Problem with ScaLAPACK_installer

[Togran]

Hello!
I trying to install ScaLAPACK on my Scientific Linux 6.1 using installer script. I already have mpich2 installed on my system.
I type command : "./setup.py --mpibindir=/usr/lib/mpich2/bin/ --mpiincdir=/usr/include/mpich2-i386/"
And then I see this message:

/home/scalapack_installer/script/utils.py:25: DeprecationWarning: The popen2 module is deprecated. Use the subprocess module.
import popen2
========================================
Setting up the framework

MPI bin dir is /usr/lib/mpich2/bin
MPI include dir is /usr/include/mpich2-i386
Install directory is... /home/scalapack_installer/install
Build directory is... /home/scalapack_installer/build
mpicc : /usr/lib/mpich2/bin/mpicc
mpif90 : /usr/lib/mpich2/bin/mpif90
mpirun is /usr/lib/mpich2/bin/mpirun
BLAS library is...
LAPACK library is... Check if included in Blas library
Checking if mpicc works... yes
Checking if mpirun works...

COMMON: mpirun not working! aborting...
error is:
****************************************

****************************************

What wrong with mpirun? I compiled examples for mpich2 and it worked successful. In my system mpirun is link to /usr/bin/mpiexec.py, but it work! Please help me! Thank's alot!

[admin]

Hi This is what the installer is doing:

Code: Select all

    def check_mpicc(self):
        """ checks if mpicc works """
        # simply generates a C program containing a couple of calls
        # to MPI routines and checks if the compilation and execution
        # are succesful
        print 'Checking if mpicc works...',
        sys.stdout.flush()
        # generate
        writefile('tmpc.c',"""
            #include \"mpi.h\"
            #include <stdio.h>
            int main(int argc, char **argv){
            int iam;
            MPI_Init( &argc, &argv );
            MPI_Comm_rank( MPI_COMM_WORLD, &iam );
            if(iam==0){fprintf(stdout, \"success\" );fflush(stdout);}
            MPI_Finalize();
            return 0;
            }\n""")

        # compile
        ccomm = self.config.mpicc+' '+self.config.ccflags+' -o tmpc '+os.path.join(os.getcwd(),'tmpc.c')
        (output, error, retz) = runShellCommand(ccomm)
     
        if(retz != 0):
            print '\n\nCOMMON: mpicc not working! aborting...'
            print 'stderr:\n','*'*40,'\n',error,'\n','*'*40
            sys.exit()
        print 'yes'
       
        """ checks if mpirun works """
        print 'Checking if mpirun works...',
        sys.stdout.flush()
        # run
        comm = self.config.mpirun + ' ./tmpc'
        (output, error, retz) = runShellCommand(comm)
        if(retz != 0):
            print '\n\nCOMMON: mpirun not working! aborting...'
            print 'error is:\n','*'*40,'\n',error,'\n','*'*40
            sys.exit()


So you should be able to find ./tmpc and see why it is not working.
Just type
/usr/bin/mpiexec.py ./tmpc
Maybe you need some arguments?

Let mw know what is the actual problem and we see if there is a way to fix it.
Julie

[Togran]

Thank you for help!
I trying to do mpiexec ./tmpc and I saw this message:

/usr/bin/mpdroot: open failed for root's mpd conf filempiexec_sclinux (__init__ 1208): forked process failed; status=255

What does this mean?

P.S. tmpc is executable file. Then I type ./tmpc, I saw "success". This is OK?

P.P.S Sorry for my English... I am from Russia...

[Togran]

OK. I find some information about mpd conf on http://www.underworldproject.org/documentation/MpichDownload.html
I create file /etc/mpd.conf and then trying to do mpiexec ./tmpc
I see "success", but then I type command : "./setup.py --mpibindir=/usr/lib/mpich2/bin/ --mpiincdir=/usr/include/mpich2-i386/" nothing has changed:

/home/scalapack_installer/script/utils.py:25: DeprecationWarning: The popen2 module is deprecated. Use the subprocess module.
import popen2
========================================
Setting up the framework

MPI bin dir is /usr/lib/mpich2/bin
MPI include dir is /usr/include/mpich2-i386
Install directory is... /home/scalapack_installer/install
Build directory is... /home/scalapack_installer/build
mpicc : /usr/lib/mpich2/bin/mpicc
mpif90 : /usr/lib/mpich2/bin/mpif90
mpirun is /usr/lib/mpich2/bin/mpirun
BLAS library is...
LAPACK library is... Check if included in Blas library
Checking if mpicc works... yes
Checking if mpirun works...

COMMON: mpirun not working! aborting...
error is:
****************************************

****************************************

What is the problem? What I can do? Please help, it's very important for me!

[admin]

Not sure, hard to tell when no message is displayed.
mpirun test is running the command mpirun ./tmpc
I would advice to use directly the SCALAPACK Makefile build system.
Download ScaLAPACK from http://www.netlib.org/scalapack/scalapack.tgz
ScaLAPACK will need LAPACK and BLAS (a Reference BLAS implementation is included inside the LAPACK package) http://www.netlib.org/lapack/lapack.tgz

[Togran]

mpirun test is running the command mpirun ./tmpc

Yes. And return "success"!!! What the next step? Maybe, error after mpirun, or setup.py can't find ./tmpc ? Need some special library?
Let's thinking together... Thank you!

[admin]

Togran,
Try installing ScaLAPACK directly and see if it works.

[Togran]

Hello again! I did the next step with your installer! Here is the log of command
./setup.py --mpibindir=/usr/lib/mpich2/bin/ --mpiincdir=/usr/include/mpich2-i386/
--blaslib=/usr/lib/libblas.so --lapacklib=/usr/lib/liblapack.so

and

mpirun -np 4 ./xCbtest

Please, help me. What am I doing wrong?

[admin]

just a quick question, Do you use two different MPICH2 implementation ? I see /usr/lib/mpich2/ and /usr/include/mpich2-i386/
Since ScaLAPACK 2.0.0, the build is actually easier than before, so you should directly go with the Makefile system.
You can just go in your build/scalapack-2.0.1 directory, look at the SLmake.inc to check your variable and type make.

Those tests xCbtest and xFbtest must fail, but they must fail just at the end when BLACS_ABORT is tested.
Is it really all the output you get or just the error messages?
Before exiting, the output should be:

Code: Select all

 RUNNING BLACS_SET/BLACS_GET TESTS
 PASSED  BLACS_SET/BLACS_GET TESTS
 
 CALL BLACS_GRIDEXIT
 DONE BLACS_GRIDEXIT
   
 The final auxiliary test is for BLACS_ABORT.
 Immediately after this message, all processes should be killed.
 If processes survive the call, your BLACS_ABORT is incorrect.