This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author rhettinger
Recipients christian.heimes, rhettinger, serhiy.storchaka, vstinner
Date 2013-08-17.22:41:01
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <>
I just made the patch a few minutes ago.  Am just starting to work on benchmarks.  I posted here so that you guys could help find the strengths and weaknesses of the approach.  

The theory is sound, but it is going to take a good deal of effort to isolate the effects (either good or bad) in realistic benchmarks.  The short, tight loops in timeit tend to put everything in cache and hide the effects  of good memory access patterns.

If would love to have you all team up with me on this one.  I could use help on timings, examining disassembly, and runs with cache grind.

This is just in the experimental phase, so it is premature to be throwing up obstacles with "show me your data" or "what does it do in situatation x".  At this point, you all could help out by running some timings designed to isolate the effects on high collision data big enough to not fit in  L2 cache.  I hope you assist in evaluating the idea with an open mind.
Date User Action Args
2013-08-17 22:41:01rhettingersetrecipients: + rhettinger, vstinner, christian.heimes, serhiy.storchaka
2013-08-17 22:41:01rhettingersetmessageid: <>
2013-08-17 22:41:01rhettingerlinkissue18771 messages
2013-08-17 22:41:01rhettingercreate