++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
From greve at lowtem.hokudai.ac.jp Thu Sep 5 09:47:42 2013
From: greve at lowtem.hokudai.ac.jp (Ralf Greve)
Date: Thu, 05 Sep 2013 16:47:42 +0900
Subject: [SICOPOLIS:22] Problem with Lis version 1.2.56 and newer
Message-ID: <5228371E.90107@lowtem.hokudai.ac.jp>
Dear colleagues,
the shallow shelf solver of SICOPOLIS uses the Library of Iterative
Solvers for Linear Systems (Lis, ) in order
to compute the horizontal velocity field of ice shelves (solver
BiCGSTAB(l), preconditioner ILU, with OpenMP). As you may know, we have
the persistent problem that things work perfectly fine with Lis versions
1.2.55 and older; however, with versions 1.2.56 and newer the solver
produces nonsense.
Please see
for an example (diagnostic computation for the Antarctic Ross ice
shelf). When Lis 1.2.55 is used, SICOPOLIS produces a nice, realistic
velocity field with maximum values of a bit more than 1000 m/yr.
However, when Lis 1.2.56 is used, the result is evidently nonsense
(values of the order of 10^13 m/yr!).
The Lis changelog reports for version 1.2.56
- Added support for lis_matrix_is_assembled() in lis_matrix_set_*()
for calling solver multiple times without allocating new memory
(thanks to Malte Thoma).
- Changed origin of Fortran subroutine lis_vector_set_size() to 1
(thanks to Kengo Tomida).
However, so far we have been unable to track down the problem. Do you
have any idea what might cause it? You may have a look at the relevant
part in the Fortran source code; please see
(lines 768-811). The strange thing is that the Lis test program test4f.F
(that comes with the package) works well on our machines for the old and
new Lis versions, so that it cannot be a general problem of the
installations.
Any help would be greatly appreciated. Thanks a lot in advance.
Best regards,
Ralf
--
Dr. Ralf Greve
Professor, Glacier and Ice Sheet Research
Institute of Low Temperature Science
Hokkaido University
Kita-19, Nishi-8, Kita-ku
Sapporo 060-0819, Japan
Phone: (+81)-(0)11-706-6891
greve at lowtem.hokudai.ac.jp
http://wwwice.lowtem.hokudai.ac.jp/~greve/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
From greve at lowtem.hokudai.ac.jp Fri Sep 6 11:20:03 2013
From: greve at lowtem.hokudai.ac.jp (Ralf Greve)
Date: Fri, 06 Sep 2013 18:20:03 +0900
Subject: [SICOPOLIS:23] Problem with Lis solved
Message-ID: <52299E43.5010200@lowtem.hokudai.ac.jp>
Dear colleagues,
with the help of Fuyuki Saito, Tatsuru Sato and Akira Nishida (Lis
developer) we have been able to track down the problem. There is
apparently a bug in the Fortran subroutine lis_vector_set_size of Lis.
We could circumvent it by not using this function, and defining the size
of the vectors with the function lis_vector_duplicate instead. Please
see
,
lines 775-779, for the detailed implementation.
SICOPOLIS should now work with recent versions (1.4.x) of Lis. The price
to pay is that is does not work with older versions any more. This is
unavoidable because, among other things, the naming convention for the
storage method of the sparse matrices was changed in Lis. So please
install a recent version of Lis if you are
planning to do coupled ice-sheet/ice-shelf simulations with the latest
revision of SICOPOLIS.
Best regards,
Ralf
--
Dr. Ralf Greve
Professor, Glacier and Ice Sheet Research
Institute of Low Temperature Science
Hokkaido University
Kita-19, Nishi-8, Kita-ku
Sapporo 060-0819, Japan
Phone: (+81)-(0)11-706-6891
greve at lowtem.hokudai.ac.jp
http://wwwice.lowtem.hokudai.ac.jp/~greve/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
From greve at lowtem.hokudai.ac.jp Tue Sep 10 10:09:39 2013
From: greve at lowtem.hokudai.ac.jp (Ralf Greve)
Date: Tue, 10 Sep 2013 17:09:39 +0900
Subject: [SICOPOLIS:24] Coming soon: Release of version 3.1
Message-ID: <522ED3C3.40408@lowtem.hokudai.ac.jp>
Dear colleagues,
even though no new physics has been implemented since version 3.0, I am
planning to release version 3.1 on the basis of the current state of
SICOPOLIS. This is because the code has seen a number of improvements
over the last months:
- Discretisation of the horizontal and vertical advection
terms in the 3-d temperature and age equations:
Improved upwind scheme implemented that uses interpolated
velocities on the main grid (ADV_HOR==3, ADV_VERT==3).
- Slightly improved boundary and transition conditions for
the computation of the age.
- Index order of 2-d and 3-d arrays in the time-slice output
files *.nc (netCDF) and *.erg (native binary) changed:
(j,i) -> (i,j), (kc/t/r,j,i) -> (i,j,kc/t/r).
Output time units changed from seconds to years.
- Routines that use Lis changed such that they are compatible
with Lis version 1.4.13 and newer.
- Solvers for systems of linear equations assembled in the
new module sico_sle_solvers.
- All subroutines and functions are now included in the main
program sicopolis following a 'contains' statement.
- Intent attributes inserted in all subroutines and
functions.
- New topography input data for Antarctica (BEDMAP2) and
Greenland (Bamber Version 2).
- New input data for the geothermal heat flux for Antarctica
and Greenland [Michael Purucker (pers. comm. 2012), based
on Fox Maule et al., (2005)].
For more details please see the commit logs at
(since
revision 341).
SICOPOLIS will be developed continuously beyond that point; this will
happen within the upcoming developmental version 3.2-dev. Please stay tuned.
Best regards,
Ralf
--
Dr. Ralf Greve
Professor, Glacier and Ice Sheet Research
Institute of Low Temperature Science
Hokkaido University
Kita-19, Nishi-8, Kita-ku
Sapporo 060-0819, Japan
Phone: (+81)-(0)11-706-6891
greve at lowtem.hokudai.ac.jp
http://wwwice.lowtem.hokudai.ac.jp/~greve/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
From greve at lowtem.hokudai.ac.jp Thu Sep 12 06:41:42 2013
From: greve at lowtem.hokudai.ac.jp (Ralf Greve)
Date: Thu, 12 Sep 2013 13:41:42 +0900
Subject: [SICOPOLIS:25] Release of SICOPOLIS version 3.1
Message-ID: <52314606.3020405@lowtem.hokudai.ac.jp>
Dear colleagues,
you have probably seen this already on the Cryolist, but just in case:
It is my pleasure to announce that the official version 3.1 of the
ice-sheet model SICOPOLIS (SImulation COde for POLythermal Ice Sheets)
has been released earlier today.
The physics is still the same as in version 3.0, but the code has seen a
number of improvements over the last months that justify the release.
For a summary please see
(scroll down to lines 468-498). For more details please see the commit
logs at
(revisions 341-399).
On the SICOPOLIS website you will find
further information, downloading instructions and documentation.
Feedbacks are of course welcome, and I'll be happy to help in case of
problems that may come up when using the program.
Best regards,
Ralf
--
Dr. Ralf Greve
Professor, Glacier and Ice Sheet Research
Institute of Low Temperature Science
Hokkaido University
Kita-19, Nishi-8, Kita-ku
Sapporo 060-0819, Japan
Phone: (+81)-(0)11-706-6891
greve at lowtem.hokudai.ac.jp
http://wwwice.lowtem.hokudai.ac.jp/~greve/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++