<p>Due to the flexible and robust features, recent advances emerging from computing intelligence in mechanical systems have led to many innovative techniques from signal processing to robotics. However, conventional designs of mechanical computing are typically manifested as deterministic functions with single-task layouts, which show limitations in versatility. This work proposes a versatile mechanical computing architecture, which can integrate both analog and logic operations into a unified robotic system. Inspired by Fourier optics, elastic wave metamaterials are introduced to spatial analog computing with robot crawling, enabling flexible spatiotemporal modulations and self-regulating locomotion. Using serial/parallel strategies, the metamaterial allows nested and cascaded designs from combinational analogs to the extended binary logics. In particular, by encoding analog signals as mechanical qubit-like states, the system can realize higher dimensional logic mappings through reprogrammable matrix operations. Beyond single-task layouts, nonlinear harmonics are also introduced to realize the parallel computing of multiple tasks or their transformation to specific tasks. The versatile strategies are shown to support the parallelism, integration and transformation from analogs to logics, which wish to offer an effective approach for incorporating mechanical intelligence in robotics.</p>