Pyston 2.1 Is Blowing Past Python 3.8/3.9 Performance

[bug77]

Excuse my ignorance, but does it perform any form of profile-driven optimization? Wikipedia mentions that it uses a "tracing JIT", which seems to imply some level of feedback.

Iirc, "tracing JIT" happens when the code is being watched/traced, so that only the more frequently accessed paths are actually JIT-ed.

[Anarchy]

I didn’t say it could replace it tomorrow, i said they should starts preparing a cpython replacement with pypy (and complete it with what’s needed for this).

Realistically, it might be better to just fork the language itself and improve the parts that make Python "slow"; pypy developers have already done some work in this department. Although even that is not optimal because if you need more performance from python, then you're using it wrong -- go and learn c/c++/rust/fortran/java/go.

[Adarion]

I was impressed at first, since e.g. portage (Gentoo) depends on python (well and a lot of other packages).
But then, in the end the article mentioned that it is not freedom software (too bad it's not open source). Uh. Seriously, could have saved me some time if that was in the beginning. :/

[hajj_3]

Julia seem to be popular for data analysis, machine learning etc: https://julialang.org/

[Kemosabe]

Julia seem to be popular for data analysis, machine learning etc: https://julialang.org/

Yeah, and as soon as you use halfway complex libraries, it takes forever to even start!

[onlyLinuxLuvUBack]

Realistically, it might be better to just fork the language itself and improve the parts that make Python "slow"; pypy developers have already done some work in this department. Although even that is not optimal because if you need more performance from python, then you're using it wrong -- go and learn c/c++/rust/fortran/java/go.

what about nim ?

[yanestra]

The Python developer community has often enough decided against the better code and instead for the easier maintainability. E.g. in case of Stackless Python.

[baryluk]

There are low-hanging fruits in the CPython that could make it perform better. It just isn't a main priority to CPython developers. Hopefully, a bit of external competition will make them revisit this and make some improvements, i.e. in memory allocator, hash tables, and other tricks. JIT in CPython is not going to happen for a really long time, because of how it can sometimes make performance worse, and because how much it would complicate code, and testing required (especially on various architectures).