https://bugs.freedesktop.org/show_bug.cgi?id=30007 --- Comment #28 from Fredrik Höglund <fredrik@xxxxxxx> 2010-09-13 16:39:13 PDT --- (In reply to comment #25) > Fredrik, thanks for the feedback. I've pushed your patch to the piglit > repository. Thanks. > (In reply to comment #24) > > Oops, I submitted before I was done. This shader would be better, because the > > loop could be unrolled even if kernelSize is different in each execution: > > Yeah, the shader could work provided the instruction scheduler generates no > more than 4 TEX indirections. I'd personally lean towards making kernelSize a > compile-time constant. So instead of calling glUniform1f(kernelSize, i), let's > call glUseProgram(shader[i]); > > Fredrik, please, would it be hard to implement this in kwin? It shouldn't be too hard, and it's actually something I've been planning to do but never got around to before the KDE 4.5 release. In the meantime I've committed a change that gets rid of the kernelSize uniform and makes the loop always do a fixed number of iterations instead (with the unused kernel weights set to zero). The MESA_GLSL=dump output shows that the loop is being unrolled, so with a bit of luck the shader should be working now. (In reply to comment #26) > Another option for kwin would be using the shader only on GL3-capable hardware, > because lots of GL2 hardware don't support indirect indexing. Working well with the Mesa drivers and the hardware our users are using is something I consider a high priority, so if we can make the shader work without indirect indexing then I think that's the preferred approach. Thanks to you and Tom for all the suggestions. -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug. _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/dri-devel
