Solver failed with code -2 when running in jupyter notebook

[shushu-qin]

Hi, I am using 0.4.0 version of the package. I tried the simple example in a python script file. It can be run without any error. However, if I run the same code in a jupyter notebook or use pypardiso in a member function of a class, it fails. Can anyone help resolve this issue?

The code I was testing is as follows: ` import pypardiso import numpy as np import scipy.sparse as sp from scipy.io import loadmat

A = sp.rand(10, 10, density=0.5, format='csr') b = np.random.rand(10) x = pypardiso.spsolve(A, b) print(x) `

And the error I received is:

[haasad]

Hi @shushu-qin,

I tried to reproduce your issue:

python3 -m venv venv
. ./venv/bin/activate
pip install pypardiso jupyter
jupyter notebook

Can you give me some more details? What OS are you using? What's the output of pip list of your virtualenv?

Error code -2 of the Pardiso solver is short for not enough memory. Is there some memory restriction in place on your system that could interfere with the jupyter kernel?

[shushu-qin]

Thank you for the reply. I am gonna give as much information as possible 😄

1. The OS is Ubuntu 20.04.2 LTS.
2. Here is the output of pip list pip_list.txt
3. CPU: 12th Gen Intel(R) Core(TM) i9-12900KF with configuration: cores=16 enabledcores=16 threads=24
4. The memory of my system is as follows. I think the memory won't be a problem. It's wired to get this error. total used free shared buff/cache available Mem: 64100 8993 8804 343 46302 54053 Swap: 2047 354 1693
5. Here is the log of statistical info: statistical_info.txt
6. I also tried to run the unit test in the pypardiso package. It gave me this error

PS: I used to use this library for the same code on an old desktop: Ubuntu 20.04 Intel i5 cores=4, threads=8. It worked. Now I am moving the code to another computer, the simple test example run successfully alone, but failed in my project and in the jupyter notebook.

[haasad]

This looks very similar to what you describe: https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/pardiso-insufficient-memory/m-p/1358741#M32717%3Fprofile.language=en

Can you try downgrading mkl and intel-openmp and see if that changes anything?

pip install mkl==2022.0.1 intel-openmp==2022.0.1

[shushu-qin]

Downgrading mkl and intel-openmp doesn't change anything. I circumvented the problem in an inefficient way: save the matrix in a file and start a new process to read the file and perform calculation...