T.Q. Viet and Y. Tsutomu (Japan)
Asynchronous Parallel Programming Model, Flat MPI, Matrix Multiplication, SUMMA, SMP cluster
Our study proposes a novel MPI-only parallel program ming model with improved performance for SMP clusters. By rescheduling tasks in a typical flat MPI solution, our model forces processors of an SMP node to work in differ ent phases, thereby avoiding unneccessary communication and computation bottlenecks. This study achieves a signif icant performance improvement with a minimal program ming effort. In comparison with a de-facto flat MPI solu tion, our algorithm can yield a 21% performance improve ment for a 16-node cluster of Xeon dual-processor SMPs while performing a distributed matrix multiplication.
Important Links:
Go Back
