some loop {
   for my $regexp ( regexps ) {
      /$regexp/
   }
}
[download]

is a very bad pattern unless the regexps are precompiled. In your case, the regexps are not precompiled since you're using a hash key (a string) as the regexp.

In the first program, you could even flatten all of %ignore into a single regexp if you didn't need to do $ignore{$regexp}++.

The performance of the regexp engine depends highly on the regexp in question, but you didn't show us the regexps.

Since you only case about one value returned by split, a regexp should be faster than split.