Archive for April, 2007

Abstract Analysis of Change

Monday, April 30th, 2007 by Neil Sculthorpe

I talked about my recent thoughts on analysing the propogation of change in data-flow networks. The main idea is to label each node in the network based on how it propogates change from its input to its output. Ideally, we want to avoid re-calculating as much of the network as possible at each time step. If we can deduce, either through static analysis or run-time checks, that the output from some nodes will not change, then there is no need to re-compute that node.