Learning Sparse Matrix Row Permutations for Efficient SpMM on GPU Architectures | IEEE Conference Publication | IEEE Xplore