I have 2 machines: M1 Mac Mini, with horrible thermals and a peak of 3.2 GHz and 8 GB of RAM (absolutely love it though) A beefy linux (Ubuntu 21.10) workstation with 64 GB of DDR5 RAM and a 12th-Gen 4.9 GHz processor with 16 threads. (8 physical cores, efficiency cores disabled) I have some code…

Name: Code Profiling and Optimization (in Julia)
Uploaded: 2022-01-26T11:30:58Z
Description: [Code Profiling and Optimization (in Julia)] shows how to profile. Profile a small solve. Don’t optimize what the profile doesn’t show.

Thank you, I wasn’t aware of this actually. Looks fantastic!

[image] ash: @ChrisRackauckas sorry for the direct ping, but would you happen to know why KLU factorization causes this issue? (tl;dr KLU factorization runs at the same speed on two machines, one fast one slow. Default algorithm is twice as fast on fast machine). Did you profile? How much tim…

I ran a little experiment to figure out the effects of multithreading on both systems. I extended the simulation time by 100 fold and ran 2 simulations, one with 1 thread and one with 16 threads on the workstation or 8 on the Mac. I also used BLAS.set_num_threads(N_threads). I additionally started u…

Could you post an annotated profile?

Embarrassingly, I don’t know much about profiling and don’t know how to annotate one. However, I profiled some code on both systems and tried UMF vs KLU. There is a weird bug where @profview from VSCode does not capture the true time it takes to run the function vs @time. Thus, the M1 results below…

[Code Profiling and Optimization (in Julia)] shows how to profile. Profile a small solve. Don’t optimize what the profile doesn’t show.

[image] ash: EDIT: all tests were done with 1 thread and OpenBLAS. [image] ash: 12th-Gen 4.9 GHz processor with 16 threads. (8 physical cores, efficiency cores disabled) 1.) Only OpenBlas v0.3.19 ( Dec 19, 2021 release ) contains: “added cpu detection for Intel Alder Lake” - so prob…

Thank you for the video, I watched (almost) all of it and definitely learned a lot. My ODE function was already nonallocating (thanks to you in another thread, actually). Here is a flame graph for UMF, a short 10s run with left-heavy mode (I don’t know how to read it otherwise honestly). [CleanSh…

Thank you for posting this, I really appreciate it. I am not surprised by this architecture’s growing pain. I will get some sleep and work on compiling Julia from source with the new version of OpenBLAS and report back.

Okay yes, that looks beautiful. So then did you play around with configurations of preconditioned GMRES like is described in Solving Large Stiff Equations · DifferentialEquations.jl ? When it’s the linear solve that’s the limiting factor, the only things you can do are (a) try a different ODE solver…

Multithreaded code on beefy computer runs just as fast as serial code on M1 Mac

General Usage Performance

ChrisRackauckas January 26, 2022, 11:30am 20

shows how to profile. Profile a small solve. Don’t optimize what the profile doesn’t show.

Topic		Replies	Views
Regarding the multithreaded performance of OpenBLAS Performance blas , multithreading	7	6032	January 31, 2022
JuMP.jl and DifferentialEquation.jl benchmarks on M1 Max Julia 1.7.0 x89 vs ARM. (spoiler: ARM is 1.5-2x faster) General Usage jump , diffeq , apple	12	2859	December 5, 2021
OpenBLAS is faster than Intel MKL on AMD Hardware (Ryzen) Performance blas , lapack	40	37258	June 19, 2020
Why doesn't multithreading help here? Performance	12	1516	August 22, 2020
Current OpenBLAS Versions (January 2022) do not support Intel gen 11 performantly? Performance linearalgebra	50	5071	April 7, 2022

Multithreaded code on beefy computer runs just as fast as serial code on M1 Mac

Related topics