Modern GPUs are designed for regular problems and suffer from load imbalance when processing irregular data. Prior to our work, a domain expert selects the best kernel to map fine-grained irregular parallelism to a GPU. We instead propose Seer, an abstraction for producing a simple, reproduceable, and understandable decision tree selector model which performs runtime kernel selection for irregular workloads. To showcase our framework, we conduct a case study in Sparse Matrix Vector Multiplication (SpMV), in which Seer predicts the best strategy for a given dataset with an improvement of 2× over the best single iteration kernel across the entire SuiteSparse Matrix Collection dataset.
@article{arxiv.2403.17017,
title = {Seer: Predictive Runtime Kernel Selection for Irregular Problems},
author = {Ryan Swann and Muhammad Osama and Karthik Sangaiah and Jalal Mahmud},
journal= {arXiv preprint arXiv:2403.17017},
year = {2024}
}