A new algorithm for dynamic programming on linear arrays is presented which uses a single data path and runs twice as fast using less than half the memory locations of the best previously known algorithm. The algorithm employs a redundant data technique in which a permuted copy of the dynamic programming table is maintained to reduce communication costs. The simplified control structure and exclusive use of shift registers for storage (i.e. no RAM is required) make the algorithm particularly suitable for VLSI implementation
Published in:
Parallel and Distributed Processing, 1994. Proceedings. Sixth IEEE Symposium on
Date of Conference: 26-29 Oct 1994