***********************************************************************************
*                                                                                 *
*                      gpmTOOL - GPU Power Modelling Tool                         *
*                                                                                 *
***********************************************************************************

1 - README

    gpmTOOL is a command line tool for modelling the power consumption of
    a GPU device. The tool implements the iterative heuristic algorithm proposed in
    [1] (https://youtu.be/ppsPx6zaC0U), to determine the unknown characteristics of
    GPU devices in order to estimate the GPU power consumption across an ample range of
    frequency and voltage configurations for the multiple GPU frequency domains.

    By providing an input file with details of the execution of different applications
    on a specific GPU device, the tool is able to model the characteristics of the
    different architecture components. The tool can also estimate how the cores voltage
    scales with their frequency. From the resulting model it is possible to estimate
    the power consumption of any new application for different frequency configurations,
    by providing the utilization of the modelled GPU components.

    To get more information of how to install or use it, you should keep reading this file.

    Also, if you use the gpmTOOL tool in a publication, please cite:

    [1] João Guerreiro, Aleksandar Ilic, Nuno Roma, Pedro Tomás. GPGPU Power Modelling
        for Multi-Domain Voltage-Frequency Scaling. 24th IEEE International Symposium on
        High-Performance Computing Architecture (HPCA), 2018.

2 - CONTACT

    If you have problems, questions, ideas or suggestions, please contact us by
    e-mail at joao.guerreiro@inesc-id.pt .

3 - VERSIONS

    Version 1.0 was originally released in January 2018 alongside the work in [1].

    Version 2.0 was released in August 2018. Major changes:

            - New microbenchmark suite.
            - Different iterative algorithm to estimate the parameters and cores
            and memory voltages.


[1] João Guerreiro, Aleksandar Ilic, Nuno Roma, Pedro Tomás. GPGPU Power Modelling
    for Multi-Domain Voltage-Frequency Scaling. 24th IEEE International Symposium on
    High-Performance Computing Architecture (HPCA), 2018.