Article - How fast is your programming language

[djholiver]

Hi,

I’m sure some will find this interesting.

The last paragraph: "Python remains one of the two most used scripting languages despite its poor performance,” the developer reflects. “Will there be a language to displace Python in the next decade? I am not optimistic.” " and lack of Julia representation is “upsetting” if nothing else.

Regards,

[stevengj]

(I set a time limit on this thread because general questions like “when will Python get replaced” otherwise cycle around endlessly.)

[Dan]

This is not a “replace Python” thread. I think this is a “language comparison benchmark” thread. The correct link would be:

GitHub - attractivechaos/plb2: A programming language benchmark

And it does include Julia. Unfortunately Julia not given a lot of love.

[jar1]

Imho the “Conclusions” section of that benchmark were written too early, before code review from all the language experts. We’ve already seen huge speedups from simple changes like fixing the loop order to match the array layout, and more changes are in the pipeline.

[DNF]

I’m a bit confused. Julia was there in the plot, with performance results basically tied with the other top performers

[Benny]

I was wondering how Pypy got into the top ranks and my hunch was right, no classes. The benchmarks didn’t need composite types that much I suppose. Still impressive how far JITs have come, though.

[Palli]

I get 1.95 sec. something like 24% speedup for the matmul if I run in the REPL (and add @simd), partial timing here:

@time a = matgen(n)
0.005693 seconds (2 allocations: 17.166 MiB)

@time c = matmul(n, a, b);
  1.932954 seconds (2 allocations: 17.166 MiB)

So it’s worth it to AOT compile at least that one. And the for loops need to be split in two, i.e. dual-for loop doesn’t take @simd but maybe should?

And I’m unclear why to I get 2, not one allocation each?