Efficient Cosynthesis from Extended Dataflow Graphs for Multimedia Applications

This thesis addresses problems in automatic code synthesis especially for multimedia applications since automatic code synthesis is a promising high level design methodolgy for rapid prototyping as system complextiy increases and fast design turn-aroud time becomes important. However, there are some problems to support multimedia applications in the high level design methodology. Multimedia applications use composite data types, such as video frame or network packet extensively which is the major consumer of scarce memory resource. Existent dataflow models have inherent difficulty of efficiently expressing the mixture of a composite data type and its constituents: for example, a video frame and macro blocks. A video frame is regarded as a unit of data sample in integer rate dataflow graphs, and should be broken down into multiple macro blocks explicitly consuming extra memory space. Thus in this thesis we propose a new dataflow extension called fractional rate dataflow (FRDF) in which fractional number of samples can be produced or consumed to specify multimedia applications. In the proposed FRDF model, a constituent data type is considered as a fraction of the composite data type. An FRDF graph can be synthesized into efficient codes with shorter latency and less buffer memory than integer rate dataflow graphs. In addition, we propose a buffer sharing technique to minimize buffer requirement for composite type data in code synthesis step. In order to share more buffers, the proposed technique separates global memory buffers from local pointer buffers: the global buffers store live samples and the local buffers store the pointers to the global buffer entries. The technique minimizes the buffer memory by sharing global buffers for data samples of different size. An embedded multimedia system is desirable to support multiple applications by dynamically reconfiguring the system functionality, which is called multimode, to cope with rapidly evolving standards and signal processing algorithms as well as to enhance the hardware utilization significantly. A multi-mode