What do you mean exactly? If you want to compare the performance of different implementations, you first need to create a representative test set. Some implementations are very sensitive to memory cache size and you might not realize this with only small data sets.
Once you have a test set, you should automate the running of these tests and the time recording so it's easy to run tests between code modifications. If you are benchmarking in order to optimize your code, using a code profiler is the way to go. Believe me, you will find bottlenecks you would never have imagined!