Evaluation of DLX Microprocessor Instructions Efficiency for Image Compression

Internet of Things (IoT) implementation usually uses a generic microprocessor, especially when the implementation is for a specific functionality where ASIC (Application-specific Integrated Circuit) has not yet available. However, a generic microprocessor originally designed for general purpose application that uses many machine instructions, which in return causes an inefficient power consumption. Although ASIC is the best choice when comes to power efficiency, ASIC is hardcoded, meaning that it is not programmable or is not flexible for future enhancement or to accommodate intelligent properties like ability to learn and grow. This research tries to find the minimum instruction set to propose an ASIP (Application-specific Instruction Set Processor) for application in high quality image compression coming from CCTV. This research uses the DLX microprocessor instruction set as a reference on Huffman Coding as part of JPEG compression. The research shows that Huffman Coding needs only 9 type of machine instruction, they are ADD, ADDI, SUBI, LW, SW, BEQZ, BNEZ, SEQ, and SGT.