On Thu, 2019-10-17 at 20:55 -0500, Jason Ekstrand wrote: > On Thu, Oct 17, 2019 at 10:39 AM Erik Faye-Lund < > erik.faye-l...@collabora.com> wrote: > > This is discussed in the merge request thread. Zink currently only > > support vertex and fragment shaders, so it's the only place this > > can occur. If someone wants to enable this for drivers that > > supports geometry or tesselation shaders, they would need to extend > > this code first. Unless I beat them to it, of course. I don't want > > to implement this blindly, which is why I left this out for now. > > > > If you hook the pass up and get it working on iris, you wouldn't be > implementing it blind. :-P >
Not the worst idea. I gave it a quick stab, but coulnt't even get the vertex-shader only path working on Iris. I'm probably breaking some expectations in the way the lowering-pass is called or something, but I couldn't figure it out quickly, and kinda lost my motivation. Here's what I did so far: https://gitlab.freedesktop.org/kusma/mesa/tree/iris-ucp-lowering I noticed that if I also call nir_shader_gather_info after the lowering (like the current iris code does), everything seems to get clipped instead... And if I do both nir_lower_global_vars_to_local and nir_lower_vars_to_ssa, I trigger this assert: nir_lower_var_copies.c:89: emit_deref_copy_load_store: Assertion `glsl_type_is_vector_or_scalar(dst_deref->type)' failed. So... I dunno. There's probably something wrong here somewhere ;) _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
