mpich 'make testing'

Marc Cozzi cozzi at
Tue Aug 7 14:34:19 PDT 2001


 Not quite the same problems but I would be interested in what procedures
other people are using to build the mpich package with PGI. I've followed
method sent to the list by David Walker will no success. It would almost
seem there is a syntax error in the sed command. After making the change to
the Makefile, doing a make clean then make mpi >& Make.log I get a failure
Configuration completed.
cd mpich-1.2.0/ && BINDIR="" LIBDIR="" sed -e 's at MPIR_HOME   =
   = $\\PGI/linux86@' Makefile > Makefile.pgi
sed: -e expression #1, char 58: Unterminated `s' command
make: *** [mpich-1.2.0] Error 1

Tried the method on the Portland web page which is about the same. I see a
lot of
unresolved references to bproc stuff. Created a pgconfig file that has the
Then ececuted it.
./configure -c++=pgCC -cc=pgcc -fc=pgf77 -cflags="-Msignextend -tp px
FOR_XDR -DHAVE_RPC_RPC_H=1" -opt=-fast
-fflags="-tp px" -c++flags="-tp px" -f90flags="-tp px" -f90=pgf90
al/mpich -comm=shared
sed -e 's at MPIR_HOME   = .*$@MPIR_HOME   = $$\\PGI/linux86@' Makefile >

This runs along quit a ways but near the end complains about the following:

/root/beompi/beompi-1.0.14/mpich-1.2.0/lib/libmpich.a(chnodename.o): In
function `
chnodename.o(.text+0x17): undefined reference to `bproc_currnode'
make[4]: *** [cpi_log] Error 1
** C application cannot be linked with logging library
** Testing if C application can be linked with tracing library
make[3]: *** [overtake] Error 1
make[2]: [linktest] Error 2 (ignored)
Could not link a C program with MPI libraries
make[2]: *** [linktest] Error 1
make[1]: *** [linktest] Error 2
make: *** [mpi] Error 2

Any suggestion you may have would be of great help. G77 is just plain
too picky and slow for the codes we need to run..

Marc Cozzi
Univ. of Notre Dame
cozzi at

-----Original Message-----
From: Nicholas York [mailto:nick at]
Sent: Monday, August 06, 2001 4:29 PM
To: Richard Walsh; beowulf at
Subject: Re: mpich 'make testing'


Figuring the problem could be with PGI, I then tried compiling with GNU CC.

./configure -cc=gcc -c++=g++ -cflags="-mcpu=i686 -O2 -Os --fast-math
" -fc=g77 -fflags="-mcpu=i686 -O2 -Os --fast-math" -f90=pgf90 -f90flags="-tp
p6 -fast"

After fixing the probems introduced by gcc2.96's va_arg(), I got it to
compile again.  However, I'm still getting the same problems with 'make
testing' (and with all the tests executed by examples/test/pt2pt/runtests)


----- Original Message -----
From: "Richard Walsh" <rbw at>
To: <beowulf at>; <nick at>
Sent: Monday, August 06, 2001 2:17 PM
Subject: Re: mpich 'make testing'

> Nick York writes:
> >I've compiled mpich-1.2.1 with the Portland Group compiler and I'm =
> >getting the error below.  I had installed NPB and had been testing it =
> >(works fine with 1 process) and had been getting the same problem when I
> >ran any test with multiple processes, so I went back and started testing
> >my mpich installation and found this problem.  I've seen a few people =
> >reporting problems with casting integers when compiling this with GNU =
> >cc, but no specific fixes to the problem.=20
> I do not believe Portland Group supports mpich-1.2.1 at
> present. There is information on getting 1.2.0 runnning
> with it on there web site, but as of July 2001, 1.2.1 was
> not yet supported.
> Regards,
> Richard Walsh
> #---------------------------------------------------
> #
> # Richard Walsh
> # Project Manager, Financial Markets
> # netASPx, Inc.
> # 1200 Washington Ave. So.
> # Minneapolis, MN 55415
> # VOX:    612-337-3467
> # FAX:    612-337-3400
> # EMAIL:  rbw at, richard.walsh at
> #
> #---------------------------------------------------
> # "What you can do, or dream you can, begin it;
> #  Boldness has genius, power, and magic in it."
> #                                  -Goethe
> #---------------------------------------------------
> # "Without mystery, there can be no authority."
> #                                  -Charles DeGaulle
> #---------------------------------------------------

Beowulf mailing list, Beowulf at
To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list