Circuit design for logic automata

Abstract

The Logic Automata model is a universal distributed computing structure which pushes parallelism to the bit-level extreme. This new model drastically differs from conventional computer architectures in that it exposes, rather than hides, the physics underlying the computation by accommodating data processing and storage in a local and distributed manner. Based on Logic Automata, highly scalable computing structures for digital and analog processing have been developed; and they are verified at the transistor level in this thesis. The Asynchronous Logic Automata (ALA) model is derived by adding the temporal locality, i.e., the asynchrony in data exchanges, in addition to the spacial locality of the Logic Automata model. As a demonstration of this incrementally extensible, clockless structure, we designed an ALA cell library in 90 nm CMOS technology and established a "pick-and-place" design flow for fast ALA circuit layout. The work flow gracefully aligns the description of computer programs and circuit realizations, providing a simpler and more scalable solution for Application Specific Integrated Circuit (ASIC) designs, which are currently limited by global constraints such as the clock and long interconnects. The potential of the ALA circuit design flow is tested with example applications for mathematical operations. The same Logic Automata model can also be augmented by relaxing the digital states into analog ones for interesting analog computations. The Analog Logic Automata (AnLA) model is a merge of the Analog Logic principle and the Logic Automata architecture, in which efficient processing is embedded onto a scalable construction.

(cont.) In order to study the unique property of this mixed-signal computing structure, we designed and fabricated an AnLA test chip in AMI 0.5[mu]m CMOS technology. Chip tests of an AnLA Noise-Locked Loop (NLL) circuit as well as application tests of AnLA image processing and Error-Correcting Code (ECC) decoding, show large potential of the AnLA structure.