On the PowerBook G4, from a fresh restart, and just for the record, TenFourFox 10.0.2 Sunspider benchmarks at 1811.6ms. Safari 4.1.3, by contrast, benchmarks at 6036.6ms. Tests on a 12" PowerBook, 1.5GHz machine, 1.25MB RAM, running Tiger with all updates.
Results on my dual G5 2.3GHz were less dramatic: 1311 in 10.0.1 and 1072 in 10.0.2 (Safari 5.0.6 is 3305).
I also have a G4 Cube (dual 500MHz 7400). On it, I can't report Sunspider scores for 10.0.1, as it was never installed on the machine, but 10.0.2 gives me 3918.1ms as opposed to Safari 4.1.3's 13566.5ms. So that makes it competitive with the dual G5 running Safari, running at nearly 5x the clock speed.
But I am just delighted to have really dramatic improvement on the G4 7450 in my Powerbook, which I do want to keep going, and which has been struggling somewhat on the web.
Someone can write good code around here — and do so way better than Apple's own engineers at this level.
Any idea why the Collusion extension doesn't work correctly? More specifically, it works but all the bubbles are pushed off to the left side of the screen. It's almost as if there is a coordinate translation issue (screen origin in center vs. side).
Thanks, that tells me it's either in the new branching code or in the sqrt library. Since branching would probably cause other errors, sqrt is the likely suspect. I appreciate the detective work. Filed as issue 134.