Representative benchmarks for string and memory routines

[David Gilbert]

Hi,
  Does anyone have good suggestions for benchmarks that actually spend
a useful amount of
time in the str* and mem* routines that typically get hand optimised?
I've been looking at
optimising the likes of memchr and strlen etc and have some very pretty graphs
of little artificial benchmarks but I'd like to know of things that
actually matter in real code.

  SPEC (2006) doesn't seem to spend long in the mem* and str*
routines; it looks like git might
actually spend a reasonable amount of time - any other offers?

Dave