AMBER Archive (2009)

Subject: Re: [AMBER] logdvdl does not work

From: steinbrt_at_rci.rutgers.edu
Date: Tue Sep 08 2009 - 12:37:31 CDT

Hi,

did you have softcore potentials turned on (ifsc=1)? logdvdl was added to
the softcore module and is not initialized without it, therefore the
segfault.

I agree (and apologize) that this is not intuitive and clearly documented.
I will change the behaviour soon, definitely for the release of Amber11
next spring. There may be a bugfix posted to the list before that.

Kind Regards,

Thomas

On Tue, September 8, 2009 12:53 pm, helde010_at_umn.edu wrote:
> Hello AMBER listserv,
>
> I am doing thermodynamic integration calculations and I would like to use
> the logdvdl option described in the AMBER 10 manual on page 106. However,
> whenever I set it equal to 1 to try to turn it on, AMBER segfaults.
>
> For this listserv post I tried to make a reproducible error, so I used the
> input files from Tutorial A6. I used the input files for the implicit
> solvent calculation part of the tutorial, taking the example for
> model_step2.mdin and changing ntx to 1 and irest to 0. In the groupfile
> model_step2.group I changed the starting coordinate file to the one
> provided in the first section of the tutorial.
>
> The calculation runs fine with only those changes. But when I add the line
> logdvdl=1 to the input file I get a segfault. Here is the output:
>
> Running multisander version of sander amber10
> Total processors = 2
> Number of groups = 2
>
> [blade286:01527] *** Process received signal *** [blade286:01528] ***
> Process received signal *** [blade286:01528] Signal: Segmentation fault
> (11) [blade286:01528] Signal code: Address not mapped (1) [blade286:01528]
> Failing at address: 0x8 [blade286:01527] Signal: Segmentation fault (11)
> [blade286:01527] Signal code: Address not mapped (1) [blade286:01527]
> Failing at address: 0x8 [blade286:01527] [ 0] /lib64/tls/libpthread.so.0
> [0x2a961d2720] [blade286:01527] [ 1]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(softcore_mp_log_dvdl_+0x38)
> [0x691580] [blade286:01527] [ 2]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(mix_frcti_+0x71d) [0x5fc41d]
> [blade286:01527] [ 3] /usr/local/amber/amber10-OMPI/exe/sander.MPI
> [0x53832b] [blade286:01527] [ 4]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(sander_+0x7187) [0x4ea56f]
> [blade286:01527] [ 5]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(MAIN__+0x1382) [0x4e3352]
> [blade286:01527] [ 6]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(main+0x32) [0x43f5fa]
> [blade286:01528] [ 0] /lib64/tls/libpthread.so.0 [0x2a961d2720]
> [blade286:01528] [ 1]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(softcore_mp_log_dvdl_+0x38)
> [0x691580] [blade286:01528] [ 2]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(mix_frcti_+0x71d) [0x5fc41d]
> [blade286:01528] [ 3] /usr/local/amber/amber10-OMPI/exe/sander.MPI
> [0x53835c] [blade286:01528] [ 4]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(sander_+0x7187) [0x4ea56f]
> [blade286:01528] [ 5]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(MAIN__+0x1382) [0x4e3352]
> [blade286:01528] [ 6]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(main+0x32) [0x43f5fa]
> [blade286:01528] [ 7] /lib64/tls/libc.so.6(__libc_start_main+0xea)
> [0x2a962f8aaa] [blade286:01528] [ 8]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(free+0x4a) [0x43f52a]
> [blade286:01528] *** End of error message *** [blade286:01527] [ 7]
> /lib64/tls/libc.so.6(__libc_start_main+0xea) [0x2a962f8aaa]
> [blade286:01527] [ 8]
> /usr/local/amber/amber10-OMPI/exe/sander.MPI(free+0x4a) [0x43f52a]
> [blade286:01527] *** End of error message *** mpirun noticed that job rank
> 0 with PID 1527 on node blade286 exited on signal 11 (Segmentation fault).
> 1 additional process aborted (not shown)
>
> [1]+ Exit 139 mpirun -np 2 $AMBERHOME/exe/sander.MPI -ng 2 -groupfile
> model_step1.group
>
> I did not compile this executable, it was compiled by a sysadmin of the
> system I am running this on. It is AMBER 10 and I know they compiled with
> open MPI. Is there a way I can make this work? For the sake of statistics
> it would be nice to have all the dvdl values, without having to print a
> huge output file.
>
> -Hugh Heldenbrand
> Graduate Student, University of Minnesota
>
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>

Dr. Thomas Steinbrecher
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber