Sunday, 19 October 2014

Timing CUDA kernels

Are you up to benchmarking your CUDA application? Are you looking for the easiest possible way on earth to time your kernels? Then, you're at the right place because in this post we're going through some code snippets to help you out. The implementation I present today is inspired by MATLAB's tic-toc simple interface. You start the timer with a tic() and get the elapsed time with a toc(). This saves a lot of lines of code and keeps you code simpler and tidy.

The latest version of the code can be found on github.

No comments:

Post a Comment