Application-Specific Deadlock Free Wormhole Routing on Multicomputers

We show that given a priori knowledge of the messages to be routed for a specific parallel application, efficient algorithms can be designed to generate low message traffic congestion and deadlock free routings at compile-time. The problem can be formulated as a graph theoretic problem. Since the problem of finding a deadlock free routing with minimal maximum message traffic congestion is shown to be NP-hard, an efficient heuristic is proposed. Performance of the heuristic for both random message distribution and for several specific applications on hypercube and torus topologies is evaluated by simulations. We show that, compared with the E-cube and XY fixed routings, the heuristic has significant improvement with respect to maximum congestion for the specific applications as well as for the nonuniform message distribution case which models communication locality. For the uniform message distribution case, the heuristic has moderate improvement.

[1]  Frank Harary,et al.  Graph Theory , 2016 .

[2]  Dharma P. Agrawal,et al.  Scheduling pipelined communication in distributed memory multiprocessors for real-time applications , 1991, [1991] Proceedings. The 18th Annual International Symposium on Computer Architecture.

[3]  William J. Dally,et al.  Performance Analysis of k-Ary n-Cube Interconnection Networks , 1987, IEEE Trans. Computers.

[4]  Daniel H. Linder,et al.  An Adaptive and Fault Tolerant Wormhole Routing Strategy for k-Ary n-Cubes , 1994, IEEE Trans. Computers.

[5]  Alfred V. Aho,et al.  The Design and Analysis of Computer Algorithms , 1974 .

[6]  Kang G. Shin,et al.  Traffic routing for multi-computer networks with virtual cut-through capability , 1990, Proceedings.,10th International Conference on Distributed Computing Systems.

[7]  C. R. Jesshope,et al.  High performance communications in processor networks , 1989, ISCA '89.

[8]  Francine Berman,et al.  On Mapping Parallel Algorithms into Parallel Architectures , 1987, J. Parallel Distributed Comput..

[9]  Charles L. Seitz,et al.  Multicomputers: message-passing concurrent computers , 1988, Computer.

[10]  John Paul Shen,et al.  Interprocessor Traffic Scheduling Algorithm for Multiple-Processor Networks , 1987, IEEE Transactions on Computers.

[11]  H. T. Kung,et al.  Supporting systolic and memory communication in iWarp , 1990, ISCA '90.

[12]  William J. Dally,et al.  Deadlock-Free Message Routing in Multiprocessor Interconnection Networks , 1987, IEEE Transactions on Computers.