Am Sonntag, den 11.06.2017, 19:21 +0200 schrieb Gert Wollny: > > > > > > > Have you measured the CPU overhead of the new code? > > So far no, I guess one would do that with the shader-db to get > reasonable complex shaders, but I only have a r600 based card so I'm > not sure whether I can run this. In any case, tomorrow I will take a > look into this.
I did runs of the shader-db/run program with valgrind/callgrind Here the original merge_registers reports 0.21% and my code reports 0.50%. If it is important to cut down on these addes 0.3%, I think, I can eliminate 0.1% by changing the implementation to not used so many dynamically allocated objects, but beyond that it will be difficult. Best, Gert _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
