Elapsed Time in Emulation Mode

I’m working in strictly emulation mode for the time being while waiting on cuda hardware and I need to see the execution time of each cuda kernel and also the cuda fft routine in my code. I have tried using the cudaEventElapsedTime() method, but no matter what I try it simply returns 0.0 in time. Does anyone know how to get the elapsed time while in emulation mode? Also, the time would need to be from the GPU standpoint instead of simply the CPU it’s being executed on.

Thanks in advance.

Uh, elapsed time in emulation has no real meaning…

So there is no way to emulate the algorithm implemented in cuda and see if the total execution time if it were on hardware? The requirements for the algorithm to work is the worst case kernel function needs to execute in less than 100 ns. The only way to test to see if it is working correctly is to buy the hardware?

In theory something like Barra: http://gpgpu.univ-perp.fr/index.php/Barra could provide an answer, but I would call that “dicey” at best.

Thanks, I’m going to give it a try.