Re: A microbenchmarking library

[mratsim]

100 cycles is the cost of a cache miss so allocating then comparing 2 strings 
for 100 cycles seems pretty reasonable.