store instructions. > > Well, Dave was saying that he might want to use texturing for > read-only images, in which case it'd be convenient to know at > set_shader_images() time as well (or, well when writing that state > out). And you'd want to generate different shader code depending on > its read-only-ness. > > Anyways, I'm perfectly happy to let you and Dave fight it out :)
Well, the fact is, we don't have to use samplers at all. GL_MAX_COMBINED_IMAGE_UNITS_AND_FRAGMENT_OUTPUTS assures that we won't get too many images + fragment outputs, thus we won't have to worry about such things. Marek _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev