1 |
% $Header: /u/gcmpack/manual/part6/mnc.tex,v 1.10 2004/07/05 16:35:32 edhill Exp $ |
2 |
% $Name: $ |
3 |
|
4 |
\section{NetCDF I/O Integration: MNC} |
5 |
\label{sec:pkg:mnc} |
6 |
|
7 |
The \texttt{mnc} package is a set of convenience routines written to |
8 |
expedite the process of creating, appending, and reading NetCDF files. |
9 |
NetCDF is an increasingly popular self-describing file format |
10 |
\cite{rew:97} intended primarily for scientific data sets. An |
11 |
extensive collection of NetCDF reference papers, user guides, |
12 |
software, FAQs, and other information can be obtained from UCAR's web |
13 |
site at: |
14 |
\begin{rawhtml} <A href="http://www.unidata.ucar.edu/packages/netcdf/"> \end{rawhtml} |
15 |
\begin{verbatim} |
16 |
http://www.unidata.ucar.edu/packages/netcdf/ |
17 |
\end{verbatim} |
18 |
\begin{rawhtml} </A> \end{rawhtml} |
19 |
|
20 |
|
21 |
\subsection{Introduction} |
22 |
|
23 |
The \texttt{mnc} package is a two-level convenience library (or |
24 |
``wrapper'') for most of the NetCDF Fortran API. Its purpose is to |
25 |
streamline the user interface to NetCDF by maintaining internal |
26 |
relations (look-up tables) keyed with strings (or names) and entities |
27 |
such as NetCDF files, variables, and attributes. |
28 |
|
29 |
The two levels of the \texttt{mnc} package are: |
30 |
\begin{description} |
31 |
|
32 |
\item[Upper level] \ |
33 |
|
34 |
The upper level contains information about two kinds of |
35 |
associations: |
36 |
\begin{description} |
37 |
\item[grid type] is lookup table indexed with a grid type name. |
38 |
Each grid type name is associated with a number of dimensions, the |
39 |
dimension sizes (one of which may be unlimited), and starting and |
40 |
ending index arrays. The intent is to store all the necessary |
41 |
size and shape information for the Fortran arrays containing |
42 |
MITgcm--style ``tile'' variables (that is, a central region |
43 |
surrounded by a variably-sized ``halo'' or exchange region as |
44 |
shown in Figures \ref{fig:communication_primitives} and |
45 |
\ref{fig:tiling-strategy}). |
46 |
|
47 |
\item[variable type] is a lookup table indexed by a variable type |
48 |
name. For each name, the table contains a reference to a grid |
49 |
type for the variable and the names and values of various |
50 |
attributes. |
51 |
\end{description} |
52 |
|
53 |
Within the upper level, these associations are not permanently tied |
54 |
to any particular NetCDF file. This allows the information to be |
55 |
re-used over multiple file reads and writes. |
56 |
|
57 |
\item[Lower level] \ |
58 |
|
59 |
In the lower (or internal) level, associations are stored for NetCDF |
60 |
files and many of the entities that they contain including |
61 |
dimensions, variables, and global attributes. All associations are |
62 |
on a per-file basis. Thus, each entity is tied to a unique NetCDF |
63 |
file and will be created or destroyed when files are, respectively, |
64 |
opened or closed. |
65 |
|
66 |
\end{description} |
67 |
|
68 |
|
69 |
\subsection{Using MNC} |
70 |
|
71 |
\subsubsection{Grid--Types and Variable--Types} |
72 |
|
73 |
As a convenience for users, the MNC package includes numerous routines |
74 |
to aid in the writing of data to NetCDF format. Probably the biggest |
75 |
convenience is the use of pre-defined ``grid types'' and ``variable |
76 |
types''. These ``types'' are simply look-up tables that store |
77 |
dimensions, indicies, attributes, and other information that can all |
78 |
be retrieved using a single character string. |
79 |
|
80 |
The ``grid types'' are a way of mapping variables within MITgcm to |
81 |
NetCDF arrays. Within MITgcm, most spatial variables are defined |
82 |
using two-- or three--dimensional arrays with ``overlap'' regions (see |
83 |
Figures \ref{fig:communication_primitives}, a possible vertical index, |
84 |
and \ref{fig:tiling-strategy}) and tile indicies such as the following |
85 |
``U'' velocity: |
86 |
\begin{verbatim} |
87 |
_RL uVel (1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy) |
88 |
\end{verbatim} |
89 |
as defined in \filelink{model/inc/DYNVARS.h}{model-inc-DYNVARS.h} |
90 |
|
91 |
The grid type is a character string that encodes the presence and |
92 |
types associated with the four possible dimensions. The character |
93 |
string follows the format |
94 |
\begin{center} |
95 |
\texttt{H0\_H1\_H2\_\_V\_\_T} |
96 |
\end{center} |
97 |
where the terms \textit{H0}, \textit{H1}, \textit{H2}, \textit{V}, |
98 |
\textit{T} can be almost any combination of the following: |
99 |
\begin{center} |
100 |
\begin{tabular}[h]{|ccc|c|c|}\hline |
101 |
\multicolumn{3}{|c|}{Horizontal} & Vertical & Time \\ |
102 |
\textbf{H0}: location & \textbf{H1}: dimensions & \textbf{H2}: halo |
103 |
& \textbf{V}: location & \textbf{T}: level \\\hline |
104 |
\texttt{-} & xy & Hn & \texttt{-} & \texttt{-} \\ |
105 |
U & x & Hy & i & t \\ |
106 |
V & y & & c & \\ |
107 |
Cen & & & & \\ |
108 |
Cor & & & & \\\hline |
109 |
\end{tabular} |
110 |
\end{center} |
111 |
A example list of all pre-defined combinations is contained in the |
112 |
file |
113 |
\begin{center} |
114 |
\texttt{pkg/mnc/pre-defined\_grids.txt}. |
115 |
\end{center} |
116 |
|
117 |
The variable type is an association between a variable type name and the |
118 |
following items: |
119 |
\begin{center} |
120 |
\begin{tabular}[h]{|l|l|}\hline |
121 |
\textbf{Item} & \textbf{Purpose} \\\hline |
122 |
grid type & defines the in-memory arrangement \\ |
123 |
\texttt{bi,bj} dimensions & tiling indices, if present \\\hline |
124 |
\end{tabular} |
125 |
\end{center} |
126 |
and is used by the \texttt{mnc\_cw\_*\_[R|W]} subroutines for reading |
127 |
and writing variables. |
128 |
|
129 |
|
130 |
\subsubsection{An Example} |
131 |
|
132 |
Writing variables to NetCDF files can be accomplished in as few as two |
133 |
function calls. The first function call defines a variable type, |
134 |
associates it with a name (character string), and provides additional |
135 |
information about the indicies for the tile (\texttt{bi},\texttt{bj}) |
136 |
dimensions. The second function call will write the data at, if |
137 |
necessary, the current time level within the model. |
138 |
|
139 |
Examples of the initialization calls can be found in the file |
140 |
\filelink{model/src/ini\_mnc\_io.F}{model-src-ini_mnc_io.F} |
141 |
where these function calls: |
142 |
{\footnotesize |
143 |
\begin{verbatim} |
144 |
C Create MNC definitions for DYNVARS.h variables |
145 |
CALL MNC_CW_ADD_VNAME('iter', '-_-_--__-__t', 0,0, myThid) |
146 |
CALL MNC_CW_ADD_VATTR_TEXT('iter',1, |
147 |
& 'long_name','iteration_count', myThid) |
148 |
|
149 |
CALL MNC_CW_ADD_VNAME('model_time', '-_-_--__-__t', 0,0, myThid) |
150 |
CALL MNC_CW_ADD_VATTR_TEXT('model_time',1, |
151 |
& 'long_name','Model Time', myThid) |
152 |
CALL MNC_CW_ADD_VATTR_TEXT('model_time',1,'units','s', myThid) |
153 |
|
154 |
CALL MNC_CW_ADD_VNAME('U', 'U_xy_Hn__C__t', 4,5, myThid) |
155 |
CALL MNC_CW_ADD_VATTR_TEXT('U',1,'units','m/s', myThid) |
156 |
CALL MNC_CW_ADD_VATTR_TEXT('U',1, |
157 |
& 'coordinates','XU YU RC iter', myThid) |
158 |
|
159 |
CALL MNC_CW_ADD_VNAME('T', 'Cen_xy_Hn__C__t', 4,5, myThid) |
160 |
CALL MNC_CW_ADD_VATTR_TEXT('T',1,'units','degC', myThid) |
161 |
CALL MNC_CW_ADD_VATTR_TEXT('T',1,'long_name', |
162 |
& 'potential_temperature', myThid) |
163 |
CALL MNC_CW_ADD_VATTR_TEXT('T',1, |
164 |
& 'coordinates','XC YC RC iter', myThid) |
165 |
\end{verbatim} |
166 |
} |
167 |
{\noindent initialize four \texttt{VNAME}s and add one or more NetCDF |
168 |
attributes to each.} |
169 |
|
170 |
The four variables defined above are subsequently written at specific |
171 |
time steps within |
172 |
\filelink{model/src/write\_state.F}{model-src-write_state.F} |
173 |
using the function calls: |
174 |
{\footnotesize |
175 |
\begin{verbatim} |
176 |
C Write dynvars using the MNC package |
177 |
CALL MNC_CW_SET_UDIM('state', -1, myThid) |
178 |
CALL MNC_CW_I_W('I','state',0,0,'iter', myIter, myThid) |
179 |
CALL MNC_CW_SET_UDIM('state', 0, myThid) |
180 |
CALL MNC_CW_RL_W('D','state',0,0,'model_time',myTime, myThid) |
181 |
CALL MNC_CW_RL_W('D','state',0,0,'U', uVel, myThid) |
182 |
CALL MNC_CW_RL_W('D','state',0,0,'T', theta, myThid) |
183 |
\end{verbatim} |
184 |
} |
185 |
|
186 |
|
187 |
\subsubsection{Parameters} |
188 |
|
189 |
Most of the MNC--related parameters are contained within a Fortran |
190 |
namelist file called \texttt{data.mnc}. If this file does not exist, |
191 |
then the MNC package will interpret that as an indication that it is |
192 |
not to be used. If the \texttt{data.mnc} file does exist, then it may |
193 |
contain the following parameters: |
194 |
|
195 |
\begin{center} |
196 |
{\footnotesize |
197 |
\begin{tabular}[htb]{|l|c|l|l|}\hline |
198 |
\textbf{Name} & \textbf{T} & |
199 |
\textbf{Default} & \textbf{Description} \\\hline |
200 |
& & & \\ |
201 |
\texttt{useMNC} & L & \texttt{.FALSE.} & |
202 |
\textbf{overall MNC ON/OFF switch} \\ |
203 |
\texttt{mnc\_echo\_gvtypes} & L & \texttt{.FALSE.} & |
204 |
echo pre-defined ``types'' (debugging) \\ |
205 |
\texttt{mnc\_use\_outdir} & L & \texttt{.FALSE.} & |
206 |
create a directory for output \\ |
207 |
\texttt{mnc\_outdir\_str} & S & \texttt{'mnc\_'} & |
208 |
output directory name \\ |
209 |
\texttt{mnc\_outdir\_date} & L & \texttt{.FALSE.} & |
210 |
embed date in the output dir name \\ |
211 |
\texttt{pickup\_write\_mnc} & L & \texttt{.FALSE.} & |
212 |
use MNC to write (create) pickup files \\ |
213 |
\texttt{pickup\_read\_mnc} & L & \texttt{.FALSE.} & |
214 |
use MNC to read pickup files \\ |
215 |
\texttt{mnc\_use\_indir} & L & \texttt{.FALSE.} & |
216 |
use a directory (path) for input \\ |
217 |
\texttt{mnc\_indir\_str} & S & \texttt{''} & |
218 |
input directory (or path) name \\ |
219 |
\texttt{snapshot\_mnc} & L & \texttt{.FALSE.} & |
220 |
write \texttt{snapshot} (instantaneous) w/MNC \\ |
221 |
\texttt{monitor\_mnc} & L & \texttt{.FALSE.} & |
222 |
write \texttt{monitor} w/MNC \\ |
223 |
\texttt{timeave\_mnc} & L & \texttt{.FALSE.} & |
224 |
write \texttt{timeave} w/MNC \\\hline |
225 |
\end{tabular} |
226 |
} |
227 |
\end{center} |
228 |
|
229 |
Additional MNC--related parameters are contained within the main |
230 |
\texttt{data} namelist file and in some of the namelist files for |
231 |
individual packages. These options are: |
232 |
\begin{center} |
233 |
{\footnotesize |
234 |
\begin{tabular}[htb]{|l|c|l|l|}\hline |
235 |
\textbf{Name} & \textbf{T} & |
236 |
\textbf{Default} & \textbf{Description} \\\hline |
237 |
\multicolumn{4}{|c|}{\ } \\ |
238 |
\multicolumn{4}{|c|}{Main namelist file: |
239 |
``\textbf{data}''} \\\hline |
240 |
\texttt{snapshot\_ioinc} & L & \texttt{.FALSE.} & |
241 |
write \texttt{snapshot} ``inclusively'' \\ |
242 |
\texttt{timeave\_ioinc} & L & \texttt{.FALSE.} & |
243 |
write \texttt{timeave} ``inclusively'' \\ |
244 |
\texttt{monitor\_ioinc} & L & \texttt{.FALSE.} & |
245 |
write \texttt{monitor} ``inclusively'' \\\hline |
246 |
\multicolumn{4}{|c|}{\ } \\ |
247 |
\multicolumn{4}{|c|}{Diagnostics namelist file: |
248 |
``\textbf{data.diagnostics}''} \\\hline |
249 |
\texttt{diag\_mnc} & L & \texttt{.FALSE.} & |
250 |
write \texttt{diagnostics} w/MNC \\ |
251 |
\texttt{diag\_ioinc} & L & \texttt{.FALSE.} & |
252 |
write \texttt{diagnostics} ``inclusively'' \\\hline |
253 |
\end{tabular} |
254 |
} |
255 |
\end{center} |
256 |
|
257 |
By default, turning on MNC for a particular output stream will result |
258 |
in turning off all the corresponding (usually, default) MDSIO or |
259 |
STDOUT output mechanisms. In other words, output defaults to being an |
260 |
exclusive selection. To enable multiple kinds of simultaneous output, |
261 |
flags of the form \texttt{NAME\_ioinc} can be used where \texttt{NAME} |
262 |
corresponds to the various MNC output flags. When a |
263 |
\texttt{NAME\_ioinc} flag is set to \texttt{.TRUE.}, then multiple |
264 |
forms of output are allowed for the \texttt{NAME} output mechanism. |
265 |
The intent of this design is that typical users will only want one |
266 |
kind of output while people debugging the code (particularly the I/O |
267 |
routines) may want simultaneous types of output. |
268 |
|
269 |
This ``inclusive'' versus ``exclusive'' design is easily applied in |
270 |
cases where three or more kinds of output may be generated. Thus, it |
271 |
can be readily extended to additional new output types (eg. HDF5). |
272 |
|
273 |
Input types are always exclusive. |