I've installed Amber 9 on an itanium cluster running Linux using the
ifort compiler and qsnet's implementation of mpich. After the
installation, I ran the serial and parallel tests. The serial tests ran
successfully, however, the parallel tests ran with mixed results.
In my initial configuration I set the 8-byte integer flag and ran with 2
and 8 processors (where applicable). In general, these tests worked well
except for the REM series of tests. These tests tended to diverge
unacceptably from the supplied results.
In the next series, I set the 4-byte integer flag, again using 2 and 8
processors. In this case, the REM tests ran well, however, several other
tests either took too long to complete or crashed.
In all cases, bintraj gave some peculiar errors and
tgtmd/change_target.ntr crashed.
The full results are listed below:
test.sander.BASIC.MPI:
tgtmd/change_target.ntr; ./Run.tgtmd
P P P C
bintraj; ./Run.bintraj
? ? ? ?
ti_ggcc; ./Run.test1;
P P C P
ti_ggcc; ./Run.test2;
P P C P
ti_ggcc; ./Run.test3;
P P C P
ti_ggcc; ./Run.test4;
P P C P
test.sander.GB.MPI:
sodium; ./Run.sodium
P P C P
pheMTI; ./Run.0
P P C P
pheMTI; ./Run.1
P P C P
pheMTI; ./Run.lambda0
P P C P
pheMTI; ./Run.lambda1
P P C P
campTI; ./Run.campTI
P P C P
test.sander.LES.MPI:
LES_TEMP; ./Run.2temp
P P C P
LES_GB; ./Run.LES
P F P P
test.sander.QMMM.MPI:
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_min
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb1
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb1_aq1
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb1_aq2
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb1_mmcorr
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb1_qmewald2
P P ? ?
qmmm2/1NLN_periodic_lnk_atoms; ./Run.1NLN_MD_ntb2
P P ? ?
qmmm2/MG_QM_water_MM_AM1_periodic; ./Run.notimaged_md_pme
P P ? ?
qmmm2/MG_QM_water_MM_AM1_periodic; ./Run.notimaged_md_pme_qmewald
P P ? ?
qmmm2/MG_QM_water_MM_AM1_periodic; ./Run.notimaged_md_pme_qmewald_lowmem
P P ? ?
test.sander.PIMD.MPI:
qmmm2/pimd_pme; ./Run.pimd
? P ? ?
test.sander.REM:
rem_gb_2rep; ./Run.rem
F F P P
rem_gb_4rep; ./Run.rem
F P P P
rem_hybrid; ./Run.rem
P P F? P
NOTE: P-passed, F-failed, C-crashed, ?-odd results/took too long, F? -
large numerical deviations
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Jul 16 2006 - 06:07:09 PDT
