Hello, This surprisingly large series fixes bugs in the data flow algorithm for copy propagation. As far as I can tell, there were a myriad of issues. The updated algorithm seems to follow the textbook and online materials much more closely.
No Piglit regressions on Ivybridge. shader-db statistics are more or less a wash: total instructions in shared programs: 1569184 -> 1569423 (0.02%) instructions in affected programs: 44961 -> 45200 (0.53%) A bunch of things are helped by a small margin, while other things are hurt by a small margin (a couple of extra MOVs). The few hurt shaders I looked into appeared to suffer from compute-to-MRF not handling control flow. There may be other reasons as well. I'd still like to land it, though, as the old algorithm seems to be pretty broken. --Ken _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
