aboutsummaryrefslogtreecommitdiff
path: root/Carpet/doc
diff options
context:
space:
mode:
authorThomas Radke <tradke@aei.mpg.de>2006-08-28 14:34:00 +0000
committerThomas Radke <tradke@aei.mpg.de>2006-08-28 14:34:00 +0000
commitb9480679de65fe9a5b7408d90e479df06f64767c (patch)
treeb5297c37333c55a66fd62faa03f4c38afcfe28bb /Carpet/doc
parentf722e634a4ac97637dbd16424a04160fd99e310c (diff)
Carpet/doc: document when allocation/deallocation and synchronisation of grid variables happens
darcs-hash:20060828143437-776a0-55c86cf90bf568d0a0abd765e4f00ebc15b71542.gz
Diffstat (limited to 'Carpet/doc')
-rw-r--r--Carpet/doc/scheduling.tex19
1 files changed, 15 insertions, 4 deletions
diff --git a/Carpet/doc/scheduling.tex b/Carpet/doc/scheduling.tex
index 016a0e879..e43371f9c 100644
--- a/Carpet/doc/scheduling.tex
+++ b/Carpet/doc/scheduling.tex
@@ -68,13 +68,22 @@ There are two major pieces of software involved in scheduling:
consistent with the \verb|BEFORE| and \verb|AFTER| clauses in
all the schedule groups. The flesh also handles repeatedly
calling scheduled routines which are scheduled with a
- \verb|WHILE| clause. The flesh does \emph{not} know anything
+ \verb|WHILE| clause.
+ In addition, the flesh determines when storage is turned
+ on/off for grid scalars, functions, and arrays and when
+ grid arrays and functions are synchronised,
+ based on the {\tt STORAGE:} and {\tt SYNC:} statements in
+ schedule blocks.
+
+ The flesh does \emph{not} know anything
about the spatial grid or grids, or how grid functions are
stored in memory or distributed across processors.
\item The driver, typically \thorn{PUGH} or \thorn{Carpet}.
The driver defines the set of allowed schedule bins.
The driver knows about the grid or grids, and how grid functions
- are stored in memory and distributed across processors.
+ are stored in memory, distributed across processors, and
+ synchronised.
+
The driver does \emph{not} know what's in \verb|schedule.ccl| files.
\end{itemize}
@@ -476,8 +485,10 @@ Synchronization and turning storage on/off happen in level mode.%%%
singlemap and local modes if there is only one
component per processor.
}%%%
-{} Boundary conditions must be selected (in the sense of thorn
-\thorn{Boundary}) in level mode. Cactus output must be done in level mode.
+{} (At any time only the ``current'' refinement level has storage
+turned on.) Boundary conditions must be selected (in the sense of
+thorn \thorn{Boundary}) in level mode. Cactus output must be done
+in level mode.
Reduction/interpolation of grid arrays and/or grid functions may be
done in either level mode (applying only to that refinement level),