Cost-Efficient SIMD ASIP Architecture for Mobile Touchscreen Controllers

This paper presented an advanced application-specific instruction-set processor (ASIP) design for touch controller systems, efficiently accelerating various signal processing algorithms for high-quality touch detection problems on resource-limited mobile devices. Mapping to the baseline ARM Cortex-M3 embedded processor, for the first time, we identify critical bottlenecks in the target touch detection algorithm by measuring the processing time of major functions. Then, an ASIP architecture is newly defined by introducing touch-aware single-instruction multiple-data (SIMD) instructions, which are specialized to the element-wise 2D matrix computing, masking, and feature extracting steps. Implementation results show that adding the proposed ASIP core to the baseline SoC solution enhances the processing efficiency of critical functions by more than 2 times while increasing the overall area costs by only 8%.