The Research And Implementation Of Multi-core Collaborative Computing Platform

The focus of modern processor architecture has shifted from the increase of clock speeds to the amount of parallelism. Multi-core processors have replaced single-core processers and have become the main route of enhancing processor performance.In terms of the processor’s structure, there are two kinds of multi-core processor. One is homogeneous multi-core processor, in which the kernels are identical and share the same position, and can be regarded as the chip of the SMP architecture. The other one is heterogeneous multi-core processor and its kernels are not identical and have different status. Because heterogeneous multi-core use a variety of kernels with different functions, each kernel can finish the task they are good at to improve the performance of the whole system and meet demands of embedded applications in different areas such as chip size, calculation performance and power dissipation. These advantages also make heterogeneous multi-core widely applied in the embedded computing.How to make full use of multi-core advantages to accomplish certain tasks collaboratively becomes the focus of current research. Different from the traditional single-core platform, multi-core platform can run different operating systems on different processor cores; requires different compilers and programming models, which brought difficulties for the software interface and information interaction between multi cores. Based on the investigation of many multi-core collaborative computing platforms, for embedded real-time application, this thesis implements a multi-core cooperative computing platform DualOSCCP on the PB11 MPCore, a four homogeneous cores development board.In order to facilitate the system transplant, DualOSCCP adopts the design idea of heterogeneous multi-core platform, regards PB11 MPCore as a heterogeneous multi-core platform, the CPU0 and the three other CPU1-3 as the different processor cores, constructed a dual operating systems in heterogeneous environment, CPU0 is running ARM Linux, the other three cores run aCoral in SMP way. aCoral is a lightweight real-time operating system with high efficiency, and can ensure tasks in real time. Linux and aCoral use different compilers to compile, and the codes running on different operating systems must be written and compiled respectively, which enhanced the portability in heterogeneous cores. For example, on the ARM+DSP platform, ARM and DSP using different compiler, porting aCoral to DSP, DualOSCCP can be easily ported to the platform. For different types of tasks, DualOSCCP can make these tasks selectively be allocated to run on different operating systems, such as a single IO task may be chose to perform on the Linux, and multi real-time computing tasks can be selected for parallel execution on aCoral, so that the overall system performance is greatly improved.