1 |
% $Header: /u/gcmpack/manual/part6/exch2.tex,v 1.25 2005/08/09 21:52:09 edhill Exp $ |
2 |
% $Name: $ |
3 |
|
4 |
%% * Introduction |
5 |
%% o what it does, citations (refs go into mitgcm_manual.bib, |
6 |
%% preferably in alphabetic order) |
7 |
%% o Equations |
8 |
%% * Key subroutines and parameters |
9 |
%% * Reference material (auto generated from Protex and structured comments) |
10 |
%% o automatically inserted at \section{Reference} |
11 |
|
12 |
|
13 |
\subsection{exch2: Extended Cubed Sphere \mbox{Topology}} |
14 |
\label{sec:exch2} |
15 |
|
16 |
|
17 |
\subsubsection{Introduction} |
18 |
|
19 |
The \texttt{exch2} package extends the original cubed sphere topology |
20 |
configuration to allow more flexible domain decomposition and |
21 |
parallelization. Cube faces (also called subdomains) may be divided |
22 |
into any number of tiles that divide evenly into the grid point |
23 |
dimensions of the subdomain. Furthermore, the tiles can run on |
24 |
separate processors individually or in groups, which provides for |
25 |
manual compile-time load balancing across a relatively arbitrary |
26 |
number of processors. |
27 |
|
28 |
The exchange parameters are declared in |
29 |
\filelink{pkg/exch2/W2\_EXCH2\_TOPOLOGY.h}{pkg-exch2-W2_EXCH2_TOPOLOGY.h} |
30 |
and assigned in |
31 |
\filelink{pkg/exch2/w2\_e2setup.F}{pkg-exch2-w2_e2setup.F}. The |
32 |
validity of the cube topology depends on the \file{SIZE.h} file as |
33 |
detailed below. The default files provided in the release configure a |
34 |
cubed sphere topology of six tiles, one per subdomain, each with |
35 |
32$\times$32 grid points, with all tiles running on a single processor. Both |
36 |
files are generated by Matlab scripts in |
37 |
\file{utils/exch2/matlab-topology-generator}; see Section |
38 |
\ref{sec:topogen} \sectiontitle{Generating Topology Files for exch2} |
39 |
for details on creating alternate topologies. Pregenerated examples |
40 |
of these files with alternate topologies are provided under |
41 |
\file{utils/exch2/code-mods} along with the appropriate \file{SIZE.h} |
42 |
file for single-processor execution. |
43 |
|
44 |
\subsubsection{Invoking exch2} |
45 |
|
46 |
To use exch2 with the cubed sphere, the following conditions must be |
47 |
met: |
48 |
|
49 |
\begin{itemize} |
50 |
\item The exch2 package is included when \file{genmake2} is run. The |
51 |
easiest way to do this is to add the line \code{exch2} to the |
52 |
\file{packages.conf} file -- see Section \ref{sect:buildingCode} |
53 |
\sectiontitle{Building the code} for general |
54 |
details. |
55 |
|
56 |
\item An example of \file{W2\_EXCH2\_TOPOLOGY.h} and |
57 |
\file{w2\_e2setup.F} must reside in a directory containing files |
58 |
symbolically linked by the \file{genmake2} script. The safest place |
59 |
to put these is the directory indicated in the \code{-mods=DIR} |
60 |
command line modifier (typically \file{../code}), or the build |
61 |
directory. The default versions of these files reside in |
62 |
\file{pkg/exch2} and are linked automatically if no other versions |
63 |
exist elsewhere in the build path, but they should be left untouched |
64 |
to avoid breaking configurations other than the one you intend to |
65 |
modify. |
66 |
|
67 |
\item Files containing grid parameters, named \file{tile00$n$.mitgrid} |
68 |
where $n$=\code{(1:6)} (one per subdomain), must be in the working |
69 |
directory when the MITgcm executable is run. These files are |
70 |
provided in the example experiments for cubed sphere configurations |
71 |
with 32$\times$32 cube sides -- please contact MITgcm support if you |
72 |
want to generate files for other configurations. |
73 |
|
74 |
\item As always when compiling MITgcm, the file \file{SIZE.h} must be |
75 |
placed where \file{genmake2} will find it. In particular for exch2, |
76 |
the domain decomposition specified in \file{SIZE.h} must correspond |
77 |
with the particular configuration's topology specified in |
78 |
\file{W2\_EXCH2\_TOPOLOGY.h} and \file{w2\_e2setup.F}. Domain |
79 |
decomposition issues particular to exch2 are addressed in Section |
80 |
\ref{sec:topogen} \sectiontitle{Generating Topology Files for exch2} |
81 |
and \ref{sec:exch2mpi} \sectiontitle{exch2, SIZE.h, and |
82 |
Multiprocessing}; a more general background on the subject |
83 |
relevant to MITgcm is presented in Section |
84 |
\ref{sect:specifying_a_decomposition} |
85 |
\sectiontitle{Specifying a decomposition}. |
86 |
\end{itemize} |
87 |
|
88 |
At the time of this writing the following examples use exch2 and may |
89 |
be used for guidance: |
90 |
|
91 |
\begin{verbatim} |
92 |
verification/adjust_nlfs.cs-32x32x1 |
93 |
verification/adjustment.cs-32x32x1 |
94 |
verification/aim.5l_cs |
95 |
verification/global_ocean.cs32x15 |
96 |
verification/hs94.cs-32x32x5 |
97 |
\end{verbatim} |
98 |
|
99 |
|
100 |
|
101 |
|
102 |
\subsubsection{Generating Topology Files for exch2} |
103 |
\label{sec:topogen} |
104 |
|
105 |
Alternate cubed sphere topologies may be created using the Matlab |
106 |
scripts in \file{utils/exch2/matlab-topology-generator}. Running the |
107 |
m-file |
108 |
\filelink{driver.m}{utils-exch2-matlab-topology-generator_driver.m} |
109 |
from the Matlab prompt (there are no parameters to pass) generates |
110 |
exch2 topology files \file{W2\_EXCH2\_TOPOLOGY.h} and |
111 |
\file{w2\_e2setup.F} in the working directory and displays a figure of |
112 |
the topology via Matlab -- figures \ref{fig:6tile}, \ref{fig:12tile}, |
113 |
and \ref{fig:24tile} are examples of the generated diagrams. The other |
114 |
m-files in the directory are |
115 |
subroutines called from \file{driver.m} and should not be run ``bare'' except |
116 |
for development purposes. \\ |
117 |
|
118 |
The parameters that determine the dimensions and topology of the |
119 |
generated configuration are \code{nr}, \code{nb}, \code{ng}, |
120 |
\code{tnx} and \code{tny}, and all are assigned early in the script. \\ |
121 |
|
122 |
The first three determine the height and width of the subdomains and |
123 |
hence the size of the overall domain. Each one determines the number |
124 |
of grid points, and therefore the resolution, along the subdomain |
125 |
sides in a ``great circle'' around each the three spatial axes of the cube. At the time |
126 |
of this writing MITgcm requires these three parameters to be equal, |
127 |
but they provide for future releases to accomodate different |
128 |
resolutions around the axes to allow subdomains with differing resolutions.\\ |
129 |
|
130 |
The parameters \code{tnx} and \code{tny} determine the width and height of |
131 |
the tiles into which the subdomains are decomposed, and must evenly |
132 |
divide the integer assigned to \code{nr}, \code{nb} and \code{ng}. |
133 |
The result is a rectangular tiling of the subdomain. Figure |
134 |
\ref{fig:24tile} shows one possible topology for a twenty-four-tile |
135 |
cube, and figure \ref{fig:12tile} shows one for twelve tiles. \\ |
136 |
|
137 |
\begin{figure} |
138 |
\begin{center} |
139 |
\resizebox{4in}{!}{ |
140 |
\includegraphics{part6/s24t_16x16.ps} |
141 |
} |
142 |
\end{center} |
143 |
|
144 |
\caption{Plot of a cubed sphere topology with a 32$\times$192 domain |
145 |
divided into six 32$\times$32 subdomains, each of which is divided |
146 |
into four tiles of width \code{tnx=16} and height \code{tny=16} for a |
147 |
total of twenty-four tiles. The colored borders of the subdomains |
148 |
represent the parameters \code{nr} (red), \code{nb} (blue), and |
149 |
\code{ng} (green). } \label{fig:24tile} |
150 |
\end{figure} |
151 |
|
152 |
\begin{figure} |
153 |
\begin{center} |
154 |
\resizebox{4in}{!}{ |
155 |
\includegraphics{part6/s12t_16x32.ps} |
156 |
} |
157 |
\end{center} |
158 |
\caption{Plot of a cubed sphere topology with a 32$\times$192 domain |
159 |
divided into six 32$\times$32 subdomains of two tiles each |
160 |
(\code{tnx=16, tny=32}). |
161 |
} \label{fig:12tile} |
162 |
\end{figure} |
163 |
|
164 |
\begin{figure} |
165 |
\begin{center} |
166 |
\resizebox{4in}{!}{ |
167 |
\includegraphics{part6/s6t_32x32.ps} |
168 |
} |
169 |
\end{center} |
170 |
\caption{Plot of a cubed sphere topology with a 32$\times$192 domain |
171 |
divided into six 32$\times$32 subdomains with one tile each |
172 |
(\code{tnx=32, tny=32}). This is the default configuration. |
173 |
} |
174 |
\label{fig:6tile} |
175 |
\end{figure} |
176 |
|
177 |
|
178 |
Tiles can be selected from the topology to be omitted from being |
179 |
allocated memory and processors. This tuning is useful in ocean |
180 |
modeling for omitting tiles that fall entirely on land. The tiles |
181 |
omitted are specified in the file |
182 |
\filelink{blanklist.txt}{utils-exch2-matlab-topology-generator_blanklist.txt} |
183 |
by their tile number in the topology, separated by a newline. \\ |
184 |
|
185 |
|
186 |
|
187 |
|
188 |
\subsubsection{exch2, SIZE.h, and Multiprocessing} |
189 |
\label{sec:exch2mpi} |
190 |
|
191 |
Once the topology configuration files are created, the Fortran |
192 |
\code{PARAMETER}s in \file{SIZE.h} must be configured to match. |
193 |
Section \ref{sect:specifying_a_decomposition} \sectiontitle{Specifying |
194 |
a decomposition} provides a general description of domain |
195 |
decomposition within MITgcm and its relation to \file{SIZE.h}. The |
196 |
current section specifies constraints that the exch2 package imposes |
197 |
and describes how to enable parallel execution with MPI. |
198 |
|
199 |
As in the general case, the parameters \varlink{sNx}{sNx} and |
200 |
\varlink{sNy}{sNy} define the size of the individual tiles, and so |
201 |
must be assigned the same respective values as \code{tnx} and |
202 |
\code{tny} in \file{driver.m}. |
203 |
|
204 |
The halo width parameters \varlink{OLx}{OLx} and \varlink{OLy}{OLy} |
205 |
have no special bearing on exch2 and may be assigned as in the general |
206 |
case. The same holds for \varlink{Nr}{Nr}, the number of vertical |
207 |
levels in the model. |
208 |
|
209 |
The parameters \varlink{nSx}{nSx}, \varlink{nSy}{nSy}, |
210 |
\varlink{nPx}{nPx}, and \varlink{nPy}{nPy} relate to the number of |
211 |
tiles and how they are distributed on processors. When using exch2, |
212 |
the tiles are stored in the $x$ dimension, and so |
213 |
\code{\varlink{nSy}{nSy}=1} in all cases. Since the tiles as |
214 |
configured by exch2 cannot be split up accross processors without |
215 |
regenerating the topology, \code{\varlink{nPy}{nPy}=1} as well. |
216 |
|
217 |
The number of tiles MITgcm allocates and how they are distributed |
218 |
between processors depends on \varlink{nPx}{nPx} and |
219 |
\varlink{nSx}{nSx}. \varlink{nSx}{nSx} is the number of tiles per |
220 |
processor and \varlink{nPx}{nPx} is the number of processors. The |
221 |
total number of tiles in the topology minus those listed in |
222 |
\file{blanklist.txt} must equal \code{nSx*nPx}. Note that in order to |
223 |
obtain maximum usage from a given number of processors in some cases, |
224 |
this restriction might entail sharing a processor with a tile that |
225 |
would otherwise be excluded because it is topographically outside of |
226 |
the domain and therefore in \file{blanklist.txt}. For example, |
227 |
suppose you have five processors and a domain decomposition of |
228 |
thirty-six tiles that allows you to exclude seven tiles. To evenly |
229 |
distribute the remaining twenty-nine tiles among five processors, you |
230 |
would have to run one ``dummy'' tile to make an even six tiles per |
231 |
processor. Such dummy tiles are \emph{not} listed in |
232 |
\file{blanklist.txt}. |
233 |
|
234 |
The following is an example of \file{SIZE.h} for the twelve-tile |
235 |
configuration illustrated in figure \ref{fig:12tile} running on one |
236 |
processor: |
237 |
|
238 |
\begin{verbatim} |
239 |
PARAMETER ( |
240 |
& sNx = 16, |
241 |
& sNy = 32, |
242 |
& OLx = 2, |
243 |
& OLy = 2, |
244 |
& nSx = 12, |
245 |
& nSy = 1, |
246 |
& nPx = 1, |
247 |
& nPy = 1, |
248 |
& Nx = sNx*nSx*nPx, |
249 |
& Ny = sNy*nSy*nPy, |
250 |
& Nr = 5) |
251 |
\end{verbatim} |
252 |
|
253 |
The following is an example for the twenty-four-tile topology in |
254 |
figure \ref{fig:24tile} running on six processors: |
255 |
|
256 |
\begin{verbatim} |
257 |
PARAMETER ( |
258 |
& sNx = 16, |
259 |
& sNy = 16, |
260 |
& OLx = 2, |
261 |
& OLy = 2, |
262 |
& nSx = 4, |
263 |
& nSy = 1, |
264 |
& nPx = 6, |
265 |
& nPy = 1, |
266 |
& Nx = sNx*nSx*nPx, |
267 |
& Ny = sNy*nSy*nPy, |
268 |
& Nr = 5) |
269 |
\end{verbatim} |
270 |
|
271 |
|
272 |
\subsubsection{Key Variables} |
273 |
|
274 |
The descriptions of the variables are divided up into scalars, |
275 |
one-dimensional arrays indexed to the tile number, and two and |
276 |
three-dimensional arrays indexed to tile number and neighboring tile. |
277 |
This division reflects the functionality of these variables: The |
278 |
scalars are common to every part of the topology, the tile-indexed |
279 |
arrays to individual tiles, and the arrays indexed by tile and |
280 |
neighbor to relationships between tiles and their neighbors. \\ |
281 |
|
282 |
Scalars: |
283 |
|
284 |
The number of tiles in a particular topology is set with the parameter |
285 |
\code{NTILES}, and the maximum number of neighbors of any tiles by |
286 |
\code{MAX\_NEIGHBOURS}. These parameters are used for defining the |
287 |
size of the various one and two dimensional arrays that store tile |
288 |
parameters indexed to the tile number and are assigned in the files |
289 |
generated by \file{driver.m}.\\ |
290 |
|
291 |
The scalar parameters \varlink{exch2\_domain\_nxt}{exch2_domain_nxt} |
292 |
and \varlink{exch2\_domain\_nyt}{exch2_domain_nyt} express the number |
293 |
of tiles in the $x$ and $y$ global indices. For example, the default |
294 |
setup of six tiles (Fig. \ref{fig:6tile}) has |
295 |
\code{exch2\_domain\_nxt=6} and \code{exch2\_domain\_nyt=1}. A |
296 |
topology of twenty-four square tiles, four per subdomain (as in figure |
297 |
\ref{fig:24tile}), will have \code{exch2\_domain\_nxt=12} and |
298 |
\code{exch2\_domain\_nyt=2}. Note that these parameters express the |
299 |
tile layout in order to allow global data files that are tile-layout-neutral. |
300 |
They have no bearing on the internal storage of the arrays. The tiles |
301 |
are stored internally in a range from \code{\varlink{bi}{bi}=(1:NTILES)} in the |
302 |
$x$ axis, and the $y$ axis variable \varlink{bj}{bj} is assumed to |
303 |
equal \code{1} throughout the package. \\ |
304 |
|
305 |
Arrays indexed to tile number: |
306 |
|
307 |
The following arrays are of length \code{NTILES} and are indexed to |
308 |
the tile number, which is indicated in the diagrams with the notation |
309 |
\textsf{t}$n$. The indices are omitted in the descriptions. \\ |
310 |
|
311 |
The arrays \varlink{exch2\_tnx}{exch2_tnx} and |
312 |
\varlink{exch2\_tny}{exch2_tny} express the $x$ and $y$ dimensions of |
313 |
each tile. At present for each tile \texttt{exch2\_tnx=sNx} and |
314 |
\texttt{exch2\_tny=sNy}, as assigned in \file{SIZE.h} and described in |
315 |
Section \ref{sec:exch2mpi} \sectiontitle{exch2, SIZE.h, and |
316 |
Multiprocessing}. Future releases of MITgcm may allow varying tile |
317 |
sizes. \\ |
318 |
|
319 |
The arrays \varlink{exch2\_tbasex}{exch2_tbasex} and |
320 |
\varlink{exch2\_tbasey}{exch2_tbasey} determine the tiles' |
321 |
Cartesian origin within a subdomain |
322 |
and locate the edges of different tiles relative to each other. As |
323 |
an example, in the default six-tile topology (Fig. \ref{fig:6tile}) |
324 |
each index in these arrays is set to \code{0} since a tile occupies |
325 |
its entire subdomain. The twenty-four-tile case discussed above will |
326 |
have values of \code{0} or \code{16}, depending on the quadrant of the |
327 |
tile within the subdomain. The elements of the arrays |
328 |
\varlink{exch2\_txglobalo}{exch2_txglobalo} and |
329 |
\varlink{exch2\_txglobalo}{exch2_txglobalo} are similar to |
330 |
\varlink{exch2\_tbasex}{exch2_tbasex} and |
331 |
\varlink{exch2\_tbasey}{exch2_tbasey}, but locate the tile edges within the |
332 |
global address space, similar to that used by global output and input |
333 |
files. \\ |
334 |
|
335 |
The array \varlink{exch2\_myFace}{exch2_myFace} contains the number of |
336 |
the subdomain of each tile, in a range \code{(1:6)} in the case of the |
337 |
standard cube topology and indicated by \textbf{\textsf{f}}$n$ in |
338 |
figures \ref{fig:12tile} and |
339 |
\ref{fig:24tile}. \varlink{exch2\_nNeighbours}{exch2_nNeighbours} |
340 |
contains a count of the neighboring tiles each tile has, and sets |
341 |
the bounds for looping over neighboring tiles. |
342 |
\varlink{exch2\_tProc}{exch2_tProc} holds the process rank of each |
343 |
tile, and is used in interprocess communication. \\ |
344 |
|
345 |
|
346 |
The arrays \varlink{exch2\_isWedge}{exch2_isWedge}, |
347 |
\varlink{exch2\_isEedge}{exch2_isEedge}, |
348 |
\varlink{exch2\_isSedge}{exch2_isSedge}, and |
349 |
\varlink{exch2\_isNedge}{exch2_isNedge} are set to \code{1} if the |
350 |
indexed tile lies on the edge of its subdomain, \code{0} if |
351 |
not. The values are used within the topology generator to determine |
352 |
the orientation of neighboring tiles, and to indicate whether a tile |
353 |
lies on the corner of a subdomain. The latter case requires special |
354 |
exchange and numerical handling for the singularities at the eight |
355 |
corners of the cube. \\ |
356 |
|
357 |
|
358 |
Arrays Indexed to Tile Number and Neighbor: |
359 |
|
360 |
The following arrays have vectors of length \code{MAX\_NEIGHBOURS} and |
361 |
\code{NTILES} and describe the orientations between the the tiles. \\ |
362 |
|
363 |
The array \code{exch2\_neighbourId(a,T)} holds the tile number |
364 |
\code{Tn} for each of the tile number \code{T}'s neighboring tiles |
365 |
\code{a}. The neighbor tiles are indexed |
366 |
\code{(1:exch2\_nNeighbours(T))} in the order right to left on the |
367 |
north then south edges, and then top to bottom on the east then west |
368 |
edges. \\ |
369 |
|
370 |
The \code{exch2\_opposingSend\_record(a,T)} array holds the |
371 |
index \code{b} of the element in \texttt{exch2\_neighbourId(b,Tn)} |
372 |
that holds the tile number \code{T}, given |
373 |
\code{Tn=exch2\_neighborId(a,T)}. In other words, |
374 |
\begin{verbatim} |
375 |
exch2_neighbourId( exch2_opposingSend_record(a,T), |
376 |
exch2_neighbourId(a,T) ) = T |
377 |
\end{verbatim} |
378 |
This provides a back-reference from the neighbor tiles. \\ |
379 |
|
380 |
The arrays \varlink{exch2\_pi}{exch2_pi} and |
381 |
\varlink{exch2\_pj}{exch2_pj} specify the transformations of indices |
382 |
in exchanges between the neighboring tiles. These transformations are |
383 |
necessary in exchanges between subdomains because a horizontal dimension |
384 |
in one subdomain |
385 |
may map to other horizonal dimension in an adjacent subdomain, and |
386 |
may also have its indexing reversed. This swapping arises from the |
387 |
``folding'' of two-dimensional arrays into a three-dimensional |
388 |
cube. \\ |
389 |
|
390 |
The dimensions of \code{exch2\_pi(t,N,T)} and \code{exch2\_pj(t,N,T)} |
391 |
are the neighbor ID \code{N} and the tile number \code{T} as explained |
392 |
above, plus a vector of length \code{2} containing transformation |
393 |
factors \code{t}. The first element of the transformation vector |
394 |
holds the factor to multiply the index in the same dimension, and the |
395 |
second element holds the the same for the orthogonal dimension. To |
396 |
clarify, \code{exch2\_pi(1,N,T)} holds the mapping of the $x$ axis |
397 |
index of tile \code{T} to the $x$ axis of tile \code{T}'s neighbor |
398 |
\code{N}, and \code{exch2\_pi(2,N,T)} holds the mapping of \code{T}'s |
399 |
$x$ index to the neighbor \code{N}'s $y$ index. \\ |
400 |
|
401 |
One of the two elements of \code{exch2\_pi} or \code{exch2\_pj} for a |
402 |
given tile \code{T} and neighbor \code{N} will be \code{0}, reflecting |
403 |
the fact that the two axes are orthogonal. The other element will be |
404 |
\code{1} or \code{-1}, depending on whether the axes are indexed in |
405 |
the same or opposite directions. For example, the transform vector of |
406 |
the arrays for all tile neighbors on the same subdomain will be |
407 |
\code{(1,0)}, since all tiles on the same subdomain are oriented |
408 |
identically. An axis that corresponds to the orthogonal dimension |
409 |
with the same index direction in a particular tile-neighbor |
410 |
orientation will have \code{(0,1)}. Those with the opposite index |
411 |
direction will have \code{(0,-1)} in order to reverse the ordering. \\ |
412 |
|
413 |
The arrays \varlink{exch2\_oi}{exch2_oi}, |
414 |
\varlink{exch2\_oj}{exch2_oj}, \varlink{exch2\_oi\_f}{exch2_oi_f}, and |
415 |
\varlink{exch2\_oj\_f}{exch2_oj_f} are indexed to tile number and |
416 |
neighbor and specify the relative offset within the subdomain of the |
417 |
array index of a variable going from a neighboring tile \code{N} to a |
418 |
local tile \code{T}. Consider \code{T=1} in the six-tile topology |
419 |
(Fig. \ref{fig:6tile}), where |
420 |
|
421 |
\begin{verbatim} |
422 |
exch2_oi(1,1)=33 |
423 |
exch2_oi(2,1)=0 |
424 |
exch2_oi(3,1)=32 |
425 |
exch2_oi(4,1)=-32 |
426 |
\end{verbatim} |
427 |
|
428 |
The simplest case is \code{exch2\_oi(2,1)}, the southern neighbor, |
429 |
which is \code{Tn=6}. The axes of \code{T} and \code{Tn} have the |
430 |
same orientation and their $x$ axes have the same origin, and so an |
431 |
exchange between the two requires no changes to the $x$ index. For |
432 |
the western neighbor (\code{Tn=5}), \code{code\_oi(3,1)=32} since the |
433 |
\code{x=0} vector on \code{T} corresponds to the \code{y=32} vector on |
434 |
\code{Tn}. The eastern edge of \code{T} shows the reverse case |
435 |
(\code{exch2\_oi(4,1)=-32)}), where \code{x=32} on \code{T} exchanges |
436 |
with \code{x=0} on \code{Tn=2}. \\ |
437 |
|
438 |
The most interesting case, where \code{exch2\_oi(1,1)=33} and |
439 |
\code{Tn=3}, involves a reversal of indices. As in every case, the |
440 |
offset \code{exch2\_oi} is added to the original $x$ index of \code{T} |
441 |
multiplied by the transformation factor \code{exch2\_pi(t,N,T)}. Here |
442 |
\code{exch2\_pi(1,1,1)=0} since the $x$ axis of \code{T} is orthogonal |
443 |
to the $x$ axis of \code{Tn}. \code{exch2\_pi(2,1,1)=-1} since the |
444 |
$x$ axis of \code{T} corresponds to the $y$ axis of \code{Tn}, but the |
445 |
index is reversed. The result is that the index of the northern edge |
446 |
of \code{T}, which runs \code{(1:32)}, is transformed to |
447 |
\code{(-1:-32)}. \code{exch2\_oi(1,1)} is then added to this range to |
448 |
get back \code{(32:1)} -- the index of the $y$ axis of \code{Tn} |
449 |
relative to \code{T}. This transformation may seem overly convoluted |
450 |
for the six-tile case, but it is necessary to provide a general |
451 |
solution for various topologies. \\ |
452 |
|
453 |
|
454 |
|
455 |
Finally, \varlink{exch2\_itlo\_c}{exch2_itlo_c}, |
456 |
\varlink{exch2\_ithi\_c}{exch2_ithi_c}, |
457 |
\varlink{exch2\_jtlo\_c}{exch2_jtlo_c} and |
458 |
\varlink{exch2\_jthi\_c}{exch2_jthi_c} hold the location and index |
459 |
bounds of the edge segment of the neighbor tile \code{N}'s subdomain |
460 |
that gets exchanged with the local tile \code{T}. To take the example |
461 |
of tile \code{T=2} in the twelve-tile topology |
462 |
(Fig. \ref{fig:12tile}): \\ |
463 |
|
464 |
\begin{verbatim} |
465 |
exch2_itlo_c(4,2)=17 |
466 |
exch2_ithi_c(4,2)=17 |
467 |
exch2_jtlo_c(4,2)=0 |
468 |
exch2_jthi_c(4,2)=33 |
469 |
\end{verbatim} |
470 |
|
471 |
Here \code{N=4}, indicating the western neighbor, which is |
472 |
\code{Tn=1}. \code{Tn} resides on the same subdomain as \code{T}, so |
473 |
the tiles have the same orientation and the same $x$ and $y$ axes. |
474 |
The $x$ axis is orthogonal to the western edge and the tile is 16 |
475 |
points wide, so \code{exch2\_itlo\_c} and \code{exch2\_ithi\_c} |
476 |
indicate the column beyond \code{Tn}'s eastern edge, in that tile's |
477 |
halo region. Since the border of the tiles extends through the entire |
478 |
height of the subdomain, the $y$ axis bounds \code{exch2\_jtlo\_c} to |
479 |
\code{exch2\_jthi\_c} cover the height of \code{(1:32)}, plus 1 in |
480 |
either direction to cover part of the halo. \\ |
481 |
|
482 |
For the north edge of the same tile \code{T=2} where \code{N=1} and |
483 |
the neighbor tile is \code{Tn=5}: |
484 |
|
485 |
\begin{verbatim} |
486 |
exch2_itlo_c(1,2)=0 |
487 |
exch2_ithi_c(1,2)=0 |
488 |
exch2_jtlo_c(1,2)=0 |
489 |
exch2_jthi_c(1,2)=17 |
490 |
\end{verbatim} |
491 |
|
492 |
\code{T}'s northern edge is parallel to the $x$ axis, but since |
493 |
\code{Tn}'s $y$ axis corresponds to \code{T}'s $x$ axis, \code{T}'s |
494 |
northern edge exchanges with \code{Tn}'s western edge. The western |
495 |
edge of the tiles corresponds to the lower bound of the $x$ axis, so |
496 |
\code{exch2\_itlo\_c} and \code{exch2\_ithi\_c} are \code{0}, in the |
497 |
western halo region of \code{Tn}. The range of |
498 |
\code{exch2\_jtlo\_c} and \code{exch2\_jthi\_c} correspond to the |
499 |
width of \code{T}'s northern edge, expanded by one into the halo. \\ |
500 |
|
501 |
|
502 |
\subsubsection{Key Routines} |
503 |
|
504 |
Most of the subroutines particular to exch2 handle the exchanges |
505 |
themselves and are of the same format as those described in |
506 |
\ref{sect:cube_sphere_communication} \sectiontitle{Cube sphere |
507 |
communication}. Like the original routines, they are written as |
508 |
templates which the local Makefile converts from \code{RX} into |
509 |
\code{RL} and \code{RS} forms. \\ |
510 |
|
511 |
The interfaces with the core model subroutines are |
512 |
\code{EXCH\_UV\_XY\_RX}, \code{EXCH\_UV\_XYZ\_RX} and |
513 |
\code{EXCH\_XY\_RX}. They override the standard exchange routines |
514 |
when \code{genmake2} is run with \code{exch2} option. They in turn |
515 |
call the local exch2 subroutines \code{EXCH2\_UV\_XY\_RX} and |
516 |
\code{EXCH2\_UV\_XYZ\_RX} for two and three-dimensional vector |
517 |
quantities, and \code{EXCH2\_XY\_RX} and \code{EXCH2\_XYZ\_RX} for two |
518 |
and three-dimensional scalar quantities. These subroutines set the |
519 |
dimensions of the area to be exchanged, call \code{EXCH2\_RX1\_CUBE} |
520 |
for scalars and \code{EXCH2\_RX2\_CUBE} for vectors, and then handle |
521 |
the singularities at the cube corners. \\ |
522 |
|
523 |
The separate scalar and vector forms of \code{EXCH2\_RX1\_CUBE} and |
524 |
\code{EXCH2\_RX2\_CUBE} reflect that the vector-handling subroutine |
525 |
needs to pass both the $u$ and $v$ components of the physical vectors. |
526 |
This swapping arises from the topological folding discussed above, where the |
527 |
$x$ and $y$ axes get swapped in some cases, and is not an |
528 |
issue with the scalar case. These subroutines call |
529 |
\code{EXCH2\_SEND\_RX1} and \code{EXCH2\_SEND\_RX2}, which do most of |
530 |
the work using the variables discussed above. \\ |
531 |
|
532 |
\subsubsection{Experiments and tutorials that use exch2} |
533 |
\label{sec:pkg:exch2:experiments} |
534 |
|
535 |
\begin{itemize} |
536 |
\item{Held Suarez tutorial, in tutorial\_held\_suarez\_cs verification directory, |
537 |
described in section \ref{sect:eg-hs} } |
538 |
\end{itemize} |