If you're building an assembler, and the syntax is line based, you'll get a speed-up by parsing each line individually, rather than as one big file.

The reason is that P::RD performs the equivalent of:

s/^$syntax_I_recognize// && do { &Action_for_syntax };
[download]

And it does this for everything, over and over again. This "nibbling" approach is great for short strings, but for long strings, there's a whole lot of string copying going on.

theDamian said that a significant speedup would have been realized by changing these to:

/\G$syntax_I_recognize/gc && do { &Action_for_syntax };
[download]

and letting pos() walk through the string, but alas never got the time to change P::RD into Parse::FastDecent (this would have been the basic change).

-- Randal L. Schwartz, Perl hacker
Be sure to read my standard disclaimer if this is a reply.