But if PGO counts for a 20% performance increase, then that's not everything that's wrong with the Linux build. Windows Firefox is 100% faster than Linux. Of course, Linux Firefox is also 80% better than Windows IE, but what else is new...
The only other thing I can find fault with is that the test obviously only tests Javascript performance. Page rendering performance, and general UI responsiveness is very important for user experience, but also difficult to test I guess.
Core 2 Duo @ 3.33GHz with 6 gigs of RAM, but compiling everything *twice* and also doing the other necessary stuff for PGO (running and testing the programs before recompiling them again) is *not* my idea of fun.
I gotta agree with that. PGO is a great optimization but it's also more complex than just adding an extra CFLAG. To invest that extra work it should be for something where those 10-20% of extra performance really makes a difference for you.