I'm not positive on the specifics of Benchmark, but since the first two (child_grepping and child_perlizing) start separate processes, it would make sense that their CPU time wouldn't be recorded by Benchmark. That's why child_grepping appears to be so much faster than the others...child_perlizing take so long (even though it starts a child) because it starts a separate perl interpreter for each interation, which is horrendously inefficient.

I think what you should take away from this is that standard regexes are the most efficient way to match stuff in general, I only use grep in certain situations where it lends itself to simpler code.

Enjoy

Trinary