Message 391560 - Python tracker

➜

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	vstinner
Recipients	corona10, erlendaasland, miss-islington, pablogsal, petr.viktorin, phsilva, rhettinger, serhiy.storchaka, shihai1991, vstinner
Date	2021-04-21.23:27:27
SpamBayes Score	-1.0
Marked as misclassified	Yes
Message-id	<1619047647.53.0.541144988327.issue40137@roundup.psfhosted.org>
In-reply-to

Content
msg391004 benchmark comparing master with Raymond's optimization (commit 139c232f3851b393798d0ea4e65f1298bfbcd9cf) compared to Python 3.9: Mean +- std dev: [py39] 38.8 ns +- 0.5 ns -> [obj_cache] 40.3 ns +- 0.1 ns: 1.04x slower For me, these numbers make no sense :-D Getting kwd_mark from the instance (Raymond's new code) must be faster than calling _PyType_GetModuleByDef() on the instance. As I wrote in my previous comment, my bet is that PGO is not reproducible and so we should not pay too much attention to differences of a few nanoseconds.

msg391004 benchmark comparing master with Raymond's optimization (commit 139c232f3851b393798d0ea4e65f1298bfbcd9cf) compared to Python 3.9:

Mean +- std dev: [py39] 38.8 ns +- 0.5 ns -> [obj_cache] 40.3 ns +- 0.1 ns: 1.04x slower

For me, these numbers make no sense :-D Getting kwd_mark from the instance (Raymond's new code) *must* be faster than calling _PyType_GetModuleByDef() on the instance.

As I wrote in my previous comment, my bet is that PGO is not reproducible and so we should not pay too much attention to differences of a few nanoseconds.

History
Date	User	Action	Args
2021-04-21 23:27:27	vstinner	set	recipients: + vstinner, rhettinger, phsilva, petr.viktorin, serhiy.storchaka, corona10, pablogsal, miss-islington, shihai1991, erlendaasland
2021-04-21 23:27:27	vstinner	set	messageid: <1619047647.53.0.541144988327.issue40137@roundup.psfhosted.org>
2021-04-21 23:27:27	vstinner	link	issue40137 messages
2021-04-21 23:27:27	vstinner	create