We need to choose some benchmarks to measure bytecode compiler speed. Profiling gmail should yield some ideas; sayrer says bz has some thoughts on this.
My main thought is that I see jsparse showing up as taking more time than the script execution in many cases. gmail it's not quite that bad, but the Tp4 version of www.sohu.com is (on that page jsparse is 3% of the total pageload; for comparison reflow is 12%).
You need to log in before you can comment on or make changes to this bug.