home

Archive for the 'openmpi' Category

GROMACS 4.5.5, OpenMPI 1.6, And FFTW 3.3.2 Compilation Under Mountain Lion (OSX 10.8) With XCode (And A Little Help From Homebrew)

Tuesday, August 28th, 2012

Minus a few glitches easily fixed with the right software, this build wasn’t bad at all (and thanks to Adam Lindsay for the title catch).

Now sitting in front of a new Core i7 MacBook Pro, one of the first compilations I wanted to have finished for new projects was GROMACS 4.5.5. As my procedure for compiling GROMACS 3.3.3 had been a highly-traveled page, I wanted to provide a brief summary of my successful 4.5.5 compilation.

A Few Piece Of Info

1. XCode

This used to be disc-download and install, now it’s available as a free download from the App Store (1.57 GB download, so plan to do something else while you wait for the download).

2. Homebrew

Having Homebrew installed in Mountain Lion made the installation of FFTW easy and OpenMPI trivial once gfortran was equally trivially installed. Therefore, to make your life easier, I can’t recommend a Homebrew installation enough. For additional install tweaks, I followed the following page: gist.github.com/1860902

Installation Procedure

1. Download gromacs 4.5.5

…and place it in your home folder (will go to Downloads most likely, drag it to your home folder for ease of building).

2. Extract into your home holder

…with a double-click, making ~/gromacs-4.5.5.

3. brew install fftw

With the install of Homebrew, you’ll simply run the following from a terminal window and produce the following output:

brew install fftw

==> Downloading http://www.fftw.org/fftw-3.3.2.tar.gz
######################################################################## 100.0%
==> ./configure --enable-single --enable-sse --enable-shared --disable-debug 
--prefix=/usr/local/Cellar/fftw/3.3.2 --enable-threads --disable-fortran
==> make install
==> make clean
==> ./configure --enable-sse2 --enable-shared --disable-debug 
--prefix=/usr/local/Cellar/fftw/3.3.2 --enable-threads --disable-fortran
==> make install
==> make clean
==> ./configure --enable-long-double --enable-shared --disable-debug 
--prefix=/usr/local/Cellar/fftw/3.3.2 --enable-threads --disable-fortran
==> make install
/usr/local/Cellar/fftw/3.3.2: 34 files, 13M, built in 2.7 minutes

4. brew install gfortran

If you don’t install gfortran FIRST and try to install OpenMPI, you’ll get the following error in Homebrew:

==> Downloading http://www.open-mpi.org/software/ompi/v1.6/downloads/openmpi-1.6.tar.bz2
######################################################################## 100.0%
Error: This formula requires a fortran compiler, but we could not find one by
looking at the FC environment variable or searching your PATH for `gfortran`.
Please take one of the following actions:

  - Decide to use the build of gfortran 4.2.x provided by Homebrew using
        `brew install gfortran`

  - Choose another Fortran compiler by setting the FC environment variable:
        export FC=/path/to/some/fortran/compiler
    Using an alternative compiler may produce more efficient code, but we will
    not be able to provide support for build errors.

So don’t. Installing gfortran will produce the following:

brew install gfortran

==> Downloading http://r.research.att.com/tools/gcc-42-5666.3-darwin11.pkg
######################################################################## 100.0%
==> Installing gfortran 4.2.4 for XCode 4.2 (build 5666) or higher
==> Caveats
Brews that require a Fortran compiler should not use:
  depends_on 'gfortran'

The preferred method of declaring Fortran support is to use:
  def install
    ...
    ENV.fortran
    ...
  end

==> Summary
/usr/local/Cellar/gfortran/4.2.4-5666.3: 86 files, 72M, built in 39 seconds

5. brew install openmpi

This is what allows you to use all cores on your machine and is not in the default XCode install.

brew install openmpi

==> Downloading http://www.open-mpi.org/software/ompi/v1.6/downloads/openmpi-1.6.tar.bz2
Already downloaded: /Library/Caches/Homebrew/open-mpi-1.6.tar.bz2
==> Using Homebrew-provided fortran compiler.
This may be changed by setting the FC environment variable.

==> ./configure --prefix=/usr/local/Cellar/open-mpi/1.6 --enable-ipv6
==> make all
==> make install
/usr/local/Cellar/open-mpi/1.6: 674 files, 21M, built in 5.9 minutes

6. cd gromacs-4.5.5

7. ./configure –enable-float –enable-mpi

You’ll produce output such as found in: 2012august29_gromacs455_configure.txt

You’ll also get two odd errors at the end of the ./configure run that do not affect the rest of the procedure:

./configure --enable-float --enable-mpi

...
./configure: line 29242: sort: No such file or directory
./configure: line 29239: sed: No such file or directory

So ignore them.

NOTE: If you’ve been going by my 3.3.3 procedure and used…

./configure --enable-mpi --enable-double

You’ll get the following error when you try to run make:

Making all in include
Making all in .
make[2]: Nothing to be done for `all-am'.
Making all in types
make[2]: Nothing to be done for `all'.

...

/bin/sh ../../libtool --tag=CC   --mode=compile mpicc -DHAVE_CONFIG_H -I. -I../../src -I/usr/include/libxml2 -I../../include -DGMXLIBDIR=\"/usr/local/gromacs/share/top\"   -O3 -fomit-frame-pointer -finline-functions -Wall -Wno-unused -msse2 -funroll-all-loops -std=gnu99 -MT genborn_sse2_double.lo -MD -MP -MF .deps/genborn_sse2_double.Tpo -c -o genborn_sse2_double.lo genborn_sse2_double.c
 mpicc -DHAVE_CONFIG_H -I. -I../../src -I/usr/include/libxml2 -I../../include -DGMXLIBDIR=\"/usr/local/gromacs/share/top\" -O3 -fomit-frame-pointer -finline-functions -Wall -Wno-unused -msse2 -funroll-all-loops -std=gnu99 -MT genborn_sse2_double.lo -MD -MP -MF .deps/genborn_sse2_double.Tpo -c genborn_sse2_double.c  -fno-common -DPIC -o .libs/genborn_sse2_double.o
genborn_sse2_double.c:931: internal compiler error: Segmentation fault: 11
Please submit a full bug report,
with preprocessed source if appropriate.
See  for instructions.
make[3]: *** [genborn_sse2_double.lo] Error 1
make[2]: *** [all-recursive] Error 1
make[1]: *** [all] Error 2
make: *** [all-recursive] Error 1

So don’t do that, either. The proper flag is the enable-float.

8. make

This will produce the output available for download at: 2012august29_gromacs455_make.txt

9. make install

This will produce the output available for download at: 2012august29_gromacs455_make_install.txt

10. make links

This will produce the short piece of output reproduced below.

cd /usr/local/gromacs/bin && programs=`ls` && cd /usr/local/bin && \
	for i in $programs; do \
	   (test ! -f $i && ln -s /usr/local/gromacs/bin/$i . ; exit 0); \
	done

And with that, you should be able to run all programs from a terminal window.

Amber 11 And AmberTools 1.5 In Ubuntu 10.04 LTS (And Related, Including A How-To For EOL 8.10)

Saturday, July 16th, 2011

Having successfully navigated serial and parallel Amber10 installs under Ubuntu 8.10, I am pleased to report that the process for Amber11 with OpenMPI (from apt-get, one doesn’t have to build from scratch) under Ubuntu 10.10 is seemingly much easier (and have it here so I don’t forget). There is a bit of persnicketiness to the order of the serial and parallel installs that must be kept track of (and I’m building in serial-to-parallel order), but the process is otherwise straightforward.

For organizational purposes, I’m building amber11 in my $HOME directory. This removes some of the PATH issues with sudo-ing aspects of the install (and can be moved into another directory after the build is complete).

1. apt-get Installs

The search for dependent programs and libraries is a long and involved one given how many programs I have installed. Therefore, instead of trying to find all of the amber-dependent installs for successful building, I’m simply providing the list of everything I have on the test machine. As hard drives are cheap and Ubuntu will warn of conflicts, I recommend simply installing the below and accepting the 100 Mb hit to NOT have to find the smallest apt-get set (yes, some of these are obviously not needed).

sudo apt-get install build-essential cmake doxygen freeglut3-dev g++-multilib gcc-multilib gettext gnuplot ia32-libs lib32asound2 lib32gcc1 lib32gcc1-dbg lib32gfortran3 lib32gomp1 lib32mudflap0 lib32ncurses5 lib32nss-mdns lib32z1 libavdevice52 libc6-dev-i386 libc6-i386 libfreeimage-dev libglew1.5-dev libopenal1 libopenexr-dev libpng12-dev libqt4-dev libssl-dev libstdc++6-4.3-dbg libstdc++6-4.3-dev libstdc++6-4.3-doc libxi-dev libxml-simple-perl libxmu-dev mercurial nfs-common nfs-kernel-server portmap python2.6-dev rpm ssh

The above said, there are some obvious most-important installs that have to be there (according to the “official” Ubuntu amber11 install summary at ambermd.org/ubuntu.html). You could try to work with only these first if you were in a diagnostic mood today:

sudo apt-get install bison csh flex fort77 g++ gcc gfortran libbz2-dev libnetcdf-dev libopenmpi-dev libxext-dev libxt-dev openmpi-bin patch tcsh xorg-dev zlib1g-dev

With that, we move onto the AmberTools 1.5 install.

2. AmberTools 1.5 (Serial)

The AmberTools build process deals with PATH specifications for both it and Amber, then walks you through patching and a successful build.

user@machine:~$ tar xjf AmberTools-1.5.tar.bz2 
user@machine:~$ cd amber11/
user@machine:~/amber11$ echo "export AMBERHOME=$PWD" >> ~/.bashrc
user@machine:~/amber11$ echo "export PATH=$PATH:$AMBERHOME/bin" >> ~/.bashrc
user@machine:~/amber11$ source ~/.bashrc
user@machine:~/amber11$ wget http://ambermd.org/bugfixes/AmberTools/1.5/bugfix.all
user@machine:~/amber11$ patch -p0 < bugfix.all
user@machine:~/amber11$ rm bugfix.all
user@machine:~/amber11$ cd AmberTools/src/
user@machine:~/amber11/AmberTools/src$ ./configure gnu
user@machine:~/amber11/AmberTools/src$ make install
user@machine:~/amber11/AmberTools/src$ cd

3. Amber 11 (Serial Install)

For the Amber build, not building the serial version first will produce the following error (which you may or may not be searching against in google presently):

Warning: Deleted feature: PAUSE statement at (1)
cpp -traditional -P  -DBINTRAJ -DMPI    svbksb.f > _svbksb.f
mpif90 -c -O3 -mtune=generic -ffree-form   -o svbksb.o _svbksb.f
cpp -traditional -P  -DBINTRAJ -DMPI    pythag.f > _pythag.f
mpif90 -c -O3 -mtune=generic -ffree-form   -o pythag.o _pythag.f
Error: a serial version of libFpbsa.a must be built before parallel build.
make[2]: *** [libFpbsa.parallel] Error 2
make[2]: Leaving directory `/home/genomebio/amber11/AmberTools/src/pbsa'
make[1]: *** [libpbsa] Error 2
make[1]: Leaving directory `/home/genomebio/amber11/src/sander'
make: *** [parallel] Error 2

The “gnu” is also important, as there appears to be some kind of formatting (fortran-specific) issue with some files in the non-gnu build attempt that produces the following error if you just blindly run a ./configure:

Error: Unclassifiable statement at (1)
constants.f:39.1:

double precision, parameter :: two       = 2.0d0                        
 1
Error: Non-numeric character in statement label at (1)
constants.f:39.1:

double precision, parameter :: two       = 2.0d0                        
 1
Error: Unclassifiable statement at (1)
constants.f:40.1:

double precision, parameter :: three     = 3.0d0                        
 1
Error: Non-numeric character in statement label at (1)
Fatal Error: Error count reached limit of 25.
make[1]: *** [constants.o] Error 1
make[1]: Leaving directory `/home/user/amber11/src/sander'
make: *** [parallel] Error 2

With that, the serial build is below, including bug fixes.

user@machine:~$ tar xfj Amber11.tar.bz2
user@machine:~$ cd $AMBERHOME
user@machine:~/amber11$ wget http://ambermd.org/bugfixes/11.0/bugfix.all
user@machine:~/amber11$ wget http://ambermd.org/bugfixes/11.0/apply_bugfix.x
user@machine:~/amber11$ chmod +x ./apply_bugfix.x
user@machine:~/amber11$ ./apply_bugfix.x bugfix.all
user@machine:~/amber11$ cd AmberTools/src/
user@machine:~/amber11/AmberTools/src$ ./configure gnu
user@machine:~/amber11/AmberTools/src$ cd $AMBERHOME
user@machine:~/amber11$ ./AT15_Amber11.py 
user@machine:~/amber11$ cd src/
user@machine:~/amber11/src$ make serial

4. Amber 11 (Parallel)

Hopefully the serial build ran non-problematically. The parallel install works just as simply provided you run the process in the order below. The key steps are the “make clean,” new ./configure, re-run of ./AT_Amber11.py, and the other “make clean.”

user@machine:~/amber11/src$ cd $AMBERHOME
user@machine:~/amber11$ cd AmberTools/src/
user@machine:~/amber11/AmberTools/src$ make clean
user@machine:~/amber11/AmberTools/src$ ./configure -mpi gnu
user@machine:~/amber11/AmberTools/src$ cd $AMBERHOME
user@machine:~/amber11$ ./AT15_Amber11.py 
user@machine:~/amber11$ cd src/
user@machine:~/amber11/src$ make clean
user@machine:~/amber11/src$ make parallel

5. Amber 11 (Tests)

Finally, testing the install. Nothing specific to be done as far as the code is concerned, simply running the tests.

user@machine:~/amber11/src$ cd ..
user@machine:~/amber11$ cd test/
user@machine:~/amber11/test$ make -f Makefile
user@machine:~/amber11/test$ 

From the out-of-the-box installation above, my test results complete as follows:

365 file comparisons passed
15 file comparisons failed
0 tests experienced errors
Test log file saved as logs/test_amber_serial/2011-07-14_11-19-47.log
Test diffs file saved as logs/test_amber_serial/2011-07-14_11-19-47.diff

The failed tests include those already mentioned by the Amber developers to fail. This list is provided at the end of the AT15_Amber11.py results:

NOTE: Because PBSA has changed since Amber 11 was released, some
tests are known to fail and others are known to quit in error. These
can be safely ignored.

Tests that error: Tests in $AMBERHOME/test/sander_pbsa_frc
   Run.argasp.min    Run.dadt.min      Run.dgdc.min
   Run.lysasp.min    Run.polyALA.min   Run.polyAT.min
   Run.argasp.min    Run.dadt.min      Run.dgdc.min
   Run.lysasp.min    Run.polyALA.min   Run.polyAT.min
   Run.argasp.min    Run.dadt.min      Run.dgdc.min
   Run.lysasp.min    Run.polyALA.min   Run.polyAT.min

Tests that produce possible FAILUREs:
   cd sander_pbsa_ipb2   && ./Run.110D.min
   cd sander_pbsa_lpb    && ./Run.lsolver.min (only some of them fail here)
   cd sander_pbsa_tsr    && ./Run.tsrb.min
   cd sander_pbsa_decres && ./Run.pbsa_decres
   mm_pbsa.pl tests 02, 03, and 05

6. Quick Summary

For ease of copy-and-paste-ing, the command list is below:

apt-get

sudo apt-get install build-essential cmake doxygen freeglut3-dev g++-multilib gcc-multilib gettext gnuplot ia32-libs lib32asound2 lib32gcc1 lib32gcc1-dbg lib32gfortran3 lib32gomp1 lib32mudflap0 lib32ncurses5 lib32nss-mdns lib32z1 libavdevice52 libc6-dev-i386 libc6-i386 libfreeimage-dev libglew1.5-dev libopenal1 libopenexr-dev libpng12-dev libqt4-dev libssl-dev libstdc++6-4.3-dbg libstdc++6-4.3-dev libstdc++6-4.3-doc libxi-dev libxml-simple-perl libxmu-dev mercurial nfs-common nfs-kernel-server portmap python2.6-dev rpm ssh

sudo apt-get install bison csh flex fort77 g++ gcc gfortran libbz2-dev libnetcdf-dev libopenmpi-dev libxext-dev libxt-dev openmpi-bin patch tcsh xorg-dev zlib1g-dev

AmberTools

tar xjf AmberTools-1.5.tar.bz2 
cd amber11/
echo "export AMBERHOME=$PWD" >> ~/.bashrc
echo "export PATH=$PATH:$AMBERHOME/bin" >> ~/.bashrc
source ~/.bashrc
wget http://ambermd.org/bugfixes/AmberTools/1.5/bugfix.all
patch -p0 < bugfix.all
rm bugfix.all
cd AmberTools/src/
./configure gnu
make install
cd

Amber 11 (Serial)

tar xfj Amber11.tar.bz2
cd $AMBERHOME
wget http://ambermd.org/bugfixes/11.0/bugfix.all
wget http://ambermd.org/bugfixes/11.0/apply_bugfix.x
chmod +x ./apply_bugfix.x
./apply_bugfix.x bugfix.all
cd AmberTools/src/
./configure gnu
cd $AMBERHOME
./AT15_Amber11.py 
cd src/
make serial

Amber 11 (Parallel)

cd $AMBERHOME
cd AmberTools/src/
make clean
./configure -mpi gnu
cd $AMBERHOME
./AT15_Amber11.py 
cd src/
make clean
make parallel

Amber Tests

cd ..
cd test/
make -f Makefile

7. And Furthermore…

I tried the above on an old linux box running Intrepid Ibex (8.10), which counts as an End-Of-Life (Obsolete) version. Running all of the apt-get installs will work despite 8.10 not existing in the standard package locations, but you have to make the following addition to /etc/apt/sources.list.

sudo pico /etc/apt/sources.list

And copy-and-paste the following (this all taken from help.ubuntu.com/community/EOLUpgrades/Intrepid):

## EOL upgrade sources.list
# Required
deb http://old-releases.ubuntu.com/ubuntu/ intrepid main restricted universe multiverse
deb http://old-releases.ubuntu.com/ubuntu/ intrepid-updates main restricted universe multiverse
deb http://old-releases.ubuntu.com/ubuntu/ intrepid-security main restricted universe multiverse

Installing And Mounting Network Drives Using NFS In Ubuntu (And Generally)

Sunday, April 12th, 2009

This is another piece in an Ubuntu puzzle that, when assembled, will describe how to set up an MPI (message passing interface) computer cluster for running parallel calculations (upcoming).  As a brief explanation of what’s going on, many of the MPI (OpenMPI, MPICH, MPICH2) set-up procedures you may stumble across online describe how to use the network file system (NFS) protocol to set up one directory on a host node (head node/server node/master node/whatever) of your cluster so that, by mounting a directory on a guest node (client node/slave node/whatever) to this network-accessible drive, the head and guest nodes all see the same work directory and executables (both MPI and your program of choice).  There are more clever ways to set the cluster up that will likely run at a slightly faster pace than NFS may allow, but we’ll ignore that at the moment.  The install procedure below is Ubuntu-specific only in the apt-get stage (NFS support is not part of the default installation).  After all of the components are installed (post-apt-get), the setup should be Linux-universal.

LEGEND

Text in black – my ramblings.

Text in bold red – things you will type in the Terminal

Text in green – text you will either see or will type into files (using pico, my preference)

Below is all I need to do in Ubuntu to do what I need it to do.  I’ll be dividing the installation procedure into HOST and GUEST sections for organizational purposes.

1. HOST Node Installation

a. sudo apt-get install nfs-kernel-server nfs-common portmap

My apt-get list is becoming gigantic as part of the cluster work, but I’m only focusing on NFS right now (if you intend on running any MPI-based code, you also need to include SSH), so we only need to deal with these three packages (they will also install libevent1, libgssglue1, libnfsidmap2, and librpcsecgss3, but that’s the beauty of letting apt-get do the dirty work.  You can see this previous post if you intend on finding yourself installing any programs in Ubuntu while [shiver] not online).  You’ll see plenty of output and, hopefully, no errors.

b. sudo dpkg-reconfigure portmap

Installing packages with NFS in the title makes sense.  What’s the deal with portmap?  NFS uses remote procedure calls (RPCs) for communication.  Portmap is a server/service that maps these RPCs to their proper services (as a translator between RPC numbers and DARPA port numbers. Yup.), thereby directing cluster traffic.

The portmap configuration file (/etc/default/portmap) looks as below:

# Portmap configuration file
#
# Note: if you manually edit this configuration file,
# portmap configuration scripts will avoid modifying it
# (for example, by running ‘dpkg-reconfigure portmap’).

# If you want portmap to listen only to the loopback
# interface, uncomment the following line (it will be
# uncommented automatically if you configure this
# through debconf).
#OPTIONS=”-i 127.0.0.1″

For the purposes of the cluster work to be done in upcoming posts, you do NOT want to uncomment the OPTIONS line (which would then bind loopback), not that you would start randomly uncommenting lines in the first place.

c. sudo /etc/init.d/portmap restart

If you make changes to the /etc/default/portmap configuration file, reconfigure portmap, etc., you’ll need to restart portmap for the changes to be implemented.  It is recommended that you run this restart upon installation regardless (especially having run dpkg-reconfigure portmap above).

d. sudo mkdir /[work_directory]

This makes the directory to be shared among all of the other cluster machines.  Call it what you will.

e. sudo chmod a+wrx /[work_directory]

We now provide carte blanche to this directory so anyone can read, write, and execute programs in this directory.

f. sudo pico /etc/exports

The last file modification step on the HOST node will mount the /[work_directory] as an NFS-accessible directory to the GUEST machines and preserve this NFS accessibility until you change the settings (or the directory), preserving the accessibility upon reboot.

# /etc/exports: the access control list for filesystems which may be exported
#               to NFS clients.  See exports(5).
#
# Example for NFSv2 and NFSv3:
# /srv/homes       hostname1(rw,sync) hostname2(ro,sync)
#
# Example for NFSv4:
# /srv/nfs4        gss/krb5i(rw,sync,fsid=0,crossmnt)
# /srv/nfs4/homes  gss/krb5i(rw,sync)
#
/[work_directory] *(rw,sync)

To translate:  * = open up to all clients; rw = read-write priviledges to specified clients; sync = commit all changes to the disk before the server responds to some request making a change to the disk.  It is also worded as “read/write and all transfers to disk are committed to the disk before the write request by the client is completed” and “this option does not allow the server to reply to requests before the changes made by the request are written to the disk,” which may or may not help to enlighten.  Basically, it makes sure the NEXT modification to some file doesn’t occur until the CURRENT operation on that file is complete.

This only scratches the surface of all things /etc/exports but is enough for my purposes (anyone can mount the drive and read/write.  If your machine is online, let SSH take care of the rest of it).

g. sudo /etc/init.d/nfs-kernel-server restart

Having made the changes to /etc/exports, we restart the NFS server to commit those changes to the operating system.

h. sudo exportfs -a

If you RTFM, you know “the exportfs command is used to maintain the current table of exported file systems for NFS. This list is kept in a separate file named /var/lib/nfs/xtab which is read by mountd when a remote host requests access to mount a file tree, and parts of the list which are active are kept in the kernel’s export table.” (see HERE for more info).

With all of that completed, you will now have a network-accessible drive /[work_directory] sitting on the HEAD node.

Before moving on to the GUEST nodes, now what?  Again, based on other MPI and cross-network installation descriptions, a completely reasonable thing to do is build ALL of your cluster-specific programs (MPI and calculation programs) into /[work_directory].  You will then mount this directory on the client machines and set the PATHs on these machines to include /[work_directory]. Everyone then sees the same programs.

2. GUEST Node Installation

a. sudo apt-get install portmap nfs-common

This installs portmap and the NFS command files (but not the server) and also “starts up” the client NFS tools.  You should be ready to mount network drives immediately.

b. sudo mkdir /[work_directory]

We make the directory that will have the network drive mounted (the same name as the directory on the HOST node).  I’ve placed it in the same location in the directory hierarchy as it exists on the HEAD node (sitting right in /, not in /mnt or whatever).

For immediate access -> c1. sudo mount HOST_MACHINE:/[work_directory] /[work_directory]

Here, HOST_MACHINE may be an IP address (often something like 192.168.nn.nn or 10.1.nn.nn if you’ve set up your cluster on a switch, the actual IP address for the HOST machine if you know it (type ifconfig on the HOST machine to find out)) or a domain name (head.campus.edu, for instance).  It’s that simple (if everything installed properly).

For long-term, automated access -> c2a. sudo pico /etc/fstab

If you want to make this connection permanent (and this is a very good idea on a cluster), you can modify /etc/fstab by adding the following line:

HOST_MACHINE:/[work_directory] /[work_directory] nfs   rw   0   0

Then type

c2b. sudo mount /[work_directory]

As for additional information and discussion, there is quite a bit online already (as you might expect), with a long Ubuntu thread on the subject at ubuntuforums.org/showthread.php?t=249889.  For a bit more technical information on the subject, check out www.troubleshooters.com/linux/nfs.htm.

www.ubuntu.com
en.wikipedia.org/wiki/Message_Passing_Interface
www.open-mpi.org
www.mcs.anl.gov/research/projects/mpi/mpich1
www.mcs.anl.gov/research/projects/mpich2
en.wikipedia.org/wiki/Network_file_system
www.debian.org/doc/manuals/apt-howto
en.wikipedia.org/wiki/Secure_Shell
www.somewhereville.com/?p=616
en.wikipedia.org/wiki/Remote_procedure_call
www.redhat.com/docs/manuals/enterprise/RHEL-5-manual/Deployment_Guide-en-US/s1-nfs-server-config-exports.html
www.centos.org/docs/5/html/Deployment_Guide-en-US/s1-nfs-server-export.html
linux.about.com/library/cmd/blcmdl8_exportfs.htm
ubuntuforums.org/showthread.php?t=249889
www.troubleshooters.com/linux/nfs.htm

Building Parallel Abinit 5.6.x With OpenMPI 1.2.x (And NOT OpenMPI 1.3.x) From Sources In Ubuntu 8.x – iofn1.F90 Problem Solved

Wednesday, March 25th, 2009

This post is an update to my previous post on building Abinit with OpenMPI in Ubuntu, with this post providing a workaround (solution?) to a run-benign but ultimately thoroughly aggravating issue with starting calculations in the abinip parallel build.

The description of the procedure, and the problem in the OpenMPI 1.3.x build, is as taken from the previous page (repeated so that the error makes its way and embeds itself a little deeper into the search engines).

To run parallel Abinit on a multi-processor box (that is, SMP.  The actual multi-node cluster setup is in progress), the command is SUPPOSED to be follows:

mpirun -np N /opt/etsf/abinit/5.6/bin/abinip < input.file >& output

Where N is the number of processors.  For mpirun, you need to specify the full path to the executable (which, for the build above, is as Abinit installs abinip when the build occurs in the /opt directory).  The input.file specification is as per the Abinit users manual so I won’t go into it here. You will also be asked to supply your password because I’ve done nothing to the setup of ssh (you are, in effect, logging into your machine to run the MPI calculation).

Now, when the above is run, this is the error that I get:

abinit : nproc,me=           4           0
ABINIT

Give name for formatted input file:
At line 127 of file iofn1.F90 (unit = 5, file = ’stdin’)
Fortran runtime error: End of file
abinit : nproc,me=           4           1
abinit : nproc,me=           4           2
abinit : nproc,me=           4           3
————————————————————————–
mpirun has exited due to process rank 0 with PID 7131 on
node terahertz-desktop exiting without calling “finalize”. This may
have caused other processes in the application to be
terminated by signals sent by mpirun (as reported here).
————————————————————————–

What is supposed to happen is that the input.file file lists the files that Abinit requires to perform the run and provides these files by name one-at-a-time as Abinit requests them upon start-up.  For some reason, the input.file file is not being read properly or is not being read at all before the job crashes.  Oddities noted in the above order of the output include

(1) the abinit : nproc,me= values are not grouped above the “Give name for formatted input file:” <- Abinit does not appear to be trying to read the text from the nproc,me lines as actual input data, as you have to provide all of the files before Abinit will crash with a wrong file name.

(2) At line 127 of file iofn1.F90 <- this is an Abinit file that is responsible for reading the contents of input.files.  So, is the problem with this fine in Abinit? Well…

(3) The serial build of Abinit (abinis) runs just fine with input.file <- which leads me to conclude that the problem is mpirun-related.  I hope to resolve this (I’m sure it will be trivial) and post my error accordingly.

What’s the work-around?  Simple.  Copy the contents of the input.file file (literally Crtl+C with the text selected) and paste it after running this command:

mpirun -np N /opt/etsf/abinit/5.6/bin/abinip

Abinit will ask for the files in order AND your Crtl+C includes the carriage returns at the end of each line, so you are effectively feeding Abinit the same content it would read from the input.file file if, in fact, it was capable of reading the input.file file.

After considerable searching for NOT the error I was having with Abinit, I discovered the following thread [Post 1, Post 2, Post 3, Post 4, Post 5, Post 6] in the OpenMPI Users Mailing List (see?  Once these lists get populated with enough content, you’re bound to find just about everything).  This doesn’t directly address the problem (the problem is related but different, the origin of the problem is the same, and googling “OpenMPI” and “stdin” was what brought it to my attention).

The solution to the problem above is to build Open-MPI 1.2.x instead of Open-MPI 1.3.x.

NOTE: If, after building Open-MPI 1.2.x, you receive the following error the first time you run mpirun:

mpirun: error while loading shared libraries: libopen-rte.so.0: cannot open shared object file: No such file or directory

Simply type the following:

sudo ldconfig

To make the proper links to libopen-rte and associated libraries.  From the man page

ldconfig creates the necessary links and cache to the most recent shared libraries found in the directories specified on the command line, in the file /etc/ld.so.conf, and in the trusted directories (/lib and /usr/lib). The cache is used by the run-time linker, ld.so or ld-linux.so. ldconfig checks the header and filenames of the libraries it encounters when determining which versions should have their links updated.

The stdin problem, I think, may remain in the entire 1.3.x build series as, and it’s not a great piece of deduction, the fix is reported in the trunk for 1.4.1 (not being big on alpha/beta testing and wanting to get Abinit running more than wanting a final answer to this problem, I did not try installing the trunk build and testing accordingly), although I have no idea how quickly things happen in OpenMPI development.

Long-story-short, Abinit 5.6.5 and OpenMPI 1.2.x works just fine at the READ *.files step and, most importantly, Abinit runs can now be properly scripted to run without my having to be by the machine to copy+paste the contents of the *.files file.

www.somewhereville.com/?p=384
www.abinit.org
www.open-mpi.org
www.ubuntu.com
www.open-mpi.org/software/ompi/v1.3
en.wikipedia.org/wiki/Symmetric_multiprocessing
www.open-mpi.org/community/lists/users/2008/12/7520.php
www.open-mpi.org/community/lists/users/2008/12/7522.php
www.open-mpi.org/community/lists/users/2008/12/7531.php
www.open-mpi.org/community/lists/users/2008/12/7536.php
www.open-mpi.org/community/lists/users/2008/12/7539.php
www.open-mpi.org/community/lists/users/2008/12/7537.php
www.open-mpi.org/community/lists/ompi.php
www.google.com
www.open-mpi.org/software/ompi/v1.2/
linux.die.net/man/8/ldconfig
www.open-mpi.org/nightly/trunk

Amber And Ubuntu Part 2. Amber10 (Parallel Execution) Installation In Ubuntu 8.10 (Intrepid Ibex) With OpenMPI 1.3… And Commentary

Sunday, March 15th, 2009

After considerable trial and building/testing errors, what follows is as simplified a complete installation and (non-X11/QM) testing of Amber10 and OpenMPI 1.3 as I think can be procedure’d in Ubuntu 8.10 (and likely previous and subsequent Ubuntu versions), dealing specifically with assorted issues with root permissions and variable definitions as per the standard procedure for Amber10 installation.

I’ll begin with the short procedure and bare minimum notes, then will address a multitude of specific problems that may (did) arise during all of the build procedures.  The purpose for listing everything, it is hoped, is to make these errors appear in google during searches so that, when you come/came across the errors, your search will have provided some amount of useful feedback (and, for a few of the problems I had with previous builds of other programs, this blog is the ONLY thing that comes up in google).

Some of the content below is an extension of the single-processor build of Amber10 I posted previously.  In the interest of keeping the reading to a minimum, the short procedure below is light on explanations that are provided in the long procedure that follows.  If you’re running on a multi-core computer (and who isn’t anymore), you likely want to take advantage of the MPI capability, so I would advise simply following the procedure below and IGNORING the previous post (as I also state at the top of the previous page).

Enough blabbering.

LEGEND

Text in black – my ramblings.

Text in bold preformatted red - things you will type in the Terminal

Text in green – text you will either see or will type into files (using pico, my preference)

Amber10, OpenMPI 1.3, Ubuntu 8.10: The Easy, Teenage New York Version

0a. I assume you’re working from a fresh installation of Ubuntu 8.10.  Some of what is below is Ubuntu-specific because of the way that Ubuntu decides to deal with the Root/Administrator/User relationship.  That said, much is the same and I’ll try to remark accordingly.

0b. As with the previous post, I am writing this with an expected audience of a non-technical Linux user, which means more detail of the steps and not just “cd $HOME and ./config -> make -> make install.”  If the steps are too obvious to you, find someone who thinks a Terminal window is one that won’t close and let them do it.  Speaking of…

0c. Everything below is done from a Terminal window.  The last icon you will use is the Terminal icon you click on.  If you’ve never had the pleasure in Ubuntu Desktop, go to Applications -> Accessories -> Terminal.  You can save yourself some time dragging around menus by hovering over the Terminal icon, right-clicking, and “Add this launcher to panel.”

0d. One more thing for the absolute newbies.  I am assuming that you’ve used Firefox to download OpenMPI 1.3, the AmberTools 1.2 and Amber10 bz2 files, and the bugfix.all files for AmberTools 1.2 and Amber10 (more on that in 0e).  That’s all you’ll need for the installation below.  Note that I’m not dealing with X11 (library specifications and other testing to do first), MKL (the Intel Math Kernel Library), or the GOTO BLAS libraries.  The default download folder for Firefox in a fresh installation of Ubuntu is the Desktop.  I will assume your files are on the Desktop, which will then be moved around in the installation procedure below.

0e. bugfix.all (Amber10) and bugfix.all (AmberTools 1.2) – Yes, the Amber people provide both of these files for the two different programs using the same name.  As I assume you’re using Firefox to download these files from their respective pages, I recommend that you RENAME the bugfix.all file for AmberTools 1.2 to bugfix_at.all, which is the naming convention I use in the patching step of the installation process (just to keep the confusion down).  In case you’ve not had the pleasure yet, instead of clicking on the bugfix.all link on each page and saving the file that loads, simply right-click and “Save Link As…”, then save the AmberTools 1.2 bugfix.all file as bugfix_at.all.

0f. This OpenMPI 1.3 build and series of Amber10 tests assumes SMP (symmetric multi-processing) only, meaning you run only with the CPUs on your motherboard.  The setup of OpenMPI for cluster-based computing is a little more complicated and is not presented below (but is in process for write-up).

0g. And, further, I am not using the OpenMPI you can install by simply sudo apt-get install libopenmpi1 linopenmpi-dev openmpi-bin openmpi-doc for two reasons.  First, that installs OpenMPI 1.2.8 (I believe), which mangles one of the tests in such a way that I feared for subsequent stability in Amber10 work I might find myself doing.  Second, I want to eventually be able to build OpenMPI around other Fortran compilers (specifically g95 or the Intel Fortran Compiler) or INSERT OPTION X and, therefore, prefer to build from source.

So, from a Terminal window, the mostly comment-free/comment-minimal and almost flawless procedure is as follows.  Blindly assume that everything I am doing, especially in regards to WHERE this build is occurring and WHAT happens in the last few steps, is for a good reason (explained in detail later)…

0. cd $HOME (if you’re not in your $HOME directory already or don’t know where it is)

1. sudo apt-get update (requires administrative password)

2. sudo apt-get install ssh g++ g++-multilib g++-4.3-multilib gcc-4.3-doc libstdc++6-4.3-dbg libstdc++6-4.3-doc flex bison fort77 netcdf-bin gfortran gfortran-multilib gfortran-doc gfortran-4.3-multilib gfortran-4.3-doc libgfortran3-dbg autoconf autoconf2.13 autobook autoconf-archive gnu-standards autoconf-doc libtool gettext patch libblas3gf liblapack3gf libgfortran2 markdown csh (this is my ever-growing list of necessary programs and libraries that are not installed as part of a fresh Ubuntu installation)

3. pico .bashrc

Add the following lines to the bottom of this file.  This will make more sense shortly…

AMBERHOME=$HOME/Documents/amber10/
export AMBERHOME

MPI_HOME=/
export MPI_HOME

Crtl-X and the Enter Key twice to Exit

4. source .bashrc

5. pico. profile

Add the following line to the bottom of this file.

PATH=”$HOME/Documents/amber10/exe:$PATH”

Crtl-X and the Enter Key twice to Exit

6. source .profile

7. mv $HOME/Desktop/Amber* $HOME/Documents/ (this assumes the downloaded files are on the desktop)

8. mv $HOME/Desktop/openmpi-1.3 $HOME/Documents/ (this assumes the downloaded files are on the desktop)

9. cd $HOME/Documents/

10. gunzip openmpi-1.3.tar.gz

11. tar xvf openmpi-1.3.tar

12. cd openmpi-1.3

13. ./configure –prefix=/ (installs the MPI binaries and libraries into /. Avoids library errors I ran across despite MPI_HOME, but may be fixable. Copious output to follow.  For my results, click HERE)

14. sudo make all install (this installs binaries and libraries into “/”. Copious output to follow.  For my results, click HERE)

15. cd $HOME/Documents/

16. tar xvjf Amber10.tar.bz2

17. tar xvjf AmberTools-1.2.tar.bz2

18. mv $HOME/Desktop/bugfix* $HOME/Documents/$AMBERHOME (moves bugfix files to the Amber directory)

19. cd $AMBERHOME

20. patch -p0 -N -r patch-rejects < bugfix_at.all (patches AmberTools)

21. patch -p0 -N -r patch-rejects < bugfix.all (patches Amber10)

22. cd src/

23. ./configure_at -noX11

24. make -f Makefile_at (copious output to follow.  For my results, click HERE)

25. cd ../bin/

26. pico mopac.sh

At the top of the file, change sh to bash

Crtl-X and the Enter Key twice to Exit

27. cd ../test/

28. make -f Makefile_at test (copious output to follow.  For my results, click HERE)

29. cd ../src/

30. ./configure_amber -openmpi gfortran

31. make parallel (copious output to follow.  For my results, click HERE)

32. cd $HOME/.ssh/ (at this step, we allow auto-login in ssh so that the multiple mpirun tests do not require that you supply your password constantly)

33. ssh-keygen -t dsa

34. cat id_dsa.pub >> authorized_keys2

35. chmod 644 authorized_keys2

36. cd $AMBERHOME/test/

37. csh

38. setenv DO_PARALLEL ‘mpirun -np N (here, N is the number of processors you wish to use on your mobo)

39. make test.parallel.MM (copious output to follow.  For my results, click HERE)

40. exit (exits the csh shell)

41. sudo cp -r $HOME/Documents/amber10 /opt/amber10

42. cd $HOME

43. rm -r $HOME/Documents/amber10 (this deletes the build directory.  Delete or keep as you like)

44. pico .bashrc

Make the following change to the bottom of this file.

AMBERHOME=/opt/amber10/

Crtl-X and the Enter Key twice to Exit

45. source .bashrc

46. pico .profile

Make the following change to the bottom of this file.

PATH=”/opt/amber10/exe:$PATH”

Crtl-X and the Enter Key twice to Exit

47. source .profile

That is it!  In theory, you should now have a complete and tested Amber10 and AmberTools 1.2 build sitting in /opt/amber10.

Amber10, OpenMPI 1.3, Ubuntu 8.10: The Notes

What follows is the complete list of problems, questions, errors, issues, and general craziness from much trial and error for the installation procedure above.  As you can guess from my constant mentioning of Ubuntu in my statements-with-qualifications, some of these problems likely will not occur in other distros.  My primary reason for the extended discussion below is so that the errors and issues make their way into google so that people searching for fixes to these problems (if they come across them) will see actual content (if they choose to read it) and maybe find a reasonable fix.

I’ll be expanding in sections by grouping numbers above.

0a – 0g Installation Preparations

I’ve not much to add here except that OpenMPI is likely not the only way to install MPI Amber10 in Ubuntu, but I think it is easier than MPICH2 to set up cross-cluster calculations on a switch’ed network.  My stronger preference for OpenMPI stems from both past positive experience with OpenMPI and GROMACS (specifically on my Macbook Pro) and eventual success with OpenMPI 1.2.X and Abinit 5.6.5.  I had hoped to use the same version of OpenMPI for GROMACS, Amber10, and Abinit, but ran into a yet-to-be-resolved issue with OpenMPI 1.3.x in the building of Abinit (the problem is, apparently, resolved in the upcoming 1.4.1 build, but I’m not much for using release candidates.  I’ll be discussing this in an upcoming Abinit installation post based on my previous Abinit installation post).

1 – 2 apt-get

My listed apt-get installation set contains many, many programs and libraries that are not necessarily needed in the OpenMPI and Amber10 build but are required for other programs.  The apt-get approach is still much cleaner than installing the entire OpenSuse or Fedora DVD, but you do find yourself scrambling the first time you try to install anything to determine what programs are missing.  You don’t know you need csh installed until the test scripts fail.  You forget about ssh until you run mpirun for the first time.  I do not yet know if MKL, GOTO, or any of the X11-based AmberTools programs require additional libraries to be installed, so the above apt-get list may grow.  Check the comments section at bottom for updates.

The list below shows essential programs and libraries AND their suggested additional installs. As long as you’re online and apt-get‘ing anyway, might as well not risk missing something for your next compiling adventure.

g++ g++-multilib  g++-4.3-multilib  gcc-4.3-doc  libstdc++6-4.3-dbg  libstdc++6-4.3-doc

gfortran gfortran-multilib gfortran-doc gfortran-4.3-multilib gfortran-4.3-doc libgfortran3-dbg

autoconf autoconf2.13 autobook autoconf-archive gnu-standards autoconf-doc libtool gettext

flex bison

ssh csh patch markdown fort77 netcdf-bin libblas3gf liblapack3gf libgfortran2

3 – 9 .bashrc and .profile Modifications, Building In Your Own Directory

For a number of programs, the procedure from source is ./configure, make, and make install, make install often only responsible for moving folders into the correct directories (specifically, directories only root has access to, such as /usr/local/bin, /lib, and /opt).  In many distributions, this final step is actually sudo make install.  This division of make and make install is not preserved in Amber, which complicates the Ubuntu build slightly.  The building in $HOME/Documents (as I’ve described the procedure above) saves you from having to constantly sudo the extraction and building process in directories you, the typical user, do not have access to write into.

Working in $HOME\Documents (or any of your $HOME folders) allows for the complete build and testing of AmberTools and Amber10.

The other benefit from doing as much in $HOME as possible is the lack of a need to define variables as the root user (specifically, AMBERHOME, MPI_HOME, and DO_PARALLEL) by setting variables in the root .bashrc and .profile files, adding lines to the Makefiles for AmberTools and Amber10, or setting variables at prompts.  This variable specification issue arises because when you run a program with sudo, you invoke the root privileges and the root variable definitions, so any specifications you make for your PATH or these Amber-specific variables are lost.

Once the build is complete in the $HOME/Documents folder, we move the entire directory into /opt, my default location for all programs I build from source (but $HOME/Documents is just fine as well once the PATH is set).

10 – 14 OpenMPI 1.3

So, why not then use OpenMPI 1.2.x?  The build of Amber10 works just fine with OpenMPI 1.2.x in Ubuntu with all of the installation specifications described in Step 2 of the short procedure (the extensive apt-get).  The problem with 1.2.x occurs for a single test after the Amber10 build that I’ve not yet figured out a workaround for, but the error is sinister enough that I decided to not risk similar errors in my own Amber work and, instead, use OpenMPI 1.3.x, which does not suffer the same error.  The only (ONLY) test to fail with OpenMPI 1.2.x is the cnstph (constant pH simulation) test, which runs perfectly well but fails at the close of the calculation (you can test this yourself by changing the nstlim value in the mdin file to any arbitrarily large number).  The failure message, which kills the test series, is below.  This job also fails if you simply try to run it independently of the pre-defined test set (not the make test.parallel.MM, but cd’ing into the cnstph directory, setting variables, and running the script).

==============================================================
cd cnstph && ./Run.cnstph
[ubuntu-desktop:18263] *** Process received signal ***
[ubuntu-desktop:18263] Signal: Segmentation fault (11)
[ubuntu-desktop:18263] Signal code: Address not mapped (1)
[ubuntu-desktop:18263] Failing at address: 0x9069d2cb0
[ubuntu-desktop:18264] *** Process received signal ***
[ubuntu-desktop:18265] *** Process received signal ***
[ubuntu-desktop:18265] Signal: Segmentation fault (11)
[ubuntu-desktop:18265] Signal code: Address not mapped (1)
[ubuntu-desktop:18265] Failing at address: 0x907031c70
[ubuntu-desktop:18264] Signal: Segmentation fault (11)
[ubuntu-desktop:18264] Signal code: Address not mapped (1)
[ubuntu-desktop:18264] Failing at address: 0x906dbdc00
[ubuntu-desktop:18263] [ 0] /lib/libpthread.so.0 [0x2b17b5bef0f0]
[ubuntu-desktop:18263] [ 1] /usr/local/lib/libopen-pal.so.0(_int_free+0x57) [0x2b17b4c10937]
[ubuntu-desktop:18263] [ 2] /usr/local/lib/libopen-pal.so.0(free+0xeb) [0x2b17b4c122bb]
[ubuntu-desktop:18263] [ 3] /home/userid/Documents/amber10/exe/sander.MPI(sander_+0x73ce) [0x4c6ec2]
[ubuntu-desktop:18264] [ 0] /lib/libpthread.so.0 [0x2abd8f5810f0]
[ubuntu-desktop:18264] [ 1] /usr/local/lib/libopen-pal.so.0(_int_free+0x57) [0x2abd8e5a2937]
[ubuntu-desktop:18264] [ 2] /usr/local/lib/libopen-pal.so.0(free+0xeb) [0x2abd8e5a42bb]
[ubuntu-desktop:18264] [ 3] /home/userid/Documents/amber10/exe/sander.MPI(sander_+0x73ce) [0x4c6ec2]
[ubuntu-desktop:18264] [ 4] /home/userid/Documents/amber10/exe/sander.MPI(MAIN__+0xf0a) [0x4bfa66]
[ubuntu-desktop:18264] [ 5] /home/userid/Documents/amber10/exe/sander.MPI(main+0x2c) [0x88309c]
[ubuntu-desktop:18264] [ 6] /lib/libc.so.6(__libc_start_main+0xe6) [0x2abd8f7ad466]
[ubuntu-desktop:18264] [ 7] /home/userid/Documents/amber10/exe/sander.MPI [0x43a649]
[ubuntu-desktop:18264] *** End of error message ***
[ubuntu-desktop:18263] [ 4] /home/userid/Documents/amber10/exe/sander.MPI(MAIN__+0xf0a) [0x4bfa66]
[ubuntu-desktop:18263] [ 5] /home/userid/Documents/amber10/exe/sander.MPI(main+0x2c) [0x88309c]
[ubuntu-desktop:18263] [ 6] /lib/libc.so.6(__libc_start_main+0xe6) [0x2b17b5e1b466]
[ubuntu-desktop:18263] [ 7] /home/userid/Documents/amber10/exe/sander.MPI [0x43a649]
[ubuntu-desktop:18263] *** End of error message ***
[ubuntu-desktop:18265] [ 0] /lib/libpthread.so.0 [0x2ba0032600f0]
[ubuntu-desktop:18265] [ 1] /usr/local/lib/libopen-pal.so.0(_int_free+0x57) [0x2ba002281937]
[ubuntu-desktop:18265] [ 2] /usr/local/lib/libopen-pal.so.0(free+0xeb) [0x2ba0022832bb]
[ubuntu-desktop:18265] [ 3] /home/userid/Documents/amber10/exe/sander.MPI(sander_+0x73ce) [0x4c6ec2]
[ubuntu-desktop:18265] [ 4] /home/userid/Documents/amber10/exe/sander.MPI(MAIN__+0xf0a) [0x4bfa66]
[ubuntu-desktop:18265] [ 5] /home/userid/Documents/amber10/exe/sander.MPI(main+0x2c) [0x88309c]
[ubuntu-desktop:18265] [ 6] /lib/libc.so.6(__libc_start_main+0xe6) [0x2ba00348c466]
[ubuntu-desktop:18265] [ 7] /home/userid/Documents/amber10/exe/sander.MPI [0x43a649]
[ubuntu-desktop:18265] *** End of error message ***
mpirun noticed that job rank 0 with PID 18262 on node ubuntu-desktop exited on signal 11 (Segmentation fault).
3 additional processes aborted (not shown)
./Run.cnstph:  Program error
make[1]: *** [test.sander.GB] Error 1
make[1]: Leaving directory `/home/userid/Documents/amber10/test’
make: *** [test.sander.GB.MPI] Error 2

Errors like this one scare me to no end, especially when the error seems to be in the proper termination of a process (such as writing final positions or data files) and you risk such errors occurring after 2 week simulations with no way to get your data back.  If you decide (if it’s already installed, for instance) to use OpenMPI 1.2.x with Amber10 but still want to test the build, I’d suggest simply commenting out (#) the line

# cd cnstph && ./Run.cnstph

from the Makefile in the ../test directory.  I can’t imagine this happens in all other distributions, but I also don’t know what the problem could be given that all of the other tests work just fine.  That said, there’s a failed test for the OpenMPI 1.3.x build as well when you forget to run the tests from csh, but that has nothing to do with OpenMPI (see below).

16 – 19 Building Amber10 and AmberTools At $HOME

As described in the 3 – 9 section above, the problem with not building in your $HOME directory is the passing of variables in the build processes.  For instance, if you set MPI_HOME in your $HOME .bashrc file and then sudo make parallel, the error you’ll see is

Starting installation of Amber10 (parallel) at Mon Mar  9 22:30:36 EDT 2009.
cd sander; make parallel
make[1]: Entering directory `/opt/amber10/src/sander’
./checkparconf
cpp -traditional -I/usr/local/include -P -xassembler-with-cpp -Dsecond=ambsecond -DBINTRAJ -DMPI  constants.f > _constants.f
/usr/local/bin/mpif90 -c -O3 -fno-range-check -fno-second-underscore -ffree-form  -o constants.o _constants.f
Cannot open configuration file /usr/share/openmpi/mpif90-wrapper-data.txt
Error parsing data file mpif90: Not found
make[1]: *** [constants.o] Error 243
make[1]: Leaving directory `/opt/amber10/src/sander’
make: *** [parallel] Error 2

because the MPI_HOME variable is not specified for root.  Performing the compilation in $HOME/Documents avoids this issue.

If you want to build Amber and AmberTools in /opt as root for some reason and do not want to deal with modifying the .bashrc and .profile files in /root, you can modify the appropriate Amber files to define the variables needed for both building and testing. For AmberTools testing, you need to define AMBERHOME, which you do at the top of Makefile_at.

cd ../test/

sudo pico Makefile_at

include ../src/config.h

AMBERHOME=/opt/amber10/
export AMBERHOME

test: is_amberhome_defined \

sudo make -f Makefile_at test

For the Amber10 build process, you would need to modify configure_amber at the top of the file to specify the MPI_HOME variable.

sudo pico configure_amber

#!/bin/sh
#set -xv

MPI_HOME=/
export MPI_HOME

command=”$0 $*”

For the Amber10 testing process, you would need to assign both AMBERHOME (so the tests know where to look for the executables) and DO_PARALLEL (so the tests know to use OpenMPI) at the top of the file.

cd ../test/

sudo pico Makefile

include ../src/config_amber.h

AMBERHOME=/opt/amber10
export AMBERHOME

DO_PARALLEL=mpirun -np 4
export DO_PARALLEL

SHELL=/bin/sh

It otherwise makes no difference at all how you choose to do things so long as the program gets built.  Much of the Ubuntu literature I’ve stumbled across attempts to make people avoid doing anything to change the root account settings, which was the approach I chose to use in the “Easy” procedure above.

20 – 21 Patching Amber10 and AmberTools 1.2

No surprises and not necessary for building.  Do it anyway.  And patch is included as one of the apt-get‘ed programs.  Simply be cognizant of the naming of the bugfix files (the directories are the same for both Amber10 and AmberTools and the patch is simply applied to the files it finds).

22 – 24 Building AmberTools

Your choices for the AmberTools build are fairly limited.  Via configure_at –help,

Usage: ./configure_at [flags] compiler

where compiler is one of:

gcc, icc, solaris_cc, irix_cc, osf1_cc

If not specified then gcc is used.

Option flags:

-mpi        use MPI for parallelization
-scalapack  use ScaLAPACK for linear algebra (utilizes MPI)
-openmp     Use OpenMP pragmas for parallelization (icc, solaris_cc,gcc(>4.2))
-opteron    options for solaris/opteron
-ultra2     options for solaris/ultra2
-ultra3     options for solaris/ultra3
-ultra4     options for solaris/ultra4
-bit64      64-bit compilation for solaris
-perflib    Use solaris performance library in lieu of LAPACK and BLAS
-cygwin     modifications for cygwin/windows
-p4         use optimizations specific for the Intel Pentium4 processor
-altix      use optimizations specific for the SGI Altix with icc
-static     create statically linked executables
-noX11      Do not build programs that require X11 libraries, e.g. xleap.
-nobintraj  Delete support for binary (netCDF) trajectory files
-nosleap    Do not build sleap, which requires unantiquated compilers.

Environment variables:
MKL_HOME    If present, will link in Intel’s MKL libraries (icc,gcc)
GOTO        If present, and MKL_HOME is not set, will use this location
for the Goto BLAS routines

We’re building gcc (default) with the above installs.  Running ./configure -noX11 with the apt-get-installed programs above should produce the following output.

Setting AMBERHOME to /home/userid/Documents/amber10

Testing the C compiler:
mpicc  -m64 -o testp testp.c
OK

Obtaining the C++ compiler version:
g++ -v
The version is ../src/configure
4.3.2
[: 520: 3: unexpected operator
OK

Testing the g77 compiler:
g77 -O2 -fno-automatic -finit-local-zero -o testp testp.f
./configure_at: 538: g77: not found
./configure_at: 539: ./testp: not found
Unable to compile a Fortran program using g77 -O2 -fno-automatic -finit-local-zero

Testing the gfortran compiler:
gfortran -O1 -fno-automatic -o testp testp.f
OK

Testing flex:
OK

Configuring netcdf; (may be time-consuming)

NETCDF configure succeeded.

The configuration file, config.h, was successfully created.

The next step is to type ‘make -f Makefile_at’

25 – 28 Testing AmberTools

As a quick head’s up, if you don’t have csh installed, your test will fail at the following step with the following error:

cd ptraj_rmsa && ./Run.rms
/bin/sh: ./Run.rms: not found
make: *** [test.ptraj] Error 127

Otherwise, you can see the output from my test set HERE.  Installing csh is much easier than modifying multiple Run scripts.

That said, the one fix I do perform is to modify the mopac.sh file only slightly in accordance with the post by Mark Williamson on the Vanderbilt University Amber Listserve (the first stumbling block a ran into during testing).

29 – 31 Building Parallel Amber10

Running ./configure_amber -openmpi gfortran should output the following:

Setting AMBERHOME to /home/userid/Documents/amber10

Setting up Amber configuration file for architecture: gfortran
Using parallel communications library: openmpi
The MKL_HOME environment variable is not defined.

Testing the C compiler:
gcc  -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -O2 -m64 -o testp testp.c
OK

Testing the Fortran compiler:
gfortran -O0 -fno-range-check -fno-second-underscore -o testp testp.f
OK

——   Configuring the netCDF libraries:   ——–

Configuring netcdf; (may be time-consuming)
NETCDF configure succeeded.
MPI_HOME is set to /

The configuration file, config_amber.h, was successfully created.

32 – 35 Automatic ssh Login

These step saves you from constantly having to input your password for the mpirun testing phase.  This strictness to password provision is because of ssh (and, because I have machines both online and accessible, I prefer to deal with setting up the ssh side right instead of not having that layer of security).  The first time you run mpirun, ssh will throw back at you the following:

The authenticity of host ‘userid-desktop (127.0.1.1)’ can’t be established.
RSA key fingerprint is eb:86:24:66:67:0a:7a:7b:44:95:a6:83:d2:a8:68:01.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added ‘terahertz-desktop’ (RSA) to the list of known hosts.

Generating the automatic login file for ssh will look like the following:

Generating public/private dsa key pair.
Enter file in which to save the key (/home/userid/.ssh/id_dsa):
Created directory ‘/home/userid/.ssh’.
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /home/userid/.ssh/id_dsa.
Your public key has been saved in /home/userid/.ssh/id_dsa.pub.
The key fingerprint is:
54:84:68:79:3b:ec:17:41:c9:96:98:10:df:4f:cc:42 userid@userid-desktop
The key’s randomart image is:
+–[ DSA 1024]—-+

36 – 40 Testing Amber10

If you’re not in csh, either this test fails…

cd rdc && ./Run.dip
if: Badly formed number.
make[1]: *** [test.sander.BASIC] Error 1
make[1]: Leaving directory `/home/userid/Documents/amber10/test’
make: *** [test.sander.BASIC.MPI] Error 2

or this one…

cd pheMTI && ./Run.lambda0
This test must be run in parallel
make: *** [test.sander.TI] Error 1

Again, re-writing scripts is far less fun than sudo apt-get install csh and forgetting about it.

41 – 47 Moving The Built Amber10 and AmberTools

The final sequence of steps moves the $HOME/Documents-built amber10 into /opt (if you choose to), removes the build from your $HOME directory, and resets your PATH and AMBERHOME variables in .bashrc and .profile, thereby completing the build process.

And Finally…

If questions are raised, comments are thought of, speed-ups identified, etc., please either send me an email or post them here.  Our concern as computational chemists should be making predictions and interpreting data, not making compilation errors and interpreting error messages.

en.wikipedia.org/wiki/X_Window_System
en.wikipedia.org/wiki/Quantum_mechanics
ambermd.org
www.open-mpi.org
www.ubuntu.com
www.google.com
www.somewhereville.com
www.somewhereville.com/?p=345
en.wikipedia.org/wiki/Message_Passing_Interface
en.wikipedia.org/wiki/Zappa_in_New_York
help.ubuntu.com/community/RootSudo
www.linux.org
help.ubuntu.com/community/UsingTheTerminal
www.mozilla.com/en-US/firefox
ambermd.org/AmberTools-get.html
www.bzip.org
www.intel.com/cd/software/products/asmo-na/eng/307757.htm
www.tacc.utexas.edu/resources/software/gotoblasfaq.php
ambermd.org/bugfixes.html
en.wikipedia.org/wiki/Symmetric_multiprocessing
en.wikipedia.org/wiki/Fortran
www.g95.org
www.intel.com/cd/software/products/asmo-na/eng/282048.htm
www.mcs.anl.gov/index.php
www.gromacs.org
www.apple.com/macbookpro
www.abinit.org
www.abinit.org/package/?text=5_6_5
www.open-mpi.org/community/lists/users/2008/12/7522.php
www.open-mpi.org/community/lists/users/2008/12/7531.php
www.open-mpi.org/community/lists/users/2008/12/7536.php
www.open-mpi.org/community/lists/users/2008/12/7539.php
www.open-mpi.org/nightly/trunk
www.opensuse.org/en
fedoraproject.org
en.wikipedia.org/wiki/C_shell
en.wikipedia.org/wiki/Secure_Shell
www.open-mpi.org/software/ompi/v1.2

Obligatory

  • CNYO

  • Sol. Sys. Amb.

  • Ubuntu 4 Nano

  • NMT Review

  • N-Fact. Collab.

  • Pres. Asn. CNY

  • T R P Nanosys

  • Nano Gallery

  • nano gallery
  • Aerial Photos

    More @ flickr.com

    Syracuse Scenes

    More @ flickr.com