espressomd-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ESPResSo-users] crash using RATTLE with MPI


From: Axel Arnold
Subject: Re: [ESPResSo-users] crash using RATTLE with MPI
Date: Fri, 5 Aug 2011 17:40:36 +0200
User-agent: KMail/1.13.5 (Linux/2.6.34.8-0.2-desktop; KDE/4.4.4; x86_64; ; )

Hi,

as you can see from the error message, your constraints require particles to 
be shifted by more than a local processor length. That can happen easily if 
you constrain a whole polymer. One one CPU, ESPResSo doesn't complain, but it 
is still physically questionable. RATTLE or most other constraints algorithms 
are good at keeping together small molecules, not large molecules. In fact, 
most codes just don't allow you to constraint more than a couple of atoms per 
molecule.

Since I don't know any constraints algorithm that will effectively allow to 
constraint large molecules, the best will be if you use tight enough harmonic 
bonds (interaction energy of 30-100kT/R^2 or so should be ok with a typical LJ 
time step).

Axel

On Friday 05 August 2011 15:44:05 Muhammad Anwar wrote:
> Hello, I am running espresso-3.0.1 with lam/MPI.
> 
>  All tests pass both in serial and parallel, but when I try to run a
> calculation with restrained bonds on more than 1 cpu, it breaks with the
> output at the end of this email.
> 
> The important lines of the script are:
> 
> set bond_l [expr 1.53]
> inter 1 rigid_bond $bond_l 1e-2 1e-2
> 
> If I run on 1 CPU, output is OK:
> .
> .
> 1.530076435924329 Warmup finished. Minimal distance now 1.530076435924329
> Step 609/200^CCtrl-C caught...
> 
> 
> If I run with the rigid_bond replaced with a harmonic one, output is OK
> in parallel or serial:
> inter 1 harmonic 1 7
> .
> .
> .
> 2.738430991400251 Warmup finished. Minimal distance now 2.738430991400251
> Step 6061/20000000      6.06000 -0.0082 66.4397  6602.3605  65.5916
> -0.253^CCtrl-C caught...
> 
> Have I found a bug, or am I doing something wrong?
> Thanks
> Best Regards,
> Muhammad Anwar
> 
> ----------------------------------------
> 
> Interactions:
>    {1 RIGID_BOND 1.53 0.01 0.01} {2 angle 134.19148789844388
> 1.9163715186897738} {3 dihedral 1 -0.969533500066257 3.623170173257985}
> {0 0 lj-gen 0.10449043857692163 4.5 9.0 0.031 0.0 12  6  1.0 2.0 0.0 }
> {0 1 lj-gen 0.1627071790768632 4.5 9.0 0.031 0.0 12  6  1.0 2.0 0.0 } {1
> 1 lj-gen 0.25317460716839746 4.5 9.0 0.031 0.0 12  6  1.0 2.0 0.0 }
> Creating vmd_start.script...
> vmd_start.script finished.
> background_errors 0 {053 RATTLE failed to converge after 1000
> iterations} {004 some particles moved more than min_local_box_l, reduce
> the time step} 1 <consent> 2 <consent> 3 {053 RATTLE failed to converge
> after 1000 iterations}
>     while executing
> "integrate 1000"
>     ("while" body line 3)
>     invoked from within
> "while { $min < 1.50} {
>     inter ljforcecap $cap
>     integrate 1000
>     set min [analyze mindist]
>     incr cap 1000
> puts -nonewline " $min "

-- 
JP Dr. Axel Arnold
ICP, Universit├Ąt Stuttgart
Pfaffenwaldring 27
70569 Stuttgart, Germany
Email: address@hidden
Tel: +49 711 685 67609



reply via email to

[Prev in Thread] Current Thread [Next in Thread]