differences in multiprocessor energy minimization results

From: Markus Metz (metma99_at_gmail.com)
Date: Wed Aug 17 2011 - 11:44:19 CDT

Next message: Chathurika Abeyrathne: "Re: vmd-l: Re: MMTools"
Previous message: Axel Kohlmeyer: "Re: MMTools"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

Dear all:

I have run a slightly modified calculation of the waterbox tutorial (only
minimization for 1000 steps) on one and 8 processors several times and
noticed that only in the former case the results are reproducible.
The results on the 8 processor job are always different. Please see below.
Has anybody else observed this?
Is there a way to avoid this from happening?
Any input is highly appreciated,
Markus

1-3-box: modified: only 1000 min steps;

-------------------------------------------------------------------------------------------------------------------------------
/NAMD_2.8b1_Linux-x86_64/charmrun /NAMD_2.8b1_Linux-x86_64/namd2 ++local +p8
./ubq_wb_eq.conf

run1:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG

ENERGY: 1000 902.6995 768.6109 221.8336
6.9052 -31264.3440 3102.3496 0.0000
0.0000 0.0000 -26261.9451 0.0000 -26261.9451
-26261.9451 0.0000 -4018.4992 -4067.9469
86856.0000 -4018.4992 -4067.9469

run2:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG

ENERGY: 1000 911.5202 759.3571 221.8352
7.0819 -31266.6410 3099.9282 0.0000
0.0000 0.0000 -26266.9183 0.0000 -26266.9183
-26266.9183 0.0000 -4155.1534 -4093.7960
86856.0000 -4155.1534 -4093.7960

run3:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG

ENERGY: 1000 913.0340 770.5699 221.8180
7.1120 -31294.0397 3121.7942 0.0000
0.0000 0.0000 -26259.7116 0.0000 -26259.7116
-26259.7116 0.0000 -4022.4225 -3996.8125
86856.0000 -4022.4225 -3996.8125

run4:
TIMING: 1000 CPU: 27.5208, 0.0274798/step Wall: 33.7872, 0.0336357/step, 0
hours remaining, 11.281601 MB of memory in use.
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG

ENERGY: 1000 901.6344 766.4679 221.8613
6.9223 -31261.1588 3101.8025 0.0000
0.0000 0.0000 -26262.4703 0.0000 -26262.4703
-26262.4703 0.0000 -4008.5786 -4070.2326
86856.0000 -4008.5786 -4070.2326

run5:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG

ENERGY: 1000 907.0763 761.6902 221.8116
6.9968 -31264.2395 3100.4343 0.0000
0.0000 0.0000 -26266.2302 0.0000 -26266.2302
-26266.2302 0.0000 -4095.4498 -4084.1115
86856.0000 -4095.4498 -4084.1115
-------------------------------------------------------------------------------------------------------------------------------
/NAMD_2.8b1_Linux-x86_64/charmrun /NAMD_2.8b1_Linux-x86_64/namd2 ++local +p1
./ubq_wb_eq.conf

run1:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG PRESSURE
GPRESSURE VOLUME PRESSAVG GPRESSAVG