High-Performance Embedded Computing

7.4 Hardware/Software Co-synthesis Algorithms 411

1. An initial processor allocation and binding is constructed by selecting for

each process the fastest PE for that process in the technology table. If the

system cannot be scheduled in this allocation, then there is no feasible solu-

tion. The initial schedule tells the algorithm the rates at which processing

elements must communicate with each other.

2. Processes are reallocated to reduce PE cost. This step is performed itera-

tively by first pairwise merging processes to try to eliminate PEs, then mov-

ing processes to balance the load. These operations are repeated until the

total cost of the hardware stops going down. At the end of this step, we have

a good approximation of the number and types ...