Parallel QR Decomposition for Electromagnetic Scattering Problems

摘要

This report introduces a new parallel QR decomposition algorithm. Test resultsare presented for several problem sizes, numbers of processors, and data from the electromagnetic scattering problem domain. The development of the algorithm marks a departure from past parallel QR algorithms. The load balancing method used considers total computational work as opposed to just balancing Givens rotations. This results in expected efficiencies which approach optimal as problem size grows relative to number of processors. The hybrid nature of the algorithm, which maximizes computation between communication and synchronization, indicates potential for good performance on distributed memory machines and networks of workstations. Implementation results on shared memory and distributed shared memory architectures show promise and track expected performance well up to 12 processors.