The final project in the course “Accelerator-based Programming” consisted of implementing the SELL-C-sigma sparse matrix storage format. I found this project very interesting but also quite challenging. For the matrix initialization, I used a matrix that is stored in compressed row storage (CRS) format and converted this to SELL-C-sigma.
SELL-C-sigma is a very intriguing format as it is designed to work well for vectorization, multicore processing and GPUs without any need of matrix format change. Therefore, the format is largely hardware-agnostic and well suited for hybrid heterogeneous systems.
To understand how SELL-C-sigma works it was helpful to first look at CRS, Jagged Diagonal Storage (JDS), ELLPACK and Sliced ELLPACK. I created the following graphics to help me understand these formats.
The following images are licenced under CC BY-NC.
