next up previous
Next: About this document ... Up: Parallel Implementation of the Previous: References

Figures


 
Figure:  The speed-up and scaling of the line opacity calculation using PIOFS for different numbers of 2 nodes used. The upper panel with the wall-clock time (in seconds) used for the test calculations, the lower panel the speed-up factors. The dotted curves give the expected maximum parallel speed-up.
\begin{figure}
\begin{center}
\leavevmode

\psfig {file=lte.fig.ps,width=12cm,angle=90}
\end{center}\end{figure}


 
Figure:  The speed-up and scaling of the NLTE parallel algorithm using Fortran-90 MODULEs and individual array allocation for different numbers of 2 nodes used. The upper panel with the wall-clock time (in seconds) used for the test calculations, the lower panel the speed-up factors. The dotted curves give the expected maximum parallel speed-up
\begin{figure}
\begin{center}
\leavevmode

\psfig {file=nlte.fig.ps,width=12cm,angle=90}
\end{center}\end{figure}


 
Figure:  The speed-up and scaling of the NLTE parallel algorithm for a practical application for different numbers of 2 nodes used. The upper panel with the wall-clock time (in seconds) used for the calculations, the lower panel the speed-up factors.
\begin{figure}
\begin{center}
\leavevmode

\psfig {file=ctctim.ps,width=12cm,angle=-90}
\end{center}\end{figure}



Peter H. Hauschildt
4/27/1999