Sparse Matrix-vector Multiplication on GPGPU Clusters: A New Storage Format and a Scalable Implementation | IEEE Conference Publication | IEEE Xplore