Ripples: Parallel Influence Maximization for CPU-GPU Distributed PlatformsAn
open source library written in C++, MPI, OpenMP and
CUDA for influence maximization on CPU-GPU
distributed platforms.
Download
source code from Github Developer:
Marco Minutoli; Contributors:
Mahantesh Halappanavar, Ananth Kalyanaraman, Antonino
Tumeo, and Maurizio Drocco (An
older implementation using C++ and OpenMP is also
available upon request) License: BSD 3-Clause license (Open Source Initiative) Our new CPU+GPU implementation ,CuRipples, achieves a speedup of 790× over a state-of-the-art serial implementation (left), while also significantly improving the approximation factor (to ?=0.13) and doubling the number of seeds (right). The input network is com-Orkut. GTC 2020: Accelerating Graph Algorithms on Exascale Systems ICS 2020 (Slides) Super Computing Conference 2020 (Slides) Protecting Essential Connections in a Tangled Web: A promotional article on this work. (PDF copy on this site) HPCWire feature article : PNNL Develops Speedy Supercomputer-Powered Network Analysis Tool. (PDF copy on this site) |