https://hgpu.org/?p=24963
Irregularity Mitigation and Portability Abstractions for Accelerated Sparse Matrix Factorization