Making Python 5x FASTER with Guido van Rossum and Mark Shannon - Talk Python To Me Ep.339

[@pradhvanbisht2177]

I watched it last night live. Thanks for organizing it and putting in all the effort. The live streams are definitely are my favorite in a way wherein the live chat you pick up questions and ask the guest. Makes the whole podcast turned into a live stream more engaging for us as the viewer.

[@mikeckennedy]

Thanks, I'm glad you're enjoying the live stream. I do like how it opens another dimension of the experience and let's you all have a voice in the episode. :)

[@pradhvanbisht2177]

@@mikeckennedy Indeed. I try to catch as many live streams as I can of Talk Python as well as of Python Bytes just so I can try to interact with the guest and you folks. Which in the previous years would be only me listening with no interaction.

[@nitin.cherian]

Thanks, Micheal. Glad to see the faces of people who are taking part in the podcast. Really like the prompt display of the relevant web pages/docs as the conversation takes place.

[@mikeckennedy]

Thanks Nitin!

[@alexanderscott2456]

Michael, I've been looking forward to this interview all week.Thank you so much for the content. Keep up the excellent work.

[@mikeckennedy]

Thanks Alexander, I appreciate it.

[@vOVERip]

Awsome! Thank you so much for this interview and in generall for the whole podcast. I'm a big fan! I'm sorry for my english, but hope you get the intention. :)

[@mikeckennedy]

Came through great, thanks so much!

[@a4e69636b]

Can't wait for the JIT compiler. Thankfully we already have Numba.

[@talkpython]

Indeed. There is a lot of room for improvement it sounds like. Happy Mark and Guido are working on it, as well as others.

[@blueman333]

Numba still requires lots of restrictions which takes the fun out of python. Recently I wrote some code which was not so fast because of large for loops. So I thought I would just plug in JIT and have good results. But complicated classes along with third party libraries, I couldn't make it any faster. Neither with numba nor with drop in JIT compiler in python 3.10

[@MagnusAnand]

What a fantastic talk

[@mikeckennedy]

Thanks Guz!

[@PaulJacobson]

Thank you for this one. I really enjoyed this.

[@mikeckennedy]

Thanks Paul, glad you liked it. :)

[@George-lt6jy]

I think this is a great idea, there should be more experimentation with CPython like this. The reference implementation has so much influence on the direction of the language. I love hearing that there is interest in improving in this direction and I look forward to see how these interesting ideas play out. From my point of view, I think they are approaching their changes in a responsible way. Maybe, down the line we can see some creative things.

[@talkpython]

Thanks George. Definitely agree. There's a lot happening around performance all of a sudden. We have this work with Guido and Mark. There is Sam Gross's work. There's Anthony Shaw's work. And the core devs are compiling cpython down to web assembly. I'm sure something cool will come from all of this and other initiatives.

[@planetarylizard1297]

Great topic, great guests!

[@muratahmetgenc6942]

it's a very long video, how do we do it, in short. cython?

[@shalokshalom]

Codon is an alternative compiler, that is still WAY faster than Python 3.11

[@ORood]

Nice 👍

[@hamkaprastika2251]

I dоwnloaded everything is okay

[@AxiomPrime69]

I would like to see Python foundation transform into a DAO built on Algorand. I think governance run by dev contributors and tokenomics could provide more resources for the foundation.

[@talkpython]

Non-profits have issues, but I think a DAO on Algorand might destabilize it more than help.

[@kyuss0x1]

nogil python is more interesting idea

[@mikeckennedy]

Hey, yes, Sam's nogil idea is interesting. Two things: 1. This has been going for over a year, Sam's nogil work was announced last week. 2. They are multiplicative. If Guido and Mark make Python 5x faster, as they say. And Sam makes Python 95% scalable for "embarrassingly parallel" problems, combining them, say on a 10 core machine would make it 48x faster. That is more interesting than both 5x and 10x (each project alone).

[@kyuss0x1]

boring

[@mikeckennedy]

If you were bored, then maybe this isn't for you then. Also, try to think of the hard work that people put into things before you insult them. You could be way less of a jerk if you said, "this was boring because x, y, and z, but would have been amazing if you did u & v." But no, you're not a builder. You're not a creator. You're just a low-life who complains and doesn't contribute.

[@kyuss0x1]

still boring

[@ChuckVenter]

Not helpful

[@biskitpagla]

this is a certified "zoomer who just started learning programming with python" moment