\relax 
\@writefile{toc}{\author{Sylvain Contassot-Vivier}{}}
\@writefile{toc}{\author{Stephane Vialle}{}}
\@writefile{toc}{\author{Jens Gustedt}{}}
\@writefile{loa}{\addvspace {10\p@ }}
\@writefile{toc}{\contentsline {chapter}{\numberline {6}Development methodologies for GPU and cluster of GPUs}{83}}
\@writefile{lof}{\addvspace {10\p@ }}
\@writefile{lot}{\addvspace {10\p@ }}
\@writefile{toc}{\contentsline {section}{\numberline {6.1}Introduction}{84}}
\newlabel{ch6:intro}{{6.1}{84}}
\@writefile{toc}{\contentsline {section}{\numberline {6.2}General scheme of synchronous code with computation/communication overlapping in GPU clusters}{84}}
\newlabel{ch6:part1}{{6.2}{84}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.2.1}Synchronous parallel algorithms on GPU clusters}{84}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.1}{\ignorespaces Native overlap of internode CPU communications with GPU computations.\relax }}{86}}
\newlabel{fig:ch6p1overlapnative}{{6.1}{86}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.2.2}Native overlap of CPU communications and GPU computations}{86}}
\newlabel{algo:ch6p1overlapnative}{{6.1}{87}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.1}Generic scheme implicitly overlapping MPI communications with CUDA GPU computations}{87}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.2}{\ignorespaces Overlap of internode CPU communications with a sequence of CPU/GPU data transfers and GPU computations.\relax }}{88}}
\newlabel{fig:ch6p1overlapseqsequence}{{6.2}{88}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.2.3}Overlapping with sequences of transfers and computations}{88}}
\newlabel{algo:ch6p1overlapseqsequence}{{6.2}{89}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.2}Generic scheme explicitly overlapping MPI communications with sequences of CUDA CPU/GPU transfers and CUDA GPU computations}{89}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.3}{\ignorespaces Overlap of internode CPU communications with a streamed sequence of CPU/GPU data transfers and GPU computations.\relax }}{90}}
\newlabel{fig:ch6p1overlapstreamsequence}{{6.3}{90}}
\newlabel{algo:ch6p1overlapstreamsequence}{{6.3}{91}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.3}Generic scheme explicitly overlapping MPI communications with streamed sequences of CUDA CPU/GPU transfers and CUDA GPU computations}{91}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.4}{\ignorespaces Complete overlap of internode CPU communications, CPU/GPU data transfers and GPU computations, interleaving computation-communication iterations\relax }}{93}}
\newlabel{fig:ch6p1overlapinterleaved}{{6.4}{93}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.2.4}Interleaved communications-transfers-computations overlapping}{93}}
\newlabel{algo:ch6p1overlapinterleaved}{{6.4}{94}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.4}Generic scheme explicitly overlapping MPI communications, CUDA CPU/GPU transfers and CUDA GPU computations, interleaving computation-communication iterations}{94}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.2.5}Experimental validation}{96}}
\newlabel{ch6:p1expes}{{6.2.5}{96}}
\newlabel{ch6:p1block-cyclic}{{6.2.5}{96}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.5}{\ignorespaces Experimental performances of different synchronous algorithms computing a dense matrix product\relax }}{97}}
\newlabel{fig:ch6p1syncexpematrixprod}{{6.5}{97}}
\@writefile{toc}{\contentsline {section}{\numberline {6.3}General scheme of asynchronous parallel code with computation/communication overlapping}{98}}
\newlabel{ch6:part2}{{6.3}{98}}
\@writefile{loa}{\contentsline {algocf}{\numberline {3}{\ignorespaces Synchronous iterative scheme\relax }}{98}}
\newlabel{algo:ch6p2sync}{{3}{98}}
\@writefile{loa}{\contentsline {algocf}{\numberline {4}{\ignorespaces Asynchronous iterative scheme\relax }}{98}}
\newlabel{algo:ch6p2async}{{4}{98}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.3.1}A basic asynchronous scheme}{99}}
\newlabel{ch6:p2BasicAsync}{{6.3.1}{99}}
\newlabel{algo:ch6p2BasicAsync}{{6.5}{100}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.5}Initialization of the basic asynchronous scheme}{100}}
\newlabel{algo:ch6p2BasicAsyncComp}{{6.6}{101}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.6}Computing function in the basic asynchronous scheme}{101}}
\newlabel{algo:ch6p2BasicAsyncSendings}{{6.7}{102}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.7}Sending function in the basic asynchronous scheme}{102}}
\newlabel{algo:ch6p2BasicAsyncReceptions}{{6.8}{103}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.8}Reception function in the basic asynchronous scheme}{103}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.3.2}Synchronization of the asynchronous scheme}{104}}
\newlabel{ch6:p2SsyncOverAsync}{{6.3.2}{104}}
\newlabel{algo:ch6p2Sync}{{6.9}{105}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.9}Initialization of the synchronized scheme}{105}}
\newlabel{algo:ch6p2SyncComp}{{6.10}{106}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.10}Computing function in the synchronized scheme}{106}}
\newlabel{algo:ch6p2SyncReceptions}{{6.11}{107}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.11}Reception function in the synchronized scheme}{107}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.3.3}Asynchronous scheme using MPI, OpenMP and CUDA}{108}}
\newlabel{ch6:p2GPUAsync}{{6.3.3}{108}}
\newlabel{algo:ch6p2AsyncSyncComp}{{6.12}{109}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.12}Computing function in the final asynchronous scheme}{109}}
\newlabel{algo:ch6p2syncGPU}{{6.13}{111}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.13}Computing function in the final asynchronous scheme}{111}}
\newlabel{algo:ch6p2FullOverAsyncMain}{{6.14}{113}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.14}Initialization of the main process of complete overlap with asynchronism}{113}}
\newlabel{algo:ch6p2FullOverAsyncComp1}{{6.15}{114}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.15}Computing function in the final asynchronous scheme with CPU/GPU overlap}{114}}
\newlabel{algo:ch6p2FullOverAsyncComp2}{{6.16}{115}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.16}Auxiliary computing function in the final asynchronous scheme with CPU/GPU overlap}{115}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.3.4}Experimental validation}{116}}
\newlabel{sec:ch6p2expes}{{6.3.4}{116}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.6}{\ignorespaces Computation times of the test application in synchronous and asynchronous modes.\relax }}{117}}
\newlabel{fig:ch6p2syncasync}{{6.6}{117}}
\@writefile{lof}{\contentsline {figure}{\numberline {6.7}{\ignorespaces Computation times with or without overlap of Jacobian updatings in asynchronous mode.\relax }}{118}}
\newlabel{fig:ch6p2aux}{{6.7}{118}}
\@writefile{toc}{\contentsline {section}{\numberline {6.4}Perspective: A unifying programming model}{119}}
\newlabel{sec:ch6p3unify}{{6.4}{119}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.4.1}Resources}{119}}
\newlabel{sec:ch6p3resources}{{6.4.1}{119}}
\newlabel{algo:ch6p3ORWLresources}{{6.17}{120}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.17}Declaration of ORWL resources for a block-cyclic matrix multiplication}{120}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.4.2}Control}{120}}
\newlabel{sec:ch6p3ORWLcontrol}{{6.4.2}{120}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.4.3}Example: block-cyclic matrix multiplication (MM)}{121}}
\newlabel{sec:ch6p3ORWLMM}{{6.4.3}{121}}
\newlabel{algo:ch6p3ORWLBCCMM}{{6.18}{121}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.18}Block-cyclic matrix multiplication, high level per task view}{121}}
\newlabel{algo:ch6p3ORWLlcopy}{{6.19}{122}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.19}An iterative local copy operation}{122}}
\newlabel{algo:ch6p3ORWLrcopy}{{6.20}{122}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.20}An iterative remote copy operation as part of a block cyclic matrix multiplication task}{122}}
\newlabel{algo:ch6p3ORWLtrans}{{6.21}{122}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.21}An iterative GPU transfer and compute operation as part of a block cyclic matrix multiplication task}{122}}
\newlabel{algo:ch6p3ORWLdecl}{{6.22}{123}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.22}Dynamic declaration of handles to represent the resources}{123}}
\newlabel{algo:ch6p3ORWLinit}{{6.23}{124}}
\@writefile{lol}{\contentsline {lstlisting}{\numberline {6.23}Dynamic initialization of access mode and priorities}{124}}
\@writefile{toc}{\contentsline {subsection}{\numberline {6.4.4}Tasks and operations}{124}}
\newlabel{sec:ch6p3tasks}{{6.4.4}{124}}
\@writefile{toc}{\contentsline {section}{\numberline {6.5}Conclusion}{125}}
\newlabel{ch6:conclu}{{6.5}{125}}
\@writefile{toc}{\contentsline {section}{\numberline {6.6}Glossary}{125}}
\@writefile{toc}{\contentsline {section}{Bibliography}{126}}
\@setckpt{Chapters/chapter6/ch6}{
\setcounter{page}{128}
\setcounter{equation}{0}
\setcounter{enumi}{4}
\setcounter{enumii}{0}
\setcounter{enumiii}{0}
\setcounter{enumiv}{21}
\setcounter{footnote}{0}
\setcounter{mpfootnote}{0}
\setcounter{part}{1}
\setcounter{chapter}{6}
\setcounter{section}{6}
\setcounter{subsection}{0}
\setcounter{subsubsection}{0}
\setcounter{paragraph}{0}
\setcounter{subparagraph}{0}
\setcounter{figure}{7}
\setcounter{table}{0}
\setcounter{numauthors}{0}
\setcounter{parentequation}{8}
\setcounter{subfigure}{0}
\setcounter{lofdepth}{1}
\setcounter{subtable}{0}
\setcounter{lotdepth}{1}
\setcounter{lstnumber}{17}
\setcounter{ContinuedFloat}{0}
\setcounter{AlgoLine}{0}
\setcounter{algocfline}{4}
\setcounter{algocfproc}{4}
\setcounter{algocf}{4}
\setcounter{proposition}{0}
\setcounter{theorem}{0}
\setcounter{exercise}{0}
\setcounter{example}{0}
\setcounter{definition}{0}
\setcounter{proof}{0}
\setcounter{lstlisting}{23}
}