Parallel Algorithms for Linear Approximation on Distributed Memory Machines

(Summary of Results) Abstract In this paper, we summarize our results in paralleliz-ing the linear approximation step on current distributed memory machines. We rst analyze the features of current distributed memory machines and the problem characteristics to understand the overheads in parallel solutions to the problem. Based on these, we propose an asynchronous algorithm which enhances processor utilization and overlaps communication with computation by maintaining algorithmic threads in each processing node. Our implementation shows that, given a 512512 image, the linear approximation task can be performed in 0.015 seconds on a SP-2 having 64 processing nodes and in 0.032 seconds on a T3D having 32 processing nodes. A serial implementation takes 0.445 seconds on a single processing node of SP-2 and 0.779 seconds on a single processing node of T3D. Experimental results on various sizes of images using 4, 8, 16, 32, and 64 processing nodes are also reported.