Sizing of communication buffers for communicating signal processes

A method is presented for the minimal sizing of communication buffers in an environment of communicating signal processes. Existing methods based upon lifetime analysis techniques are mainly suited for sizing buffers in a network of processes with non-conditional interactions. The authors consider a more general problem where the execution of each process depends inherently on complex interactions with other processes as well as on its internal conditional behavior. To address the buffer sizing for this general case, the use of implicit state enumeration techniques is proposed. A dynamic data flow graph (DDF) is used as the model to describe the entire system of communicating processes. This DDF is translated into a network of finite state machine that forms state space on which methods based on implicit state space exploration are applied. To reduce the size of the state space, abstraction techniques are used to produce a simplified version in which irrelevant information is removed. The feasibility of the proposed approach is demonstrated by a practical signal processing example