# Does Debian's BenchmarkGames show representative performance?

**URL:** https://discourse.julialang.org/t/does-debians-benchmarkgames-show-representative-performance/80018
**Category:** Community
**Tags:** benchmark
**Created:** [April 25, 2022, 8:48am UTC](https://discourse.julialang.org/t/does-debians-benchmarkgames-show-representative-performance/80018 "2022-04-25T08:48:46Z")
**Posts on this page:** 1
**Page:** 3

<div class="post-metadata">

### Author: ![Palli](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/palli/32/3380_2.png) [@Palli](https://discourse.julialang.org/u/Palli)
#### Post date: [August 18, 2022, 11:04pm UTC](https://discourse.julialang.org/t/does-debians-benchmarkgames-show-representative-performance/80018/41 "2022-08-18T23:04:44Z")

</div>

A.  
Debian’s benchmark game doesn’t allow showing Julia’s fastest speed, but its fork does (or so far only enabled for a few benchmarks, I convinced the guy to allow julia/aot, and thanks @kristoffer.carlsson for PackageCompiler.jl), and there Julia/aot is 3rd, ahead of Rust and C, only 9% slower than fastest entry C++:

> **[nsieve - benchmarks, Which programming language or compiler is faster](https://programming-language-benchmarks.vercel.app/problem/nsieve)**
>
> benchmarks,benchmark,performance,Acton,C,C#,C++,Chapel,Codon,Crystal,D,Dart,Go,Haxe,Java,Javascript,Julia,Kotlin,Lua,Nim,OCaml,Odin,Perl,Php,Python,Ruby,Rust,Swift,Typescript,V,Wasm,Zig,Acton lang,C lang,C# lang,C++ lang,Chapel lang,Codon...

What does it take to eliminate the 117ms “time(sys)” (fully or mostly), as with the other languages? Then Julia would be 19% faster than the current top entry.

[https://github.com/hanabi1224/Programming-Language-Benchmarks/issues/291?notification\_referrer\_id=NT\_kwDOAHonKLI0MTg4NDEyNjcyOjgwMDU0MTY&notifications\_query=reason%3Aparticipating#issuecomment-1216960434](https://github.com/hanabi1224/Programming-Language-Benchmarks/issues/291?notification_referrer_id=NT_kwDOAHonKLI0MTg4NDEyNjcyOjgwMDU0MTY&notifications_query=reason%3Aparticipating#issuecomment-1216960434)

Without AoT, Julia is only 14th, 41% slower than fastest (but not really, since Julia’s fixed overhead is misleading).

B.  
I would still like Julia faster without AoT, the extra 174 ms are compilation overhead that might be possible to reduce. For scripting (and benchmarking), a separate non-default sysimage, e.g. stripping out LinearAlgebra might be useful. I’ve been asked if any alternative sysimage is official, so could Julia include a such JuliaLite sysimage, that would just be taking up space until you enable it?

[Previous page](https://discourse.julialang.org/t/does-debians-benchmarkgames-show-representative-performance/80018.md?page=2)