Dataset for the paper Software-defined PMC for Runtime Power Management System of a Many-core Neuromorphic Platform

Published: 30-10-2017| Version 1 | DOI: 10.17632/37wctb99rw.1
Contributor:
Indar Sugiarto

Description

This is the dataset of three non-Spiking Neural Network applications running on the SpiNNaker platform: JPEG image encoding, JPEG image decoding, and edge detection (referred to as A1, A2, and A3 respectively in the paper). The application A1 and A2 are the first applications that are developed by considering the impact of DMA Full Counter (DFC) for performance improvement. Both application retrieve/store data from/to SDRAM using DMA mechanism. The application A3 uses the old mechanism (without involving DFC), in which a master core is assigned with a task for coordinating DMA among cores in a SpiNNaker chip. The applications A1, A2, and A3 were run alternately whilst changing the governor. During the application execution, we measure the energy consumption and temperature of SpiNNaker chips using a SpiNNaker profiler program.

Files

Steps to reproduce

The data were gathered using an in-house data acquisition system (DAS) developed in the group of Advanced Processor Technology (APT) in the School of Computer Science University of Manchester. This DAS is not commercially available and is operationally maintained by APT. In addition of using the DAS, the three SpiNNaker programs (referred to as A1, A2, and A3 in the paper) are needed. These programs are free and available/downloadable through the github. Please contact the author of the paper (Indar Sugiarto) for the detailed step-by-step procedure to use the DAS as well as to compile the program for SpiNNaker machine in order to reproduce the data.