Industry Articles

Wayne Wolf
EETimes (6/21/2011 12:08 AM EDT)

Following on Part 1 and Part 2 in this series, we now will reconsider hardware/software co-design for more general multiprocessor architectures.

Many useful systems can be designed by hardware/software partitioning algorithms based on the CPU+ accelerator template. Hardware/software partitioning can also be used to design PEs that are part of larger multiprocessors.

But if we want to design a complete application-specific multiprocessor system, we need to use more general co-synthesis algorithms that do not rely on the CPU+ accelerator template.

In the most general case, all these tasks are related. Different partitions of functionality into processes clearly changes scheduling and allocation. Even if we chose a partitioning of functions, scheduling, allocating, and binding are closely related.

We want to choose the processing element for a process - both the general allocation and the binding to a specific type - based on the overall system schedule and when that process has to finish.

But we can't determine the schedule and the completion time of a process until we at least choose an allocation and most likely a binding. This is the Gordian knot that co-synthesis designers must face-the set of intertwined problems that must somehow be unraveled.