Automated generation of machine instruction decoders

This paper proposes a method of automated generation of machine instruction decoders for various processor architectures, mainly microcontrollers. Only minimal, high-level input from user is required: a set of assembly instruction templates and a list of register names. The method utilises the target architecture assembler to reveal the mapping of assembly-level instructions onto their binary encodings by mutating variables in the templates. The recovered mapping is then used as the central part of the architecture-independent decoder. The developed tools allow to significantly simplify the support of a large number of different processor architectures, since the proposed file format does not require high skill of the operator. At the same time, automated generation of decoders is performed much faster than manual or semi-automatic (description of the command character encodings in a certain language manually) development of a corresponding tool. A system based on the proposed method has been implemented and tested over a set of four microcontroller architectures: PIC16F877A, AVR, Tricore, H8/300H. The speed of decoding of our system is in all cases higher than that of standard tools that are in the public domain