Still, iterating over a 10_000_000 element list more than 160 thousand times a second isn't too shabby a performance...

In fact, it's not shabby enough. That sort of speed should have made you take another look at your benchmark.

Try changing your cmpthese() call to

cmpthese( -5, {
                for  => sub { by_for(\%h) }, 
                grep => sub { by_grep(\%h) },
              }
);
[download]

-sauoq
"My two cents aren't worth a dime.";