Roughly a year ago we talked about Vennsa’s OnPoint tool for identifying what went wrong during verification when something goes wrong. I got an update at DAC recently, where they talked about two concepts they’ve brought to their technology in order to make it easier to decide what to fix when there’s a problem.
The first is that of triage, which automatically tries to combine different failures if they appear to have the same root cause. Prior to this, you would manually ask the tool for possible root causes for each failure; now this evaluation and “binning” (more or less) occurs automatically.
The thing is, any given failure may have more than one candidate root cause, and there may be several failures, some of whose candidate root causes overlap. Understanding this picture and the conclusions the tool draws from it falls into the second new concept, which they call causality analysis. This involves a more precise understanding of all root cause candidates and the trajectories of the possible fixes – that is, the knock-on implications of each fix.
In particular, where a given fix candidate is a candidate for several failures, that fix will get a higher ranking than if it only solves a single failure.
In evaluating which fixes to suggest and how to rank them, it also takes into account the complete set of facts regarding which checkers may have blessed or frowned on various parts of the simulation; any available information regarding the desired state of the system at the failed point; whether a given fix would break something else; and whether it would be better to apply a fix further upstream or downstream in the logic.
One key goal of the improvements is to make it easier to figure out who owns a fix. If a set of failures gets grouped during triage, it’s more evident that there’s an owner for the entire group.
More info in their release…