This project introduces sTiles, an accelerated computational framework for sparse factorizations of structured matrices, built to enhance the INLA method using HPC tools.
Highlights
📄 Papers
-
sTiles: An Accelerated Computational Framework for Sparse Factorizations of Structured Matrices
Focused on the sTiles solver. Accepted at ISC High Performance 2025. -
GPU-Accelerated Parallel Selected Inversion for Structured Matrices Using sTiles
Focused on the selected inversion. Currently under review. -
Elevating INLA: Next-Level Speed with sTiles
A broader application-focused paper. Writing in progress.
🛠️ Code & Installation
Find the full codebase and installation instructions here: GitHub Repository
🎤 Presentations on sTiles
-
From Single Core to Many Cores to GPUs
INLA Workshop, University of Glasgow, Scotland (2025)
[View Slides PDF] -
sTiles: An Accelerated Computational Framework for Sparse Factorizations of Structured Matrices
ISC High Performance 2025, Hamburg, Germany
[View Slides PDF]
While there is some overlap between the two slide decks, the presentation at the INLA Workshop is tailored for statisticians, whereas the ISC talk targets an HPC audience. For a complete understanding, we recommend consulting the published papers above.
📬 Stay Connected
If you're interested in joining the sTiles email list to receive updates about the project, or if you'd like to contribute to the development of a Python wrapper, please feel free to contact me.
I'm also happy to receive test matrices that perform poorly with the current solver, we can explore customizations to improve performance and robustness.