-
Notifications
You must be signed in to change notification settings - Fork 43
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
Installing Slycot (openmp) breaks numpy #230
Comments
How are you installing the various packages? All with conda? Or other ways? |
Thanks for the quick response. I detailed the steps in the original post above. |
Given that you are reporting a clash with numpy and openmp binaries from conda forge, then I think you should open the issue at one or both of these locations: https://github.com/conda-forge/openmp-feedstock https://github.com/conda-forge/numpy-feedstock I don't see a slycot specific issue (yet). |
I agree that it's not a true issue of slycot. But maybe slycot could move to a more modern version of libflang. I have little hope, that anyone at openmp is going to fix a 6 year old version. I could frankenstein together a working environment by ignoring the dependencies and installing:
|
Ah, ok I understand now. slycot forces an old libflang and thus forces you to install an old openmp, which has an incompatibility with numpy. So yes, it would be helpful to move slycot to a newer libflang. It doesn't look like this is something we explicitly set: https://github.com/conda-forge/slycot-feedstock/blob/main/recipe/meta.yaml It may mean that conda forge requires libfang<6 from the |
It looks locked to libflang 5 on windows across all of conda-forge: https://github.com/conda-forge/conda-forge-pinning-feedstock/blob/main/recipe/conda_build_config.yaml#L30 |
Yes we try to follow the conda-forge pinning for flang. See also conda-forge/slycot-feedstock#65 and conda-forge/blas-feedstock#106 |
Thank you for your insights. I filed the issue with numpy and openmp feedstock. |
There is a similar problem under Debian. Using the installed python3-numpy package (numpy version 1.24.2 in trixie) pip install slycot doesn't work correctly (building stops). The only way to install slycot ist to force the removing of python3-numpy sudo dpkg -P --force-all python3-numpy) and install numpy using pip (in this case the installed version is numpy-1.26.4).After this step pip install spycot works. |
Could you post a build log with the error message, please? |
Tested on a docker after clonng Slycot in git, but it correclty works after updating numpy. In the pyproject.toml the version of rerquired numpy is numpy >= 1.23.1 an in Debian we have 1.24.2. root@debian:/psc/Slycot# pip install --break-system-packages . �� Building wheel for slycot (pyproject.toml) did not run successfully.
note: This error originates from a subprocess, and is likely not a problem with pip. |
After installing slycot in a conda environment, a kernel crash occurs when trying to solve a linear system with a singular matrix.
I narrowed the issue down to the following:
Openmp places its "libiomp5md.dll" in Library\bin.
The version from numpy is either in Library\bin (channel main) or directly in bin (conda-forge).
Actual Behavior
Using numpy.linalg.solve on singular matrix with dimension over 149 crashes the Kernel.
Expected Behavior
Expected a "LinAlgError: Singular matrix" as it is shown with matrices up to the dimension of 149.
Steps to Reproduce
mamba create -n slycot_crash python=3.11 numpy ipython -c conda-forge
mamba activate slycot_crash
ipython
exit
mamba install openmp=5 -c conda-forge
The text was updated successfully, but these errors were encountered: