Hacker News Reader

Comment by biomcgary | Hacker News Reader

Python performance, both single core and concurrent, seem to be its primary Achilles heel. However, Python is unmatched for developer speed in compute heavy AI / ML since it uses C, etc., under the hood. I don't really like Python but it's impossible to avoid in data science because you don't have to re-invent the wheel.

Several languages have persisted beyond their natural shelf life due to their library ecosystems. For a period of time, CPAN was one of the best arguments for using Perl. More recently, Java's ecosystem is also a pretty good reason to use Java right now, even if you don't like the language.

Languages seem to thrive early due to initial ease of use (Python, particularly, but also relative uptake of Go vs Rust), but can wane due to a thousand cuts (or sigils in Perl's case).

View on HN

> However, Python is unmatched for developer speed in compute heavy AI / ML since it uses C, etc., under the hood.

John Carmack seems to think otherwise.

> "That means that in the time that Python can perform a single FLOP, an A100 could have chewed through 9.75 million FLOPS"

https://twitter.com/id_aa_carmack/status/1503844580474687493...

Am I missing something?

The OP's comment is that while Python's native interpreter is slow, Python code as a whole can still be fast since people will delegate the hot spots to non-Python libraries optimized for speed.

To refute that, you linked to John Carmack quoting Horace He saying that Python's native interpreter does CPU-based math slower than a GPU. But...that's why people writing Python use libraries to delegate math-intensive work to a GPU, which is the OP's point.

If one is forced into writing code that needs to be fast in a language other than Python, then I'm not sure I understand why Python gets to claim any credit?

Slight nitpick: the credit that was being claimed is that Python allows fast developer speed while not sacrificing overall runtime performance.

The code that is written in a language other than Python is often not written by the developer writing the Python code.

As an example, I've been fiddling with inferring automatic data extraction rules for data in scraped webpages.

The meat of the algorithm and its heuristics is in Python. The heavy lifting of parsing HTML and evaluating CSS selectors is done by a C parser that I did not write. For me, prototyping the heuristics in C would have been quite a bit slower.

I generally respect Carmack, but it's clear he wasn't considering Jax or even numpy, but just python flops. Jax code running on TPUs often reaches order of magnitude of the hardware's capabilities.

Reply on news.ycombinator.com