Message10087

So XML and text processing is slow. 2to3 is likely especially slow because of heavy use of regexes in tokenization, and certain regexes have been problematic in Jython's port of sre.

On the other hand, we are somewhat faster on more algorithmic benchmarks, such as fannkuch. We should also pay extra attention to min times here, since that shows JIT warmup. nqueens shows that we are more or less the same performance in our implementation of generators.

Note that I did not run this with --rigorous, but instead --fast, given that I wanted to see some immediate results. As we work on performance improvements, we can look at this in more detail, especially on a per benchmark basis.