Fix written, but it looks like your test case isn't stable on a CPU:
wrapping first mol.: -106.78485 75.50831 130.78422
NSTEP = 419 TIME(PS) = 100.655 TEMP(K) = 310.24 PRESS =
38.3
Etot = -821073.3783 EKtot = 197974.6983 EPtot =
-1019048.0766
BOND = 7225.8198 ANGLE = 20282.5049 DIHED =
25288.9107
1-4 NB = 7993.3767 1-4 EEL = 64395.7655 VDWAALS =
113322.4023
EELEC = -1257556.8565 EHBOND = 0.0000 RESTRAINT =
0.0000
EKCMT = 85727.6418 VIRIAL = 83112.9747 VOLUME =
3163587.3994
Density =
1.0130
Ewald error estimate: 0.7254E-05
------------------------------------------------------------------------------
vlimit exceeded for step 419; vmax = 31.2712
Coordinate resetting cannot be accomplished,
deviation is too large
iter_cnt, my_bond_idx, i and j are : 5 17724 35719 35720
On the GPU with the XH4 (X 'cuz it's cool) SHAKE code, I got to step 1016
before I terminated, it looked like:
NSTEP = 1016 TIME(PS) = 101.588 TEMP(K) = 310.21 PRESS =
-50.5
Etot = -820346.7580 EKtot = 197959.3125 EPtot =
-1018306.0705
BOND = 7186.5762 ANGLE = 20059.4155 DIHED =
25244.7360
1-4 NB = 7950.3435 1-4 EEL = 64571.7679 VDWAALS =
112082.6139
EELEC = -1255401.5236 EHBOND = 0.0000 RESTRAINT =
0.0000
EKCMT = 85654.1222 VIRIAL = 89107.1805 VOLUME =
3168254.7911
Density =
1.0115
------------------------------------------------------------------------------
Since in both cases step 1 looks like:
NSTEP = 1 TIME(PS) = 100.002 TEMP(K) = 303.16 PRESS =
49.8
Etot = -831389.1144 EKtot = 193459.4531 EPtot =
-1024848.5675
BOND = 6650.7913 ANGLE = 18819.2775 DIHED =
25178.4869
1-4 NB = 7963.5507 1-4 EEL = 64631.2441 VDWAALS =
113969.4007
EELEC = -1262061.3186 EHBOND = 0.0000 RESTRAINT =
0.0000
EKCMT = 84204.6709 VIRIAL = 80808.8977 VOLUME =
3158971.8866
Density =
1.0144
------------------------------------------------------------------------------
It seems this system is pretty unstable in general...
What happens on a CPU at your end?
On Thu, Jan 12, 2012 at 7:53 AM, Ben Roberts <ben.roberts.geek.nz> wrote:
> Thanks Scott! I agree that NH4+ simulation is not a common problem, though
> it's not inherently stupid. I appreciate your help in tracking this issue
> down and fixing it :-)
>
> B.
>
> On 12/01/2012, at 10:47 AM, Scott Le Grand wrote:
>
> > Fix checked in... Off to adding NH4+ SHAKE support now (in 3 years no
> one
> > ever complained about this: Until now)...
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 12 2012 - 17:00:03 PST
