rv-sparse: Open-source RISC-V Vector accelerated sparse linear algebra library (RISC-V Mentorship)

  • Internship
  • Remote

Mentors: Talha Ahmed and Shayan Hassan Baig

Company: Micro Electronics Research Lab – UIT (MERL-UIT)

Sparse linear algebra is fundamental to many domains including scientific computing, machine learning, graph analytics, and large-scale simulations. In modern machine learning workloads, especially pruned neural networks, sparsity is intentionally introduced to reduce computation and memory usage. While several specialized hardware accelerators exist for efficiently executing sparse matrix operations in such workloads, these solutions rely on dedicated hardware and are not always available in general-purpose systems.

Within the RISC-V ecosystem, optimized libraries such as OpenBLAS provide highly efficient implementations for dense linear algebra operations. However, similar support for sparse matrix multiplication optimized for the RISC-V Vector Extension remains limited.

This project aims to develop a lightweight C library that implements key sparse linear algebra primitives particularly sparse matrix multiplication using RISC-V vector intrinsics. The project will handle CSR and CSC sparse storage formats, vectorization strategies, and memory-efficient traversal techniques to better exploit data-level parallelism on RVV-enabled processors. The outcome will include a clean API for sparse matmul operations, optimized implementations using vector intrinsics, and well-documented source code. The goal is to provide an accelerator-independent software solution that enables efficient sparse computation on general-purpose RISC-V vector processors.

Repository URL: https://github.com/merledu/rv-sparse

Learning Objectives:

  • Familiarity with open-source workflows using Git, GitHub and Linux development environments.
  • Contribution to open source projects via pull requests, issue tracking, and collaborative code reviews.
  • Utilization of RISC-V vector intrinsics in C code.
  • Writing RISC-V vector kernels.

Coding Challenge:

https://docs.google.com/document/d/15LcAv0bXG6-J-n7p6rQUkhout0UnwftpRGL43cbOrUg/edit?usp=sharing

To apply for this job please visit mentorship.lfx.linuxfoundation.org.