diff --git a/paper/Compilation.tex b/paper/Compilation.tex index 40b7b4263329dacbb77017c66dbe4ae9d2195c5c..4d09e5d317711a9ab095b258f2443d21aaecfa89 100644 --- a/paper/Compilation.tex +++ b/paper/Compilation.tex @@ -253,7 +253,7 @@ new thread, using the Posix pthreads library. The {\tt wait} intrinsic is implemented using {\tt pthread\_join}. %------------------------------------------------------------------------------ -\subsubsection{VISC Runtime} +\subsubsection{\NAME{} Runtime} \label{sec:compiler:flow} %------------------------------------------------------------------------------ @@ -280,7 +280,7 @@ calls to the accelerator runtime in the generated native binary. These data copies are expensive and critical to application performance. It may happen that such a copy is unnecessary because the data is already present on the device. This would happen because the data was brought in the device memory -by a prior node executing on the device. Thus, as an optimization, the VISC +by a prior node executing on the device. Thus, as an optimization, the \NAME{} runtime incorporates a feature we call the ``memory tracker,'' which keeps track of the latest copy of data arrays to avoid unnecessary copies to and from the accelerator. @@ -302,7 +302,7 @@ copies to and from the accelerator. %\end{itemize} %\end{enumerate} -%The VISC compilation flow can be broken down into the following steps - +%The \NAME{} compilation flow can be broken down into the following steps - %\begin{enumerate} %\item The front end would parse the source files and generate the hierarchical