Dear all, we are experiencing extremely high first call latency times with Julia 1.6 when running our geophysical flows multiphysics finite element PDE solver (on the order of 15 hours on a supercomputer node with 192 GBytes of mem; on a local laptop it is killed after some hours by the OS due to m…

For the records, we could already “solve” (bypass) the issue. See https://github.com/JuliaLang/julia/issues/43206#issuecomment-983474073 for more details.

FWIW, things get “stuck” for me when the time stepper function is first called, i.e. https://github.com/gridapapps/GridapGeosciences.jl/blob/master/test/Williamson2ThetaMethodFullNewtonTests.jl#L37

I tried to run your MWE and was stymied by the non-minimal requirements of GridapPardiso.jl. As a longer-term fix, you may want to instead target Pardiso.jl , which should automatically grab BinaryBuilder.jl-provided MKL binaries. Depending on how your cluster’s filesystem is set up, you may be runn…

Ok. The same here. Which version of Julia are you using? 1.6.3?

Thanks for that. I was aware of the BinaryBuilder.jl provided MKL binaries, but did not have the time to set them up yet in GridapPardiso. Definitely we have to leverage that in GridapPardiso as well, much easier. I have modified the commit sha in my original post to point to a commit in which Grid…

Have you tried running this on 1.7 (or a recent master)?

Not yet. But I will definitely give it a try and let you know the outcome. Thanks!

I tried 1.7 and it seems there is the same issue, it gets “stuck” at the same place as with Julia 1.6

You should open an issue, even if you don’t have a minimal example. Maybe one thing to do before, if you find time, is to check whether compiling a julia of the branch in above linked PR resolves this issue: https://github.com/JuliaLang/julia/pull/42156

When I’m tune my model on Pluto.jl with revise.jl, I found that sometimes, maybe once or twice a day, when I change some parameter of my model, and rerun the cell, it get stucked for several minutes. I don’t really know what it is doing, even the first time only takes several seconds, but it could …

Extremely high first call latency Julia 1.6 versus 1.5 with multiphysics PDE solver

General Usage Performance

lmiq November 18, 2021, 12:12am 2

Maybe this thread gives you some insight: Understanding and optimizing compiler time (just a bit)

Topic		Replies	Views
22 seconds to 3 and now more: Let's fix all of the DifferentialEquations.jl + universe compile times! Performance precompilation	6	5106	August 17, 2021
Help reducing compilation and inference time Performance compilation , latency	71	1622	July 11, 2024
How to cut down compile time when inference is not the problem? Performance	4	1295	November 26, 2021
Shouldn't 1.8.0 be faster than Julia 1.7? Performance	30	2742	September 16, 2022
Understanding and optimizing compiler time (just a bit) General Usage question	21	8775	May 14, 2024

Extremely high first call latency Julia 1.6 versus 1.5 with multiphysics PDE solver

Related topics