Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optionally avoid incref/decref immortality checks on 3.12+ #1044

Open
JukkaL opened this issue Dec 30, 2023 · 0 comments
Open

Optionally avoid incref/decref immortality checks on 3.12+ #1044

JukkaL opened this issue Dec 30, 2023 · 0 comments
Labels

Comments

@JukkaL
Copy link
Collaborator

JukkaL commented Dec 30, 2023

Python 3.12 added support for immortal objects (python/cpython#19474). These add immortality checks to incref/decref operations, which add some overhead. Some programs don't get any advantage from immortal objects, and these could benefit from skipping these checks (enabled with a mypyc optimization flag). Skipping the checks seems a safe thing to do, based on this comment in Include/object.h in Python 3.12.0:

In 64+ bit systems, an object will be marked as immortal by setting all of the
lower 32 bits of the reference count field, which is equal to: 0xFFFFFFFF

Using the lower 32 bits makes the value backwards compatible by allowing
C-Extensions without the updated checks in Py_INCREF and Py_DECREF to safely
increase and decrease the objects reference count. The object would lose its
immortality, but the execution would still be correct.

Reference count increases will use saturated arithmetic, taking advantage of
having all the lower 32 bits set, which will avoid the reference count to go
beyond the refcount limit. Immortality checks for reference count decreases will
be done by checking the bit sign flag in the lower 32 bits.

Immortality checks in the Python runtime, stdlib and C extensions will still be present, but if most time is spent in code compiled using mypyc, skipping the checks could improve performance somewhat. In mypy self check I saw a 1.9% performance improvement from skipping them. I wouldn't be surprised if some use cases would get a 5-10% performance improvement. The self check improvement is small enough that it doesn't seem essential to use this in mypy wheels.

@JukkaL JukkaL added the speed label Dec 30, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant