That's what we have seen on at least one comparison benchmark (from intel, I
believe, comparing their mpi implementation to several others on a large
cluster). I have never supported openmpi for pmemd because 1) when it first
came out, the release date was very close to an amber release and the
available documentation on openmpi was not very thorough, 2) later
indications were it is relatively slow. This may not remain the case
forever, but it has been true in the last couple of releases that there were
really no benefits associated with moving from the mpi implementations we do
support to openmpi.
Regards - Bob Duke
----- Original Message -----
From: "Giorgos Lamprinidis" <lamprinidis.pharm.uoa.gr>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Tuesday, July 14, 2009 3:29 AM
Subject: Re: [AMBER] Need help for PMEMD 10 installation
> Ross,
>
> You mean that using mvapich i get faster calculations compared to openmpi,
> when runing pmemd or sander.MPI ?
>
>
> Regards
>
> George Lamprinidis PhD
> School of Pharmacy
> University of Athens
> Greece
>
>
>
>
> ----- Original Message -----
> From: "Ross Walker" <ross.rosswalker.co.uk>
> To: <jun.strubi.ox.ac.uk>; "'AMBER Mailing List'" <amber.ambermd.org>
> Sent: Monday, July 13, 2009 10:18 PM
> Subject: RE: [AMBER] Need help for PMEMD 10 installation
>
>
>> Hi Jun,
>>
>> I don't think there was ever an openmpi configure script for PMEMD,
>> largely
>> because the openMPI performance is pretty awful compared infiniband tuned
>> libraries such as MVAPICH. Nevertheless it should be fairly easy to get
>> to
>> work. Just open your config.h file and make the edits I suggested in the
>> earlier mail.
>>
>> 1) Change all ifort to mpif90, all icc to mpcc
>>
>> 2) Remove the references to the mpi libraries since these will be handled
>> by
>> mpif90.
>>
>> 3) make clean; make
>>
>> This should like okay this time.
>>
>> All the best
>> Ross
>>
>>> -----Original Message-----
>>> From: amber-bounces.ambermd.org [mailto:amber-bounces.ambermd.org] On
>>> Behalf Of Jun Dong
>>> Sent: Monday, July 13, 2009 11:11 AM
>>> To: AMBER Mailing List
>>> Subject: RE: [AMBER] Need help for PMEMD 10 installation
>>>
>>> Dear Ross,
>>> I recompile openmpi-1.3.2 using ifort. Then use its mpif90 and mpicc to
>>> compile pmemd,
>>> ./configure linux_em64t ifort mpich
>>> make
>>> /lib/cpp -traditional -P -I/raid/1/software/AMBER/openmpiifort/include
>>> -DPUBFFT -DMPI -DSLOW_NONBLOCKING_MPI -DDIRFRC_EFS -DDIRFRC_COMTRANS -
>>> DDIRFRC_NOVEC -DMKL -DFFTLOADBAL_2PROC get_cmdline.fpp get_cmdline.f90
>>> mpif90 -c -auto -tpp7 -xP -ip -O3 get_cmdline.f90
>>> ifort: command line remark #10148: option '-tp' not supported
>>> /lib/cpp -traditional -P -I/raid/1/software/AMBER/openmpiifort/include
>>> -DPUBFFT -DMPI -DSLOW_NONBLOCKING_MPI -DDIRFRC_EFS -DDIRFRC_COMTRANS -
>>> DDIRFRC_NOVEC -DMKL -DFFTLOADBAL_2PROC master_setup.fpp
>>> master_setup.f90
>>> mpif90 -c -auto -tpp7 -xP -ip -O3 master_setup.f90
>>> ifort: command line remark #10148: option '-tp' not supported
>>> /lib/cpp -traditional -P -I/raid/1/software/AMBER/openmpiifort/include
>>> -DPUBFFT -DMPI -DSLOW_NONBLOCKING_MPI -DDIRFRC_EFS -DDIRFRC_COMTRANS -
>>> DDIRFRC_NOVEC -DMKL -DFFTLOADBAL_2PROC pmemd.fpp pmemd.f90
>>> mpif90 -c -auto -tpp7 -xP -ip -O3 pmemd.f90
>>> ifort: command line remark #10148: option '-tp' not supported
>>> /lib/cpp -traditional -P -I/raid/1/software/AMBER/openmpiifort/include
>>> -DPUBFFT -DMPI -DSLOW_NONBLOCKING_MPI -DDIRFRC_EFS -DDIRFRC_COMTRANS -
>>> DDIRFRC_NOVEC -DMKL -DFFTLOADBAL_2PROC erfcfun.fpp erfcfun.f90
>>> mpif90 -c -auto -tpp7 -xP -ip -O3 erfcfun.f90
>>> ifort: command line remark #10148: option '-tp' not supported
>>> mpif90 -o pmemd gbl_constants.o gbl_datatypes.o state_info.o
>>> file_io_dat.o mdin_ctrl_dat.o mdin_ewald_dat.o prmtop_dat.o
>>> inpcrd_dat.o dynamics_dat.o img.o parallel_dat.o parallel.o
>>> gb_parallel.o pme_direct.o pme_recip_dat.o pme_slab_recip.o
>>> pme_blk_recip.o pme_slab_fft.o pme_blk_fft.o pme_fft_dat.o fft1d.o
>>> bspline.o pme_force.o pbc.o nb_pairlist.o nb_exclusions.o cit.o
>>> dynamics.o bonds.o angles.o dihedrals.o extra_pnts_nb14.o runmd.o
>>> loadbal.o shake.o prfs.o mol_list.o runmin.o constraints.o
>>> axis_optimize.o gb_ene.o veclib.o gb_force.o timers.o pmemd_lib.o
>>> runfiles.o file_io.o bintraj.o pmemd_clib.o pmemd.o random.o degcnt.o
>>> erfcfun.o nmr_calls.o nmr_lib.o get_cmdline.o master_setup.o
>>> pme_alltasks_setup.o pme_setup.o ene_frc_splines.o gb_alltasks_setup.o
>>> nextprmtop_section.o
>>> /raid/1/software/linux/compiler/mkl/9.1.023/lib/em64t/libmkl_em64t.a -
>>> L/raid/1/software/linux/compiler/mkl/9.1.023/lib/em64t -lguide -
>>> lpthread -L/raid/1/software/AMBER/openmpiifort/lib -lmpich -limf -lsvml
>>> -Wl,-
>>> rpath=/raid/1/software/linux/compiler/mkl/9.1.023/lib/em64t:/raid/1/sof
>>> tware/AMBER/openmpiifort/lib:/raid/1/software/linux/compiler/lfc64/lib
>>> ld: cannot find -lmpich
>>> make[1]: *** [pmemd] Error 1
>>> make[1]: Leaving directory
>>> `/raid/1/software/AMBER/amber10/src/pmemd/src'
>>> make: *** [all] Error 2
>>>
>>> I just wonder if there is a precompiled pmemd for linux 64 bit with
>>> openmpi that can be downloaded. Since it is not so easy to compile.
>>> Many thanks,
>>> Jun
>>>
>>>
>>> _______________________________________________
>>> 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
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jul 14 2009 - 10:08:48 PDT
