[gmx-users] 3 questions of gromacs 4.5.2 for both CPU and GPU

[Rossen Apostolov]

Hi,

On 11/24/10 2:59 AM, lin hen wrote:
> hi,
>
> I am testing the dhfr benchmarks with gromacs 4.5.2
>
> 1. for the benchmarks: dhfr-solv-RF-1nm.bench and 
> dhfr-solv-RF-2nm.bench, set the steps = 10000,
>     The GPU version running results shows as following:
>    ./mdrun-gpu
>
>  Pre-simulation ~15s memtest in progress...done, no errors detected
> starting mdrun 'Protein in water'
> 10000 steps,     20.0 ps.
> Writing final coordinates.
> Back Off! I just backed up confout.gro to ./#confout.gro.2#
> Post-simulation ~15s memtest in progress...done, no errors detected
>         OpenMM run - timing based on wallclock.
> *  NODE (s)   Real (s)      (%)
>        Time:     90.796     90.796    100.0
>                        1:30
>                (Mnbf/s)   (MFlops)   (ns/day)  (hour/ns)
> Performance:      0.000      0.060     19.034      1.261*
> gcq#100: "Jesus Can't Save You, Though It's Nice to Think He Tried" 
> (Black Crowes)
>
>      The CPU version running results shows as following:
>     ./mdrun
>
> Writing final coordinates.
> Back Off! I just backed up confout.gro to ./#confout.gro.3#
>  Average load imbalance: 4.7 %
>  Part of the total run time spent waiting due to load imbalance: 2.7 %
>  Steps where the load balancing was limited by -rdd, -rcon and/or 
> -dds: X 0 % Y 0 % Z 0 %
>
>         Parallel run - timing based on wallclock.
> *      NODE (s)   Real (s)      (%)
>        Time:     86.176     86.176    100.0
>                        1:26
>                (Mnbf/s)   (GFlops)   (ns/day)  (hour/ns)
> Performance:    741.634     35.849     20.054      1.197*
> gcq#0: Thanx for Using GROMACS - Have a Nice Day
>
>
> The gpu version looks *slower *than cpu version, does it make sense? 
> Is there anyway I could optimize?
>

In the default build, the CPU version will use all available cores on 
the machine and it's faster.

>
> 2.  For the CPU benchmark dhfr-impl-1nm.bench, when I run :
>      ./mdrun
>      Back Off! I just backed up md.log to ./#md.log.9#
> Reading file topol.tpr, VERSION 4.5.1-dev-20100917-b1d66 (single 
> precision)
> Starting 16 threads
> -------------------------------------------------------
> Program mdrun, VERSION 4.5.2
> Source code file: domdec.c, line: 6428
> Fatal error:
> There is no domain decomposition for 16 nodes that is compatible with 
> the given box and a minimum cell size of 1.02425 nm
> Change the number of nodes or mdrun option -rcon or -dds or your LINCS 
> settings
> Look in the log file for details on the domain decomposition
> For more information and tips for troubleshooting, please check the 
> GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> -------------------------------------------------------
> Thanx for Using GROMACS - Have a Nice Day
>
>
> Where could I modify this nodes#?

Give the option -nt to mdrun to specify the number of threads you want 
to launch, e.g.

$ mdrun -nt 8

>
> 3.  for CPU benchmark dhfr-impl-inf.bench
>     I modified cpu-imp-RF-inf.mdp:  nsteps  10000
>     grompp -f cpu-imp-RF-inf.mdp -o topol.tpr
>     it will show:
>     NOTE 1 [file cpu-imp-RF-inf.mdp]:
>   The Berendsen thermostat does not generate the correct kinetic energy
>   distribution. You might want to consider using the V-rescale thermostat.
> Generated 2278 of the 2278 non-bonded parameter combinations
> Generating 1-4 interactions: fudge = 0.5
> Generated 2278 of the 2278 1-4 parameter combinations
> Excluding 3 bonded neighbours molecule type 'Protein'
> turning all bonds into constraints...
> NOTE 2 [file topol.top, line 23568]:
>   System has non-zero total charge: -1.100000e+01
>
> GB parameter(s) missing or negative for atom type 'H0'
> -------------------------------------------------------
> Program grompp, VERSION 4.5.1
> Source code file: grompp.c, line: 870
> Fatal error:
> Can't do GB electrostatics; the forcefield is missing 1 values for
> atomtype radii, or they might be negative
> .
> For more information and tips for troubleshooting, please check the 
> GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> -------------------------------------------------------
> "Everybody's Good Enough For Some Change" (LIVE)
>
> and when I run ./mdrun
> the steps still -1, infinite time
> How could I genete the right topol.tpr?
>
Did you search the mailing lists?

Copy and paste the error message in 
http://www.gromacs.org/Support/Mailing_Lists/Search
and you will get a top hit
http://lists.gromacs.org/pipermail/gmx-users/2010-September/053764.html

You are using grompp 4.5.1 and mdrun 4.5.2 !? Remove all gromacs 
installs that you have, make a clean 4.5.3 install and try again.

Rossen

>
> Thanks a lot.
>
>
> YY

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20101124/5b79f3a2/attachment.html>