Conversation

Notices

1. Embed this notice
   mppf (mppf@hachyderm.io)'s status on Wednesday, 31-Jan-2024 08:27:25 JST mppf

   I've created a #blog post comparing the performance of standard library 'sort' calls in different languages. Chapel's sort uses composable parallelism to be 10x faster than other popular languages for a test sort on my PC.

   https://chapel-lang.org/blog/posts/std-sort-performance/

   #chapellang #rustlang #julialang #golang #radixsort

   • Embed this notice
     phiofx (phiofx@hachyderm.io)'s status on Wednesday, 31-Jan-2024 08:50:46 JST phiofx

     in reply to

     @mppf not a totally fair comparison but drives the point across :-). What would performance be like (compared e.g. to C++) if one were to use only a single cpu?

   • Embed this notice
     mppf (mppf@hachyderm.io)'s status on Wednesday, 31-Jan-2024 09:09:52 JST mppf

     in reply to

     • phiofx

     @phiofx hi yes the blog article covers that but it is a bit of an aside; the 1-core Chapel sort clocked at 64 million elements sorted per second, which is a bit faster than he Rust sort. I did put it in the chart etc because it’s not the point. The point is, we need parallelism.

   • Embed this notice
     phiofx (phiofx@hachyderm.io)'s status on Wednesday, 31-Jan-2024 09:32:27 JST phiofx

     in reply to

     @mppf yes, end of Moore's law etc. But I always thought parallelism for the masses wont be easy, so looking for a catch :-). Maybe after forty years of #HPC it could be made easy - at least for some important standard use cases. That would be really interesting. Intrigued enough by your post to learn more about #chapellang