Abstract
A novel approach to the operation scheduling problem in an automated data path synthesis system is presented. After the start time and the required time of each operation is obtained by the ASAP (as soon as possible) and ALAP (as late as possible) method, a linear integer programming approach is formulated to fully utilize the hardware resources, i.e. to minimize the requirement of function units under the given timing constraint. The formulation is generalized to deal with multicycle operations, multiple operations per cycle, pipelined data paths and mutually exclusive operations. Experimental results are presented for example problems, illustrating the performance of the LIP model. The problems were solved using the LINDO package, which uses a branch-and-bound strategy, on the VAX-11/8800. LIP yielded optimal solutions in seconds for all of the available benchmarks.