Message12027

I think this requires proper Java profiling to be answered. Such investigation is a long standing todo for me. Anyway, we should move performance discussion to a separate issue.
The new results are comparable to what I observed with SymPy and are the reason why I carefully question every bit of additional performance regression. At least until we understand this better. Jython should be faster than CPython as it as a JIT, no? I.e. something is wrong and we will find out what it is :)