PyPy: Faster Python with Minimal Effort

#1

Post by endlesstrax » Wed, Nov 11, 2020, 2:18 PM UTC

PyPy: Faster Python with Minimal Effort
realpython.com

Re: PyPy: Faster Python with Minimal Effort

#2

Post by est » Wed, Nov 11, 2020, 3:15 PM UTC

Easier pypy with minimal effort

https://github.com/fijal/jitpy

Re: PyPy: Faster Python with Minimal Effort

#3

Post by davelpz » Wed, Nov 11, 2020, 3:21 PM UTC

est wrote: post #2

Easier pypy with minimal effort https://github.com/fijal/jitpy

That project hasn't been updated in 6 years

Re: PyPy: Faster Python with Minimal Effort

#4

Post by blackrobot » Wed, Nov 11, 2020, 3:28 PM UTC

est wrote: post #2

Easier pypy with minimal effort https://github.com/fijal/jitpy

The author of the repo says to just use pypy. https://github.com/fijal/jitpy/issues/7#issuecomment-3335573...

Re: PyPy: Faster Python with Minimal Effort

#5

Post by alquemist » Wed, Nov 11, 2020, 3:29 PM UTC

est wrote: post #2

Easier pypy with minimal effort https://github.com/fijal/jitpy

This is amazing, and obvious in retrospect. Gradual compilation can be adopted, well, gradually, without breaking existing code. 100% upside, 0% downside. PyPy team should consider making this delivery mode the #1 priority. Even worth reconsidering the object management impedance as well, and go for 100% compatibility with CPython object model. I suspect this would have a small runtime cost, but a would be a huge boon for smooth risk-free adoption.

Re: PyPy: Faster Python with Minimal Effort

#6

Post by ed25519FUUU » Wed, Nov 11, 2020, 3:54 PM UTC

It’s probably a testament to the type of software I typically develop with Python, but I’ve never felt the urge to use PyPy. Even if it’s faster with zero effort, it’s still different.

Any time performance has been a concern that application was in Java or Go (for me).

Re: PyPy: Faster Python with Minimal Effort

#7

Post by est » Wed, Nov 11, 2020, 5:08 PM UTC

alquemist wrote: post #5

est wrote: post #2

Easier pypy with minimal effort https://github.com/fijal/jitpy

This is amazing, and obvious in retrospect. Gradual compilation can be adopted, well, gradually, without breaking existing code. 100% upside, 0% downside. PyPy team should consider making this delivery mode the #1 priority. Even worth reconsidering the object management impedance as well, and go for 100% compatibility with CPython object model. I suspect this would have a small runtime cost, but a would be a huge boo…

Back in Python2 days, you just slam "import psyco" and your code magically goes faster. The pysco authors went ahead and invented pypy.

Ambitious project, but less popularity.

Re: PyPy: Faster Python with Minimal Effort

#8

Post by carapace » Wed, Nov 11, 2020, 5:19 PM UTC

(Pypy also supports Python 2 and will for the foreseeable future.)

Re: PyPy: Faster Python with Minimal Effort

#9

Post by sockpuppet_12 » Wed, Nov 11, 2020, 7:50 PM UTC

Why’d they name it PyPy, it sounds phonetically exactly like PyPi. aaaahhh this is going to be so confusing

Re: PyPy: Faster Python with Minimal Effort

#10

Post by moralsupply » Wed, Nov 11, 2020, 8:02 PM UTC

How compatible is it nowadays with Numpy, Scipy and the ML toolkits?