I know it's not super detailed at this point, but now is not a good
time for JS to expend a lot of effort forming detailed plans.
* trace recursion
Real code hits this, and it will also help our benchmarketing.
* improved ARM performance
It looks like we could do a lot to improve the code generated for this
* cleaner, safer code -> moving from jschar* buffers to c++ vectors,
JS now has C++ facilities available that will improve the readability
and safety of our code.
* nanojit compilation speed improvements
Some shortcomings in the nanojit design have become apparent as we've
gotten more experience with it. We can fix them to improve the
performance of the trace compilation phase.
* GC scalability improvements
We should do better in situations where lots of garbage is generated,
such as animations etc.
* ES 3.1 major features in testing, including Object.defineProperty et
al, strict mode
This standard looks ready to implement, finally.
* improve interpreter performance
We should continue to improve performance where we fall off trace, by
generating the interpreter loop as LIR.
We'll be able to really flesh these out at the upcoming work weeks.
Where would jsd2 fit in these plans?