Video Not Available
Details
Abstract
In this paper, the NoCAPA is described with a pipelined structure for fully-connected Ising machine with a NoC based communication and computing paradigm, and a distributed data storage strategy. To enhance the convergence efficiency, a novel parallel multi-spin update algorithm is also proposed, achieving an iteration speed up by maximum 32X. The implementation results demonstrate that our system achieves high scalability, high operating frequency and parallel update, while consuming much less hardware resources on the FPGA prototype.