一种矩阵相乘的并行算法实现与性能评测

摘要

With traditional serial algorithm for matrix multiplication , the result will be con-strained by matrix scale , CPU frequency , and memory size and storage space .The most ef-fective way to solve the limitation was parallel algorithm .Therefore, this paper completed the matrix multiplication with SPMD model and MPI in cluster computing environment .Experi-mental results showed that this algorithm under a certain size of the matrix has a good speed-up and efficiency .%用传统的串行算法进行矩阵相乘运算会受到矩阵规模、单机的CPU主频、内存大小和存储器空间等方面的限制。而使用并行算法是解决上述限制的最有效途径。为此，在集群计算环境下，使用SPMD计算模型和基于MPI消息传递技术设计实现了矩阵相乘的并行算法。实验表明，此并行算法在一定矩阵规模下具有较好的加速比和并行效率。