1 |
<!DOCTYPE ARTICLE PUBLIC "-//OASIS//DTD DocBook V4.1//EN"> |
2 |
|
3 |
<article id="MITgcm-Development-HOWTO"> |
4 |
|
5 |
<!-- |
6 |
Build commands: |
7 |
db2ps -d ldp.dsl devel_HOWTO.sgml |
8 |
db2pdf -d ldp.dsl devel_HOWTO.sgml |
9 |
db2html -d ./ldp.dsl devel_HOWTO.sgml |
10 |
db2html -u -d ./ldp.dsl devel_HOWTO.sgml |
11 |
--> |
12 |
|
13 |
<articleinfo> |
14 |
<title>MITgcm Development HOWTO</title> |
15 |
|
16 |
<author> |
17 |
<firstname>Ed</firstname> |
18 |
<surname>Hill III</surname> |
19 |
<affiliation> |
20 |
<address><email>eh3@mit.edu</email></address> |
21 |
</affiliation> |
22 |
</author> |
23 |
|
24 |
<revhistory> |
25 |
<revision> |
26 |
<revnumber>0.01</revnumber> |
27 |
<date>2003-08-07</date> |
28 |
<authorinitials>eh3</authorinitials> |
29 |
<revremark> |
30 |
Initial version. |
31 |
</revremark> |
32 |
</revision> |
33 |
</revhistory> |
34 |
|
35 |
<abstract> |
36 |
<para>This document describes how to develop software for the |
37 |
MITgcm project.</para> |
38 |
</abstract> |
39 |
</articleinfo> |
40 |
|
41 |
<sect1 id="intro"> |
42 |
<title>Introduction</title> <para>The purpose of this document is |
43 |
to help new developers get "up to speed" with MITgcm |
44 |
development.</para> |
45 |
<sect2> |
46 |
<title>New Versions of This Document</title> <para>You can |
47 |
obtain the latest version of this document <ulink |
48 |
url="http://mitgcm.org/dev_docs/devel_HOWTO/">online</ulink> in |
49 |
various formats.</para> |
50 |
</sect2> |
51 |
<sect2> |
52 |
<title>Feedback and corrections</title> <para>If you have |
53 |
questions or comments about this document, please feel free to |
54 |
<ulink url="mailto:MITgcm-support@mitgcm.org">contact the |
55 |
authors</ulink>. |
56 |
</para> |
57 |
</sect2> |
58 |
</sect1> |
59 |
|
60 |
<sect1 id="background"> |
61 |
<title>Background</title> |
62 |
|
63 |
<sect2> |
64 |
<title>User Manual</title> |
65 |
|
66 |
<para>Before jumping into |
67 |
development, please familiarize yourself with the MITgcm user |
68 |
manual which is available <ulink |
69 |
url="http://mitgcm.org/">on the main web page</ulink>. This |
70 |
document contains volumes of useful information and is included |
71 |
here by reference.</para> |
72 |
|
73 |
<para>Also, a "snapshot" or<ulink |
74 |
url="http://mitgcm.org/dev_docs/">development version</ulink> of |
75 |
the user manual may be available, though this is only put on the |
76 |
web for testing purposes.</para> |
77 |
</sect2> |
78 |
|
79 |
<sect2> |
80 |
<title>Prerequisites</title> <para>To develop for MITgcm project |
81 |
you will need a UNIX or UNIX-like set of build tools including |
82 |
the following:</para> |
83 |
<blockquote> |
84 |
<simplelist type="inline"> |
85 |
<member>CVS client</member> |
86 |
<member>make or (preferably) GNU make</member> |
87 |
<member>FORTRAN compiler</member> |
88 |
<member>C compiler</member> |
89 |
<member>[ba]sh and [t]csh shells</member> |
90 |
<member>PERL</member> |
91 |
<member>LaTeX and LaTeX2HTML</member> |
92 |
</simplelist> |
93 |
</blockquote> |
94 |
<para>Essentially all of the work described here has been tested |
95 |
on recent versions of Red Hat Linux (eg. 7.3 through 9). Except |
96 |
where noted, all shell commands will be provided using bash |
97 |
syntax. |
98 |
</para> |
99 |
</sect2> |
100 |
|
101 |
</sect1> |
102 |
|
103 |
<sect1 id="cvs"> |
104 |
<title>CVS Repository</title> |
105 |
<sect2> |
106 |
<title>Layout</title> |
107 |
|
108 |
<para>Unlike many open source projects, the MITgcm CVS tree does |
109 |
not follow a simple "src", "docs", "share", and "test" directory |
110 |
layout. Instead, there are multiple higher-level directories |
111 |
that each, to some extent, depend upon the presence of the |
112 |
others. The tree currently resembles:</para> |
113 |
|
114 |
<programlisting>gcmpack/ |
115 |
MITgcm-contrib contributed code |
116 |
CS-regrid goes into utils |
117 |
cvspolicy.html -save- |
118 |
CVSROOT -save- |
119 |
development experimental stuff |
120 |
manual -save- |
121 |
misc -?- |
122 |
|
123 |
MITgcm code |
124 |
adjoint fold into genmake |
125 |
bin stub for ecco build |
126 |
compare01 old from 20th century |
127 |
diags timeave f77 in pkgs now |
128 |
doc tags -- connect to real docs? |
129 |
eesupp cnh? |
130 |
exe ecco user build |
131 |
*- jobs runtime shell scripts for |
132 |
| various platforms |
133 |
| lsopt line search |
134 |
m| model main dynamics (core) |
135 |
e| optimization_drivers ? |
136 |
r| optim line search interface |
137 |
g| pkg alternate and optional numerics, etc. |
138 |
e*- tools |
139 |
?| tutorial_examples documented tests |
140 |
| only populated on release1 branch |
141 |
| and not validated during "testscript" |
142 |
*- utils |
143 |
verification std tests |
144 |
|
145 |
|
146 |
mitgcmdoc -> manual -remove- |
147 |
mitgcm.org build web site |
148 |
models -?- |
149 |
packages -?- |
150 |
preprocess -?- |
151 |
tmp -?- |
152 |
</programlisting> |
153 |
|
154 |
<para>Efforts are underway to reduce the complexity.</para> |
155 |
|
156 |
</sect2> |
157 |
|
158 |
<!-- |
159 |
<sect2> |
160 |
<title>Releases</title> <para>Currently, there are two main |
161 |
branches:</para> |
162 |
<itemizedlist mark="bullet"> |
163 |
<listitem> |
164 |
<para>Development</para> |
165 |
<itemizedlist mark="bullet"> |
166 |
<listitem> |
167 |
<para>MAIN</para> |
168 |
</listitem> |
169 |
<listitem> |
170 |
<para>ecco-branch</para> |
171 |
</listitem> |
172 |
</itemizedlist> |
173 |
</listitem> |
174 |
<listitem> |
175 |
<para>Production</para> |
176 |
<itemizedlist mark="bullet"> |
177 |
<listitem> |
178 |
<para>Release1</para> |
179 |
</listitem> |
180 |
<listitem> |
181 |
<para>Release2</para> |
182 |
</listitem> |
183 |
</itemizedlist> |
184 |
</listitem> |
185 |
</itemizedlist> |
186 |
</sect2> |
187 |
--> |
188 |
|
189 |
<sect2> |
190 |
<title>Branches</title> |
191 |
|
192 |
<para>As shown in the online <ulink |
193 |
url="http://dev.mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/doc/tag-index?graph=1.174">ViewCVS-generated |
194 |
tree</ulink>, the MITgcm codebase is split into to two branches |
195 |
or "lines" under which development proceeds. These two lines |
196 |
are referred to as the "MAIN" and "ecco" versions of the code. |
197 |
While not identical, the bulk of the MAIN and ecco lines are |
198 |
composed of files from the same codebase. |
199 |
</para> |
200 |
|
201 |
<para>Periodically, a "Release" branch is formed from the "MAIN" |
202 |
development branch. This is done in order to create a |
203 |
relatively stable reference point for both users and developers. |
204 |
The intent is that once a relese branch has been created, only |
205 |
bug-fixes will be added to it. Meanwhile, development (which |
206 |
might "break" or otherwise render invalid the documentation, |
207 |
tutorials, and/or examples contained within a release branch) is |
208 |
allowed to continue along the MAIN and ecco lines.</para> |
209 |
</sect2> |
210 |
|
211 |
<sect2> |
212 |
<title>Tagging</title> |
213 |
|
214 |
<para>The intent of tagging is to create "known-good" |
215 |
checkpoints that developers can use as references. |
216 |
Traditionally, MITgcm tagging has maintained the following |
217 |
conventions:</para> |
218 |
|
219 |
<orderedlist> |
220 |
<listitem> |
221 |
<para>Developer checks out code into a local CVS-managed |
222 |
directory, makes various changes/additions, tests these |
223 |
edits, and eventually reaches a point where (s)he is |
224 |
satisfied that the changes form a new "useful" point in the |
225 |
evolution of the code.</para> |
226 |
</listitem> |
227 |
|
228 |
<listitem> |
229 |
<para>The developer then runs the <ulink |
230 |
url="http://dev.mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/verification/testscript">testscript</ulink> |
231 |
shell script to see if any problems are introduced. While |
232 |
not intended to be exhaustive, the test cases within the |
233 |
verification directory do provide some indication whether |
234 |
gross errors have been introduced. |
235 |
</para> |
236 |
</listitem> |
237 |
|
238 |
<listitem> |
239 |
<para>Having satisfied him- or herself that the changes are |
240 |
ready to be committed to the CVS repository, the developer |
241 |
then:</para> |
242 |
<orderedlist> |
243 |
<listitem> |
244 |
<para>adds a "checkpointXY_pre" comment (where X is a |
245 |
checkpoint number and Y is a letter) to the <ulink |
246 |
url="http://dev.mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/doc/tag-index">tag-index</ulink> |
247 |
file and checks it into the CVS repository</para> |
248 |
</listitem> |
249 |
<listitem> |
250 |
<para>submits the set of changes to the CVS repository |
251 |
and adds comments to <filename>tag-index</filename> |
252 |
describing what the changes are along with a matching |
253 |
"checkpointXY_post" entry</para> |
254 |
</listitem> |
255 |
</orderedlist> |
256 |
</listitem> |
257 |
</orderedlist> |
258 |
|
259 |
<para>The result of this tagging procedure is a sequence of |
260 |
development checkpoints with comments which resembles:</para> |
261 |
|
262 |
<programlisting> |
263 |
checkpoint50e_post |
264 |
o make KPP work with PTRACERS |
265 |
- fix gad_calc_rhs to call new routine kpp_transport_ptr, which is |
266 |
nearly a copy of kpp_transport_s |
267 |
- there is no analogue to SurfaceTendencyS, so I have to use |
268 |
gPtr(of the surface layer) instead |
269 |
o add a new platform SunFire+mpi (SunFire 15000) to genmake |
270 |
checkpoint50e_pre |
271 |
|
272 |
checkpoint50d_post |
273 |
o change kpp output from multiple-record state files to single-record state |
274 |
files analogous to write_state.F |
275 |
o reduce the output frequency of cg3d-related stuff to the monitor frequency, |
276 |
analogous to the cg2d-related output. |
277 |
o fix small problem with in ptracers_write_checkpoint.F: len(suff)=512, |
278 |
so that writing to internal file fn (with length 512) fails. |
279 |
checkpoint50d_pre |
280 |
</programlisting> |
281 |
|
282 |
<para>This information can be used to refer to various stages of |
283 |
the code development. For example, bugs can be traced to |
284 |
individual sets of CVS checkins based upon their first |
285 |
appearance when comparing the results from different |
286 |
checkpoints.</para> |
287 |
|
288 |
</sect2> |
289 |
</sect1> |
290 |
|
291 |
|
292 |
<sect1 id="documentation"> |
293 |
<title>Editing the Documentation</title> |
294 |
|
295 |
<sect2 id="documentation_getting"> |
296 |
<title>Getting the Docs and Code</title> |
297 |
|
298 |
<para>The first step towards editing the documentation is to |
299 |
checkout a copy of code, docs, and build scripts from the CVS |
300 |
server using:</para> |
301 |
|
302 |
<screen> |
303 |
$ export CVS_RSH=ssh |
304 |
$ export CVSROOT=':ext:auden.lcs.mit.edu:/u/u3/gcmpack' |
305 |
$ mkdir scratch |
306 |
$ cvs co MITgcm manual mitgcm.org |
307 |
</screen> |
308 |
|
309 |
<para>These commands extract the necessary information from the |
310 |
CVS server and create a temporary (called |
311 |
<filename>scratch</filename>) directory for the storage of the |
312 |
HTML and other files that will be created. Please note that you |
313 |
must either create <filename>scratch</filename> as shown or edit |
314 |
the various <filename>Makefile</filename>s and scripts used to |
315 |
create the documentation.</para> |
316 |
</sect2> |
317 |
|
318 |
<sect2> |
319 |
<title>Editing</title> |
320 |
|
321 |
<para>The documentation is contained in the |
322 |
<filename>manual</filename> directory in a raw LaTeX format. |
323 |
The main document is <filename>manual.tex</filename> and it uses |
324 |
<command>\input{}</command>s to include the chapters and |
325 |
subsections.</para> |
326 |
|
327 |
<para>Since the same LaTeX source is used to produce PostScript, |
328 |
PDF, and HTML output, care should be taken to follow certain |
329 |
conventions. Two of the most important are the usage of the |
330 |
<command>\filelink{}{}</command> and |
331 |
<command>\varlink{}{}</command> commands. Both of these |
332 |
commands have been defined to simplify the connection between |
333 |
the automatically generated ("code browser") HTML and the HTML |
334 |
version of the manual produced by LaTeX2HTML. They each take |
335 |
two arguments (corresponding to the contents of the two sets of |
336 |
curly braces) which are the text that the author wishes to be |
337 |
"wrapped" within the link, and a specially formatted link thats |
338 |
relative to the <filename>MITgcm</filename> directory within the |
339 |
CVS tree.</para> |
340 |
|
341 |
<para>The result is a command that resembles either</para> |
342 |
|
343 |
<orderedlist> |
344 |
<listitem> |
345 |
<para>a reference to a variable or subroutine name such as |
346 |
<command>\varlink{tRef}{tRef}</command>, or </para> |
347 |
</listitem> |
348 |
|
349 |
<listitem> |
350 |
<para>a reference to a file such as |
351 |
<command>\varlink{tRef}{path-to-the-file_name.F}</command> |
352 |
where the absolute path to the file is of the form |
353 |
<filename>/foo/MITgcm/path/to/the/file_name.F</filename></para> |
354 |
<para>(please note how the leading "/foo/MITgcm" |
355 |
component of the path is dropped leaving the path |
356 |
<emphasis>relative</emphasis> to the head of the code |
357 |
directory and each directory separator "/" is turned |
358 |
into a "-")</para> |
359 |
</listitem> |
360 |
</orderedlist> |
361 |
|
362 |
|
363 |
|
364 |
</sect2> |
365 |
|
366 |
<sect2> |
367 |
<title>Building</title> <para>Given the directory structure of |
368 |
<xref linkend="documentation_getting">, the entire documentation |
369 |
for the web site can be built using:</para> |
370 |
|
371 |
<screen> |
372 |
$ cd mitgcm.org/devel/buildweb |
373 |
$ make All |
374 |
</screen> |
375 |
|
376 |
<para>Which builds the PDF from the LaTeX source, creates the |
377 |
HTML output from the LaTeX source, parses the FORTRAN code base |
378 |
to produce a hyperlinked HTML version of the source, and then |
379 |
determines the cross-linking between the various HTML |
380 |
components.</para> |
381 |
|
382 |
<para>If there are no errors, the result of the build process |
383 |
(which can take 30+ minutes on a P4/2.5Ghz) will be contained |
384 |
within a single directory called |
385 |
<filename>scratch/dev_docs</filename>. This is a freshly built |
386 |
version of the entire on-line users manual. If you have the |
387 |
correct permissions, it can be directly copied to the web server |
388 |
area:</para> |
389 |
|
390 |
<screen> |
391 |
$ mv scratch/dev_docs /u/u0/httpd/html |
392 |
</screen> |
393 |
|
394 |
<para>and the update is complete.</para> |
395 |
|
396 |
</sect2> |
397 |
|
398 |
</sect1> |
399 |
|
400 |
<sect1 id="coding"> |
401 |
<title>Coding</title> |
402 |
|
403 |
<sect2 id="packages"> |
404 |
<title>Coding Packages</title> |
405 |
|
406 |
<para>Optional parts of code have been separated from the MITgcmUV |
407 |
core driver code and organised into packages. The packaging |
408 |
structure provides a mechanism for maintaining suites of code, |
409 |
specific to particular classes of problems, in a way that is |
410 |
cleanly separated from the generic fluid dynamical |
411 |
engine.</para> |
412 |
|
413 |
<para>The MITgcmUV packaging structure is described below using |
414 |
generic package names ${pkg}. A concrete examples of a package |
415 |
is the code for implementing GM/Redi mixing. This code uses the |
416 |
package name</para> |
417 |
|
418 |
</sect2> |
419 |
|
420 |
</sect1> |
421 |
|
422 |
<sect1> |
423 |
<title>Chris's Notes...</title> |
424 |
|
425 |
<programlisting> |
426 |
MITgcmUV Packages |
427 |
================= |
428 |
|
429 |
Optional parts of code are separated from |
430 |
the MITgcmUV core driver code and organised into |
431 |
packages. The packaging structure provides a mechanism for |
432 |
maintaining suites of code, specific to particular |
433 |
classes of problem, in a way that is cleanly |
434 |
separated from the generic fluid dynamical engine. |
435 |
|
436 |
The MITgcmUV packaging structure is describe |
437 |
below using generic package names ${pkg}. |
438 |
A concrete examples of a package is the code |
439 |
for implementing GM/Redi mixing. This code uses |
440 |
the package name |
441 |
* ${PKG} = GMREDI |
442 |
* ${pkg} = gmredi |
443 |
* ${Pkg} = gmRedi |
444 |
|
445 |
Package states |
446 |
============== |
447 |
|
448 |
Packages can be any one of four states, included, |
449 |
excluded, enabled, disabled as follows: |
450 |
|
451 |
included(excluded) compile time state which |
452 |
includes(excludes) package |
453 |
code and routine calls from |
454 |
compilation/linking etc... |
455 |
|
456 |
enabled(disabled) run-time state which |
457 |
enables(disables) package code |
458 |
execution. |
459 |
|
460 |
Every call to a ${pkg}_... routine from outside the package |
461 |
should be placed within both a |
462 |
#ifdef ALLOW_${PKG} ... block and a |
463 |
if ( use${Pkg} ) ... then block. |
464 |
Package states are generally not expected to change during |
465 |
a model run. |
466 |
|
467 |
Package structure |
468 |
================= |
469 |
|
470 |
o Each package gets its runtime configuration |
471 |
parameters from a file named "data.${pkg}" |
472 |
Package runtime config. options are imported |
473 |
into a common block held in a header file |
474 |
called "${PKG}.h". |
475 |
|
476 |
o The core driver part of the model can check |
477 |
for runtime enabling or disabling of individual packages |
478 |
through logical flags use${Pkg}. |
479 |
The information is loaded from a |
480 |
global package setup file called "data.pkg". |
481 |
The use${Pkg} flags are not used within |
482 |
individual packages. |
483 |
|
484 |
o Included in "${PKG}.h" is a logical flag |
485 |
called ${Pkg}IsOn. The "${PKG}.h" header file can be imported |
486 |
by other packages to check dependencies and requirements |
487 |
from other packages ( see "Package Boot Sequence" section). |
488 |
NOTE: This procedure is not presently implemented, |
489 |
----- neither for kpp nor for gmRedi. |
490 |
|
491 |
CPP Flags |
492 |
========= |
493 |
|
494 |
1. Within the core driver code flags of the form |
495 |
ALLOW_${PKG} are used to include or exclude |
496 |
whole packages. The ALLOW_${PKG} flags are included |
497 |
from a PKG_CPP_OPTIONS block which is currently |
498 |
held in-line in the CPP_OPTIONS.h header file. |
499 |
e.g. |
500 |
|
501 |
Core model code ..... |
502 |
|
503 |
#include "CPP_OPTIONS.h" |
504 |
: |
505 |
: |
506 |
: |
507 |
|
508 |
#ifdef ALLOW_${PKG} |
509 |
if ( use${Pkg} ) CALL ${PKG}_DO_SOMETHING(...) |
510 |
#endif |
511 |
|
512 |
2. Within an individual package a header file, |
513 |
"${PKG}_OPTIONS.h", is used to set CPP flags |
514 |
specific to that package. It is not recommended |
515 |
to include this file in "CPP_OPTIONS.h". |
516 |
|
517 |
|
518 |
Package Boot Sequence |
519 |
===================== |
520 |
|
521 |
Calls to package routines within the core code timestepping |
522 |
loop can vary. However, all packages follow a required |
523 |
"boot" sequence outlined here: |
524 |
|
525 |
1. S/R PACKAGES_BOOT() |
526 |
: |
527 |
CALL OPEN_COPY_DATA_FILE( 'data.pkg', 'PACKAGES_BOOT', ... ) |
528 |
|
529 |
|
530 |
2. S/R PACKAGES_READPARMS() |
531 |
: |
532 |
#ifdef ALLOW_${PKG} |
533 |
if ( use${Pkg} ) |
534 |
& CALL ${PKG}_READPARMS( retCode ) |
535 |
#endif |
536 |
|
537 |
2. S/R PACKAGES_CHECK() |
538 |
: |
539 |
#ifdef ALLOW_${PKG} |
540 |
if ( use${Pkg} ) |
541 |
& CALL ${PKG}_CHECK( retCode ) |
542 |
#else |
543 |
if ( use${Pkg} ) |
544 |
& CALL PACKAGES_CHECK_ERROR('${PKG}') |
545 |
#endif |
546 |
|
547 |
3. S/R PACKAGES_INIT() |
548 |
: |
549 |
#ifdef ALLOW_${PKG} |
550 |
if ( use${Pkg} ) |
551 |
& CALL ${PKG}_INIT( retCode ) |
552 |
#endif |
553 |
|
554 |
|
555 |
Description |
556 |
=========== |
557 |
|
558 |
- ${PKG}_READPARMS() |
559 |
is responsible for reading |
560 |
in the package parameters file data.${pkg}, and storing |
561 |
the package parameters in "${PKG}.h". |
562 |
-> called in INITIALISE_FIXED |
563 |
|
564 |
- ${PKG}_CHECK() |
565 |
is responsible for validating |
566 |
basic package setup and inter-package dependencies. |
567 |
${PKG}_CHECK can import other package parameters it may |
568 |
need to check. This is done through header files "${PKG}.h". |
569 |
It is assumed that parameters owned by other packages |
570 |
will not be reset during ${PKG}_CHECK(). |
571 |
-> called in INITIALISE_FIXED |
572 |
|
573 |
- ${PKG}_INIT() |
574 |
is responsible for completing the |
575 |
internal setup of a package. This routine is called after |
576 |
the core model state has been completely initialised |
577 |
but before the core model timestepping starts. |
578 |
-> called in INITIALISE_VARIA |
579 |
|
580 |
Summary |
581 |
======= |
582 |
|
583 |
- CPP options: |
584 |
----------------------- |
585 |
* ALLOW_${PKG} include/exclude package for compilation |
586 |
|
587 |
- FORTRAN logical: |
588 |
----------------------- |
589 |
* use${Pkg} enable package for execution at runtime |
590 |
-> declared in PARAMS.h |
591 |
* ${Pkg}IsOn for package cross-dependency check |
592 |
-> declared in ${PKG}.h |
593 |
N.B.: Not presently used! |
594 |
|
595 |
- header files |
596 |
----------------------- |
597 |
* ${PKG}_OPTIONS.h has further package-specific CPP options |
598 |
* ${PKG}.h package-specific common block variables, fields |
599 |
|
600 |
- FORTRAN source files |
601 |
----------------------- |
602 |
* ${pkg}_readparms.F reads parameters from file data.${pkg} |
603 |
* ${pkg}_check.F checks package dependencies and consistencies |
604 |
* ${pkg}_init.F initialises package-related fields |
605 |
* ${pkg}_... .F package source code |
606 |
|
607 |
- parameter file |
608 |
----------------------- |
609 |
* data.${pkg} parameter file |
610 |
</programlisting> |
611 |
|
612 |
</sect1> |
613 |
|
614 |
|
615 |
</article> |