Hey, I used cuda events to measure kernel execution time (and to make sure kernels were running concurrently). The concurrent kernel example is at http://wiki.tiker.net/PyCuda/Examples/KernelConcurrency. Hope that helps.
Jesse On Mon, Jun 13, 2011 at 3:10 AM, Francis <fccaba...@gmail.com> wrote: > Hi guys and gals, > > I'm wondering how I can measure (and afterwards study and improve) the > execution time of the kernel function/s in the device as well as the host > functions/code using PyCUDA? Any suggestions? > Right now I'm just using the *datetime* function of Python and then > subtracting the time before and after the kernel calls. In our Intel Xeon > setup this is about several hundred microseconds, but I want to be sure if > this is an accurate way of getting the run time. :) > > Regards, > > ./francis > > _______________________________________________ > PyCUDA mailing list > PyCUDA@tiker.net > http://lists.tiker.net/listinfo/pycuda > >
_______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda