Overall, my performance is still not as good as Source1, but it is much closer now. Here is my instruments stack of me sitting in the fountain.Ĭongrats, the Ambient Occlusion and Specular Lighting are both much faster than they were on my Nvidia 670! Maybe they perform better under linux/windows than OS X?
My ignorant guess would be that source2 is expecting better performance out of these thread locks from all of the threads that it is using. The biggest chunk of time is CThreadSyncObject:Wait(unsigned int) call where it performs the necessary calls to _pthread_mutex_lock, _pthread_mutex_unlock. Where the OpenGL rendering thread only consumes about 11.5% the main source 2 client thread only consumes about 6.4%. It seems as if there are a lot of threads that are switching priority every frame (switch_pri) and consuming about 36.4% of my CPU process. I did some Instruments sleuthing on it and it appears that there might be some significant thread waiting/switching occurring that is bogging down the FPS. Looking at the CPU in Activity Monitor, makes me believe that it is CPU limited as my quad-core CPU is grinding at 500-600%.
I've tried V-Sync, window-type, resolution and it seems that they are still all suffering from the same CPU consumption. Can confirm that fps is much lower than in source 1 and reducing the graphics only has a marginal improvement in FPS.
0 kommentar(er)
