SuperLU¶
About¶
SuperLU contains a set of subroutines to solve a sparse linear system A*X=B. It uses Gaussian elimination with partial pivoting (GEPP). The columns of A may be preordered before factorization; the preordering for sparsity is completely separate from the factorization.
Supported versions¶
To check which SuperLU versions and build types are currently supported on Discoverer, execute on the login node:
module avail superlu
Note
The SuperLU builds available in the public software repository contain their own BLAS library (provided as shared library). Be sure that library is the one in use by the executable code (use library prelinking if you expect some other BLAS installation to interfere).
If you need to go over the recipe we used to create the SuperLU build installed in the public software repository, visit:
https://gitlab.discoverer.bg/vkolev/recipes/-/tree/main/superlu/5
User-supported versions¶
Users are welcome to bring, compile, and use their own builds of SuperLU, but those builds will not be supported by the Discoverer HPC team.
Loading/Compiler affinity¶
To obtain access to the latest SuperLU load the environment module that matches your compiler requirements:
Intel oneAPI build¶
module load superlu/latest-intel
or specify the particular version of BLAS you need.
NVIDIA HPC SDK build¶
module load superlu/latest-nvidia
or specify the particular version of BLAS you need.
Getting help¶
See Getting help