Re: [AMBER] Parallel installation of AMBER 11 on MacOSX

From: George Tzotzos <gtzotzos.me.com>
Date: Wed, 02 Jun 2010 19:48:26 +0200

Hi Jason,

Indeed, I didn't see this part of your message. Apologies

I'll try. Hope this works.

Many thanks

George

On Jun 2, 2010, at 7:28 PM, Jason Swails wrote:

> Hello,
>
> On Wed, Jun 2, 2010 at 1:14 PM, George Tzotzos <gtzotzos.me.com> wrote:
>
>> Hi everybody and apologies for coming back on this one.
>>
>> I thought that the problem may had been caused by previous installations of
>> openmpi.
>>
>> So I erased by hard disc (!!!) and started from scratch
>>
>> The serial installation results in:
>>
>> Finished serial test suite for Amber 11 at Wed Jun 2 18:12:03 CEST 2010.
>>
>> 407 file comparisons passed
>> 0 file comparisons failed
>> 0 tests experienced errors
>> Test log file saved as logs/test_amber_serial/2010-06-02_17-42-09.log
>> No test diffs to save!
>>
>>
>> I installed MPICH2 in /opt as per David Watson's suggestion
>>
>> george1$ ls /opt/mpich2/
>> bin etc include lib sbin share
>>
>> For the installation, I used Jason Swails' instructions
>>
>> ./configure FC=gfortran F77=gfortran F90=gfortran CC=gcc CXX=g++
>> --prefix=/usr
>>
>
> The --prefix=/usr creates a makefile that will automatically put the
> binaries/libraries/includes in /usr (/usr/bin, /usr/lib, and /usr/include)
> when you run make install. If you want to install it in /opt/mpich2, you
> should change this to --prefix=/opt/mpich2.
>
>
>> make
>> sudo make install
>>
>> In /amber11/src, I used make clean
>>
>> make parallel give now the following error
>>
>> cp ../../AmberTools/src/netcdf/include/*.mod .
>> cp ../../AmberTools/src/netcdf/bin/ncdump
>> /Users/george1/Programs/amber11/bin
>> cpp -traditional -P -DBINTRAJ -DMPI constants.f > _constants.f
>> mpif90 -c -O3 -mtune=generic -ffree-form -o constants.o _constants.f
>> --------------------------------------------------------------------------
>> Unfortunately, this installation of Open MPI was not compiled with
>> Fortran 90 support. As such, the mpif90 compiler is non-functional.
>>
>
> This is the message that you get when you try to use /usr/bin/mpif90 that
> comes included with Xcode tools. The mpif90 that you want to use is
> /opt/mpich2/bin/mpif90, so make sure that /opt/mpich2/bin comes in your path
> before /usr/bin. You should probably also put /opt/mpich2/lib in your
> LD_LIBRARY_PATH, especially since building static executables is not
> recommended for macs (since mpif90 will properly link the required libraries
> at compile time). Do this by putting the following in your .bashrc (or
> .bash_profile)
>
> export PATH=/opt/mpich2/bin\:$PATH
> export LD_LIBRARY_PATH=/opt/mpich2/lib\:$LD_LIBRARY_PATH
>
> This is why I suggested building in /usr (that has always worked for me). I
> started building mpich2 in /mpi/mpich2/ since I decided to install multiple
> MPIs to test (mpich2 and openmpi for now), and with multiple MPIs, each one
> will need its own directory so they don't trod on each others' feet.
>
> Another option is to replace mpif90 in your config.h file everywhere with
> /opt/mpich2/bin/mpif90, to force the correct mpif90 be used regardless of
> what comes in the path first. Note that you'll have to do this for both
> config.h files in amber11/AmberTools/src and amber11/src, since I believe
> they are copies of one another, not hard (or soft) links.
>
> Hope this helps.
>
> Good luck!
> Jason
>
>
>> --------------------------------------------------------------------------
>> make[1]: *** [constants.o] Error 1
>> make: *** [parallel] Error 2
>>
>> Your help in this will be greatly appreciated
>>
>> George
>>
>>
>> On May 31, 2010, at 8:19 PM, Jason Swails wrote:
>>
>>> Did you make clean in the amber11/src directory before trying to build in
>> parallel?
>>>
>>> It looks like you may have some left-over pieces from a previous
>> compilation (in serial).
>>>
>>> Good luck!
>>> Jason
>>>
>>> --
>>> Jason Swails
>>> Quantum Theory Project,
>>> University of Florida
>>> Ph.D. Graduate Student
>>> 352-392-4032
>>>
>>> On May 31, 2010, at 10:52 AM, George Tzotzos <gtzotzos.me.com> wrote:
>>>
>>>> Hi Jason
>>>>
>>>> Thanks for the latest instructions to compile MPICH2.
>>>>
>>>> However, I encountered yet another problem. Hopefully it's the last one.
>>>>
>>>> ./configure -mpi gnu in $AMBERHOME/AmberTools/src works fine
>>>>
>>>> The configuration file, config.h, was successfully created.
>>>>
>>>> If you have amber11, the most common next step is to
>>>> 'cd ../../src; make clean; make parallel'. See the Users' Manual
>>>> for information on building a parallel version of AmberTools
>>>>
>>>>
>>>> Trying to do a "make parallel" in amber11/src gives the error below
>>>>
>>>> _schlegel_dg_ in schlegel_dg.o
>>>> _schlegel_dg_ in schlegel_dg.o
>>>> _evb_ntrfc_ in evb_ntrfc.o
>>>> _evb_ntrfc_ in evb_ntrfc.o
>>>> _evb_ntrfc_ in evb_ntrfc.o
>>>> _evb_ntrfc_ in evb_ntrfc.o
>>>> _evb_ntrfc_ in evb_ntrfc.o
>>>> ___ncsu_abmd_hooks_MOD_on_force in ncsu-abmd-hooks.o
>>>> ___ncsu_abmd_hooks_MOD_on_sander_init in ncsu-abmd-hooks.o
>>>> ___ncsu_abmd_hooks_MOD_on_sander_init in ncsu-abmd-hooks.o
>>>> ___ncsu_abmd_hooks_MOD_on_sander_init in ncsu-abmd-hooks.o
>>>> ___ncsu_sander_hooks_MOD_rem_preinit in ncsu-sander-hooks.o
>>>> ___ncsu_sander_hooks_MOD_rem_preinit in ncsu-sander-hooks.o
>>>> _qm2_load_params_and_allocate_ in
>> libsqm.a(qm2_load_params_and_allocate.o)
>>>> ld: symbol(s) not found
>>>> collect2: ld returned 1 exit status
>>>> make[1]: *** [sander.MPI] Error 1
>>>> make: *** [parallel] Error 2
>>>>
>>>> Any ideas?
>>>>
>>>> Thanks in advance
>>>>
>>>> George
>>>>
>>>>
>>>>
>>>> On May 31, 2010, at 4:01 PM, Jason Swails wrote:
>>>>
>>>>> Ah, it appears as though you have g77 installed on your machine? I
>> think
>>>>> that the configure script is trying to use g77 as the Fortran 77
>> compiler,
>>>>> when it should be using gfortran. Change the configure line to this:
>>>>>
>>>>> ./configure FC=gfortran F77=gfortran F90=gfortran CC=gcc CXX=g++
>>>>> --prefix=/usr
>>>>> make
>>>>> sudo make install
>>>>>
>>>>> This should make sure that gfortran and gcc are used as the compilers
>> for
>>>>> all fortran and c code (and g++ for c++ code).
>>>>>
>>>>> Good luck!
>>>>> Jason
>>>>>
>>>>> On Mon, May 31, 2010 at 9:47 AM, George Tzotzos <gtzotzos.me.com>
>> wrote:
>>>>>
>>>>>> Hi Jason,
>>>>>>
>>>>>> Many thanks for the prompt reply.
>>>>>>
>>>>>> I downloaded MPICH2 as you suggested and tried to install. However,
>> I'm
>>>>>> unable to proceed with the installation.
>>>>>>
>>>>>> See the error message below
>>>>>>
>>>>>> Any suggestions?
>>>>>>
>>>>>> Many thanks in advance
>>>>>>
>>>>>> George
>>>>>>
>>>>>> checking whether g77 accepts the FLIBS found by autoconf... yes
>>>>>> checking whether C can link with
>>>>>> -L/usr/local/lib/gcc/i686-apple-darwin8.8.1/3.4.0
>>>>>> -L/usr/local/lib/gcc/i686-apple-darwin8.8.1/3.4.0/../../.. -lfrtbegin
>>>>>> -lg2c... yes
>>>>>> checking for linker for Fortran main programs... Use Fortran to link
>>>>>> programs
>>>>>> checking whether Fortran 77 and C objects are compatible... no
>>>>>> checking for file... file
>>>>>> configure: error: **** Incompatible Fortran and C Object File Types!
>> ****
>>>>>> F77 Object File Type produced by "g77 -O2" is : : Mach-O object i386.
>>>>>> C Object File Type produced by "gcc -O2" is : : Mach-O 64-bit object
>>>>>> x86_64.
>>>>>> make: *** No targets specified and no makefile found. Stop.
>>>>>> make: *** No rule to make target `install'. Stop.
>>>>>>
>>>>>> On May 31, 2010, at 3:18 PM, Jason Swails wrote:
>>>>>>
>>>>>>> Hi George,
>>>>>>>
>>>>>>> Once you have the compilers built for Mac OS X 10.6, installing Amber
>>>>>> itself
>>>>>>> is quite simple (all you need is a simple ./configure gnu, make
>> serial,
>>>>>> and
>>>>>>> make serial in the amber11/AmberTools/src and amber11/src
>> directories).
>>>>>>> Luckily, building the compilers from macports or fink is a one-time
>>>>>>> necessity, so from now on Amber11 should be quite easy to build.
>>>>>>>
>>>>>>> Amber11 in parallel requires, once your system is properly set up,
>> only
>>>>>> for
>>>>>>> you to type "./configure -mpi gnu" in the amber11/AmberTools/src
>>>>>> directory
>>>>>>> and then make parallel instead of make serial as described above.
>> The
>>>>>> only
>>>>>>> additional thing you need is an MPI installation. You can either use
>>>>>>> macports or fink to build the MPI, or you can build your own from
>> source.
>>>>>> I
>>>>>>> would suggest building your own, since it's quite easy.
>>>>>>>
>>>>>>> First thing to do is pick an MPI and download the source code. As
>> has
>>>>>> been
>>>>>>> suggested many times before, I would suggest MPICH2 (you can download
>> the
>>>>>>> source code from
>>>>>>>
>>>>>>
>> http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/1.2.1p1/mpich2-1.2.1p1.tar.gz
>>>>>>> Untar the archive in a directory, then cd into that directory and
>> type
>>>>>>> "./configure --prefix=/usr; make; make install".
>>>>>>>
>>>>>>> This will configure it to be put in /usr/bin and /usr/lib and
>>>>>> /usr/include
>>>>>>> for the binaries, libraries, and include files, respectively. This
>>>>>> should
>>>>>>> successfully overwrite the fortran-deficient MPI installation that
>> comes
>>>>>>> with xcode tools.
>>>>>>>
>>>>>>> After you've done this, the 3 simple commands described above (and in
>> the
>>>>>>> Amber11 manual) should be sufficient to build a parallel version.
>>>>>> However,
>>>>>>> you have to make clean before you do this, since you don't want to
>> use
>>>>>> the
>>>>>>> serial installation as a starting point.
>>>>>>>
>>>>>>> Mengjuei has also set up a very helpful "amber on mac" weblog that
>> you
>>>>>> can
>>>>>>> find here http://amberonmac.blogspot.com/.
>>>>>>>
>>>>>>> Good luck!
>>>>>>> Jason
>>>>>>>
>>>>>>> On Mon, May 31, 2010 at 5:09 AM, George Tzotzos <gtzotzos.me.com>
>> wrote:
>>>>>>>
>>>>>>>> Hi folks,
>>>>>>>>
>>>>>>>> I've made a successful serial installation of AMBER11 on a quad core
>> MAC
>>>>>>>> (Intel Core i7). My installation was done pretty much on the lines
>>>>>> suggested
>>>>>>>> in the message of Erik Zuiderweg dated May 24.
>>>>>>>>
>>>>>>>> My question is whether there's a way to take this as a starting
>> point
>>>>>> for a
>>>>>>>> parallel installation or does one need to recompile the program from
>>>>>>>> scratch? If the former option is possible any suggestions on how to
>>>>>> proceed
>>>>>>>> from there would be most useful. I went through the mail archive but
>>>>>> somehow
>>>>>>>> I'm getting lost in the detail.
>>>>>>>>
>>>>>>>> Your help would be greatly appreciated
>>>>>>>>
>>>>>>>> George
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> AMBER mailing list
>>>>>>>> AMBER.ambermd.org
>>>>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Jason M. Swails
>>>>>>> Quantum Theory Project,
>>>>>>> University of Florida
>>>>>>> Ph.D. Graduate Student
>>>>>>> 352-392-4032
>>>>>>> _______________________________________________
>>>>>>> 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
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Jason M. Swails
>>>>> Quantum Theory Project,
>>>>> University of Florida
>>>>> Ph.D. Graduate Student
>>>>> 352-392-4032
>>>>> _______________________________________________
>>>>> 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
>>
>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> 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 Wed Jun 02 2010 - 11:00:12 PDT
Custom Search