AMBER Archive (2006)

Subject: Re: AMBER: RE: PMEMD compilation problems

From: Robert Duke (rduke_at_email.unc.edu)
Date: Fri May 19 2006 - 11:20:52 CDT


Simon -
The "global name too long" message is not a real problem - the compiler just
truncates the repetitive end of really long names generated by fusion of the
module and local name. The problem, most likely, is that your mpi libraries
were not compiled with the fortran interfaces being built by ifort, or
perhaps not built at all. Did you build mpich yourself, specifying ifort as
the fortran compiler? If you are using the mpi libraries that perhaps came
installed on the machine with lam, they won't work; there is an assumption
in that stuff that you are using gnu compilers and gnu linkage conventions.
You could be getting other libraries (like lam libraries) if you are not
providing the correct path to the mpi libs (I expect this is not what you
are doing, but without the machine sitting in front of me, I try to think of
the possibilities). There are various notes about this sort of stuff on
both the amber website (some contributed by me) and in the pmemd README.
Once you get it working, there are also various notes (by me) on the website
about all the hoops you need to jump through to get a gigabit ethernet
cluster working half way decently.
- Bob
----- Original Message -----
From: "Simon Whitehead" <SJW084_at_bham.ac.uk>
To: <amber_at_scripps.edu>
Sent: Friday, May 19, 2006 11:37 AM
Subject: AMBER: RE: PMEMD compilation problems

> Bob,
>>
>> i am trying to compile pmemd in amber 8 on a p4 linux cluster with mpich
>> =
>> 1.2.6. I have tried the tarballed files but even though it gets further =
>> it ends with the following error;
>>
>> parallel_dat.o(.text+0x14c): In function =
>> `parallel_dat_module_mp_cit_mpi_allgathervec_':
>> : undefined reference to `mpi_allgatherv_'
>> parallel_dat.o(.text+0x3e4): In function =
>> `parallel_dat_module_mp_wrapped_cit_mpi_allgathervec_':
>> : undefined reference to `mpi_allgatherv_'
>> parallel_dat.o(.text+0x666): In function =
>> `parallel_dat_module_mp_cit_mpi_gathervec_':
>> : undefined reference to `mpi_gatherv_'
>> parallel_dat.o(.text+0x90f): In function =
>> `parallel_dat_module_mp_wrapped_cit_mpi_gathervec_':
>> : undefined reference to `mpi_gatherv_'
>> mdin_ctrl_dat.o(.text+0x58a1): In function =
>> `mdin_ctrl_dat_module_mp_bcast_mdin_ctrl_dat_':
>> : undefined reference to `mpi_bcast_`
>
> etc etc followd by;
>
>> make[1]: *** [pmemd] Error 1
>> make[1]: Leaving directory =
>> `/home/filer2/bcm/sjw084/amber8_MPI/amber8/src/pmemd/src'
>> make: *** [install] Error 2
>>
>> I also noticed a warning appearing several times during compilation, i
>> have listed an example of the error below;
>>
>> fortcom: Warning: _alltasks_setup_.f90, line 1: Global name too long, =
>> shortened from: =
>> alltasks_setup_module_mp_parallel_environment_startup_$BLK.parallel_dat_m=
>> odule_mp_mpi_status_ignore_ to: =
>> asks_setup_module_mp_parallel_environment_startup_$BLK.parallel_dat_modul=
>> e_mp_mpi_status_ignore_
>>
>> Could this be something to do with it?
>>
>> Regards
>>
>> simon
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu