News for hackers. Stuff that matters.
Re:Our Plan for Python 3.13 by vb-8448 on Thursday June 15, 2023 @09:52AM (#36340522)
> If no-GIL has a 10% single thread performance hit, that means that essentially all my existing python code would be that much worse.

Maybe in a 100% CPU bound code, most of the code is I/O bound and no one will notice the change, just my opinion.

Re:Our Plan for Python 3.13 by wongarsu on Thursday June 15, 2023 @10:16AM (#36340863)
Maybe that's just my bubble, but I see much more python in data science projects than in web servers. And in (python) data science even your file reading/writing code quickly gets CPU bound.
Re:Our Plan for Python 3.13 by baq on Thursday June 15, 2023 @10:58AM (#36341427)
There was a time like 5-10 years ago where Python was really popular for grassroots web projects. Nowadays this is mostly node looks like.
Re:Our Plan for Python 3.13 by coldtea on Thursday June 15, 2023 @11:27AM (#36341908)
Not in pure Python, that's in specialized libs, like numpy, pandas and co, done in C.

So, the hit on the Python interpreter wouldn't translate to a hit on those.

Re:Our Plan for Python 3.13 by regularfry on Thursday June 15, 2023 @11:29AM (#36341937)
That's going to be CPU-bound in numpy's C extensions rather than Python itself, one would hope. The worst of all worlds is that we get a 10% perf cut to python execution and numpy breaks because the C API is ripped up.
Re:Our Plan for Python 3.13 by gpderetta on Thursday June 15, 2023 @12:03PM (#36342511)
That's because you are doing it wrong. You'll need to split every step of your data science pipeline into a microservice, then put it in the cloud for resilience. Then the application will be so fast that it is no longer CPU bound but I/O bound.
Re:Our Plan for Python 3.13 by burnished on Thursday June 15, 2023 @10:06AM (#36340708)
It might be more helpful to think of it in terms of supported use cases, rather than just pure volume.
Re:Our Plan for Python 3.13 by celeritascelery on Thursday June 15, 2023 @11:52AM (#36342275)
Maybe, but if your code is I/O bound, then multi-threading isn't going to help you either.
Re:Our Plan for Python 3.13 by umanwizard on Thursday June 15, 2023 @05:50PM (#36347070)
But the GIL doesn't need to be held in I/O-bound code anyway, so why does it matter?