In general, the server your app or db is running on is not a tricked-out dual-Xeon. I've seen performance boosts of 4-10x just by moving servers. Can you provide a similar boost from 2.5 weeks of coding, especially with no new bugs and no additional maintenance burden?

I've provided 100x speed increases in perl code with just a day or two of work in the past. Never underestimate the power of the right optimisation, applied carefully.

In the OT's case, I doubt that inlining his functions will be *quite* as productive, but it is at least worth trying, and he has, apparently, done at least some basic profiling of his code. He might be able to play tricks with his source code using the C pre-processor to fake up macros. Read perldoc perlrun's warnings about it, and I suggest running the code through cpp seperately instead of using -P.