So I’ve been using cudaEventElapsedTime to determine the framerate of an application I am writing. Under my development at work on a WindowXP 32 bit system (NVCC 3.0, Driver Version 197.68 32bit), I get valid results for framerate as I get reasonable values returned to the “*ms” variable which you pass as a parameter.
Recently however, I wanted to try the code on my home machine (Windows 7 x64, NVCC 3.1, Driver Version 258.96 x64 & 260.61 x64) as it has a faster processor, better video card, etc… But now I’m always getting a return of 0.0030 (milliseconds) from the call to cudaEventElapsedTime, which corresponds to a frame rate of about 300,000… The code is good, but not THAT good…
So I’m wondering if anyone else has had similar issues, and how they might have resolved them.
Thanks,
Razgriz