网络并行计算中矩阵QR分解的并行算法

随着高速网络技术（如ＡＴＭ）的出现，网络并行计算系统（ＮＯＷ）已成为并行处理的主要平台，由于它的高通信延迟，某些在并行机上实现的细粒度并行算法已不适合在该环境下运行。为此，有必要对算法重新进行任务划分，研究它在网络环境中的并行实现。基于这一点，本文对矩阵的ＱＲ分解提出了一种新的任务划分策略，并由此得到了它的一种粗粒度并行算法，实验结果表明，设计的并行算法在网络并行计算环境中具有较高的加速比。

THE PARALLEL ALGORITHM OF MATRIX QR DECOMPOSITION ON NETWORK OF WORKSTATIONS

With the advances in the high speed computers network technologies such as ATM, a network of workstations is becoming the main environment for parallel processing. Due to its high communication latency, some fine granularity algorithms that have been realized in parallel computers are not fit to NOW, so it is necessary to re partition the task and research its parallel implement on NOW. Based on this point, this paper proposes a new task partition scheme for matrix QR decomposition, and a new coarse granularity parallel algorithm is given. The experiment result show that the designed parallel algorithm has high speedup on NOW.