|
The ALADIN phasing
This important operation is completed twice a
year to "phase" the ALADIN code with its "mother"
code ARPEGE.
The ALADIN code is following the ARPEGE one in its
permanent evolution (see history of
ALADIN cycles). This is due mainly to the technical evolution of
calculating machines (CDC, CRAY 2, CRAY C90, FUJITSU, ...) which
imposes a new adaptation of the code to the new machine properties.
The evolution of the number of observations over the
world dictates also a code development if one wants to take new
observations into account.
In addition to these two external constraints, code
must contain the newest coding rules if one wants to have a permanent
portable one, ...
Code optimisation, as bugs correction, compas bugs
correction, computation time and memory winning, ... is also another
constraint.
And of course, the evolution of the code is also
imposed by the new developments (a development usually involves
another one !).
A Phasers'guide
defines the rules of these phasings :
The code can be modified for the
above-mentioned purposes by a limited number of known developers or
phasers to avoid complexity and misunderstanding consequences. Thus,
ALADIN code can be touched only by GMAP team, ALADIN partners, some
SCEM teams (COMPAS) and of course persons who come to GMAP for
training.
The software validation after
modification is an absolute must. The phaser or developers must
intensively validate its work. After the entire merge of
modifications, all the system configurations must be tested and
validated. And before using the new software for operational
purpose, a double suite is strongly recommended.
So a phaser must keep in mind that he has to respect all
coding rules, to check all interactions with the remaining code at
each step. He must also phase his scripts and namelists and keep
himself informed about all what is being performed into the whole
code at the same time.
During the 4th Assembly of
ALADIN Partners (Lisbon, December the 6th, 1999), stricter rules for
participation to the maintenance effort inside the ALADIN partnership
were proposed by Dominique Giard and accepted :
The participation of each ALADIN
team to phasing must be at least 1 person once a year. It can be
more for the largest teams (more than 4 full-time equivalent persons
for instance). The minimum size of a phasing team is 8 persons for a
full cycle, less for intermediate cycles if any.
Only one inexperienced person per
phasing team is accepted. This newcomer must of course have some
basic knowledge of ALADIN.
An ordered list of all ALADIN
scientists will be elaborated. Persons will be invited following
this list in "rotation mode", so that everyone will
contribute but not very often. Each team will send its own part of
the list, mentioning constraints if any, to the Toulouse support
team. Otherwise the list of quarterly contributions will be used.
Phasing is centralised in Toulouse
since a very close co-operation between ALADIN and ARPEGE scientists
is required for this crucial exercise. The stays are usually 6 weeks
long, costs may be supported by Météo-France. Deported
validation of cycles are accepted as contributions to phasing
exercises exceptionally, for technical options that cannot be
validated in Toulouse.
See history of the phasing teams
It was also admitted to dedicate 10%
of the ALADIN effort on the maintenance. Maintenance includes not
only phasing, but also code optimisation and cleaning, code
documentation and the preparation of databases for the use of all
partners.
|