Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Parallel matrix multiplication on transputer networks: Performance analysis and estimation

Technical Report ·
OSTI ID:5361225
The paper presents a parallel matrix multiplication algorithm and discusses the studies on its performance analysis and estimation. The algorithm is implemented on a network of transputers connected in a ring topology. It optimizes execution time and memory space and achieves near-ideal speedup for large matrices by a combination of efficient partitioning and overlapped computation and communication. The execution time of the algorithm has been modelled by simple equations by analyzing its computation and communication characteristics. This has enabled estimation of its performance for a variable number of processors and for different matrix sizes.
Research Organization:
Centre for Development of Advanced Computing, Bangalore (India)
OSTI ID:
5361225
Report Number(s):
PB-91-207886/XAB
Country of Publication:
United States
Language:
English