As supercomputers evolve, can parallel computing models keep pace? This paper evaluates the performance of the LogPQ model, a practical approach to parallel computation designed for massively parallel computers. LogPQ builds upon the LogP model by incorporating communication queues, aiming to enhance the efficiency of parallel algorithms. The research focuses on a parallel matrix multiplication algorithm implemented on Cray T3E, comparing its performance against the older CM-5 machine. The results reveal that T3E's communication network exhibits superior buffering behavior, reducing the need for extra buffering. However, the effect of message size persists, highlighting the importance of overhead and gap relative to message size. The implications of this study are significant for the design and optimization of parallel computing systems. By identifying the strengths and limitations of LogPQ, the research provides valuable insights for developing efficient parallel algorithms. This work contributes to the ongoing effort to harness the full potential of massively parallel computers.
Published in the International Journal of Foundations of Computer Science, this paper aligns with the journal's focus on theoretical foundations and practical applications of computer science. By evaluating the performance of a parallel computer model, it contributes to the advancement of parallel computing techniques and fits within the journal's scope of exploring fundamental concepts in computer science.