LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Kevin R Bryant (kevin.bryant_at_[hidden])
Date: 2006-10-10 10:35:22


The easiest thing you could do would be to just install your head node w/
32bit linux, not 64bit linux. (unless for whatever reason you need 64bit
libraries on the head node).

If 32bit head node is not an option, then you could always setup command
alias's to auto add in -m32. The only downside to this would be for users
that actually want to compile some 64bit code....

Kevin R Bryant <kevin.bryant_at_[hidden]>
IBM ATS Americas
Linux Cluster Architect
Cell Phone: (301) 524-1182
"Unix is user friendly. It is only careful selecting its friends."

                                                                           
             "Andrew D. Fant"
             <andrew.fant_at_tuft
             s.edu> To
             Sent by: lam_at_[hidden]
             lam-bounces_at_lam-m cc
             pi.org
                                                                   Subject
                                       LAM: Forcing addition flags to be
             10/10/2006 10:08 added in mpicc, et al
             AM
                                                                           
                                                                           
             Please respond to
              General LAM/MPI
               mailing list
                                                                           
                                                                           

Good Morning,
   I am working on setting up LAM-MPI on a Xeon EMT system running a 64-bit
distro of Linux that will be the head node for a cluster of 32-bit Linux
servers. Thankfully, we don't need to worry about jobs running on both the
head
node and the compute nodes, but a requirement has been imposed on me that
all
compiles on this system need to generate 32-bit x86 code by default. I've
sucessfully built LAM-MPI 7.1.1 on this system and when I insert -m32 on
the
mpicc command line, I generate code that runs happily (even with shared
libraries) on our 32-bit compute nodes.

   My problem is trying to add additional flags to the mpicc/mpiCC/mpif77
programs (-m32 and -march=pentium3, in particular) to automatically get
this
result. I added the options to the CFLAGS environment variable on the
configure
command line and I tried playing with the --with-extra-ld-flags option to
configure, but mpicc -showme indicates that the architecture flags are just
getting ignored.

   Is there an easy way to force the compiler wrapper programs to add
additional
flags to the default set provided?

Thanks,
             Andy

--
Andrew Fant           | The lion and the calf shall lie   | Disclaimer:
andrew.fant_at_[hidden] | down together, but the calf won't | Do you REALLY
UIT/USG               | get much sleep.                   | think I can
Tufts University      |                  W. Allen         | speak for
Tufts?
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/




graycol.gif
pic24398.gif
ecblank.gif