Performance Analysis for Streaming Applications on heterogeneous SoC architectures: Reconfigurable and Multi-cored
Speaker: Jun Zhu
We present our work in two parts on performance analysis at design time for heterogeneous SoCs. Our work targets on two SoC architectures respectively: run-time reconfigurable FPGA and the hybrid multi-core CPU/FPGA.
On the run-time reconfigurable FPGA architecture, we propose a design optimization framework for adaptive real-time streaming applications. The main contribution is a hybrid approach for performance analysis combining formal analysis and simulation using a twophase framework. We formulate the scheduling problem of adaptive streaming applications with ILP analysis, and use the simulation based on the synchronous model of computation to ensure throughput guarantees. We finally illustrate the capabilities of our methodology by experiments.
On the hybrid multi-core CPU/FPGA architecture, we address the problem of real-time streaming applications scheduling with optimized buffer requirement. The main contribution is a two-step approach to minimize the buffer requirement for streaming applications with throughput guarantees. A novel declarative way of constraint based scheduling for real-time hybrid SW/HW systems is proposed, while the application throughput is guaranteed by periodic phases in execution. We use a voice-band modem application to exemplify the scheduling capabilities of our method. The experimental results show the advantages of our techniques in both less buffer requirement and higher throughput guarantees compared to the traditional PAPS method.
In the future, we aim at leveraging our performance analysis techniques (i.e. system level analysis, simulation on model of computation, constraint based scheduling, etc), to provide design space exploration frameworks for the critical design issues (e.g. throughput guarantees, buffer requirement, energy dissipations, etc) on state of the art SoC architectures.
