The tests were run 10 times, and I used the median value. There wasn't much variance between runs, so I don't think disk caching played much of a role here.

Being a garbage collected language with a runtime, Go certainly cannot match the performance of C, and it was never my point to prove otherwise - obviously, for the same algorithm, the C implementation would be faster. Instead, I was exploring Go to highlight how simple it is to write safe, concurrent code in it.