➜

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 methane
Recipients BTaskaya, bmerry, josh.r, methane, pitrou, remi.lapeyre, serhiy.storchaka, vstinner
Date 2020-01-02.05:35:12
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1577943313.29.0.10051633763.issue36051@roundup.psfhosted.org>
In-reply-to
Content 
> In the single-threaded case, the benchmark seems to show that for 64K+, performance is improved by dropping the GIL (which I'm guessing must be statistical noise, since there shouldn't be anything contending for it), which is my reasoning behind the 65536 threshold.

Yes, single-threaded case shouldn't affected by this change.  Dropping GIL is just a very little overhead.

So you need to focus on multi-threaded case when considering threshold.
As far as seeing your result, 65536 threshold seems too small.
See 512*8192 case.  It shows 1479->980 regression.
But I don't think your script shows real performance.

I updated your patch, and I saw small improvement on 512K and significant improvement on 1M.
History
Date User Action Args
2020-01-02 05:35:13methanesetrecipients: + methane, pitrou, vstinner, serhiy.storchaka, josh.r, remi.lapeyre, BTaskaya, bmerry
2020-01-02 05:35:13methanesetmessageid: <1577943313.29.0.10051633763.issue36051@roundup.psfhosted.org>
2020-01-02 05:35:13methanelinkissue36051 messages
2020-01-02 05:35:12methanecreate