The Specific-Function-Oriented Programmable Architecture Investigation And Practices In Advanced Video Decoding Chip

With the development of multimedia and wideband technology, video decoder will get more and more widely used. However, faced with this fast growing technique, the classic block-based hybrid coding method has never been changed in essence. Even though the fast and optimized arithmetic for the single-function oriented submodule has been emerging constantly, the way to realize its basic function and the general idea remain the same.To keep up with the fast growth of multimedia and IC design, current development cycle of VLSI must be shortened in order to be more competitive. With the same thing happening to the multimedia hardware video decoder, generally speaking, as the emergence of some advanced video decoding standard, correspondent hardware video decoder will come out soon. Though development cycle is short, its structural design greatly depends on the certain standard. With the subtle improvement on the standard's arithmetic or the increasing update of video decoding standard, such design closely targeted at the standard may fail in compatibility or long lifecycle, and be knocked out easily.Therefore, specific function oriented generality study may not only shorten the R&D cycle of video decoding standard for certain function subunit, but also facilitate the utilization of systems designed on the principle of generality, which is in line with the basic rule of reusablility.According to the algorithmic characteristics of video coding, general video decoder can be classified by three classic tasks. First is the low-leveled task for simple control, which focuses on calculation. It is called dense-calculation task when the requirement for calculation is higher. Second is the data-dependence task for simple calculation but sophisticate control, which belongs to middle-level task. More sub-states exist inside the task and the transition of the sub-states depends on input data and lacks of rules. Third is the simple-system task, part of advanced tasks, with simple calculation but no rules, the complexity of which mainly lies in the control of multi-tasks to complete certain function coordinately.These three tasks generally constitute the whole video decoding system. Starting from principle of the specific function oriented generality, with the three tasks classified by the algorithmic characteristics of video coding as objects, this article makes general structural designand study of the classic representatives of these tasks-sub-pixel interpolation, variable lengthdecoding and motion compensation system from the angle of structure design and methodology. The article covers the general structural design of dense-calculation task, classic structure of variable length decoding, improvement of general structure for updated standard compatible and general design rules for simple system. The programmable abstract FSM and the IP-oriented interconnection structure are investigated preliminarily for the more general purpose.