Re: [AMBER] Problem running test.pmemd.AMOEBA

From: Robert Duke <rduke.email.unc.edu>
Date: Fri, 13 Mar 2009 14:33:55 +0000

Hi Marta,
It looks to me like you are probably okay. But everyone that uses
pmemd.amba please note all the following information:

1) Read the README that ships with pmemd.amba. This code is basically pmemd
9 with amoeba support added. In order to actually get amoeba support, you
must hack your config.h, adding -DAMOEBA to the F90_DEFINES = line.
2) All the *.save files in the 5 amoeba tests were produced using sander.
Unfortunately, there were some minor output corrections in pmemd.amba that
never got backported into sander. So sander deviates from pmemd.amba in
these 2 ways: a) there is some nonsense in the prmtop and mdin input that
is never corrected; this mostly deals with the fact that standard amber ff
bonded interaction info is not used for the amoeba ff, but sections for it
are still present in the prmtop (evaluation is turned off by forcing ntf to
1, which is why there is a discrepancy in that value), and b) In sander,
ensembles (NVT, NVE) that don't use pressure-related variables nonetheless
still report them; this is inconsistent with other amber operation, and
misleading (and pmemd.amba doesn't do it).

So if you look at the energies produced by sander vs. pmemd.amba for the 5
amoeba tests, they match. I have attached correct outputs for the five
amoeba tests from pmemd.

I apologize for this mess, but it originates from amoeba being a bit of an
orphan. Darden sort of dropped working on it. I did the pmemd port, saw
how hopelessly slow things were (and ascertained that there was no
reasonable way to get them faster), and decided I could not afford to put
more effort into amoeba within the current constraints (mind you, I think
that pmemd.amba is by far the fastest amoeba implementation out there,
especially the parallel version, but it is still very slow compared to one
of the other amber ff's, and there are all sorts of issues that insure it is
not going to get much faster). I think Scott would typically update the
test stuff, but we did not manage to close the loop as we were probably both
mired in other stuff. So the test situation is confusing, but the code
basically works. Hopefully, folks can at least compare to the files I am
releasing here if they want to use amoeba.

Best Regards - Bob Duke

----- Original Message -----
From: <mloureiro.cesga.es>
To: <amber.ambermd.org>
Sent: Friday, March 13, 2009 7:45 AM
Subject: [AMBER] Problem running test.pmemd.AMOEBA


> Hello,
>
> I have installed AMBER10 and I have run the tests included.
>
> I have a bit problem with tests that use AMOEBA.
>
> -test.sander.AMOEBA --> all tests passed
> -test.pmemd.AMOEBA --> all tests possible FAILURE
> and if you check the file .out.dif you can see the differences are slight.
>
> I have copied the beginnig of the file amoeba_wat1.out.dif to show the
> problem.
> Are right the results of test.pmemd.AMOEBA? Or maybe there is a problem
> with
> the compilation stage? Should I worry about that?
>
> 35,36c35,36
> < NATOM = 648 NTYPES = 1 NBONH = 1 MBONA = 1
> < NTHETH = 1 MTHETA = 1 NPHIH = 1 MPHIA = 1
> ---
>> NATOM = 648 NTYPES = 1 NBONH = 0 MBONA = 0
>> NTHETH = 0 MTHETA = 0 NPHIH = 0 MPHIA = 0
> 38,39c38,39
> < NBONA = 1 NTHETA = 1 NPHIA = 1 NUMBND = 1
> < NUMANG = 1 NPTRA = 1 NATYP = 1 NPHB = 1
> ---
>> NBONA = 0 NTHETA = 0 NPHIA = 0 NUMBND = 0
>> NUMANG = 0 NPTRA = 0 NATYP = 1 NPHB = 0
> 56c56
> < ntf = 8, ntb = 1, igb = 0, nsnb =
> 25
> ---
>> ntf = 1, ntb = 1, igb = 0, nsnb =
> 25
> -------------------------------------------------------------------------------->
> 3. ATOMIC COORDINATES AND VELOCITIES
>>
> -------------------------------------------------------------------------------->
> water box
>> begin time read from input coords = 0. ps
> 88,89c87
> < num_pairs_in_ee_cut,size_dipole_dipole_list = 39766 49707
> < NSTEP = 1 TIME(PS) = 0.001 TEMP(K) = 287.20 PRESS =
> -1471.6
> ---
>> NSTEP = 1 TIME(PS) = 0.001 TEMP(K) = 287.20 PRESS =
>> 0.94d91
> < EKCMT = 0. VIRIAL = 0. VOLUME = 6479.5881
> 97d93
> < Density =
> 0.9970
> 99c95
> < NSTEP = 2 TIME(PS) = 0.002 TEMP(K) = 291.07 PRESS =
> -1466.8
> ---
>> NSTEP = 2 TIME(PS) = 0.002 TEMP(K) = 291.07 PRESS =
>> 0.104d99
> < EKCMT = 0. VIRIAL = 0. VOLUME = 6479.5881
> 107d101
> < Density =
> 0.9970
> 109c103
> < NSTEP = 3 TIME(PS) = 0.003 TEMP(K) = 295.74 PRESS =
> -128.2
> ---
>> NSTEP = 3 TIME(PS) = 0.003 TEMP(K) = 295.74 PRESS =
>> 0.114d107
> < EKCMT = 0. VIRIAL = 0. VOLUME = 6479.5881
> 117d109
> < Density =
> 0.9970
> 119c111
> < NSTEP = 4 TIME(PS) = 0.004 TEMP(K) = 294.16 PRESS =
> 2130.7
> ---
>> NSTEP = 4 TIME(PS) = 0.004 TEMP(K) = 294.16 PRESS =
>> 0.124d115
> < EKCMT = 0. VIRIAL = 0. VOLUME = 6479.5881
> 127d117
> < Density =
> 0.9970
>
> Thanks,
> Marta
>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



Received on Sun Mar 15 2009 - 01:09:30 PDT
Custom Search