Julia vs. Chapel performance

I saw this Julia vs. Chapel performance comparison in the benchmarksgame website. It seems that for all the programs, Chapel is very close to or much faster than Julia. In addition, the Chapel codes are as easy to understand as Julia codes. Also, for some programs, the Julia versions have very low CPU load. Can we say Chapel is faster than Julia?

https://benchmarksgame-team.pages.debian.net/benchmarksgame/faster/julia-chapel.html

There has already been a long thread about the benchmarks game:

TLDR: “The Julia versions are mostly naive high-level versions. There’s a lot of room for improvement there.”

A small amount of effort has been put into faster Julia versions of their benchmarks (https://github.com/KristofferC/BenchmarksGame.jl), but it’s not clear when or whether these will get incorporated into the “official” benchmarks.

So, taken as a whole they currently seem to be of limited utility in comparing Julia with other languages, except as an invitation for flamewars. If you have a specific small Julia program whose performance you think should be better, or specific features of Chapel that you would like to see in Julia, it would be more productive to start a separate, more focused, discussion on those topics.

(Note, however, that Chapel is a statically typed language — it’s well known that statically typed languages can be compiled to fast code and there are many choices in this space. Julia, on the other hand, is a dynamic language, hence suitable for interactive as well as offline use, and there are fewer choices for high-performance computing in that category.)

9 Likes

If the program authors wish those programs to be incorporated into the “official” benchmarks then the program authors should contribute them.