Configurable folded array for FIR filtering

The synthesis of configurable bit-plane processing array for FIR filtering is described in this paper. Possibilities for configuration are explored and encompassed by application of folding technique. The proposed folded architecture supports on-the-fly configuration of number of taps and coefficient length. This is achieved by dynamic operations mapping on the different hardware units in array structure. Dynamic operations mapping, involved in application of folding technique, allows recognition of user defined parameters, such as number of coefficients and coefficient length on implemented array size. The architecture provides flexible computations and offers the possibility of increasing the folded system throughput, by reducing the number of operations performed on a single functional unit, at cost of decreasing the coefficient number or length. Effects of folding technique application to architecture configuration capabilities are presented. The configurable folded FIR filter synthesis process is presented in detail. The obtained folded system architecture is described by block diagram, DFG, functional block diagram and the data flow diagram. The method of operation and operations mapping on the processing units are described. The algorithms for data reordering are given. With the aim to illustrate the functionality, configuration capabilities, and “trade-offs” relating to occupation of the chip resources and achieved throughputs of synthesized folded architecture, we present results of FPGA prototyping. The proposed configurable folded array is used for H.264/AVC deblocking filter implementation with extremely low-gate count that is achieved at the cost of time, but the design meets the requirement for real-time deblocking in mobile embedded computing platforms.