Software Efforts
The members of the HPCA group have been mainly involved in the development of parallel routines and, sometimes, extense parallel libraries mainly for numerical linear algebra and the solution of control problems. As a result we can offer a variety of software packages. Some of this software is also offered as part of public libraries as LAPACK and libFLAME, and licensed libraries as SLICOT.
See our GitHub
Ongoing Software Efforts
PMLib: Power Measurement Library
PMLib is a modular open-source power measurement library to investigate power and energy usage of high-performance computing applications.
Visit the PMLib GitHub.
PyDTNN: Python Distributed Training of Neural Networks
PyDTNN is a light-weight framework developed at Universitat Jaume I for distributed Deep Learning training and inference that offers an initial starting point for interaction with deep neural networks.
Visit the PyDTNN GitHub
Proteo: a framework for the generation and evaluation of malleable MPI applications
Proteo is a configurable framework that allows the design of benchmarks to study the effect of malleability on a system, and also to incorporate malleability into real applications. It consists of SAM, an emulator, and MaM, a malleability library.
Visit the Proteo GitHub
Past Software Efforts
FLAME: Formal Linear Algebra Method Environment
High performance linear algebra libraries for current desktop processors, multicore architectures, and SMP platforms. This is a collaboration with The University of Texas at Austin.
Visit the FLAME web site.
ParILUPACK
Parallel preconditioning based on the incomplete LU factorization for multicore processors. This is a collaboration with TU-Braunschweig.
Please contact mvaya@uji.es to learn more about this effort.
MAGMA-sparse
MAGMA-sparse is a library for the solution of sparse linear systems via iterative methods on GPUs. It is developed in collaboration with Karlsruhe Institute of Technology and ICL-Tennessee.
Visit GLT web site.
Contact flegar@uji.es to learn more about this effort.
The GLT Library
Generic Lightweight Thread (GLT) is a common API that joins all the functionality offered by several lightweight thread solutions.
Visit GLT web site.
Contact adcastel@uji.es to learn more about this effort.
H2Lib+OmpS
H2Lib+OmpSs is a task-parallel version of the H2Lib libraries for dense linear algebra operations on hierarchical matrices. This is a collaboration with the University of Kiel and Barcelona Supercomputing Center.
Visit GLT web site.
Contact rcarrata@uji.es to learn more about this effort.
HPG-Aligner
HPG-Aligner is a parallel pipeline for RNA and DNA sequencing on clusters of multicore processors.
Visit GLT web site.
Contact martineh@uji.es to learn more about this effort.
GPGPU BLAS and LAPACK
Tuned implementation of the BLAS-3 and part of the LAPACK libraries for platforms equipped with one or more NVIDIA GPUs.
Visit GPGPU BLAS and LAPACK web site.
Contact figual@icc.uji.es to learn more about this effort.
EnergySaving Cluster
Power Saving System for HPC Clusters by Selective Activation/Deactivation of Nodes.
Visit EnergySaving web site (sorry, this information is only provided in Spanish).
Contact dolzm@icc.uji.es to learn more about this effort.
The rCUDA™ Framework
rCUDA™ is a Framework that allows a remote GPU usage using CUDA. It has been developed in a joint collaboration with the Grupo de Arquitecturas Paralelas from the Universidad Politécnica de Valencia. For further information, go to the rCUDA™ web page on this site.
COPABIB: Automatic Construction and Parallelization of Scientific Libraries
Codes for the solution of mathematical problems arising in linear algebra, discrete mathematics, and optimization on heterogeneous parallel computers, multi-core processors, and GPUs.
Visit the COPABIB web site.
PLiCMR: Parallel Library for Model Reduction
Codes for model reduction of large-scale time-invariant linear control systems in state-space form on parallel distributed computers.
Visit the PLiCMR Library Documentation.
Contact gquintan@icc.uji.es to get the lattest release of the PLiCMR Library.
SpaRed: A Parallel Library for Model Reduction of Large Sparse Linear Systems
Codes for model reduction of large-scale time-invariant linear control systems in state-space form, with a sparse or banded state matrix, on parallel distributed computers.
Visit the SpaRed page, with a more recent version of the library.
COMPARHE: Parallel Computing on Heterogeneous Platforms
Codes for the solution of mathematical problems arising in linear algebra, discrete mathematics, and optimization on heterogeneous parallel computers.
Visit the COMPARHE web site.
PELICANUM: Parallelization and Extension of the GNU Scientific Library – Numerical Algorithms
GNU-like codes for sparse linear algebra computations.
Visit the PELICAN page for more information.
PLiC Library: Parallel Library for Control
Codes for solving problems arising in the analysis and design of large-scale time-invariant linear control systems in state-space form on parallel distributed computers.
Visit the PLiC Library Documentation.
Contact gquintan@icc.uji.es to get the lattest release of the PLiC Library.
BandSolve
Implementations of band routines for the BLAS (Basic Linear Algebra Subprograms) and LAPACK factorization routines for the solution of linear systems on high performance processors and multicore architectures.
Please contact remon@icc.uji.es to learn more about this effort.
PLiCOC: Parallel Library for Optimal Control
Codes for optimal control of large-scale time-invariant linear control systems in state-space form on parallel distributed computers.
Visit the PLiCOC Library Documentation.
Contact gquintan@icc.uji.es to get the lattest release of the PLiCOC Library.
PQZ Library
Serial Fortran-77 and Matlab codes for computing the periodic QZ decomposition of a generalized product of matrices. This is a beta release.
Download Fortran-77 PQZ Routines.
Download Matlab PQZ Routines.
QP3 Library
Fortran-77 code for computing QR factorization with column pivoting using BLAS-3 operations. The serial version of this code is included in LAPACK 3.0.
Download the QP3 codes.
There are also ScaLAPACK and PLAPACK versions of the code that can be obtained by request from the authors (gquintan@icc.uji.es).
RRQR Library
Fortran-77 codes for computing Rank-Revealing QR Factorizations of matrices using BLAS-3. The methods in this library are both faster and more efficient than QP3 and have been included in the Boeing BCSLIB Library.
Download RRQR Library.
There is also a ScaLAPACK version of the codes that can be obtained by request from the authors (gquintan@icc.uji.es).