pact.tex

\documentclass[times, 10pt,twocolumn]{article}
\usepackage{latex8}
\usepackage{times}
\usepackage{psfig}

\usepackage{etex}
\usepackage[utf8]{inputenc}

\usepackage[left=1.00in,top=1.00in,bottom=1.00in,right=1.00in]{geometry}
\renewcommand{\baselinestretch}{1.00}
\pagestyle{empty}

\usepackage[%
backend=biber,
style=ieee,
% style=numeric-comp,
% style=numeric-comp,  % numerical-compressed
sorting=none,        % nty,nyt,nyvt,anyt,anyvt,ynt,ydnt,none
sortcites=true,      % sort \cite{b a d c}: true,false
block=none,          % space between blocks: none,space,par,nbpar,ragged
indexing=false,      % indexing options: true,false,cite,bib
citereset=none,      % don't reset cites
isbn=false,          % print ISBN?
url=true,            % print URL?
doi=false,           % print DOI?
natbib=true,         % natbib compatability
]{biblatex}

% Reduce the font size of the bibliography:
\renewcommand{\bibfont}{\normalfont\footnotesize}

\addbibresource{refs.bib}

\usepackage{adjustbox}
\usepackage{graphicx}
\usepackage{wrapfig}

\usepackage{subcaption}
\expandafter\def\csname ver@subfig.sty\endcsname{}

%-------------------------------------------------------------------------
\begin{document}
  \title{
  \vspace{-0.8in}
  Autotuning OpenCL Workgroup Sizes
  \vspace{-0.2in}
  }
  \author{Chris Cummins, Hugh Leather, University of Edinburgh
  }

  \vspace{-3em}
  \maketitle
  \vspace{-3em}
  \thispagestyle{empty}

  The physical limitations of microprocessor design have forced the
  industry towards increasingly heterogeneous designs to extract
  performance, with an an increasing pressure to offload traditionally
  CPU based workloads to the GPU. This trend has not been matched with
  adequate software tools; the popular languages OpenCL and CUDA provide
  a very low level model with little abstraction above the
  hardware. Programming at this level requires expert knowledge of both
  the domain and the target hardware, and achieving performance requires
  laborious hand tuning of each program. This has led to a growing
  disparity between the availability of parallelism in modern hardware,
  and the ability for application developers to exploit it.

  The goal of this work is to bring the performance of hand tuned
  heterogeneous code to high level programming, by incorporating
  autotuning into \textit{Algorithmic Skeletons}. Algorithmic Skeletons
  simplify parallel programming by providing reusable, high-level,
  patterns of computation. However, achieving performant skeleton
  implementations is a difficult task; skeleton authors must attempt to
  anticipate and tune for a wide range of architectures and use
  cases. This results in implementations that target the general case
  and cannot provide the performance advantages that are gained from
  tuning low level optimization parameters for individual programs and
  architectures. Autotuning offers promising performance benefits by
  tailoring parameter values to individual cases, but the high cost of
  searching the optimization space limits the practicality of autotuning
  for real world programming~\cite{Ansel2013,Nugteren2015}. We believe
  that performing machine learning-based autotuning at the skeletal
  level of the can overcome these issues.

  In this work, we present \textit{OmniTune} --- an extensible and
  distributed framework for autotuning optimization parameters in
  algorithmic skeletons at runtime. OmniTune enables a collaborative
  approach to performance tuning, in which machine learning training
  data is shared across a network of cooperating systems, amortizing the
  cost of exploring the optimization space. We demonstrate the
  practicality of OmniTune by autotuning the OpenCL workgroup size of
  stencil skeletons in SkelCL. SkelCL~\cite{Steuwer2011} is an
  Algorithmic Skeleton framework which abstracts the complexities of
  OpenCL programming, exposing a set of data parallel skeletons for high
  level heterogeneous programming in C++. Selecting an appropriate
  OpenCL workgroup size is critical for the performance of programs, and
  requires knowledge of the underlying hardware
  (Figure~\ref{fig:motivation-arch}), the data being operated on, and
  the program implementation (Figure~\ref{fig:motivation-prog}).

  \begin{figure}
    \centering
    \adjustbox{valign=t}{%
    \begin{minipage}{.48\textwidth}
      \centering
      \begin{subfigure}[h]{.48\columnwidth}
        \centering
        \includegraphics[width=1.0\columnwidth]{motivation_1}
        \vspace{-2em} % Shrink vertical padding
        \caption{}
        \label{fig:motivation-1}
      \end{subfigure}
      ~%
      \begin{subfigure}[h]{.48\columnwidth}
        \centering
        \includegraphics[width=1.0\columnwidth]{motivation_2}
        \vspace{-2em} % Shrink vertical padding
        \caption{}
        \label{fig:motivation-2}
      \end{subfigure}
      \caption{%
      Optimization space a single stencil on two different devices:
      (\subref{fig:motivation-1}) Intel CPU,
      (\subref{fig:motivation-2}) NVIDIA GPU.%
      }
      \label{fig:motivation-arch}
    \end{minipage}%
    }%
    \hspace{2.5mm}
    \adjustbox{valign=t}{%
    \begin{minipage}{.48\textwidth}
      \centering
      \begin{subfigure}[h]{.48\columnwidth}
        \centering
        \includegraphics[width=1.0\columnwidth]{motivation_3}
        \vspace{-2em} % Shrink vertical padding
        \caption{}
        \label{fig:motivation-3}
      \end{subfigure}
      ~%
      \begin{subfigure}[h]{.48\columnwidth}
        \centering
        \includegraphics[width=1.0\columnwidth]{motivation_4}
        \vspace{-2em} % Shrink vertical padding
        \caption{}
        \label{fig:motivation-4}
      \end{subfigure}
      \caption{%
      Optimization space for a single device executing two different
      stencil programs.%
      }
      \label{fig:motivation-prog}
    \end{minipage}%
    }
    \vspace{-1.5em}
  \end{figure}

  Our autotuning approach employs the novel application of linear
  regressors for classification of workgroup size, extracting 102
  features at runtime describing the program, device, and dataset, and
  predicting optimal workgroup sizes based on training data collected
  using synthetically generated stencil benchmarks.

  In an empirical study of 429 combinations of programs, architectures,
  and datasets, we find that OmniTune provides a median $3.79\times$
  speedup over the best possible fixed workgroup size, achieving 94\% of
  the maximum performance. Our results demonstrate that autotuning at
  the skeletal level --- when combined with sophisticated machine
  learning techniques --- can raise the performance above that of human
  experts, without requiring any effort from the user. By introducing
  OmniTune and demonstrating its practical utility, we hope to
  contribute toward increasing the programmability of heterogeneous
  systems.

  \vspace{-1em}
  \printbibliography

\end{document}