More CUDA performance by pushing a key !?

I am not sure if its just happening on my computer but I figured out the following:

Calling CUDA from GL limits your framerate to about 60 frames/second (like while running the fluidsGL SDK example with low detail).

Now if I run the demo and start dragging the console window across the OpenGL window, the framerate suddenly increases to 100 fps !!??

I guess NVidia uses an event-based synchronization which is accelerated by any kind of events.

In my case (I’m working on a raycaster), I can increase the framerate simply by keeping a key pushed. Then it jumps from 30 to 45 frames per second.

Does anybody already had a problem like this ?
Is there any solution like creating a timer that continuously sends windows events?

Its continued from my previous post - but as the topic is very different now, I created a post which is more general.

You didn’t state which OS you’re using, however if you’re getting exactly 60 FPS, then it sounds like you have sync-to-vblank turned on in your display configuration. This has no relationship to CUDA.

Ah, I just found the solution - it wasnt a CUDA issue

(even my timeGetTime benchmark made it look like this)

It was simply that the swapintervall had to be set in the idle function :)

void idle()





Sorry for bothering everybody - but I hope it can also be helpful.