If anything, that post is evidence of how tricky optimization is, and how easy it is to fool yourself about what matters. It's probably best to be skeptical about mmap() as a performance optimization over reading into a buffer unless evidence demonstrates otherwise. Most OS's do a pretty good job of caching at the filesystem level, and under the hood paging is essentially reading into a buffer anyway. mmap() might make the code simpler, but it's hard to imagine it makes it faster. If it does, I'd like to understand why.

EDIT: The post was http://geoff.greer.fm/2012/08/25/the-silver-searcher-benchma...