Message390769
| Author |
rhettinger |
| Recipients |
corona10, miss-islington, pablogsal, petr.viktorin, phsilva, rhettinger, serhiy.storchaka, shihai1991, vstinner |
| Date |
2021年04月11日.05:08:38 |
| SpamBayes Score |
-1.0 |
| Marked as misclassified |
Yes |
| Message-id |
<1618117719.18.0.473869674983.issue40137@roundup.psfhosted.org> |
| In-reply-to |
| Content |
FWIW, I've only looked at the lru_cache() code. Someone should go through all the PyModule_GetState calls to see if they are on a critical path. AFAICT a huge number of these changes were applied without regard to whether or not they occurred on an existing fast path.
This is a considerable number of clock cycles for a type lookup that used to be almost cost free. Whether it is high or low impact greatly depends on whether the surrounding code was doing a lot of work or very little work. In thin functions, the incremental costs will be higher. |
|