⚠ INFORMATION
This page was translated by an AI (LLM) with a cursory human check and is awaiting full review.
This page was translated by an AI (LLM) with a cursory human check and is awaiting full review.
Scientific Libraries
🔁 Sequential Libraries
| Library | Usage |
|---|---|
| ARPACK | Fortran77 routines for finding eigenvalues and eigenvectors of sparse matrices. |
| BLAS | Set of functions performing basic linear algebra operations. |
| GSL | Numerical library offering a wide range of mathematical routines. |
| LAPACK | Fortran routines for solving dense linear algebra problems. |
| METIS | Sequential graph partitioning library. |
🔀 Sequential and Parallel Libraries, CPU and GPU
| Library | Usage |
|---|---|
| CUDA Toolkit | Set of scientific libraries for GPU including CuFFT, CuBLAS, CuSPARSE, CuSOLVER and more. |
| FFTW | Sequential and parallel routines for computing the discrete Fourier transform. |
| HYPRE | Linear solvers including multigrid methods. |
| INTEL MKL | Mathematical routines including BLAS, LAPACK, ScaLAPACK, sparse solvers and FFT. |
| MATIO | C library for reading and writing MAT MATLAB files. |
| MAGMA | Set of linear algebra libraries specific to GPU similar to LAPACK. |
| MUMPS | Parallel multifrontal solver for large sparse linear systems. |
| ParMETIS | Parallel graph partitioning library. |
| PARPACK | Parallel version of the ARPACK library for computing eigenvalues and/or eigenvectors. |
| PETSc | Data structures and routines for parallel solution of linear systems. |
| ScaLAPACK | Fortran routines for parallel solution of dense linear algebra problems. |
| Suite Sparse | Collection of algorithms for sparse matrices. |
| Trilinos | Set of algorithms in a C++ object-oriented software framework for solving scientific problems. |
Command idr_module_search 🔎
The idr_module_search command developed by IDRIS allows you to know if the library you are looking for is installed on Jean Zay and, if so, in which version(s) and for which architecture(s) (CPU, V100, A100 and/or H100).
idr_module_search [options] <my_library>important
Initially, this command was implemented for searching for Python packages in our conda environments. The options provided do not all make sense in the case of searching for an HPC library. We present here only the useful options in this use case: --soft and --arch.
| Option | Description |
|---|---|
--soft | Activates the soft mode. By activating this mode, you can list several library names within a single command. |
--arch <architecture> | Limits the search to a specific architecture. The accepted values are cpu, v100, a100 and h100 (case insensitive). By default, all architectures are considered. |
-h [--help] | Displays the help and exits. |
Examples of usage
- Simple search:
idr_module_search petsc - Search for multiple libraries:
idr_module_search --soft petsc metis - Search filtered by architecture:
idr_module_search --arch a100 petsc
Example
idr_module_search --soft --arch a100 petsc metisModule metis/5.1.0 A100: metis==5.1.0Module metis/5.1.0-int32-real64 A100: metis==5.1.0-int32-real64Module metis/5.1.0-int64-real64 A100: metis==5.1.0-int64-real64Module petsc/3.22.1-mpi-cuda A100: petsc==3.22.1-mpi-cudaModule petsc/3.23.2-mpi-cuda A100: petsc==3.23.2-mpi-cudaPossible errors
- Invalid version: adding a version constraint is only possible for searching Python packages (
idr_module_search "numpy>=1.20.0"for example). The functionality is not adapted to searching for HPC libraries. - Invalid package name: names must comply with the standard format.