
Errmmm....Moz point out a flaw in his cunning plan.
Justin Dolske says... January 5, 2010 7:18 AM
It's fantastic that you're excited about the early 3.7 development work, but there are a few issues with your post that I think should be clarified...
1) There's no need to compile your own 3.7pre build -- we make nightly builds available (see http://nightly.mozilla.org/). It helps if people use these, because these builds keep themselves up-to-date automatically, and will report crash data to us (opt-in, natch) so we can catch problems early.
2) Doing your own compile for performance comparisons usually isn't a good idea, we've already spent a lot of time tweaking how things are built (eg https://bugzilla.mozilla.org/show_bug.cgi?id=409803). Your inclusion of "--enable-optimize=-O2" has, in fact, likely made your build slower than the official builds! If you're really keen on doing your own build, the steps at https://developer.mozilla.org/En/Simple_build are really all that's needed. We don't recommend people put extra stuff in their mozconfig unless there's a very specific reason for it.
3) The Electrolysis work to date has largely focused on backend work, mobile (Fennec) work, and getting plugins (eg Flash) running in a separate process. The work on, err, electrolyzing Firefox itself is just ramping up. So, unfortunately you're not really testing Electrolysis yet.
4) Electrolysis itself won't really help with things like Sunspider -- speedups there are due to improvements in the JS engine itself. Electrolysis is more about stability and security, and performance when running multiple tasks on a multicore box.
5) As other comments noted, multiprocess != multithreaded. Firefox is already multithreaded (eg, top reports 18 threads in my particular case), and web sites can already make use of that with web workers (see http://hacks.mozilla.org/2009/07/working-smarter-not-harder/).
In any case, your basic point is correct. There's a lot of exciting things in the Firefox pipeline, and 2010 is going to be a very awesome year. :-)
FAIL