In a less trivial example...

I was benchmarking a module I wrote against one I found on CPAN solving the same problem. I decided to see how using 5.10 compared to using 5.8 at the same time.

My module (which was already faster) saw a 10% improvement with 5.10. The CPAN module saw a negligible improvement (about 1%) using 5.10.

using 5.10 didn't hurt and in some circumstances added a nice boost. Depends on how you are using perl.

Maybe later I'll dig a little deeper and see what the CPAN module did differently to get so little benefit out of 5.10.