Universal Wireless Signal Multithreaded Parallel Processing Technology On GPP Platform

With the rapid development of mobile access technologies, in order to meet the increasing demand, and respond to the increasingly fierce competition in the mobile market, Third Generation Partnership Project has been on long-term evolution of mobile communication systems technology for a number of in-depth research and promotion. The real-time communication requirements have higher dependence on throughput and latency on PUSCH link process. Meanwhile, driven by Moore’s Law, the general purpose processor computing power is growing, multi-core technology, and innovative chip architecture also make wireless baseband signal realtime processing on GPP possible.Firstly, the study analyzes the main difficulties of multicore computing and architectures of multi-core CPU, so clearly the goals of design and optimizatio. Then introduce the multicore processors of Intel universal development platform to build the methods of the design and optimization of the software development process. Then according to the characteristics of PUSCH link to Proposed a series of multi-threaded implementation of progressive improvement program:(1) Research Turbo transmission block structures and processes characteristic of all decoded code blocks, this paper presents a parallel scheduling program based on Pthread:all of the code block being into multiple threads to the parallel processing of code-block-level, and achieve the common tasks scheduling for each code block.(2) According to the idea of the pipeline processing, the paper proposes to divided a single sub-frame of PUSCH into multiple task groups, each task group with one or more threads to achieve, then support the task groups to be parallel processing between each sub-frame.(3) According to the characteristics of OpenMP dynamic scheduling, propose an optimization scheme which enable to achieve presents a symbol-level task force to achieve scalable parallel processing in the symbol-level task groups to support adaptive migration based Bogomips in different servers.The main subject of the above generic multi-threaded progressive parallel processing algorithms, based on a lot of Pthread and OpenMP writing code and optimization, while taking advantage of the VTune Performance Analyzer for the thread liquidity and CPU usage analysis. Thus completing a comprehensive validation of the PUSCH link optimized parallel scheduling policy, and the performance of the test work. The results show that reducing the total delay of a single sub-frame processing to less than lms, is good to meet the LTE physical layer data decoding system for real-time requirements that must be within2ms, allow enough time to the subsequent processing system. It is possible to improve parallel processing strategy for the versatility of various LTE transmission configuration, flexible scheduling and scalability, while maintaining a high level of CPU usage.