https://hgpu.org/?p=18960
IA-SpGEMM: An Input-aware Auto-tuning Framework for Parallel Sparse Matrix-Matrix Multiplication