摘要: In FPGAs, the internal connections in a cluster of lookup tables (LUTs) are often fully-connected like full crossbar. Such high degree connectivity makes routing easier, but has significant area overhead. This paper explores use sparse crossbars as switch matrix inside clusters between inputs and LUT inputs. We have reduced densities these matrices by 50% or more saved from 10 to 18% with no degradation critical-path delay. To compensate for loss routability, increased compute time spare required. Further investigation may yield modest delay reductions.