Software-Defined Radio Demodulation With FPGA-GPU Task Partitioning

The patent combines CPU, FPGA, and GPU into a hybrid processing system. The CPU handles control and coordination, the FPGA performs parallel signal processing operations, and the GPU executes computationally intensive algorithms. This merging of different processing architectures resolves the contradiction by achieving both hardware flexibility (through FPGA programmability) and sufficient computing power (through GPU parallel processing capability).

The SDR system implements multi-functionality by enabling the same hardware platform to perform various modulation schemes and signal processing algorithms through software configuration. The FPGA can be reprogrammed for different signal processing tasks, while the GPU provides universal parallel processing capability for multiple algorithms, thus achieving both adaptability and computing power.

The patent segments the signal processing tasks by assigning different functions to different processing units. The FPGA handles real-time serial processing operations with deterministic timing, while the GPU handles computationally intensive parallel operations. This segmentation resolves the contradiction by ensuring real-time performance requirements are met by the FPGA while leveraging GPU computing power for non-time-critical calculations.

The CPU acts as an intermediary that coordinates between the FPGA and GPU, managing data flow and task distribution. It assigns real-time processing tasks to the FPGA and computationally intensive tasks to the GPU, ensuring that both processing units work efficiently together to meet real-time requirements while utilizing available computing power.