Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Performance evaluation for Julia v1.11 #1916

Open
sloede opened this issue Apr 24, 2024 · 1 comment
Open

Performance evaluation for Julia v1.11 #1916

sloede opened this issue Apr 24, 2024 · 1 comment
Labels

Comments

@sloede
Copy link
Member

sloede commented Apr 24, 2024

As discussed at yesterday's meeting, LoopVectorization.jl, Octavian.jl, and JuliaSIMD in general will be deprecated for Julia v1.11. That is, code is likely to continue to work for the time being, but bugs might not get fixed anymore. It is also to be expected that we will see a performance hit in areas where we heavily rely on LV.jl & friends, especially with the @turbo macro.

Some parts of Trixi.jl that are likely to be adversely affected by this have already been identified:

  1. multiply_dimensionwise! for mortar and AMR projections/interpolations (see interpolation.jl)
  2. Computation of metric terms for P4estMesh/StructuredMesh (e.g., here)
  3. DGMulti solver: Several areas where Octavian.jl's matmul! (see also Remove dependence on Octavian.jl for Julia 1.11 #1906), some explicit uses of @turbo
  4. Manual SIMD optimizations for the EC performance paper (these might not work at all anymore, e.g., here)

The decision yesterday was to first proceed with switching CI testing to Julia v1.10 (#1562 - @DanielDoehring is on it!), and to continue monitoring CI testing with v1.11 (#1904 will help). Once v1.11.0 is released, we will revisit this issue and benchmark the aforementioned cases, before we make decisions on how to proceed.

@sloede sloede added performance We are greedy upstream labels Apr 24, 2024
@ranocha
Copy link
Member

ranocha commented Oct 10, 2024

LoopVectorization.jl has been updated but performance regressions like JuliaLang/julia#55009 still remain in Julia v1.11.0

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants