* If you want to update the article please login/register
As the iteration progresses, we discuss the use of a communication-reduction system for the PageRank algorithm that dynamically adapts the data access's precision to the algorithm's numerical specifications. Our numerical experiments on an NVIDIA V100 GPU and a server with two Intel Xeon Platinum 8168 CPUs reveal that, relative to a standard IEEE double-precision setup, the CPMS-based PageRank completes about 10% faster, while reduced output accuracy is acceptable, with reduced output accuracy being acceptable.
Source link: https://doi.org/10.1145/3380934
* Please keep in mind that all text is summarized by machine, we do not bear any responsibility, and you should always check original source before taking any actions