Exploring parallelization for medium access schemes on many-core software defined radio architecture

Abstract

As multi-standard devices and high speed communication standards are emerging, timeliness requirements and flexibility for both baseband modem and medium access schemes are becoming essential. Software Defined Radios (SDRs), in this context, aim at offering the desired flexibility while satisfying the real-time constraints. An SDR architecture consisting of many-core homogeneous computing elements provides easy protocol implementation, a high level of portability and extension possibilities. It does not require architecture specific program code which is needed by the popular heterogeneous SDR architectures. Therefore, in this paper, we explore how a homogeneous SDR architecture is used for efficient realization and execution of Medium Access Control (MAC) protocols. In particular, we investigate the performance of two broad classes of MAC schemes on the Platform 2012 (P2012) many-core programmable computing fabric. We provide a toolchain which utilizes the characteristics of P2012 for MAC parallelization, runtime scheduling, and execution. Our results indicate that by using the supporting toolchain, reconfigurable MAC implementations are able to exploit the computational power offered by the platform and adhere to the timeliness constraints. Computationally intensive algorithms for MAC layer parameter optimization show an improvement of up to 85% in the convergence time as compared to using a single-core architecture.

Paper

“…Traditionally, the use of multi-threading and multi-processing has been confined to desktop computers and scientific computing for handling demanding computations in a timely manner. Recently, these parallelization techniques have started to gain popularity in the domain of resource constrained wireless systems due to emergence of low-power and cost-efficient multi-core/many-core processors, and increasing computing demands that need to be met [15]. Multi-threading allows an application to be partitioned into multiple semi-independent strands for parallel execution in the same address space.…”