A Hybrid Design Space Exploration Approach for a Coarse-Grained Reconfigurable Accelerator

Multitude parameters involved in the design process of a reconfigurable accelerator which is exploited in embedded systems brings about a remarkable complexity and large design space. One effective technique is design space exploration which is capable to find a right balance between the different design parameters. Quantitative design approach is an alternative which uses the data collected from applications; however it is time consuming and highly depends on designer observations and analyses and might not conclude to an optimal design. In this paper, a hybrid approach is introduced which uses an analytical approach to explore the design space for a reconfigurable accelerator and determine a wise design point based on the quantitative data collected from the targeted applications. It also provides flexibility for applying new design constraints as well as new applications characteristics. Furthermore, this approach is a methodological approach which reduces the design time and results in a design which satisfies the design goals. Experimental results show the efficacy of the hybrid approach. Keyword Design Space Exploration, Extensible Processor, Data Flow Graph, Coarse Grain Accelerator