I'm not sure the conclusion that vector instructions are responsible for the speed-up is correct. Both implementations seem to use ERMS (using REP MOVSB instructions)[0]. Looking at the profiles, the syscall implementation spends time in the [xfs] driver (even in the long test), while the mmap implementation does not. It appears the real speed-up is related to how memory-mapped pages interact with the buffer cache.

I might be misunderstanding things. What is really going on here?

[0] Lines 56 and 180 here: http://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_...