Swift vs Python vs Eiffel

[Liberty Lover]

I wonder what it would look like in Eiffel on the metal? 80x faster than Python?

88,419 views Apr 5, 2026

Apple Silicon's unified memory architecture is arguably the best hardware for data science workloads, but almost nobody in the data science world takes it seriously. The reason? Python. Python is the lingua franca of data science, but it's an interpreted, single-threaded language that can't take advantage of Apple's multi-core hardware. Running Python on a Mac is like buying a Lamborghini and never leaving first gear. In this video, I make the case that Apple's real barrier isn't hardware: it's the missing library ecosystem. Swift, as a compiled language with native access to Metal GPU acceleration, can unlock performance that Python simply can't match. To prove it, I built a direct comparison: Swift with Metal acceleration was more than 80x faster than Python on the same analytical workload. In CFO terms, a three-day Python job finishes in one hour. If Swift gets even a fraction of Python's library coverage running natively on Apple Silicon, it changes the economics of data processing entirely: local hardware instead of cloud compute, real-time analytics instead of overnight batch jobs, and deterministic processing instead of unpredictable agents. The hardware is ready. The language is ready. The ecosystem just needs to be built.

https://youtu.be/6ZGlseSqar0?si=lsp30UWVk8L57D4j

[Ian Joyner]

Hi Larry. Do you mean metal as in hardware or Metal as in the Apple product to run on the GPU.

Alas Eiffel on Mac is a bit of a sore point, and Apple went Swift, which leaves me somewhat cold.

Ian

Sent from my iPad

On 8 Aug 2026, at 8:27 am, Liberty Lover <rix....@gmail.com> wrote:

I wonder what it would look like in Eiffel on the metal? 80x faster than Python?

88,419 views Apr 5, 2026

Apple Silicon's unified memory architecture is arguably the best hardware for data science workloads, but almost nobody in the data science world takes it seriously. The reason? Python. Python is the lingua franca of data science, but it's an interpreted, single-threaded language that can't take advantage of Apple's multi-core hardware. Running Python on a Mac is like buying a Lamborghini and never leaving first gear. In this video, I make the case that Apple's real barrier isn't hardware: it's the missing library ecosystem. Swift, as a compiled language with native access to Metal GPU acceleration, can unlock performance that Python simply can't match. To prove it, I built a direct comparison: Swift with Metal acceleration was more than 80x faster than Python on the same analytical workload. In CFO terms, a three-day Python job finishes in one hour. If Swift gets even a fraction of Python's library coverage running natively on Apple Silicon, it changes the economics of data processing entirely: local hardware instead of cloud compute, real-time analytics instead of overnight batch jobs, and deterministic processing instead of unpredictable agents. The hardware is ready. The language is ready. The ecosystem just needs to be built.

https://youtu.be/6ZGlseSqar0?si=lsp30UWVk8L57D4j

--
You received this message because you are subscribed to the Google Groups "Eiffel Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to eiffel-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/eiffel-users/7f6b5afd-0c9b-4860-af4b-c29f9b1b8060n%40googlegroups.com.

[Liberty Lover]

I copied that directly from the YouTube video description. So, I am not sure what THEY mean by Metal.

[Ian Joyner]

Yes,  if they are talking about Swift, they’d mean Metal for direct GPU programming. I have never read up about Metal, but I think it was Apple’s response to games developers moaning and complain that they could not control hardware directly, but Apple provided APIs to do so (with the code the games programmers would have written for themselves).

I can see two words badly used from C-think. First ‘metal’ where the word is bandied about to say “C is close to the metal”, meaningless and it really isn’t, and you don’t want to be anyway, except for very rare hardware-oriented programming, for which I propose very small languages doing what both C and assemblers do, but with the hardware semantics in the language and in the compiler rather than in defines or inline assembler. This is rather the Burroughs approach and still the best.

Secondly, what I have observed is misunderstanding of the word and concept of ‘abstraction’ , which they interpret to mean software layers, since Apple’s API approach is called an ‘abstraction layer’. This detracts from the real meaning of abstraction, which is just a way of thinking about things (what all of computing is about).

(Sorry for the rave… not really!)

Ian

Sent from my iPad

On 9 Aug 2026, at 1:54 am, Liberty Lover <rix....@gmail.com> wrote:

I copied that directly from the YouTube video description. So, I am not sure what THEY mean by Metal.

To view this discussion visit https://groups.google.com/d/msgid/eiffel-users/1688ce5b-d3b5-4862-a10f-8b700e4cd292n%40googlegroups.com.