Characterizing energy and performance of soft-core-based homogeneous multiprocessor systems

Farshid Samsami Khodadad,Hamid Noori

The Journal of Supercomputing(2022)

引用 0|浏览8
暂无评分
摘要
Multiprocessor systems offer numerous configurations in terms of a different number of cores and frequency levels that may become optimal with respect to energy, performance, or other metrics. On FPGAs, a convenient solution for designing and building a multiprocessor system is the use of soft-core processors. The soft-core processor configuration and frequency are customizable and configurable at design time and according to the FPGA capacity, the number of cores and its configuration can be changed. In this research, different workloads have been studied and results shown that the amount of speedup would be different for each workload due to their behavior in an MPSoC fashion and it was revealed that in a multiprocessor system based on soft-core on an FPGA platform, increasing the number of processors and their operating frequency will not always improve the system energy-delay product (EDP). Hence, identifying an optimal configuration with respect to a given metric such as EDP is a complex process due to a large number of workloads and configurations. To achieve this, we use the power consumption and execution time information to identify the optimal configuration for different workloads with respect to EDP. We also perform an extensive workload characterization using performance counters available on the target platform. Using these performance counters, a vast amount of characterization data for each workload was collected. Then, we used this characterization data to choose the optimal configuration for each workload. This paper proposes a characterization method for parallel workloads that can be used to determine the optimal core and frequency configuration of an FPGA-based homogenous soft-core multiprocessor system with respect to EDP as a function of the workload (It is a datatype. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s).).
更多
查看译文
关键词
FPGA platform,Multiprocessor,Systems-on-chip,Soft-core,Performance,EDP,Power,Performance counters
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要