[gmx-developers] [Fwd: coulomb SR in 3.3cvs]

[Michel Cuendet]

Hi,

I subitted this message already on the gmx-users mailing list, but got no
feedback. I think it's an important issue for the CVS version 3.3. In the
meantime, I could reproduce the problem on xeon cpus, so it's not itanium specific.

Here is the original message :
--------------------------------------------------------------------------------

I compiled the 3.3 CVS version of 27.05.2005 on itanium2 with the intel
8.1 compiler. (for those who don't know, the cvs version brings a
terrific speedup on itaniums). Tests were performed with both versions
in double precision.

I compared the different energies obtained with 3.2.1 and 3.3, by taking
the same trr (from a 3.2.1 equilibration, 85000 atoms, PME) and running
it with both versions for about 12ps. From the tables below, one sees
that there is a gap between Coulomb (SR) of 3450 KJ/mol, and between
potential energies of  3280 KJ/mol.

A close inspection of the energy files reveals that the version 3.3
reproduces the same energy as 3.2.1 at the first step, but the system
relaxes very quickly (0.5ps) to a lower energy with 3.3.

Rerunning with 3.3 the trajectory produced by 3.2.1 surprisingly gives
the correct energies.

Any explaination for that ?

Thanks,
Michel


3.2.1 :

Energy                      Average       RMSD     Fluct.      Drift  Tot-Drift

-------------------------------------------------------------------------------

G96Bond                      6723.8    140.873    138.612   -6.69756   -87.0817

G96Angle                    10213.6    173.746    173.579    2.02771    26.3642

Proper Dih.                 4446.16    89.4129    84.2042    8.01199    104.172

Improper Dih.               3264.55     78.886    78.3318   -2.48685    -32.334

LJ-14                       852.608     79.621      79.52     1.0681    13.8874

Coulomb-14                  62903.9    126.041    122.079   -8.35397   -108.618

LJ (SR)                      166906    1099.22     1099.1   -4.19132   -54.4955

Coulomb (SR)             -1.2043e+06    1629.38    1629.34    -3.3074   -43.0028

Coulomb (LR)                -188686     91.952    90.6651    4.08445     53.106

Potential                -1.13768e+06    1031.78    1031.12   -9.84429   -127.995

3.3 CVS

Energy                      Average       RMSD     Fluct.      Drift  Tot-Drift

-------------------------------------------------------------------------------

G96Bond                     6486.57    122.559    120.492   -7.32269   -77.6351

G96Angle                    10003.9    152.062    152.054  -0.496268   -5.26144

Proper Dih.                 4391.87    83.2508    80.8574    6.47554    68.6537

Improper Dih.               3194.22    78.6896    78.6495   0.819815    8.69168

LJ-14                       833.553    76.0076    75.8776   -1.45206   -15.3947

Coulomb-14                    62899    124.539     123.82   -4.36711   -46.3001

LJ (SR)                      167720    1182.33    1157.55    78.6782    834.146

Coulomb (SR)             -1.20775e+06    1897.76    1850.85   -137.014   -1452.62

Coul. recip.                -188746     100.09    100.079    0.48938    5.18841

Potential                -1.14096e+06    1149.25    1132.33   -64.1897   -680.539


-- 
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Michel Cuendet, Ph.D. student
Laboratory of Computational Biochemistry and Chemistry
Swiss Federal Institute of Technology in Lausanne (EPFL)
CH-1015 Lausanne						
Switzerland                            Phone : +41 21 693 0324
lcbcpc21.epfl.ch/~mitch	                       +41  1 633 4194
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++



-- 
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Michel Cuendet, Ph.D. student
Laboratory of Computational Biochemistry and Chemistry
Swiss Federal Institute of Technology in Lausanne (EPFL)
CH-1015 Lausanne						
Switzerland                            Phone : +41 21 693 0324
lcbcpc21.epfl.ch/~mitch	                       +41  1 633 4194
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++