[gmx-users] PME in water

[Lianqing Zheng]

Thanks Erik! I recompiled FFTW with gcc. PME is working for water now. But
I still need to set fr->tabscale_exp=fr->tabscale to use Buckingham
potential. 

However, I got another problem: the simulation cell exploded at the first
time step:

   Energies (kJ/mol)
     BuckingHam   Coulomb (SR)   Coulomb (LR)      Potential    Kinetic
En.
    8.67478e+04   -4.14156e+07   -1.27296e+06   -4.26018e+07
8.90925e+07
   Total Energy    Temperature Pressure (bar)
    4.64907e+07    4.65377e+06   -1.09662e+06

Step 1  Warning: pressure scaling more than 1%, mu: 0.967101 0.967101 0.967101
Large VCM(group rest):    -10.41395,     -2.04940,    -33.29021, ekin-cm:
18777794.00000

I used Berendsen NPT ensemble and change tau_t, tau_p from 0.1 to 1.0, but
it didn't make much difference. EM did not help either. But it runs fine
with Cut-off Coulomb option.

Thank you very much!

Lianqing

On Mon, 5 May 2003, Erik Lindahl wrote:

>Does it work when you are not using PME?
>
>If yes, how did you compile FFTW? Try using gcc and a low optimization  
>level there too!
>
>Cheers,
>
>Erik
>
>On Monday, May 5, 2003, at 18:59 America/Los_Angeles, Lianqing Zheng  
>wrote:
>
>> Hi, Erik:
>>
>> Thanks for you suggestions! I change CC to gcc and re-compile Gromacs.
>> Something did change. Although I did get arithmetic exception this  
>> time, a
>> segmentation fault error occurred in low_constrain (). And the value  
>> of d2
>> was still strangely large sometimes even though no arithmetic exception
>> was displayed.
>>
>> For the SiO2 system, I didn't get arithmetic exception either, but
>> tabscale for EXP tern was still zero:
>> Generated table with 900 data points for Ewald.
>> Tabscale = 500 points/nm
>> Generated table with 900 data points for LJ6.
>> Tabscale = 500 points/nm
>> Generated table with 900 data points for EXPMIN.
>> Tabscale = 0 points/nm
>>
>> With coulombtype = Ewald, I got Coulomb (SR)=NaN. If I add a line  
>> before
>> the use of init_table():
>> fr->tabscale_exp=fr->tabscale;
>> It produced a number, which was different from the one from Moldy or my
>> code. For PME, it got a segmentation fault in do_inputrec ().
>>
>> Any insight? Thank you very much!
>>
>> Lianqing
>>
>> On Mon, 5 May 2003, Erik Lindahl wrote:
>>
>>> Hi,
>>>
>>> This could be a compiler and/or optimization problem. I would try to
>>> rule this out by:
>>>
>>> 1. Compiling without optimization. See the FAQ online for setting your
>>> own CFLAGS.
>>> 2. Use gcc instead of the Sun compiler.
>>>
>>> If either of these options make it work we'll have to search for a bug
>>> with the Sun compiler...
>>>
>>> Cheers,
>>>
>>> Erik
>>>
>>> On Monday, May 5, 2003, at 15:49 America/Los_Angeles, Lianqing Zheng
>>> wrote:
>>>
>>>> Dear Gromacs pals:
>>>>
>>>> I didn't succeed to use PME in the water system under the "tutor"
>>>> directory. The only modification I did in water.mdp was adding a PME
>>>> line:
>>>> coulombtype         = PME
>>>>
>>>> The error was arithmetic exception (again) occurring at solve_pme()  
>>>> in
>>>> pme.c. When kx=0, ky=0, kz=5, I got the following d1 and d2:
>>>> d1=0.238266,d2=-113907695932561209218061626031210496.000000
>>>>
>>>> Have you got the similar error? My machine is Sun Solaris2.8. Gromacs
>>>> version is 3.1.4.
>>>>
>>>> Thank you very much for your kind help!
>>>>
>>>> Lianqing
>>>>
>>>> _______________________________________________
>>>> gmx-users mailing list
>>>> gmx-users at gromacs.org
>>>> http://www.gromacs.org/mailman/listinfo/gmx-users
>>>> Please don't post (un)subscribe requests to the list. Use the
>>>> www interface or send it to gmx-users-request at gromacs.org.
>>>>
>>> ---------------------------------------------------------------------- 
>>> --
>>> -----
>>> Erik Lindahl, MSc, PhD     <lindahl at stanford.edu>
>>> D109, Fairchild Building
>>> Dept. Structural Biology, Stanford University School of Medicine
>>> Tel. 650-7250754    Fax. 650-7238464
>>>
>>> _______________________________________________
>>> gmx-users mailing list
>>> gmx-users at gromacs.org
>>> http://www.gromacs.org/mailman/listinfo/gmx-users
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-users-request at gromacs.org.
>>>
>>
>> _______________________________________________
>> gmx-users mailing list
>> gmx-users at gromacs.org
>> http://www.gromacs.org/mailman/listinfo/gmx-users
>> Please don't post (un)subscribe requests to the list. Use the
>> www interface or send it to gmx-users-request at gromacs.org.
>>
>------------------------------------------------------------------------ 
>-----
>Erik Lindahl, MSc, PhD     <lindahl at stanford.edu>
>D109, Fairchild Building
>Dept. Structural Biology, Stanford University School of Medicine
>Tel. 650-7250754    Fax. 650-7238464
>
>_______________________________________________
>gmx-users mailing list
>gmx-users at gromacs.org
>http://www.gromacs.org/mailman/listinfo/gmx-users
>Please don't post (un)subscribe requests to the list. Use the 
>www interface or send it to gmx-users-request at gromacs.org.
>