Julia 1.11.1 is out

[Also Python 3.13.0 was just released, and it has a JIT compiler, but you need to opt into it, so might run with it only, or show both ways?]

Just letting you know so you can update, and likely and hopefully faster across the board.

EDIT: The startup-cost at least is faster now. it was faster in RC4, but there's a regression: https://github.com/JuliaLang/julia/issues/56072 Feel free to run anyway just to see, but not post inferior results, then wait for 1.11.1, and see if then fixed.

I know you use -O3, and it might not be necessary, even actually worse, because then compilation takes longer, and you pay for it at runtime. Maybe just skip it?

Then -O2 is used as the default, if you can run both says to see if -O3 is actually somtimes faster, and if you can show results from 1.10 also (at least if faster..) or report here, that would be great too.

You also run all benchmarks with threads enabled, but some benchmarks are single-threaded and I think it (still) hurts them. It might be a past issue.

Note, 1.11.1 is also upcoming, and I think rather soon. 1.11.0 already has some minor problems, but not speed-related (that I know of), or affecting you.

Does this refer to a date "24.09 Benchmarks Game" and then inferred that you update next on October 24th.?

[FYI: there's also now official juliac, for AOT compilation (and smallish binaries), that would speed up further (keeping all Julia functionality, requiring very minimal annotations (for at least a main function), but it's still only on 1.12-DEV. I suppose you're ok with that, though I understand you were against AOT since somehow not the default way, and it's already possible in an unofficial way with other tools, that should work for most of your benchmarks, though then with more non-idiomatic Julia code, and even tiny(er) binaries. And a third AOT compilation way with PackageCompiler.jl with rather large binaries, but also faster code, and startup, I believe supporting all code.]