I was playing around with eio and specifically their worker pool example . I wrote various trivial worker functions to burn some CPU, just so I could see that it genuinely was using multiple cores. The intrec function below behaved as you would expect, but the floatrec one actually got slower wit…

Maybe try batching elements per core with big arrays to avoid the boxed floats overhead?

I was expecting the floating point run to be slower. The thing I can’t understand is that the timings are something like: intrec, 1 core = 1 second elapsed, 1 second user cpu interc, 2 cores = 0.5 second elapsed, 1 second user cpu floatrec, 1 core = 5 seconds elapsed, 5 second user cpu floatrec, 2…

Aha! it’s nothing to do with inter-thread scheduling or anything like that: master »~/dev/ocaml/eio_tests $ ./_build/default/bin/multicore_6.exe d 1 d 1 +Worker 1 : 3429680500000 +Elapsed time = 13.1 s master »~/dev/ocaml/eio_tests $ parallel -j 2 -n 2 ./_build/default/bin/multicore_6.exe -- d 1 d…

The PI4 which should be a Cortex-A72(?) spec says it has one FPU per core, in 64bit mode at least. Check /proc/cpuinfo and wikipedia. Cortex-A72 The Cortex-A72 processor is an efficient high-performance processor that implements the Armv8-A architecture.

I can’t reproduce on an Intel core i7 (4 physical cores), so I’d wager on a hardware peculiarity of the Pi. Perf results: https://paste.osau.re/Qel#eyJhbGciOiJBMTI4Q0JDIiwiZXh0Ijp0cnVlLCJrIjoiMHMyUFI3WnZvRkxuTHBGVl9IQ2hXQSIsImtleV9vcHMiOlsiZW5jcnlwdCIsImRlY3J5cHQiXSwia3R5Ijoib2N0In0= Edit.: the fa…

Ah but it did happen with “parallel” which doesn’t rule out ocaml, but does mean it’s nothing to do with the domain/eio threading stuff. I shall (a) try it out on an intel laptop later and (b) see if I can get another language to show a similar slowdown. I suspect I’ll see the same results as you o…

Ah but it did happen with “parallel” Ah, yes, I read too fast.

Eio/multicore puzzling performance

Learning

yawaramin November 2, 2023, 2:32am 2

It might be float boxing. See OCaml speed comparison - calculating pi with Leibniz - optimize? - #2 by copy

Topic		Replies	Views
Multicore OCaml: December 2021 and the Big PR Community multicore , multicore-monthly	7	3507	January 10, 2022
Multicore OCaml: November 2021 with results of code review Community multicore , multicore-monthly	30	7318	January 7, 2022
Understanding performance of a multi-core program Learning multicore , multiprocessing	13	1718	October 30, 2022
About Multicore Ecosystem multicore , compiler , faq	59	13774	April 10, 2018
OCaml 5 performance Ecosystem multicore , performance , profiling , eio	30	3644	September 11, 2024

Eio/multicore puzzling performance

Related topics