Perl optimizations are pretty basic, and it compiles fast - which is great for utility scripts, but sometimes it feels like a waste of potential. A lot of Perl code actually runs persistently, either behind mod_perl, plack server, mojolicious server or just as a daemon.
I've been thinking - what if we had a switch (executable flag?) that would turn on additional set of compile time optimizations? It could hypothetically be a big performance win, and not affect programs that need to boot fast. I have close to no experience in XSUB, so that's a big if.
For example, one of my projects has I think close to 100 .pm files and compiles in 800 ms. It will then run for a long time (in production) and require performance. Other than testing and development, I wouldn't mind spending 10 seconds extra if that would mean some crucial parts would run 50% faster. Or maybe I could care less about using small subroutines in tight loops, as they would get inlined.
Comments? Was it attempted already?