Projects | Pierre Ramet

Chameleon

Mon, 05 Jan 2026 00:00:00 +0000

Chameleon: A dense linear algebra solver

Chameleon is a scientific C library providing high-performance routines to solve dense linear systems (general, symmetric positive definite, least squares) using LU, Cholesky, QR and LQ factorizations, as well as singular value (SVD) and eigenvalue (EVD) decompositions. Real and complex arithmetic are supported in both single and double precision.

Chameleon is designed to fully exploit modern heterogeneous and distributed supercomputers. Relying on a tile matrix layout, algorithms are expressed as task graphs and executed dynamically by innovative runtime systems:

StarPU (enabling multi-GPU acceleration through cuBLAS or hipBLAS, and distributed-memory clusters with MPI),
PaRSEC,
OpenMP,
QUARK.

Get Chameleon

The latest official release is v1.4.0 (January 2026). All releases, source archives, and changelogs are available on .

To use the latest development version of Chameleon, clone the repository with submodules:

# Clone with submodules
git clone --recursive https://gitlab.inria.fr/solverstack/chameleon.git
cd chameleon

Available Features & Algorithms

Matrix factorizations and linear solvers:
- Cholesky: POTRF, POTRI, POTRS
- QR / LQ: GEQRF, GELQF, ORMQR, UNMQR
- LU factorization: GETRF, GESV, GETRS (with tile and partial pivoting)
- Singular Value & Eigenvalue Decomposition: GESVD, SYEV, HEEV
- BLAS-3 operations: GEMM (including A-stationary algorithms), SYMM, HEMM, TRSM, SYRK, HERK
Precision support: Single (s), Double (d), Complex single (c), Complex double (z).
Supported Architectures:
- Multicore CPUs (x86_64, ARM, PowerPC)
- NVIDIA GPUs via cuBLAS / TCUDA
- AMD GPUs via HIP / hipBLAS
- Distributed-memory clusters via MPI (using 2D block-cyclic tile distributions)
High-Level Interfaces: C, Fortran 90, Python.

Documentation

Comprehensive online documentation is available at .

Installation

Build and install with CMake

Chameleon can be built using (minimum version 3.12):

BLAS and LAPACK (e.g. OpenBLAS, Intel MKL)
HWLOC
StarPU (version >= 1.4 for multi-GPU & MPI), PaRSEC, or OpenMP
CUDA / CuBLAS or HIP / hipBLAS (optional, for GPU acceleration)
MPI (optional, for distributed memory)
Python (optional, for code generation and bindings)

mkdir build && cd build
cmake .. -DCMAKE_INSTALL_PREFIX=/path/to/install \
         -DCHAMELEON_USE_MPI=ON \
         -DCHAMELEON_USE_CUDA=ON \
         -DCHAMELEON_SCHED_STARPU=ON
make -j
make install

Chameleon is also available through package managers:

Guix-HPC: guix install chameleon
Spack: spack install chameleon
Homebrew: brew install chameleon

Get Involved

Reporting an issue

Users are encouraged to use the issue tracker to report bugs, suggest features, or ask questions:

Contributions

Authors

The development of Chameleon is conducted within the SolverStack and TOPAL / Storm research teams at Inria Bordeaux Sud-Ouest, Bordeaux University, CNRS (LaBRI), and Bordeaux INP, in close collaboration with the MORSE associate team (ICL University of Tennessee, KAUST):

Mathieu Faverge, PI
Florent Pruvost
Samuel Thibault
Philippe Virouleau
Alycia Lisito
Pierre Ramet
and the MORSE / SolverStack contributors.

License

Chameleon is distributed under the CeCILL-C open-source license (BSD-compatible):

SolverStack

Thu, 01 Jan 2026 00:00:00 +0000

SolverStack@Inria Bordeaux

SolverStack aims at providing a coherent, high-performance (HPC) linear algebra solver stack. It provides a comprehensive collection of numerical solvers, partitioning tools, and runtime systems designed for modern supercomputers operating on dense and sparse matrices.

The ecosystem includes direct, iterative (Krylov), and hybrid direct/iterative methods, with advanced preconditioners and low-rank compression techniques, ensuring portability of performance from multicore laptops to petascale and exascale supercomputers.

👉 Official Website:

Software Ecosystem

Sparse Direct Methods: ,
Dense Direct Methods:
Sparse Matrix Package:
Iterative & Krylov Methods:
Hybrid Direct/Iterative Solvers:
Randomized & Low-Rank Decompositions:
Task-Based Runtime Systems:
Communication Libraries:
Partitioning & Ordering:

Packaging & Distribution

SolverStack software components are systematically packaged and maintained for high-performance computing environments:

Guix-HPC: guix install <package> ( )
Spack: spack install <package> ( )
Homebrew: brew install <package> ( )

Partner Teams & Institutions

SolverStack software is developed by researchers and engineers in joint project-teams at Inria Bordeaux Sud-Ouest, Bordeaux University, CNRS ( ), and Bordeaux INP:

Contact:

PaStiX

Fri, 12 Jul 2024 00:00:00 +0000

PaStiX: A sparse direct solver

PaStiX (Parallel Sparse matriX package) is a scientific library that provides a high-performance parallel solver for very large sparse linear systems based on direct methods. Numerical algorithms are implemented in single or double precision (real or complex) using LLt, LDLt and LU with static pivoting (for non-symmetric matrices having a symmetric pattern). This solver also provides low-rank compression methods (BLR and HODLR) to reduce the memory footprint and/or the time-to-solution, as well as mixed-precision arithmetic.

Get PaStiX

The latest official release is v6.4.0 (July 2024). All releases, source tarballs, and changelogs are available on .

To use the latest development version of PaStiX, please clone the master branch. Note that PaStiX contains git submodules (spm and morse_cmake):

# Clone with submodules
git clone --recursive https://gitlab.inria.fr/solverstack/pastix.git
cd pastix

Available Features

Features available in shared memory with POSIX threads on multicore architectures:

	Seq	Static	Dyn	StarPU	PaRSEC
POTRF (Cholesky)	FR/LR	FR/LR	FR/LR	FR/LR/GPU	FR/LR/GPU
PXTRF (LL^t for complex)	FR/LR	FR/LR	FR/LR	FR/LR/GPU	FR/LR/GPU
HETRF (LDL^h)	FR/LR	FR/LR	FR/LR	FR/LR/GPU	FR/LR/GPU
SYTRF (LDL^t)	FR/LR	FR/LR	FR/LR	FR/LR/GPU	FR/LR/GPU
GETRF (LU)	FR/LR	FR/LR	FR/LR	FR/LR/GPU	FR/LR/GPU
TRSM	FR/LR	FR/LR	FR/LR	FR/LR	-
DIAG	FR/LR	FR/LR	FR/LR	FR/LR	-

Features available in hybrid shared/distributed memory with MPI between processes and POSIX threads within a process:

	Seq	Static	Dyn	StarPU	PaRSEC
POTRF (Cholesky)	FR	FR	FR	FR/GPU	FR/GPU
PXTRF (LL^t for complex)	FR	FR	FR	FR/GPU	FR/GPU
HETRF (LDL^h)	FR	FR	FR	FR/GPU	FR/GPU
SYTRF (LDL^t)	FR	FR	FR	FR/GPU	FR/GPU
GETRF (LU)	FR	FR	FR	FR/GPU	FR/GPU
TRSM	FR	FR	FR	FR	-
DIAG	FR	FR	FR	FR	-

FR means Full-Rank computations without compression techniques.
LR means Low-Rank compression technique to reduce the memory footprint and/or the time-to-solution.
WARNING: GPU kernels are not available on compressed supernodes.
WARNING: Low-rank compression and Schur complement are not yet available with MPI.

Documentation

The latest Doxygen documentation is available .

The and of the solver are described. Some are also provided.

Installation

Build and install with CMake

PaStiX can be built using . The build requires several library dependencies to be installed on the system:

BLAS (MKL, OpenBLAS, …) and CBLAS (sequential version required)
LAPACK and LAPACKE (sequential version required, with TMG enabled for testing)
HWLOC (highly recommended)
SCOTCH (optional)
METIS (optional)
STARPU runtime support (optional)
PARSEC runtime support (optional)
CUDA/CuBLAS to enable GPU functionality with runtime support (optional)
EZTRACE to enable tracing support (optional)
Python and Fortran compiler for wrappers and examples (optional)

For instance, on Debian-like systems, dependencies can be installed with the following command:

  sudo apt-get install cmake gcc gfortran libhwloc-dev libscotch-dev libopenblas-dev liblapacke-dev python-numpy

The main options to configure the PaStiX build are:

Standard CMake options:
- CMAKE_BUILD_TYPE: Debug, RelWithDebInfo, Release, MinSizeRel; we recommend using Release or RelWithDebInfo for best performance.
- CMAKE_INSTALL_PREFIX: Specify the prefix directory to install the library.
- BUILD_SHARED_LIBS=[OFF]: Enable building shared libraries (required for the Python wrapper).
Integer type:
- PASTIX_INT64[=ON]: Enable/disable int64_t for integer arrays.
Ordering libraries:
- Ordering libraries must match the integer type chosen for integer arrays in PaStiX.
- PASTIX_ORDERING_SCOTCH[=ON]: Enable/disable Scotch library support for ordering.
- PASTIX_ORDERING_METIS[=OFF]: Enable/disable Metis library support for ordering (Metis 5.1 required).
External schedulers:
- PASTIX_WITH_PARSEC[=OFF]: Enable/disable PaRSEC runtime support. Requires installing PaRSEC tag pastix-<releasenumber> (mymaster for master branch) from the repository that includes patches on top of the original PaRSEC runtime system. PaRSEC needs to be compiled with option -DPARSEC_WITH_DEVEL_HEADERS=ON.
- PASTIX_WITH_STARPU[=OFF]: Enable/disable StarPU runtime support (requires StarPU 1.3).
Distributed memory:
- PASTIX_WITH_MPI=[OFF]: Enable/disable distributed memory support (see above for details). If used with the PaRSEC library, MPI should be enabled or disabled consistently in both libraries.
External SpM library:
- PASTIX_WITH_EXTERNAL_SPM=[OFF]: Enable/disable the use of an external SpM library in favor of the internal one.
Documentation:
- BUILD_DOCUMENTATION[=OFF]: Enable Doxygen documentation generation.

Get involved

Reporting an issue

We strongly recommend all users to use the issue tracker to report any problems with the software, or to submit feature requests. We will do our best to answer them in a timely manner.

Contributions

Authors

The following people contribute or contributed to the development of PaStiX:

Mathieu Faverge, PI
Pierre Ramet, PI
David Goudin
Mathias Hastaran
Pascal Henon
Xavier Lacoste
François Pellegrini
Grégoire Pichon, Low-rank solver
Florent Pruvost, CMake and Spack
Theophile Terraz
Tony Delarue, MPI implementation
Brieuc Nicolas, Mixed precision
Alycia Lisito
Clément Richefort, Algebraic multigrid and mixed precision

If we missed your name, please let us know so we can update the list.

Citing PaStiX

Feel free to use the following publications to reference PaStiX:

Original paper that initiated PaStiX:
- Pascal Hénon, Pierre Ramet, Jean Roman. PaStiX: A High-Performance Parallel Direct Solver for Sparse Symmetric Definite Systems. Parallel Computing, Elsevier, 2002, 28 (2), pp.301–321.
Parallel incomplete factorization implemented in PaStiX:
- Pascal Hénon, Pierre Ramet, Jean Roman. On finding approximate supernodes for an efficient ILU(k) factorization. Parallel Computing, Elsevier, 2008, 34, pp.345–362.
Reordering strategy for blocking optimization in PaStiX:
- Grégoire Pichon, Mathieu Faverge, Pierre Ramet, Jean Roman. Reordering Strategy for Blocking Optimization in Sparse Linear Solvers. SIAM Journal on Matrix Analysis and Applications, Society for Industrial and Applied Mathematics, 2017, SIAM Journal on Matrix Analysis and Applications, 38 (1), pp.226 - 248.
On the use of low rank approximations in PaStiX:
- Grégoire Pichon, Eric Darve, Mathieu Faverge, Pierre Ramet, Jean Roman. Sparse supernodal solver using block low-rank compression: Design, performance and analysis. International Journal of Computational Science and Engineering, Inderscience, 2018, 27, pp.255 - 270.

License

SPM

Tue, 09 Jul 2024 00:00:00 +0000

SPM: SParse Matrix package

SPM (SParse Matrix package) is a scientific library that provides comprehensive basic routines to manipulate sparse matrices in CSC, CSR, and IJV formats. The functionalities covered are:

Sparse matrix -by- dense matrix products
Sparse matrix -by- vector products
Norm computations
Matrix and vector scaling
Sort functions
Graph symmetrization and merge of duplicate entries
Random generators for right-hand sides
Verification routines for linear solvers
In-place format conversion routines
Drivers to read matrices from files (MatrixMarket, Harwell-Boeing/RSA, IJV, …)
Laplacian generators for stencils
Multi-DOF and variadic DOF support

Python, Julia, and Fortran 90 wrappers are included in the package.

Warning: This package does not currently aim for peak performance, but rather provides a portable API and a simple interface to manipulate sparse matrices in both shared and distributed memory. This is conceptually similar to what was proposed in the module for shared memory.

Get SPM

The latest official release is v1.2.4 (July 2024). All releases, source tarballs, and changelogs are available on .

To use the latest development version of SPM, please clone the master branch. Note that SPM contains the morse_cmake git submodule. To obtain the source code, please use the following commands:

# Clone with submodules
git clone --recursive https://gitlab.inria.fr/solverstack/spm.git
cd spm

Documentation

The latest Doxygen documentation is available .

The main functionalities are listed .

Installation

Build and install with CMake

SPM can be built using . The build requires several library dependencies to be installed on the system:

BLAS (MKL, OpenBLAS, …) and CBLAS (sequential version required)
LAPACK and LAPACKE
Python and Fortran compiler for wrappers and examples (optional)

For instance, on Debian-like systems, dependencies can be installed with the following command:

  sudo apt-get install cmake gcc gfortran libopenblas-dev liblapacke-dev python-numpy

The main options to configure the SPM build are:

Standard CMake options:
- CMAKE_BUILD_TYPE: Debug, RelWithDebInfo, Release, MinSizeRel; we recommend using Release or RelWithDebInfo for best performance.
- CMAKE_INSTALL_PREFIX: Specify the prefix directory to install the library.
- BUILD_SHARED_LIBS=[OFF]: Enable building shared libraries (required for the Python wrapper).
Integer type:
- SPM_INT64[=ON]: Enable/disable int64_t for integer arrays.
Documentation:
- BUILD_DOCUMENTATION[=OFF]: Enable Doxygen documentation generation.

Get involved!

Reporting an issue

We strongly recommend all users to use the issue tracker to report any problems with the software, or to submit feature requests. We will do our best to answer them in a timely manner.

Contributions

Authors

The following people contribute or contributed to the development of SPM:

Mathieu Faverge
Matthieu Kuhn
Xavier Lacoste
Selmane Lebdaoui
Grégoire Pichon
Florent Pruvost
Pierre Ramet
Theophile Terraz
Tony Delarue

If we missed your name, please let us know so we can update the list.