prn
05-14-2004, 08:34 PM
Hello,
I have mailed essentially the same question to support, but in case someone else has seen similar problems, I am going to ask here on the forum too.
Here at Ball State University, we just ordered and have received a copy of Lahey LF95 Express for Linux. I have registered the product.
I installed it on (the head node of) a Linux cluster running under Red Hat 8.0. I installed it in the default /usr/local/lf9562 directory exactly as described in the User's Guide (Revision D). I performed the test in /usr/local/lf9562/examples/fortran, as the guide specified, and it passed.
I extracted mpich-1.2.5 from the cd to /home/local/test/mpich-1.2.5 (The /home filesystem is mounted across all the nodes of the cluster and is also much larger, so it seemed like a reasonable place to do some testing.) I ran configure using the command from line 387 of the README file:
./configure -fc=lf95 -f90=lf95
I then ran make and changed to the /home/local/test/mpich-1.2.5/examples/basic directory, where I ran:
[root@ccncluster basic]# make fpi
/home/local/test/mpich-1.2.5/bin/mpif77 -o fpi fpi.o
fpi.o(.data+0x20): undefined reference to `pmpi_wtime_'
fpi.o(.data+0x24): undefined reference to `pmpi_wtick_'
make: *** [fpi] Error 1
The env command shows (among other things):
LD_LIBRARY_PATH=/usr/local/lf9562/lib:/home/local/lib::/opt/scali/lib
PATH=/usr/local/lf9562/bin:/home/local/bin:/home/local/matlab6p5/bin:/opt/intel/compiler70/ia32/bin:/usr/kerberos/sbin:/usr/kerberos/bin:/opt/intel/compiler70/ia32/bin:/home/local/bin:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin:/usr/X11R6/bin:/opt/scali/bin:/opt/scali/sbin:/opt/scali/contrib/pbs/bin:/root/bin:/usr/local/fei5/bin
In fact, when I try to run "make testing" in /home/local/test/mpich-1.2.5/examples/test, I get output ending in:
--------------- Extracted output -----------------
....
**** Testing error.c ****
ccncluster : Fri May 14 10:32:15 EST 2004
/home/local/test/mpich-1.2.5/bin/mpicc -c status.c
/home/local/test/mpich-1.2.5/bin/mpicc -o status status.o
**** Testing status.c ****
FORTRAN TESTS
ccncluster : Fri May 14 10:32:16 EST 2004
/home/local/test/mpich-1.2.5/bin/mpif77 -o fmisc fmisc.f
Encountered 0 errors, 0 warnings in file fmisc.f.
fmisc.o: In function `MAIN__':
fmisc.o(.text+0x219): undefined reference to `mpi_file_open_'
fmisc.o(.text+0x24d): undefined reference to `mpi_file_write_'
fmisc.o(.text+0x261): undefined reference to `mpi_file_sync_'
fmisc.o(.text+0x27d): undefined reference to `mpi_file_get_amode_'
fmisc.o(.text+0x2d3): undefined reference to `mpi_file_get_atomicity_'
fmisc.o(.text+0x329): undefined reference to `mpi_file_set_atomicity_'
fmisc.o(.text+0x345): undefined reference to `mpi_file_get_atomicity_'
[...more of similar errors...]
fmisc.o(.data+0x54): undefined reference to `mpi_file_write_'
fmisc.o(.data+0x58): undefined reference to `mpi_file_open_'
fmisc.o(.data+0x7c): undefined reference to `pmpi_wtime_'
fmisc.o(.data+0x80): undefined reference to `pmpi_wtick_'
make[2]: *** [fmisc] Error 1
Could not build executable fmisc; aborting tests
make[1]: [testing] Error 1 (ignored)
End of testing in directory io
Running tests in directory command
End of testing in directory command
--------------- End of Extracted output -----------------
In other words, it appears that the test involving fmisc.f fails. It appears that all the tests above that were tests of mpich-1.2.5 using the C language, not FORTRAN and the first FORTRAN test failed.
I followed the instructions as closely as possible. I am guessing that the instructions for configuring and building mpich-1.2.5 are incomplete in some way. Can anyone give me any advice?
I notice that the README files from mpich (various versions) suggest using a "device" as in
configure --with-device=ch_p4
Was I supposed to include that in the configure command? Or is the problem something involving installation of the compiler and/or libraries?
Thank you.
Paul Neubauer
prn@bsu.edu
I have mailed essentially the same question to support, but in case someone else has seen similar problems, I am going to ask here on the forum too.
Here at Ball State University, we just ordered and have received a copy of Lahey LF95 Express for Linux. I have registered the product.
I installed it on (the head node of) a Linux cluster running under Red Hat 8.0. I installed it in the default /usr/local/lf9562 directory exactly as described in the User's Guide (Revision D). I performed the test in /usr/local/lf9562/examples/fortran, as the guide specified, and it passed.
I extracted mpich-1.2.5 from the cd to /home/local/test/mpich-1.2.5 (The /home filesystem is mounted across all the nodes of the cluster and is also much larger, so it seemed like a reasonable place to do some testing.) I ran configure using the command from line 387 of the README file:
./configure -fc=lf95 -f90=lf95
I then ran make and changed to the /home/local/test/mpich-1.2.5/examples/basic directory, where I ran:
[root@ccncluster basic]# make fpi
/home/local/test/mpich-1.2.5/bin/mpif77 -o fpi fpi.o
fpi.o(.data+0x20): undefined reference to `pmpi_wtime_'
fpi.o(.data+0x24): undefined reference to `pmpi_wtick_'
make: *** [fpi] Error 1
The env command shows (among other things):
LD_LIBRARY_PATH=/usr/local/lf9562/lib:/home/local/lib::/opt/scali/lib
PATH=/usr/local/lf9562/bin:/home/local/bin:/home/local/matlab6p5/bin:/opt/intel/compiler70/ia32/bin:/usr/kerberos/sbin:/usr/kerberos/bin:/opt/intel/compiler70/ia32/bin:/home/local/bin:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin:/usr/X11R6/bin:/opt/scali/bin:/opt/scali/sbin:/opt/scali/contrib/pbs/bin:/root/bin:/usr/local/fei5/bin
In fact, when I try to run "make testing" in /home/local/test/mpich-1.2.5/examples/test, I get output ending in:
--------------- Extracted output -----------------
....
**** Testing error.c ****
ccncluster : Fri May 14 10:32:15 EST 2004
/home/local/test/mpich-1.2.5/bin/mpicc -c status.c
/home/local/test/mpich-1.2.5/bin/mpicc -o status status.o
**** Testing status.c ****
FORTRAN TESTS
ccncluster : Fri May 14 10:32:16 EST 2004
/home/local/test/mpich-1.2.5/bin/mpif77 -o fmisc fmisc.f
Encountered 0 errors, 0 warnings in file fmisc.f.
fmisc.o: In function `MAIN__':
fmisc.o(.text+0x219): undefined reference to `mpi_file_open_'
fmisc.o(.text+0x24d): undefined reference to `mpi_file_write_'
fmisc.o(.text+0x261): undefined reference to `mpi_file_sync_'
fmisc.o(.text+0x27d): undefined reference to `mpi_file_get_amode_'
fmisc.o(.text+0x2d3): undefined reference to `mpi_file_get_atomicity_'
fmisc.o(.text+0x329): undefined reference to `mpi_file_set_atomicity_'
fmisc.o(.text+0x345): undefined reference to `mpi_file_get_atomicity_'
[...more of similar errors...]
fmisc.o(.data+0x54): undefined reference to `mpi_file_write_'
fmisc.o(.data+0x58): undefined reference to `mpi_file_open_'
fmisc.o(.data+0x7c): undefined reference to `pmpi_wtime_'
fmisc.o(.data+0x80): undefined reference to `pmpi_wtick_'
make[2]: *** [fmisc] Error 1
Could not build executable fmisc; aborting tests
make[1]: [testing] Error 1 (ignored)
End of testing in directory io
Running tests in directory command
End of testing in directory command
--------------- End of Extracted output -----------------
In other words, it appears that the test involving fmisc.f fails. It appears that all the tests above that were tests of mpich-1.2.5 using the C language, not FORTRAN and the first FORTRAN test failed.
I followed the instructions as closely as possible. I am guessing that the instructions for configuring and building mpich-1.2.5 are incomplete in some way. Can anyone give me any advice?
I notice that the README files from mpich (various versions) suggest using a "device" as in
configure --with-device=ch_p4
Was I supposed to include that in the configure command? Or is the problem something involving installation of the compiler and/or libraries?
Thank you.
Paul Neubauer
prn@bsu.edu