Re: [Bf-committers] Cycles benchmark as promised.
Yay, I am done. http://download.shadowrom.de/Benchmark.pdf While doing this benchmark I tried to replace the math library used with the AMD libm implementation. This was actually slower on all systems (even the amd system) so I left these results out. The more I read about MSVC the more I have the feeling that the compiler doesnt optimize math functions properly. GCC seems to be way better on that. /Jürgen -Ursprüngliche Nachricht- Von: bf-committers-boun...@blender.org [mailto:bf-committers-boun...@blender.org] Im Auftrag von Jürgen Herrmann Gesendet: Dienstag, 4. Juni 2013 21:29 An: bf-committers@blender.org Betreff: [Bf-committers] Cycles benchmark as promised. Alright I said I will do a benchmark, so here it is (partial) I am not done yet but wanted to share the first draft with you guys before I go to sleep. Ill add in another test system tomorrow and refine the visualization. You can download the pdf file here: http://download.shadowrom.de/Benchmark.pdf This is really getting interesting *g* /Jürgen ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
Re: [Bf-committers] Cycles benchmark as promised.
hi from these results it seems the fastest in cpu and gpu is mingw (with one exception for msvc). i think as a freelance 30-40% of render time reduction is a big advantage, has anyone tested the mingw openmp fix? Regards Yousef Harfoush ba...@msn.com From: shadow...@me.com To: bf-committers@blender.org Date: Wed, 5 Jun 2013 21:57:11 +0200 Subject: Re: [Bf-committers] Cycles benchmark as promised. Yay, I am done. http://download.shadowrom.de/Benchmark.pdf While doing this benchmark I tried to replace the math library used with the AMD libm implementation. This was actually slower on all systems (even the amd system) so I left these results out. The more I read about MSVC the more I have the feeling that the compiler doesn’t optimize math functions properly. GCC seems to be way better on that. /Jürgen -Ursprüngliche Nachricht- Von: bf-committers-boun...@blender.org [mailto:bf-committers-boun...@blender.org] Im Auftrag von Jürgen Herrmann Gesendet: Dienstag, 4. Juni 2013 21:29 An: bf-committers@blender.org Betreff: [Bf-committers] Cycles benchmark as promised. Alright… I said I will do a benchmark, so here it is (partial) I am not done yet but wanted to share the first draft with you guys before I go to sleep. I’ll add in another test system tomorrow and refine the visualization. You can download the pdf file here: http://download.shadowrom.de/Benchmark.pdf This is really getting interesting *g* /Jürgen ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers
[Bf-committers] Cycles benchmark as promised.
Alright I said I will do a benchmark, so here it is (partial) I am not done yet but wanted to share the first draft with you guys before I go to sleep. Ill add in another test system tomorrow and refine the visualization. You can download the pdf file here: http://download.shadowrom.de/Benchmark.pdf This is really getting interesting *g* /Jürgen ___ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers