ch17

[book_gpu.git] / BookGPU / Chapters / chapter6 / PartieAsync.tex

diff --git a/BookGPU/Chapters/chapter6/PartieAsync.tex b/BookGPU/Chapters/chapter6/PartieAsync.tex
index 3365b41039ab07ff2b0b8232e4f438ceb9a04d64..d8da9a4f8e5d2ea56424b8329721102530eb1aca 100644 (file)
--- a/BookGPU/Chapters/chapter6/PartieAsync.tex
+++ b/BookGPU/Chapters/chapter6/PartieAsync.tex
@@ -6,7 +6,7 @@ In the previous section, we have seen how to efficiently implement overlap of
 computations (CPU and GPU) with communications (GPU transfers and internode
 communications).  However, we have previously shown that for some parallel
 iterative algorithms, it is sometimes even more efficient to use an asynchronous
 computations (CPU and GPU) with communications (GPU transfers and internode
 communications).  However, we have previously shown that for some parallel
 iterative algorithms, it is sometimes even more efficient to use an asynchronous

-scheme of iterations\index{iterations!asynchronous} \cite{HPCS2002,ParCo05,Para10}.  In that case, the nodes do
+scheme of iterations\index{asynchronous iterations} \cite{HPCS2002,ParCo05,Para10}.  In that case, the nodes do

 not wait for each other but they perform their iterations using the last
 external data they have received from the other nodes, even if this
 data was produced \emph{before} the previous iteration on the other nodes.
 not wait for each other but they perform their iterations using the last
 external data they have received from the other nodes, even if this
 data was produced \emph{before} the previous iteration on the other nodes.


@@ -139,7 +139,7 @@ communication libraries such as MPI are not systematically performed in parallel
 the computations~\cite{ChVCV13,Hoefler08a}.  So, the logical and classical way
 to implement such an overlap is to use three threads: one for
 computing, one for sending, and one for receiving. Moreover, since
 the computations~\cite{ChVCV13,Hoefler08a}.  So, the logical and classical way
 to implement such an overlap is to use three threads: one for
 computing, one for sending, and one for receiving. Moreover, since

-the communication is performed by threads, blocking synchronous communications\index{MPI!communication!blocking}\index{MPI!communication!synchronous}
+the communication is performed by threads, blocking synchronous communications\index{MPI!blocking}\index{MPI!synchronous}

 can be used without deteriorating the overall performance.
 
 In this basic version, the termination\index{termination} of the global process is performed
 can be used without deteriorating the overall performance.
 
 In this basic version, the termination\index{termination} of the global process is performed


@@ -621,7 +621,7 @@ execution. They are similar to the mechanism used for managing the end messages
 at the end of the entire process.  Line~23 directly updates the
 number of other nodes that are in local convergence by adding the
 received state of the source node. This is possible due to the encoding that is used to
 at the end of the entire process.  Line~23 directly updates the
 number of other nodes that are in local convergence by adding the
 received state of the source node. This is possible due to the encoding that is used to

-represent the local convergence (1) and the non convergence (0).
+represent the local convergence (1) and the nonconvergence (0).

 
 %\begin{algorithm}[H]
 %  \caption{Reception function in the synchronized scheme.}
 
 %\begin{algorithm}[H]
 %  \caption{Reception function in the synchronized scheme.}


@@ -648,7 +648,7 @@ while(!Finished){
       case tagState: // Management of local state messages
        // Actual reception of the message
        MPI_Recv(&recvdState, 1, MPI_CHAR, status.MPI_SOURCE, tagState, MPI_COMM_WORLD, &status); 
       case tagState: // Management of local state messages
        // Actual reception of the message
        MPI_Recv(&recvdState, 1, MPI_CHAR, status.MPI_SOURCE, tagState, MPI_COMM_WORLD, &status); 

-       // Updates of numbers of stabilized nodes and received state msgs 
+       // Updates of numbers of stabilized nodes and recvd state msgs 

        nbOtherCVs += recvdState;
        nbStateMsg++;
        // Unlocking of the computing thread when states of all other 
        nbOtherCVs += recvdState;
        nbStateMsg++;
        // Unlocking of the computing thread when states of all other 


@@ -887,7 +887,7 @@ the CPU may vary depending on the application. For example, when processing data
 streams (pipelines), pre-processing of the next data item and/or post-processing
 of the previous result can be done on the CPU while the GPU is processing the current
 data item.  In other cases, the CPU can perform \emph{auxiliary}
 streams (pipelines), pre-processing of the next data item and/or post-processing
 of the previous result can be done on the CPU while the GPU is processing the current
 data item.  In other cases, the CPU can perform \emph{auxiliary}

-computations\index{computation!auxiliary}
+computations\index{computation auxiliary}

 that are not absolutely required to obtain the result but that may accelerate
 the entire iterative process.  Another possibility would be to distribute the
 main computations between the GPU and CPU. However, this
 that are not absolutely required to obtain the result but that may accelerate
 the entire iterative process.  Another possibility would be to distribute the
 main computations between the GPU and CPU. However, this