I built a 2x faster lexer, then discovered I/O was the real bottleneck (modulovalue.com)

>Memory-mapping the files made things worse due to the per-file mmap/munmap overhead

Could you elaborate on this?