Giter VIP home page Giter VIP logo

Comments (13)

lucascolley avatar lucascolley commented on June 9, 2024

hmm, nothing obvious sticks out to me in the git history. Maybe gh-20337 @ev-br ?

from scipy.

ev-br avatar ev-br commented on June 9, 2024

Maybe gh-20337?

No way it can explain the runtime change from 15min to 4 hrs.

Something else must have changed. I'd first verify the "new" run time is stable, then try to replicate locally. Then bisect.

from scipy.

KelSolaar avatar KelSolaar commented on June 9, 2024

from scipy.

KelSolaar avatar KelSolaar commented on June 9, 2024

Confirming it is coming from 1.13.0, I banned the version and did a fresh build in ~15mins using 1.12.0: https://github.com/colour-science/colour/actions/runs/8890511725/job/24410682802

from scipy.

ev-br avatar ev-br commented on June 9, 2024
  1. Do you see it on other platforms, or can you confirm it's windows specific?

  2. Can you reproduce this locally?

from scipy.

KelSolaar avatar KelSolaar commented on June 9, 2024

Hey @ev-br,

It is Windows only and I cannot reproduce locally as I'm developing on macOs.

from scipy.

ev-br avatar ev-br commented on June 9, 2024

Just to confirm: with scipy 1.13 the Windows CI run takes 4hrs and MacOS and Linux runs take 15 mins, correct?

Also, with scipy 1.12, all of them are 15 mins?


It's likely much faster to debug in a Windows VM. If only to see if the issue is specific to the GH runners or windows in general.

from scipy.

KelSolaar avatar KelSolaar commented on June 9, 2024

Correct yes!

from scipy.

ilayn avatar ilayn commented on June 9, 2024

Is this a wheel install or you are building it yourself in the CI?

from scipy.

KelSolaar avatar KelSolaar commented on June 9, 2024

It is a wheel install, Scipy is one of our dependencies.

from scipy.

lucascolley avatar lucascolley commented on June 9, 2024

Right. To clarify, the installation is fast, but the test suite takes 4 hours.

from scipy.

tylerjereddy avatar tylerjereddy commented on June 9, 2024

There was a substantial bump in vendored OpenBLAS version between 1.12.0 and 1.13.0 of course. I suppose that's also worth considering, especially since the 0.3.26.dev version specifically addressed a concurrency issue on Windows. Having a way to bisect with a simple reproducer as Evgeni alluded to would of course be great.

from scipy.

rgommers avatar rgommers commented on June 9, 2024

Agreed that OpenBLAS version is the prime candidate. We have nightlies that were just upgraded from 0.3.26.dev to 0.3.27, with a known regression that can cause this kind of performance regression fixed. Worth trying with those: https://anaconda.org/scientific-python-nightly-wheels/scipy

from scipy.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.