GPU optimization and implementation of Gaussian belief propagation algorithm
Zahra Hosseini-Doust, Dennis D. Giannacopoulos, Warren J. Gross · 2016
We report the utilization of the parallel resources of the graphic processing unit (GPU) to solve sparse systems by optimizing and implementation of a variant of Gaussian belief propagation algorithm for sparse matrices on a Tesla 2070M GPU with the Fermi architecture. The implementation was verified with finite element method data and achieved up to 4× improvement in execution time compared to serial CPU implementation.