This looks like a problem with the Intel compilers and Red Hat 9.
Unfortunately, there is not much we (the LAM developers) can do - it is
quite clear from your error messages that the compiler is not
functioning properly. As the compiler made it to through a number of
steps, my guess is that only parts of the system headers, compiler
headers, and your older glibc interact poorly. There are well known
problems (see the list archives) with the Intel compilers, Red Hat 8 or
9, and System V semaphore and shared memory support. I'm guessing the
thread library issue is because of a libpthread / libc mismatch - the
two libraries are intricately tied together on Linux. At present, it
looks like you have a limited set of options:
* Roll your cluster back to a version of Red Hat supported by the
Intel compilers
* Find a way to hack the Intel compilers into working with Red Hat 9
* Use GCC instead of the Intel compilers
Unfortunately, the LAM team does not have the ability to provide
support with hacking the Intel compilers to work with Red hat 9. I
would recommend doing a web search - you may be able to find something.
Hope this helps,
Brian
On Friday, July 25, 2003, at 04:11 PM, Manuel Tiglio wrote:
> Hi,
> I am trying to install lam 7.0 on a machine with two Xeon processors,
> RH9.0
> installed and the intel compilers. I get a bunch of errors, which I
> describe next:
>
> 1) When I configure lam without further options (except for defining
> the
> environment variables in order to use the intel compilers) I get
>
> checking whether to disable -pthread check... no
> checking for thr_create in -lthread... no
> checking for pthread_create in -lpthread... no
> checking for pthread_create... no
> checking for type of thread support... none found
>
> ***********************************************************************
> ******
>
> LAM/MPI was unable to find threading support on your system. In the
> near
> future, the LAM development team is considering requiring threading
> support
> for proper LAM execution. This is in part because we are not aware of
> any
> users that do not have thread support - so we need you to e-mail us at
> lam_at_[hidden] and let us know about this problem.
>
> To build this version of LAM/MPI without thread support, re-run
> configure with the '--without-threads' option.
>
> ***********************************************************************
> ******
>
> configure: error: "*** Can not continue."
>
> 2) Then I configured with the options --without-threads, and also with
> --with-rpi=sysv, and got (strange, since the machine does have shared
> memory)
>
>
> checking for C optimization flags... -O -axW -tpp7 -align -autodouble
> -d10
> checking for sys/sem.h... no
> checking for SYSV semaphore support... no
> checking for sys/ipc.h... no
> checking for sys/shm.h... no
> checking for SYSV shared memory support... no
>
> *** sysv RPI setup
> checking for TCP short message length... 65536 bytes
> configure: WARNING: *** sysv RPI cannot be build on a machine without
> configure: WARNING: *** SysV semaphores or shared memory.
> configure: error: *** Aborting building sysv RPI.
> configure: /bin/sh './configure' *failed* for share/ssi/rpi/sysv
> configure: >>> SSI module rpi:sysv will *not* be compiled
>
> configure: WARNING: SSI module "sysv" failed to configure properly
> configure: WARNING: This module was selected as the default
> configure: error: Cannot continue
>
> 3) So I decided to configure skipping the --with-rpi=sysv option,
> apparently in lam 7.0 I can choose it at run time. The configuration
> finished, but make gave:
>
> /usr/include/bits/ipctypes.h(33): error: "__ipc_pid_t" has already been
> declared in the current scope
> typedef int __ipc_pid_t;
> ^
>
> compilation aborted for registry.c (code 2)
>
>
> I found this message regarding an installation with RH9.0 on Xeon's
> with
> the intel 7.1 compilers,
> http://www.lam-mpi.org/MailArchives/lam/msg06465.php
> but apparently I have some even more basic problems that are not
> discussed
> there?
>
> In order to make the intel 7.1 compilers work with RH9.0 I am linking
> with
> an old glib library. Without mpi at least, the compiler seems to be
> working
> fine, but it could still be that I messed up something there?
>
> (A couple of weeks ago I installed lam on a cluster with RH7.3, older
> versions of the intel compilers, and lam 6.5.9, and it is working fine)
>
> Any help is greatly appreciated
>
> Manuel
>
>
>
>
>
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
|
