Data path allocation based on bipartite weighted matching

Abstract

Proposes a graph-theoretic approach for the data path allocation problem. The problem is decomposed into three subproblems: (1) register allocation, (2) operation assignment, and (3) connection allocation. The first two subproblems are modeled as two bipartite weighted matching problems and solved using the Hungarian method. The third subproblem is solved using a greedy method. It is shown that, by taking the other into consideration while solving one, equally satisfactory results can be obtained, regardless of the order in which (1) and (2) are performed. Two programs, LYRA and ARYL, are implemented which solve the subtasks in different orders. For register allocation, the approach is the first one to guarantee minimal use of registers while being able to take the interconnection cost into account. This research has demonstrated that the bipartite weighted matching algorithm is indeed a good solution for the data path allocation problem.<>