This paper presents a benchmark suite for evaluating a configurable computing system's infrastructure, both tools and architecture. A novel aspect of this work is the use of stressmarks, benchmarks that focus on a specific characteristic or property of interest. This is in contrast to traditional approaches that utilize functional benchmarks, benchmarks that emphasize measuring end-to-end execution time. This suite can be used to assess a broad range of configurable computing systems, including single configurable devices, multiple configurable devices, and mixed architectures, such as fixed-plus-variable devices and hybrid systems. In addition, aspects that are particularly relevant to the domain of configurable computing, such as run-time reconfiguration and variable precision arithmetic, are considered. The paper provides an overview of the benchmark suite, presents some implementation results on an Annapolis Micro Systems WILDFORCE board, reflects on the benchmark suite developed, and briefly describes future work.
Heterogeneous configurable computing systems containing general-purpose processors, special-purpose processors, field-programmable gate arrays (FTGAs), and application-specific integrated circuits (ASICs) present several programming and development challenges. The ADAPTERS effort is addressing these challenges by focusing on three technology areas: 1) a unified data flow programming environment, 2) a hardwarehoftware partitioning and mapping tool, along with a system level modeling tool, and 3) a dynamic reconfiguration run-time environment for FPGAs. To investigate automatic mapping techniques, a constant false alarm rate (CFAR) application has been implemented on a combined PChVILDFORCE platform. The combined platform containing FPGAs provided a 20% improvement in the end-to-end latency compared to a software solution running on a PC. An image compression example has also been developed to explore mode-based dynamic FPGA reconfiguration. The dynamic reconfiguration technology being developed can be used to adapt the behavior of systems, such as satellites, to changing internal and external conditions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.