Hi Lijiang,
I have made bugfix.34 to fix this. It basically corrects for the fact that bugfix.31 was generated against code that was not up to date. Hopefully this will address the issue. Take the attached bugfix. Save it in AMBERHOME/src/pmemd/src/ then do
patch -p0 <bugfix.34
cd ../
make
And see if that fixes the problem.
A lesson for everybody here "Version control systems exist for a reason! - Use them."
Thanks for your debugging.
All the best
Ross
> -----Original Message-----
> From: amber-bounces.ambermd.org [mailto:amber-bounces.ambermd.org] On
> Behalf Of Robert Duke
> Sent: Thursday, April 22, 2010 7:01 AM
> To: AMBER Mailing List
> Subject: Re: [AMBER] A possible pmemd bug
>
> HI Lijiang,
> I am guessing that you got burned by some munging of the bugfix file
> 31, and
> your fix is correct. The problem does not exist in my original fixed
> source, but looking again at bugfix 31, next_mult_fac is declared but
> not
> defined in the common block. SO everyone, for now, if you use extra
> points,
> you may want to hand patch prmtop_dat.fpp to include next_mult_fac in
> the
> prmtop_int common block. My apologies to all on this. I did the
> bugfix
> independent of work going on for adding charmm to pmemd in another
> group; I
> did not have the charmm code (if available, not updated); the original
> fix
> was correct. A generous third party did a hand merge of my source
> after I
> submitted working code for the extra points fix, but it clearly was a
> complicated-enough blending to get things screwed up. Ross, I don't
> know if
> Dave is still stuck in Germany by the volcano, but if you have time to
> fix
> the patchfile, that would be great. I don't have access to the web
> files
> and am traveling today myself in any case (hopefully not toward any
> soon-to-be active volcanoes). Lijiang, thank much for the report and
> correct analysis.
> Best Regards - Bob Duke
> ----- Original Message -----
> From: "Lijiang Yang" <lyang.mail.chem.tamu.edu>
> To: <amber.ambermd.org>
> Sent: Thursday, April 22, 2010 7:59 AM
> Subject: [AMBER] A possible pmemd bug
>
>
> > Hi Dr. Duke,
> > After I applied bugfix 31, the parallel version pmemd always
> crashes
> > and the error message is like: MPI_BCAST: Message truncated ... .
> Through
> > debugging, I found that the variable 'next_mult_fac' which is
> introduced
> > in bugfix 31 is not declared in the 'prmtop_int' common block.
> Therefore,
> > it always equals to 0 and thus induces the message buffer error at
> line
> > #1405 'call mpi_bcast(gbl_natex, next * next_mult_fac, mpi_integer,
> 0,
> > mpi_comm_world, err_code_mpi)'. Adding the variable 'next_mult_fac'
> to the
> > common block solved the problem. I don't know whether it happens to
> anyone
> > else.
> > BTW, my system is CentOS 5.4 x86_64, kernel 2.6.18-164.15.1.el5,
> the
> > compiler is intel compiler 11.1.056 and the parallel library is mpich
> > 1.2.7p1. The configure command I used: ./configure linux_em64t ifort
> mpich
> > pubfft bintraj.
> >
> > Best,
> >
> > Lijiang Yang
> > Department of Chemistry
> > Peking University
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> >
>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
- application/octet-stream attachment: bugfix.34
Received on Thu Apr 22 2010 - 17:00:02 PDT
