Research On Key Technology Of Multi - Core Processor

Along with semiconductor manufacturing process reaching nearer the bottleneck, single-core processor’s performance has limited space to enhance. Multi-core processors become the mainstream research direction to enhance the performance of the processor. For mobile communications applications, multi-core processors have high flexibility, shorter development cycles, and compatible with more communication standards. Recently, the industry has introduced8-core processor for communication applicatoins. We can see multi-core platform is a promising solution in the field of communications.This paper studies the LTE applications key modules, including the algorithm of channel estimation and MIMO detection and the corresponding physical implementation based on multicore processors, constructs a comparative analysis of the advantages and disadvantages of multi-core processors, and proposes reliable methods to improve the performance of multi-core processors for communications applications.The main work of this dissertation include:1) This dissertation analyzes the characteristics of the wireless channel and the non-ideal factors affecting the performance of the communication system, and introduces a spatial diversity MIMO channel model.2) It constructs a comparative performance analysis of the least squares estimation, linear estimation, Wiener filter estimation. This work uses the adaptive algorithm with flexible filter coefficients and order according to the channel conditions. This self-adaptive algorithm makes a valid compromise between performance and computational complexity.3) This channel estimator based on the multi-core processor takes full advantage of the parallelism of the algorithm, including the utilization of SIMD features, task-level parallelism, pipelined architecture, execution array. Inter-core communication between is another important factor affecting system performance. This estimator is highly efficient by reasonable mapping of each task, selecting the appropriate inter-core communication method.4) MIMO detection is the key to modern communication systems’exponentially increasing throughput. This dissertation analyzes the main difference among metric first, breadth-first, depth-first, and summarizes the key constraints of various algorithms that could affect the compromise between performance and complexity. This work propose a more practical method to implement MIMO detector on multi-core by adding different Euclidean distance constraints and maximum expand number of nodes on the the basis of FDBF algorithm which effectively reduces the computational complexity.5) FDBF is an algorithm executed in sequential order, pipeline structure is the efficient way to improve performance while task-level parallelism and SIMD is not appropriate in this situation. It achieves high performance by partitioning the algorithm by balanced computational load of the sub-tasks. This MIMO detector’s performance is enhanced further by using Cordic unit and demap unit in execution array. The test results show that the array unit has significantly accelerated the performance of LTE applications.