| 1 |
heimbach |
1.1 |
|
| 2 |
|
|
25-Sep-2008 |
| 3 |
|
|
heimbach@mit.edu, gforget@mit.edu, dmenemenlis@gmail.com, hzhang@caltech.edu |
| 4 |
|
|
|
| 5 |
|
|
# The following is a description of the CS510 |
| 6 |
|
|
# adjoint-based state estimation setup. |
| 7 |
|
|
# It is based on checkpoint61d. |
| 8 |
|
|
# A prototype setup ran successfully in a 360-proc. config. |
| 9 |
|
|
# on beagle.darwinproject.mit.edu |
| 10 |
|
|
# Stable run for 10-days. sensitivity growth for 1 month |
| 11 |
|
|
# (but not much tweaking with viscosity etc.) |
| 12 |
|
|
# |
| 13 |
|
|
# Default setup for columbia set up to use 450 processors of tile size 34x34. |
| 14 |
|
|
|
| 15 |
|
|
# 0. Set some variables |
| 16 |
|
|
set mitgcm_basedir = [ for example ~/MITgcm_c61d ] |
| 17 |
|
|
|
| 18 |
|
|
# 1. Check out clean MITgcm checkpoint from CVS |
| 19 |
|
|
cvs co -d MITgcm_c61d -r checkpoint61d MITgcm |
| 20 |
|
|
|
| 21 |
|
|
# 2. Check out config. files for this setup: |
| 22 |
|
|
cvs co MITgcm_contrib/high_res_cube/cs510_adjoint |
| 23 |
|
|
|
| 24 |
|
|
# 3. (I prefer it that way, but a matter of taste), |
| 25 |
|
|
\cp -r MITgcm_contrib/high_res_cube/cs510_adjoint $mitgcm_basedir/verification/. |
| 26 |
|
|
|
| 27 |
|
|
# 5. Prepare build; we now need cyrus-makedepend |
| 28 |
|
|
cd $mitgcm_basedir/tools/cyrus-imapd-makedepend |
| 29 |
|
|
./configure |
| 30 |
|
|
make |
| 31 |
|
|
|
| 32 |
|
|
# 6. Make sure to use default topology (or use different one) |
| 33 |
|
|
cd $mitgcm_basedir/verification/cs510_adjoint/code_ad/ |
| 34 |
|
|
\cp -f topol_450_34x34/* . |
| 35 |
|
|
|
| 36 |
|
|
# 7. Build model |
| 37 |
|
|
cd $mitgcm_basedir/verification/cs510_adjoint/build_ad/ |
| 38 |
|
|
../../../tools/genmake2 "-of" "../../../tools/build_options/linux_ia64_ifort+mpi_altix_nas" \ |
| 39 |
|
|
"-makedepend" "../../../tools/cyrus-imapd-makedepend/makedepend" \ |
| 40 |
|
|
"-mods" "../code_ad" |
| 41 |
|
|
make depend |
| 42 |
|
|
make adtaf |
| 43 |
|
|
make -j adall |
| 44 |
|
|
|
| 45 |
|
|
|