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

OpenBLAS 0.3.13 ex-shift patch not working #833

Closed
pablosanjose opened this issue Mar 31, 2021 · 3 comments · Fixed by JuliaPackaging/Yggdrasil#2769
Closed

OpenBLAS 0.3.13 ex-shift patch not working #833

pablosanjose opened this issue Mar 31, 2021 · 3 comments · Fixed by JuliaPackaging/Yggdrasil#2769
Labels
upstream The issue is with an upstream dependency, e.g. LLVM

Comments

@pablosanjose
Copy link
Contributor

There must have been some kind of problem with the upgrade to OpenBLAS 0.3.13 in JuliaLang/julia#39216. That PR included a patch called openblas-exshift.patch intended to bring in the fix in OpenMathLib/OpenBLAS#3091 without having to wait for OpenBLAS 0.3.14, but it seems the patch does not work. I don't know why, but the following test, discussed in the original issue Reference-LAPACK/lapack#475, still fails in Julia 1.6 and nightly

A1 = ComplexF64[0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0; 3.7796350217469814 -3.3125635598133054 0.0 0.0 0.0 0.0 0.0 0.0 6.418270043493963 -6.625127119626611 0.0 0.0 0.0 0.0 0.0 -1.0; -3.312563559813306 3.779635021746982 0.0 0.0 0.0 0.0 0.0 0.0 -6.625127119626612 6.418270043493964 -1.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 3.7796350217469814 0.0 0.0 -3.3125635598133054 0.0 0.0 0.0 -1.0 6.418270043493963 0.0 0.0 -6.625127119626611 0.0 0.0; 0.0 0.0 0.0 3.779635021746982 -3.312563559813306 0.0 0.0 0.0 0.0 0.0 0.0 6.418270043493964 -6.625127119626612 0.0 -1.0 0.0; 0.0 0.0 0.0 -3.3125635598133054 3.7796350217469814 0.0 0.0 0.0 0.0 0.0 0.0 -6.625127119626611 6.418270043493963 -1.0 0.0 0.0; 0.0 0.0 -3.312563559813306 0.0 0.0 3.779635021746982 0.0 0.0 0.0 0.0 -6.625127119626612 0.0 -1.0 6.418270043493964 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 3.7796350217469814 -3.3125635598133054 0.0 0.0 0.0 -1.0 0.0 0.0 6.418270043493963 -6.625127119626611; 0.0 0.0 0.0 0.0 0.0 0.0 -3.312563559813306 3.779635021746982 -1.0 0.0 0.0 0.0 0.0 0.0 -6.625127119626612 6.418270043493964]
B1 = ComplexF64[1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -3.7796350217469814 3.312563559813306 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.3125635598133054 -3.779635021746982 0.0 0.0 0.0 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -3.7796350217469814 0.0 0.0 3.312563559813306 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -3.779635021746982 3.3125635598133054 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.312563559813306 -3.7796350217469814 0.0 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.3125635598133054 0.0 0.0 -3.779635021746982 0.0 0.0; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -3.7796350217469814 3.312563559813306; 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.3125635598133054 -3.779635021746982]

Then, on Intel + macOS

julia> using LinearAlgebra

julia> eigen(A1, B1)  # or schur(A1, B1)
ERROR: LAPACKException(12)

The failure is platform dependent (other pencils that fail for other platforms are available in Reference-LAPACK/lapack#475), and seems similar to #832.

Also, using MKL (Julia 1.5.3 + MKL.jl) never fails.

I'm clueless about the whole Yggdrasil BinaryBuilder machinery, so I cannot fathom why the patch doesn't work (it did work when I compiled locally with it back in the day). However, perhaps the best thing is to wait for OpenBLAS 0.3.14 to be released with the fix and bake that in instead.

@ViralBShah
Copy link
Member

I don't see this patch in Yggdrasil: https://github.com/JuliaPackaging/Yggdrasil/tree/master/O/OpenBLAS/OpenBLAS%400.3.13/bundled/patches

So it just seems like it needs to be added there.

@ViralBShah
Copy link
Member

Fix in JuliaPackaging/Yggdrasil#2769

@ViralBShah
Copy link
Member

JuliaLang/julia#40343 should pick up the new binaries in master.

ViralBShah referenced this issue in JuliaLang/julia Apr 4, 2021
* Use 0.3.13+4 openblas binaries to fix #40279
* Add tests and doctest fixes for #40343 (#40346)
* Update checksums
* changed a doctest to have nicer output

Co-authored-by: Pablo San-Jose <lekand@gmail.com>
KristofferC referenced this issue in JuliaLang/julia Apr 6, 2021
* Use 0.3.13+4 openblas binaries to fix #40279
* Add tests and doctest fixes for #40343 (#40346)
* Update checksums
* changed a doctest to have nicer output

Co-authored-by: Pablo San-Jose <lekand@gmail.com>
(cherry picked from commit 58fba2b)
ElOceanografo referenced this issue in ElOceanografo/julia May 4, 2021
* Use 0.3.13+4 openblas binaries to fix #40279
* Add tests and doctest fixes for JuliaLang#40343 (JuliaLang#40346)
* Update checksums
* changed a doctest to have nicer output

Co-authored-by: Pablo San-Jose <lekand@gmail.com>
antoine-levitt referenced this issue in antoine-levitt/julia May 9, 2021
* Use 0.3.13+4 openblas binaries to fix #40279
* Add tests and doctest fixes for JuliaLang#40343 (JuliaLang#40346)
* Update checksums
* changed a doctest to have nicer output

Co-authored-by: Pablo San-Jose <lekand@gmail.com>
johanmon referenced this issue in johanmon/julia Jul 5, 2021
* Use 0.3.13+4 openblas binaries to fix #40279
* Add tests and doctest fixes for JuliaLang#40343 (JuliaLang#40346)
* Update checksums
* changed a doctest to have nicer output

Co-authored-by: Pablo San-Jose <lekand@gmail.com>
@KristofferC KristofferC transferred this issue from JuliaLang/julia Nov 26, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
upstream The issue is with an upstream dependency, e.g. LLVM
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants