A multimedia SOC micro-State low power design methodology (1)

0 introduction

Currently, mobile multimedia applications in consumer electronics becoming more and more important, however due to battery life, the power consumption issues have become increasingly prominent.

How to find energy efficiency and the balance between quality of service, has become the current SOC system design of a hotspots.

In SoC design, designers reuses existing proven mature IP core for designing a complex system and ensure their time to market of great significance.

Currently, many professional IP vendor provides a number of available to designers chosen IP core, designers need according to application requirements, choose the appropriate IP core and determine each IP configuration should be checked. For Mobile Multimedia SoC design, in order to achieve system-level optimization, SOC system designers of one of the core objectives is to guarantee the quality of multimedia services while making system costs (chip area and power consumption).

This article on multimedia application for encoding video and load characteristic features, from system design and optimization of hierarchical, will power management module embedded into multimedia SOC System.

At the same time, the system's operating state according to different IP configuration information into a series of State, in the work of their predecessors on the basis of the load by F-ARIMA model, while leveraging multimedia applications to measure quality of service indicators of important — deadline missing rate (deadline miss rate, DMR) as the feedback control information, a combination of both, real-time adjustment of Multimedia SoC system running status, implementation of Mobile Multimedia SOC design optimization of power consumption in the process.

1 common video encoding standards, as well as load analysis

In all of the video compression algorithm, MPEG-x and H.26y standards are gradually dominate.

These video compression algorithm, bring higher transfer efficiency, but also brought greater operation Terminal. According to the multimedia application features and coding standard, it is not difficult to find, not all multimedia video frames need decoding time are all consistent. In the frame of the MPEG compression standard, for example, the total of the three types of frame components, namely: internal frames (intra), bi-directional frame (bidirectional) and predictable frame (predictive). The three different types of frames that have different decoding complexity. Even the same type of frame, its decoding complexity is also a large difference. It is obvious that all the decoding tasks real-time constraints are set to the same criteria will cause the system to always work in the worst branch, thus paying unnecessary power consumption cost.

At present, in Multimedia SOC design process, for its important video application load characteristics for power optimization is a very hot research problems.

Common low-power design skills mainly in dynamic voltage scaling (scale, dynamic voltage DVS) and dynamic power management (dynamic power management, DPM) technology. In DVS technology, maintaining service quality at the same time for different computational tasks run on different operating voltage and frequency; while DPM technology is running, some dynamic closed system idle module. Among the many for multimedia applications for low-power design of the study, its main ideas can be divided into two categories. One is the multimedia application load as a random process, and then using Markov or semi-Markov model to predict the load, then according to the forecast results to adjust the system running. As in literature [4], a regression equation, by way of the work of the current "system" and "idle" time predicted the coming of the "work" and "idle" time. Literature [5] respectively using discrete Markov time series and continuous Markov sequence algorithm to predict system load. These are effectively reduce system power consumption cost, but its biggest disadvantage of multimedia applications and the content encoding is varied and cannot find a suitable model to suit all multimedia applications. On the other hand, there is also a study on utilization of real-time feedback controlled manner, to adjust the current system running mode to reduce power consumption in multimedia systems. As the literature [6], the author presents a show cache usage as the feedback control information, to reduce power consumption.

Currently, the most widely used video compression standard for MPEG-x (x = 1, 2, 4) series and H.26y, (y = 1, 2, 3, 4) series.

In General, video encoder consecutive images compressed into I, B, P three different frame types. I frame compression rates greater than p, P-frame compression ratio is greater than B frames. Continuous two I all frames between frames (not including the last I-frame) constitute an image group (GOP). A GOP consists of I frames interframe N, and P frames interframe M two parameters. And this corresponds, for the purposes of the decoder to decode load had I > > B the rule P.

Even if the same encoding for different scenarios, multimedia applications load also has a large difference here to H.264 video compression standard. for example, give out some common formats for video applications in the workload, see table 1.

According to table 1 of the experimental results, the figure will more figuratively l describing the work of the decoder.

From Figure 1, we can see that the decoder has only two States, that is, high speed work status and idle. Since the decoder running energy consumption calculations for

[1] [2]