FLORA - A DATA PATH ALLOCATOR BASED ON BRANCH-AND-BOUND SEARCH

We propose for the data path allocation problem a new approach based on a branch-and-bound search algorithm. A novel target architecture originated from RISC research is used to reduce both the storage requirement and the clock cycle length. We vertically divide the problem into three subtasks: operation assignment, storage allocation, and data transfer binding. The branch-and-bound search is applicable to every subtask. To speed up the search, we derive cost functions which accurately estimate both the storage and interconnection requirements. The proposed approach has been implemented in a tool called FLORA. The strength of our approach is demonstrated through experiments.