LAM: Runtime failure on Leopard

I just re-installed LAM on my Leopard cluster and after a
lot of  
difficulties in the configure and compile I've reached the
following  
error

[mymachine:lam-7.1.4/examples/hello] myname% mpirun -np 4
hello
mpirun (set_stdio): Bad file descriptor

Any hints?

Laminfo output

              LAM/MPI: 7.1.4
               Prefix: /opt/local
         Architecture: i386-apple-darwin9.2.0
        Configured by: myname
        Configured on: Thu Mar 27 14:38:43 EDT 2008
       Configure host: mymachine
       Memory manager: darwin7malloc
           C bindings: yes
         C++ bindings: yes
     Fortran bindings: yes
           C compiler: /usr/local/bin/gcc
         C++ compiler: /usr/local/bin/g++
     Fortran compiler: /usr/local/bin/gfortran
      Fortran symbols: underscore
          C profiling: yes
        C++ profiling: yes
    Fortran profiling: yes
       C++ exceptions: no
       Thread support: yes
        ROMIO support: yes
         IMPI support: no
        Debug support: no
         Purify clean: no
             SSI boot: globus (API v1.1, Module v0.6)
             SSI boot: rsh (API v1.1, Module v1.1)
             SSI boot: slurm (API v1.1, Module v1.0)
             SSI coll: lam_basic (API v1.1, Module v7.1)
             SSI coll: shmem (API v1.1, Module v1.0)
             SSI coll: smp (API v1.1, Module v1.2)
              SSI rpi: crtcp (API v1.1, Module v1.1)
              SSI rpi: lamd (API v1.0, Module v7.1)
              SSI rpi: sysv (API v1.0, Module v7.1)
              SSI rpi: tcp (API v1.0, Module v7.1)
              SSI rpi: usysv (API v1.0, Module v7.1)
               SSI cr: self (API v1.0, Module v1.0)



_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

Hi,

Leopard has its own mpirun.
Please check which one you are actually using. That one from
LAM or  
from the system.
How did you compile 'hello' program.

Regards,
Artur Tyliszczak.

On 2008-03-27, at 21:42, Gregory John Orris wrote:

> I just re-installed LAM on my Leopard cluster and after
a lot of
> difficulties in the configure and compile I've reached
the following
> error
>
> [mymachine:lam-7.1.4/examples/hello] myname% mpirun -np
4 hello
> mpirun (set_stdio): Bad file descriptor
>
> Any hints?
>
> Laminfo output
>
>              LAM/MPI: 7.1.4
>               Prefix: /opt/local
>         Architecture: i386-apple-darwin9.2.0
>        Configured by: myname
>        Configured on: Thu Mar 27 14:38:43 EDT 2008
>       Configure host: mymachine
>       Memory manager: darwin7malloc
>           C bindings: yes
>         C++ bindings: yes
>     Fortran bindings: yes
>           C compiler: /usr/local/bin/gcc
>         C++ compiler: /usr/local/bin/g++
>     Fortran compiler: /usr/local/bin/gfortran
>      Fortran symbols: underscore
>          C profiling: yes
>        C++ profiling: yes
>    Fortran profiling: yes
>       C++ exceptions: no
>       Thread support: yes
>        ROMIO support: yes
>         IMPI support: no
>        Debug support: no
>         Purify clean: no
>             SSI boot: globus (API v1.1, Module v0.6)
>             SSI boot: rsh (API v1.1, Module v1.1)
>             SSI boot: slurm (API v1.1, Module v1.0)
>             SSI coll: lam_basic (API v1.1, Module
v7.1)
>             SSI coll: shmem (API v1.1, Module v1.0)
>             SSI coll: smp (API v1.1, Module v1.2)
>              SSI rpi: crtcp (API v1.1, Module v1.1)
>              SSI rpi: lamd (API v1.0, Module v7.1)
>              SSI rpi: sysv (API v1.0, Module v7.1)
>              SSI rpi: tcp (API v1.0, Module v7.1)
>              SSI rpi: usysv (API v1.0, Module v7.1)
>               SSI cr: self (API v1.0, Module v1.0)
>
>
>
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

Artur,

Without going into the gory details, the native mpicc does
not include  
a fortran compiler. I thought I'd install LAM, since I used
it under  
Tiger and was more than satisfied. The compiling of the
software was  
done to ensure that the correct set of compilers. Sorry but,
if it  
were that easy I'd not sent out the query. I did notice that
several  
years ago a similar problem was found on the Itaniums.
Wondering if  
the problems are related since they seem to be internal to
LAM?

Regards,
Greg

On Mar 28, 2008, at 4:41 AM, Artur Tyliszczak wrote:

> Hi,
>
> Leopard has its own mpirun.
> Please check which one you are actually using. That one
from LAM or
> from the system.
> How did you compile 'hello' program.
>
> Regards,
> Artur Tyliszczak.
>
> On 2008-03-27, at 21:42, Gregory John Orris wrote:
>
>> I just re-installed LAM on my Leopard cluster and
after a lot of
>> difficulties in the configure and compile I've
reached the following
>> error
>>
>> [mymachine:lam-7.1.4/examples/hello] myname% mpirun
-np 4 hello
>> mpirun (set_stdio): Bad file descriptor
>>
>> Any hints?
>>
>> Laminfo output
>>
>>             LAM/MPI: 7.1.4
>>              Prefix: /opt/local
>>        Architecture: i386-apple-darwin9.2.0
>>       Configured by: myname
>>       Configured on: Thu Mar 27 14:38:43 EDT 2008
>>      Configure host: mymachine
>>      Memory manager: darwin7malloc
>>          C bindings: yes
>>        C++ bindings: yes
>>    Fortran bindings: yes
>>          C compiler: /usr/local/bin/gcc
>>        C++ compiler: /usr/local/bin/g++
>>    Fortran compiler: /usr/local/bin/gfortran
>>     Fortran symbols: underscore
>>         C profiling: yes
>>       C++ profiling: yes
>>   Fortran profiling: yes
>>      C++ exceptions: no
>>      Thread support: yes
>>       ROMIO support: yes
>>        IMPI support: no
>>       Debug support: no
>>        Purify clean: no
>>            SSI boot: globus (API v1.1, Module
v0.6)
>>            SSI boot: rsh (API v1.1, Module v1.1)
>>            SSI boot: slurm (API v1.1, Module v1.0)
>>            SSI coll: lam_basic (API v1.1, Module
v7.1)
>>            SSI coll: shmem (API v1.1, Module v1.0)
>>            SSI coll: smp (API v1.1, Module v1.2)
>>             SSI rpi: crtcp (API v1.1, Module v1.1)
>>             SSI rpi: lamd (API v1.0, Module v7.1)
>>             SSI rpi: sysv (API v1.0, Module v7.1)
>>             SSI rpi: tcp (API v1.0, Module v7.1)
>>             SSI rpi: usysv (API v1.0, Module v7.1)
>>              SSI cr: self (API v1.0, Module v1.0)
>>
>>
>>
>> _______________________________________________
>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>

_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

Hi Greg,

I had a lot of problems installing LAM on my MacBookPro
(Intel  
processor) under Leopard. Using Tiger everything was fine
but with  
Leopard......... Anyway, finally I managed to compile (Intel
fortran  
compiler) and install LAM 7.1.4, then I could also compile
my programs  
without any problems. BUT when I tried to start whatever,
even the  
simplest "Hello world" some strange errors were
reported. If you are  
interested you can try to find my thread on LAM forum.

Instead of LAM I installed OpenMPI (this has nothing common
with Open  
MP transfer protocol) which uses MPI standard. Now
everything works  
perfectly.

You can try to do the same.

Regards,
Artur.

On 2008-03-28, at 17:07, Gregory John Orris wrote:

> Artur,
>
> Without going into the gory details, the native mpicc
does not include
> a fortran compiler. I thought I'd install LAM, since I
used it under
> Tiger and was more than satisfied. The compiling of the
software was
> done to ensure that the correct set of compilers. Sorry
but, if it
> were that easy I'd not sent out the query. I did notice
that several
> years ago a similar problem was found on the Itaniums.
Wondering if
> the problems are related since they seem to be internal
to LAM?
>
> Regards,
> Greg
>
> On Mar 28, 2008, at 4:41 AM, Artur Tyliszczak wrote:
>
>> Hi,
>>
>> Leopard has its own mpirun.
>> Please check which one you are actually using. That
one from LAM or
>> from the system.
>> How did you compile 'hello' program.
>>
>> Regards,
>> Artur Tyliszczak.
>>
>> On 2008-03-27, at 21:42, Gregory John Orris wrote:
>>
>>> I just re-installed LAM on my Leopard cluster
and after a lot of
>>> difficulties in the configure and compile I've
reached the following
>>> error
>>>
>>> [mymachine:lam-7.1.4/examples/hello] myname%
mpirun -np 4 hello
>>> mpirun (set_stdio): Bad file descriptor
>>>
>>> Any hints?
>>>
>>> Laminfo output
>>>
>>>            LAM/MPI: 7.1.4
>>>             Prefix: /opt/local
>>>       Architecture: i386-apple-darwin9.2.0
>>>      Configured by: myname
>>>      Configured on: Thu Mar 27 14:38:43 EDT
2008
>>>     Configure host: mymachine
>>>     Memory manager: darwin7malloc
>>>         C bindings: yes
>>>       C++ bindings: yes
>>>   Fortran bindings: yes
>>>         C compiler: /usr/local/bin/gcc
>>>       C++ compiler: /usr/local/bin/g++
>>>   Fortran compiler: /usr/local/bin/gfortran
>>>    Fortran symbols: underscore
>>>        C profiling: yes
>>>      C++ profiling: yes
>>>  Fortran profiling: yes
>>>     C++ exceptions: no
>>>     Thread support: yes
>>>      ROMIO support: yes
>>>       IMPI support: no
>>>      Debug support: no
>>>       Purify clean: no
>>>           SSI boot: globus (API v1.1, Module
v0.6)
>>>           SSI boot: rsh (API v1.1, Module
v1.1)
>>>           SSI boot: slurm (API v1.1, Module
v1.0)
>>>           SSI coll: lam_basic (API v1.1, Module
v7.1)
>>>           SSI coll: shmem (API v1.1, Module
v1.0)
>>>           SSI coll: smp (API v1.1, Module
v1.2)
>>>            SSI rpi: crtcp (API v1.1, Module
v1.1)
>>>            SSI rpi: lamd (API v1.0, Module
v7.1)
>>>            SSI rpi: sysv (API v1.0, Module
v7.1)
>>>            SSI rpi: tcp (API v1.0, Module
v7.1)
>>>            SSI rpi: usysv (API v1.0, Module
v7.1)
>>>             SSI cr: self (API v1.0, Module
v1.0)
>>>
>>>
>>>
>>>
_______________________________________________
>>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>
>> _______________________________________________
>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>
>
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

FWIW, I have found that the -lmpi used by LAM's wrapper
compilers may  
actually find /usr/lib/libmpi rather than your LAM libmpi. 
I'm not  
enough of an OS X expert to chase this further.

If I manually compiled an MPI app with the result of
"mpicc -- 
showme" (using the full pathname to LAM's libmpi
instead of -lmpi),  
then LAM 7.1.4 worked fine for me on my Leopard MBP.

But as suggested, Open MPI is a better bet -- that's where
we're all  
spending our time these days (a slightly earlier version of
Open MPI  
is what ships in Leopard, but since Apple doesn't ship a
fortran  
compiler, fortran support was disabled).



On Mar 28, 2008, at 12:26 PM, Artur Tyliszczak wrote:
> Hi Greg,
>
> I had a lot of problems installing LAM on my MacBookPro
(Intel
> processor) under Leopard. Using Tiger everything was
fine but with
> Leopard......... Anyway, finally I managed to compile
(Intel fortran
> compiler) and install LAM 7.1.4, then I could also
compile my programs
> without any problems. BUT when I tried to start
whatever, even the
> simplest "Hello world" some strange errors
were reported. If you are
> interested you can try to find my thread on LAM forum.
>
> Instead of LAM I installed OpenMPI (this has nothing
common with Open
> MP transfer protocol) which uses MPI standard. Now
everything works
> perfectly.
>
> You can try to do the same.
>
> Regards,
> Artur.
>
> On 2008-03-28, at 17:07, Gregory John Orris wrote:
>
>> Artur,
>>
>> Without going into the gory details, the native
mpicc does not  
>> include
>> a fortran compiler. I thought I'd install LAM,
since I used it under
>> Tiger and was more than satisfied. The compiling of
the software was
>> done to ensure that the correct set of compilers.
Sorry but, if it
>> were that easy I'd not sent out the query. I did
notice that several
>> years ago a similar problem was found on the
Itaniums. Wondering if
>> the problems are related since they seem to be
internal to LAM?
>>
>> Regards,
>> Greg
>>
>> On Mar 28, 2008, at 4:41 AM, Artur Tyliszczak
wrote:
>>
>>> Hi,
>>>
>>> Leopard has its own mpirun.
>>> Please check which one you are actually using.
That one from LAM or
>>> from the system.
>>> How did you compile 'hello' program.
>>>
>>> Regards,
>>> Artur Tyliszczak.
>>>
>>> On 2008-03-27, at 21:42, Gregory John Orris
wrote:
>>>
>>>> I just re-installed LAM on my Leopard
cluster and after a lot of
>>>> difficulties in the configure and compile
I've reached the  
>>>> following
>>>> error
>>>>
>>>> [mymachine:lam-7.1.4/examples/hello]
myname% mpirun -np 4 hello
>>>> mpirun (set_stdio): Bad file descriptor
>>>>
>>>> Any hints?
>>>>
>>>> Laminfo output
>>>>
>>>>           LAM/MPI: 7.1.4
>>>>            Prefix: /opt/local
>>>>      Architecture: i386-apple-darwin9.2.0
>>>>     Configured by: myname
>>>>     Configured on: Thu Mar 27 14:38:43 EDT
2008
>>>>    Configure host: mymachine
>>>>    Memory manager: darwin7malloc
>>>>        C bindings: yes
>>>>      C++ bindings: yes
>>>>  Fortran bindings: yes
>>>>        C compiler: /usr/local/bin/gcc
>>>>      C++ compiler: /usr/local/bin/g++
>>>>  Fortran compiler: /usr/local/bin/gfortran
>>>>   Fortran symbols: underscore
>>>>       C profiling: yes
>>>>     C++ profiling: yes
>>>> Fortran profiling: yes
>>>>    C++ exceptions: no
>>>>    Thread support: yes
>>>>     ROMIO support: yes
>>>>      IMPI support: no
>>>>     Debug support: no
>>>>      Purify clean: no
>>>>          SSI boot: globus (API v1.1, Module
v0.6)
>>>>          SSI boot: rsh (API v1.1, Module
v1.1)
>>>>          SSI boot: slurm (API v1.1, Module
v1.0)
>>>>          SSI coll: lam_basic (API v1.1,
Module v7.1)
>>>>          SSI coll: shmem (API v1.1, Module
v1.0)
>>>>          SSI coll: smp (API v1.1, Module
v1.2)
>>>>           SSI rpi: crtcp (API v1.1, Module
v1.1)
>>>>           SSI rpi: lamd (API v1.0, Module
v7.1)
>>>>           SSI rpi: sysv (API v1.0, Module
v7.1)
>>>>           SSI rpi: tcp (API v1.0, Module
v7.1)
>>>>           SSI rpi: usysv (API v1.0, Module
v7.1)
>>>>            SSI cr: self (API v1.0, Module
v1.0)
>>>>
>>>>
>>>>
>>>>
_______________________________________________
>>>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>>
>>>
_______________________________________________
>>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>>
>>
>> _______________________________________________
>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/


-- 
Jeff Squyres
Cisco Systems

_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

On Mar 28, 2008, at 12:08 PM, Jeff Squyres wrote:
> FWIW, I have found that the -lmpi used by LAM's wrapper
compilers may
> actually find /usr/lib/libmpi rather than your LAM
libmpi.  I'm not
> enough of an OS X expert to chase this further.
>
> If I manually compiled an MPI app with the result of
"mpicc --
> showme" (using the full pathname to LAM's libmpi
instead of -lmpi),
> then LAM 7.1.4 worked fine for me on my Leopard MBP.

By default, OS X's linker will look through the entire
library search  
path for a dynamic library, then look through the entire
search path  
again for a matching static library.  Since LAM/MPI builds
static  
libraries by default (libmpi.a) and Open MPI shipped with OS
X builds  
with dynamic libraries (libmpi.dylib), Open MPI's
libmpi.dylib will  
always be preferred to LAM/MPI's.

There's an option to OS X's linker, -search_paths_first,
that restores  
the more common search a directory for both dynamic and
shared  
libraries then move on to the next directory search order. 
Due to the  
confusion being caused on Leopard, I'm going to have the
wrapper  
compilers automatically add this option on OS X 10.5 and
above.

More information coming on the problem that started this
thread, but  
need to do some more checking before I can definitely say
it's solved.

Brian

-- 
   Brian Barrett
   LAM/MPI Developer
   Make today a LAM/MPI day!


_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/

On Mar 27, 2008, at 2:42 PM, Gregory John Orris wrote:
> I just re-installed LAM on my Leopard cluster and after
a lot of
> difficulties in the configure and compile I've reached
the following
> error
>
> [mymachine:lam-7.1.4/examples/hello] myname% mpirun -np
4 hello
> mpirun (set_stdio): Bad file descriptor
>
> Any hints?


You're unfortunately running into a bug in the most
unportable part of  
LAM/MPI, our file descriptor passing for standard input. 
There are a  
couple of standard ways to do file descriptor passing and
for the most  
part, Apple supports the more recent way.  But one of their
macros  
(CMSG_DATA) returns a bogus value on 64 bit Intel builds on
Leopard  
(but on the other three platforms, the macro works fine and
all four  
platforms work fine on Tiger).  They have a tech note on
file  
descriptor passing changes between Tiger and Leopard and
from that I  
was able to figure out how they wanted structures laid out
and fix the  
issue.

I've posted a beta tarball of LAM/MPI 7.1.5b1, which
includes this fix  
as well as some other changes including the work-around for
the  
libmpi.dylib / libmpi.a problem Jeff spoke of in a previous
e-mail.   
It can be downloaded from :

   http://www.lam-mpi.org/b
eta/

With 7.1.5b1, I was able to build in 64-bit mode using
Apple's GCC and  
hpc.sf.net's gfortran without any problems.  Please let me
know if you  
have any issues.


Brian

-- 
   Brian Barrett
   LAM/MPI Developer
   Make today a LAM/MPI day!


_______________________________________________
This list is archived at http://www.l
am-mpi.org/MailArchives/lam/