HiCMA (Hierarchical Computations on Manycore Architectures) library presented in Euro-Par'18 paper
2018-08-28T10:38:55Z (GMT) by
This file contains the HiCMA library and scripts for reproducing the results presented in the Euro-Par 2018 conference paper entitled "Exploiting Data Sparsity for Large-Scale Matrix Computations".
The aim of this paper was to use the Hierarchical matrix Computations on Manycore Architectures (HiCMA) library as the basis to address the data sparsity structure of dense matrices on shared and shared to distributed-memory systems. HiCMA was extended to provide a high-performance implementation of the Cholesky factorization on distributed-memory systems.
A complete list of HiCMA features can be found at: https://github.com/ecrc/hicma.
The dataset is compressed in .gz format and can be uncompressed by standard compression utilities. The data are split into three archives, distmem, shmem and hicma-dev.
A PDF README file is included which provides installation instructions for an Ubuntu-based system.