Given that timing I suspect the trie optimization was not used. If the regexp is too long, perl gives up on the trie optimization and resorts to standard alternation. I ran into this a while ago and split the patterns into a set of regexps that where small enough to do the optimization. I then ran that set one after the other. The regexp debug output will tell you whether you are actually getting the tries.
in reply to Re: Efficient matching with accompanying data
in thread Efficient matching with accompanying data