From 902b20392dce7b49eb91e0c38ed91e5f9c1ae5e8 Mon Sep 17 00:00:00 2001 From: jthorn Date: Sun, 21 Nov 2004 11:14:08 +0000 Subject: document requirement that driver::ghost_size be set to a suitable value if derivatives are taken git-svn-id: http://svn.einsteintoolkit.org/cactus/EinsteinBase/ADMMacros/trunk@83 b1d164ef-f17a-46e7-89d4-021c7118ef4e --- doc/documentation.tex | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/doc/documentation.tex b/doc/documentation.tex index 343bd86..494d3a8 100644 --- a/doc/documentation.tex +++ b/doc/documentation.tex @@ -122,14 +122,16 @@ By default, the macros use centered 2nd~order finite differencing, with 3-point finite difference molecules. That is, when finite differencing the the grid-point indices $\texttt{i} \pm 1$, $\texttt{j} \pm 1$, and $\texttt{k} \pm 1$ -must also be valid. +must also be valid, and \texttt{driver::ghost\_size} must be set +to at least $1$. Some of the macros also support centered 4th~order finite differencing; This is selected with the parameter \verb|spatial_order|. This may be set to either~$2$ or~$4$; it defaults to~$2$. If it's set to~$4$, then 5-point finite difference molecules are used, so the grid-point indices $\texttt{i} \pm 2$, $\texttt{j} \pm 2$, and $\texttt{k} \pm 2$ -must also be valid. The only save way to be certain which macros support +must also be valid, and \texttt{driver::ghost\_size} must be set +to at least $2$. The only save way to be certain which macros support 4th~order finite differencing is to check the source code; the macros which don't support it simply hard-code 2nd~order finite differencing and ignore the \texttt{spatial\_order} parameter. -- cgit v1.2.3