Bit-width selection for data-path implementations

Specifications of data computations may not necessar-ily describe the ranges of the intermediate results that can be generated. However, such information is critical to de-termine the bit-widths of the resources required for a data-path implementation. In this paper, we present a novel ap-proach based on interval computations that provides, not only guaranteed range estimates that take into account de-pendencies between variables, but estimates of their prob-ability density functions that can be used when some trun-cation must be performed due to constraints in the specifi-cation. Results show that interval-based estimates are ob-tained in reasonable times and are more accurate than those provided by independent range computation, thus leading to substantial reductions in area and latency of the corre-sponding data-path implementation. 1