Friday saw primarily continued work on existing tasks. Saturday will probably be the big news day, with a bunch of projects going into the trunk together. However, a bit of news for Friday.
The new exception work continued with Richard and Georg doing all the heavy lifting, and I was running the tests and tracking down the test failures for them to resolve. It nearly went in before we left, but the Windows portion of the code required significant work. The testing and development happened on Mac and Linux, so that's not surprising. It'll definitely go in to the trunk in good time on Saturday.
Continued work with the strings has produced fantastic results. Byte strings seem to have been speed up by over 3x, according to the string benchmark. Unicode strings have seen over a 4-fold performance increase. Good going Andrew and Fredrik.
Tim Peters improved performance of string to integer conversion quite dramatically. I'm not sure that happened yesterday though, I may have missed it one of the other days.
John Benediktsson has been working on speeding up various operations under Psyco, and has gotten huge performance improvements. One of the smallest speed improvements was 10 times. "sum(range(100))" was sped up over 500 times.
Georg did some testing and was surprised that GCC 3.4.6 to 4.1 gives around a 7% speedup. With a major and minor change in there, I'm not surprised. Version 4 was a pretty big change as I recall.
Over all, Pybench is now showing around 5% performance improvement from 2.4.3 to the trunk. Remember, that this is around 15% better than where we started from the trunk.
Tune in tomorrow for the final results.comments powered by Disqus