Compiling Devices: A Structure-Based Approach

We present an approach for compiling a device into a propositional sentence (in negation normal form) and then show how to answer interesting diagnostic queries based on such a device compilation. We present two key results about our compilation approach, which makes it of interest to the diagnostic community. First, the complexity of answering diagnostic queries | which includes nd-ing minimum-cardinality diagnoses | is linear in the size of the device compilation. Second , and more importantly, the size of the device compilation depends on the device structure in a precise way. In particular, if the device structure contains no cycles, the compilation is shown to be linear in the size of the device. In general though, the size of the device compilation is shown to be exponential only in w : the width of the graph representing the device structure. Therefore, our approach seems to be among the rst model-based approaches for compiling devices that ooer formal guarantees on the size of the device compilation.