On 16/4/2010, at 5:51 p.m., Ben Roberts wrote:
>
> On 16/4/2010, at 11:28 a.m., Ben Roberts wrote:
>
>> Hi all,
>>
>> On my Mac (Snow Leopard 10.6.2, Intel compilers 11.1.084), I get this now:
>>
And finally, for the Amber parallel tests:
Sander.RISM.MPI is now generated, and the 3D-RISM tests pass.
With 2 CPUs:
385 file comparisons passed
2 file comparisons failed
0 tests experienced errors
possible FAILURE: check evbout.dif
/Applications/Amber11/2010-04-14/amber11/test/evb/malon_dbonds_umb_dg_UFF_3DG_qi_full_corrF
158c158
< 0.00396164 19.38711526 -515.49226027
> 0.00395165 19.38711410 -515.49223224
### Maximum absolute error in matching lines = 9.99e-06 at line 158 field 1
### Maximum relative error in matching lines = 2.53e-03 at line 158 field 1
Plus the customary TIP5P diff.
With 4 CPUs:
343 file comparisons passed
1 file comparison failed
0 tests experienced errors
where the failure is the TIP5P diff.
With 8 CPUs:
Finished parallel test suite for Amber 11 at Fri 16 Apr 2010 20:21:36 EDT.
Some tests require 4 threads to run, while some will not
run with more than 2. Please run further parallel tests with the
appropriate number of processors. See README for details.
315 file comparisons passed
3 file comparisons failed
18 tests experienced errors
The errors included:
cd PIMD/full_cmd_water/equilib && ./Run.full_cmd
cd PIMD/full_cmd_water/restart && ./Run.full_cmd
cd PIMD/full_nmpimd_water && ./Run.full_pimd
cd PIMD/full_pimd_amoeba && ./Run.full_pimd_amoeba
cd PIMD/full_pimd_ntp_water && ./Run.full_pimd_ntp
These ones behaved strangely - they started running in some kind of endless loop. For example, the first one ran for 45 minutes before I intervened and killed it from the command line. After that, I got considerably less patient, and would kill them after a minute or two.
cd qmmm2/xcrd_build_test/ && ./Run.ortho_qmewald0
cd qmmm2/xcrd_build_test/ && ./Run.truncoct_qmewald0
These two failed with this error:
* NB pairs <num> <number> exceeds capacity ( 185750) 3
SIZE OF NONBOND LIST = 185750
SANDER BOMB in subroutine nonbond_list
Non bond list overflow!
check MAXPR in locmem.f
(where <num> is a three-digit number and <number> is a six-digit number marginally bigger than 185,750)
The ncsu/smd2 test (which contributed two of the failures) failed with "Must have more residues than processors!". In this case, with seven residues present, I guess that means a maximum of six processors. I've prepared a patch for the run.sh script ($AMBERHOME/test/ncsu/smd2/run.sh), which I've attached.
The other failure is our old friend TIP5P.
Cheers,
Ben
--- run.sh.bak 2010-04-16 21:14:14.000000000 -0400
+++ run.sh 2010-04-16 21:16:40.000000000 -0400
.. -2,6 +2,16 ..
. ../common.sh
+#
+# Test to make sure there are at most six processors.
+#
+numprocs=`${DO_PARALLEL} echo "Testing number of processors" | wc -l`
+if [ "$numprocs" -gt 6 ] ; then
+ echo "This test requires six or fewer processors. Skipping."
+ exit
+fi
+
+
set_SANDER sander.MPI
set_SANDER_CMD 1
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Apr 16 2010 - 18:30:03 PDT
