Author vstinner
Recipients Dennis Sweeney, Zeturic, ammar2, josh.r, pmpp, serhiy.storchaka, tim.peters, vstinner
Date 2020-10-13.07:13:44
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1602573224.78.0.9016562552.issue41972@roundup.psfhosted.org>
In-reply-to
Content
random_bench.py has really weird timings. Example:

(...)
Run 15: 1 warmup, 3 values, 2^15 loops
- warmup 1: 3.05 us (+504%)
- value 1: 42.8 ns (-92%)
- value 2: 1.18 us (+133%)
- value 3: 293 ns (-42%)
(...)
Run 21: 1 warmup, 3 values, 2^15 loops
- warmup 1: 167 us (+1073%)
- value 1: 162 ns (-99%)
- value 2: 39.1 us (+174%)
- value 3: 3.53 us (-75%)
needle=3: Mean +- std dev: 17.2 us +- 39.6 us

How is it possible that in the same process (same "Run"), timing changes between 42 ns and 1 180 ns? The std dev is really large.
History
Date User Action Args
2020-10-13 07:13:44vstinnersetrecipients: + vstinner, tim.peters, pmpp, serhiy.storchaka, josh.r, ammar2, Dennis Sweeney, Zeturic
2020-10-13 07:13:44vstinnersetmessageid: <1602573224.78.0.9016562552.issue41972@roundup.psfhosted.org>
2020-10-13 07:13:44vstinnerlinkissue41972 messages
2020-10-13 07:13:44vstinnercreate