HiCMA Library: Hierarchical Matrix Computations for Manycore Architectures

HiCMA is a library for hierarchical matrix computations on manycore architectures, presented at the Euro-Par 2018 conference. The repository contains the library and scripts for reproducing results from the paper 'Exploiting Data Sparsity for Large-Scale Matrix Computations'. It includes a high-performance implementation of Cholesky factorization for distributed-memory systems.

Use Cases

• Benchmarking hierarchical matrix algorithms based on the library's implementation of Cholesky factorization.
• Reproducing performance results for large-scale matrix computations on manycore architectures as described in the paper.
• Developing extensions for exploiting data sparsity in dense matrices on shared to distributed-memory systems.

Strengths

• Includes a high-performance implementation of Cholesky factorization for distributed-memory systems.
• Provides scripts for reproducing the results from a peer-reviewed Euro-Par 2018 conference paper.
• The library is designed to exploit the data sparsity structure of dense matrices.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Kadir Akbudak

Collection Method

Software library and reproduction scripts for a research paper.