LAM: mpicc does not work where gcc work

Hello List,

I am writing a program involving a couple of libraries,
and as LAM is among them I compile the all with mpicc.
I get a trouble that I can reproduce on a single example C
source.

I can compile this C source with both gcc and mpicc :
but when I run the executable obtained with mpicc
I get a `Floating point exception' .

The example C source is the `examples/sphere.c'
distributed with the `gts' source tarball.

Any idea ?

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

Hello List,

I think I have isolated my problem further:
my current guess is that it is related to the use or not of
exception handling:
how can we manage it quickly ?
is ther any workaround ?

Thanks in advance,
Jerome

Jerome BENOIT wrote:
> Hello List,
> 
> I am writing a program involving a couple of libraries,
> and as LAM is among them I compile the all with mpicc.
> I get a trouble that I can reproduce on a single
example C source.
> 
> I can compile this C source with both gcc and mpicc :
> but when I run the executable obtained with mpicc
> I get a `Floating point exception' .
> 
> The example C source is the `examples/sphere.c'
> distributed with the `gts' source tarball.
> 
> Any idea ?
> 
> Thanks,
> Jerome
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
> 

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

Please see section 4.3.3 of the LAM/MPI Installation Guide
-- "C++  
and Exceptions" (I assume that you are working with LAM
v7.1.2 -- you  
did not specify).

If this is not the problem that you are running in to,
please provide  
all the information specified on http://www.lam
-mpi.org/using/support/.




On Dec 16, 2006, at 8:40 AM, Jerome BENOIT wrote:

> Hello List,
>
> I think I have isolated my problem further:
> my current guess is that it is related to the use or
not of  
> exception handling:
> how can we manage it quickly ?
> is ther any workaround ?
>
> Thanks in advance,
> Jerome
>
> Jerome BENOIT wrote:
>> Hello List,
>>
>> I am writing a program involving a couple of
libraries,
>> and as LAM is among them I compile the all with
mpicc.
>> I get a trouble that I can reproduce on a single
example C source.
>>
>> I can compile this C source with both gcc and mpicc
:
>> but when I run the executable obtained with mpicc
>> I get a `Floating point exception' .
>>
>> The example C source is the `examples/sphere.c'
>> distributed with the `gts' source tarball.
>>
>> Any idea ?
>>
>> Thanks,
>> Jerome
>> _______________________________________________
>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>
>
> -- 
> Jerome BENOIT
> jgmbenoit_at_mailsnare_dot_net
> _______________________________________________
> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/


-- 
Jeff Squyres
Server Virtualization Business Unit
Cisco Systems

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

Hello,

thanks for your reply.

Jeff Squyres wrote:
> Please see section 4.3.3 of the LAM/MPI Installation
Guide -- "C++ and 
> Exceptions" (I assume that you are working with
LAM v7.1.2 -- you did 
> not specify).

Indeed, I am working with LAM v7.1.2 on a Etch Debian box
(intel Pentium M):
my LAM is the one distributed by Debian.

> 
> If this is not the problem that you are running in to,
please provide 
> all the information specified on http://www.lam
-mpi.org/using/support/.

My trouble can be reproduce with the sphere C source file
cited below.
If I do

gcc -o sphere sphere.c $(pkg-configure gts --cflags --libs)
-lm 

then I obtain an executable that works.

Bu, if I compile with mpicc instead:

mpicc -o sphere sphere.c $(pkg-configure gts --cflags
--libs) -lm

then I get an executable which give me only `Floating point
exception'.
I tried with both the options -fexception  and
-fno-exception but
I get the same error message.
I used the GTS (gts.sourceforge.net) packages distributed by
Debian.

Just before to receive you email,
my plan was to make a static linking rather that a dynamic
one:
may this resolve the issue.

Thanks in advance,
Jerome

> 
> 
> 
> 
> On Dec 16, 2006, at 8:40 AM, Jerome BENOIT wrote:
> 
>> Hello List,
>>
>> I think I have isolated my problem further:
>> my current guess is that it is related to the use
or not of exception 
>> handling:
>> how can we manage it quickly ?
>> is ther any workaround ?
>>
>> Thanks in advance,
>> Jerome
>>
>> Jerome BENOIT wrote:
>>> Hello List,
>>>
>>> I am writing a program involving a couple of
libraries,
>>> and as LAM is among them I compile the all with
mpicc.
>>> I get a trouble that I can reproduce on a
single example C source.
>>>
>>> I can compile this C source with both gcc and
mpicc :
>>> but when I run the executable obtained with
mpicc
>>> I get a `Floating point exception' .
>>>
>>> The example C source is the `examples/sphere.c'
>>> distributed with the `gts' source tarball.
>>>
>>> Any idea ?
>>>
>>> Thanks,
>>> Jerome
>>> _______________________________________________
>>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
>>>
>>
>> --Jerome BENOIT
>> jgmbenoit_at_mailsnare_dot_net
>> _______________________________________________
>> This list is archived at http://www.l
am-mpi.org/MailArchives/lam/
> 
> 
> --Jeff Squyres
> Server Virtualization Business Unit
> Cisco Systems
> 
> 

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

On Dec 16, 2006, at 1:54 PM, Jerome BENOIT wrote:

> My trouble can be reproduce with the sphere C source
file cited below.
> If I do
>
> gcc -o sphere sphere.c $(pkg-configure gts --cflags
--libs) -lm
>
> then I obtain an executable that works.
>
> Bu, if I compile with mpicc instead:
>
> mpicc -o sphere sphere.c $(pkg-configure gts --cflags
--libs) -lm

I'm not quite sure why you would be linking a non-MPI
application  
with mpicc...?  (you stated that "gcc ..." would
product a working  
executable, so I'm not sure why you need mpicc to
compile/link it...?)

You should add "--showme" to the command line and
see what flags  
mpicc is adding to the overall command.  Perhaps something
there  
conflicts with whatever sphere or gts is doing...?

> then I get an executable which give me only `Floating
point  
> exception'.
> I tried with both the options -fexception  and
-fno-exception but
> I get the same error message.
> I used the GTS (gts.sourceforge.net) packages
distributed by Debian.

I am unfamiliar with these packages, and I don't know what
sphere.c  
is.  You should probably try to single-step through this
executable  
with a debugger (recompile with -g) and see where the
problem  
occurs.  Perhaps there is an inadvertent symbol clash
somewhere...?

> Just before to receive you email,
> my plan was to make a static linking rather that a
dynamic one:
> may this resolve the issue.

I somewhat doubt it.  LAM defaults to building static by
default.

-- 
Jeff Squyres
Server Virtualization Business Unit
Cisco Systems

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

Hello,

thanks for the reply.

Jeff Squyres wrote:
> On Dec 16, 2006, at 1:54 PM, Jerome BENOIT wrote:
> 
>> My trouble can be reproduce with the sphere C
source file cited below.
>> If I do
>>
>> gcc -o sphere sphere.c $(pkg-configure gts --cflags
--libs) -lm
>>
>> then I obtain an executable that works.
>>
>> Bu, if I compile with mpicc instead:
>>
>> mpicc -o sphere sphere.c $(pkg-configure gts
--cflags --libs) -lm
> 
> I'm not quite sure why you would be linking a non-MPI
application with 
> mpicc...?  (you stated that "gcc ..." would
product a working 
> executable, so I'm not sure why you need mpicc to
compile/link it...?)


The code in sphere.c is very similar to the one I use in my
own software
which is a MPI application. I noticed that I can reproduce
the same trouble
with C source file sphere.c . So something weird happens
outside my code:
at least this is a point. 



> 
> You should add "--showme" to the command line
and see what flags mpicc 
> is adding to the overall command.

I did it: when the lam libs are removed everything is fine. 

  Perhaps something there conflicts 
> with whatever sphere or gts is doing...?

I will look in this direction now.


> 
>> then I get an executable which give me only
`Floating point exception'.
>> I tried with both the options -fexception  and
-fno-exception but
>> I get the same error message.
>> I used the GTS (gts.sourceforge.net) packages
distributed by Debian.
> 
> I am unfamiliar with these packages, and I don't know
what sphere.c is.  
> You should probably try to single-step through this
executable with a 
> debugger (recompile with -g) and see where the problem
occurs.  Perhaps 
> there is an inadvertent symbol clash somewhere...?

I will check this.

Anyhow, I will let you know what I get.

Thanks again,
Jerome

> 
>> Just before to receive you email,
>> my plan was to make a static linking rather that a
dynamic one:
>> may this resolve the issue.
> 
> I somewhat doubt it.  LAM defaults to building static
by default.
> 
> --Jeff Squyres
> Server Virtualization Business Unit
> Cisco Systems
> 
> 

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

Dear all:

I installed lam-7.1.2 on our cluster. Each node of our
cluster has a 
Pentium D 945 3.4G duo-core CPU (em64t).
I installed lam to a NFS shared directory.

=============some command results=================
gbliuClusterServer:/cluster/soft/MPI/lamtests-7.1.2>
lamboot -v hf

LAM 7.1.2/MPI 2 C++/ROMIO - Indiana University

n-1<10074> ssi:boot:base:linear: booting n0
(ClusterServer)
n-1<10074> ssi:boot:base:linear: booting n1 (n23)
n-1<10074> ssi:boot:base:linear: booting n2 (n24)
n-1<10074> ssi:boot:base:linear: finished
gbliuClusterServer:/cluster/soft/MPI/lamtests-7.1.2>
cat hf
ClusterServer cpu=2
n23
n23
n24
n24
gbliuClusterServer:/cluster/soft/MPI/lamtests-7.1.2>
lamnodes
n0      ClusterServer.cluster.t02:2:origin,this_node
n1      n23.cluster.t02:2:
n2      n24.cluster.t02:2:
====================================================

It seems that lamboot was done correctly.
But when I use lamtests-7.1.2, problems occure.
Under the top dir of lamtests-7.1.2, configure and make goes
successfully.
And then I do "make -k check", it hangs up at the
first test and stops 
there.
the output is as follow:
-----------------------output----------------------
gbliuClusterServer:/cluster/soft/MPI/lamtests-7.1.2>
make -k check
Making check in reporting
make[1]: Entering directory
`/cluster/soft/MPI/lamtests-7.1.2/reporting'
make[1]: Nothing to be done for `check'.
make[1]: Leaving directory
`/cluster/soft/MPI/lamtests-7.1.2/reporting'
Making check in ccl
make[1]: Entering directory
`/cluster/soft/MPI/lamtests-7.1.2/ccl'
Making check in intercomm
make[2]: Entering directory
`/cluster/soft/MPI/lamtests-7.1.2/ccl/intercomm'
make  check-TESTS
make[3]: Entering directory
`/cluster/soft/MPI/lamtests-7.1.2/ccl/intercomm'
mpirun -x TEST -ssi cr none -s h C -ssi rpi crtcp 
/cluster/soft/MPI/lamtests-7.1.2/ccl/intercomm/./allgather_i
nter
MPI_Comm_accept: unclassified: Bad address (rank 0, comm 4)
Rank (0, MPI_COMM_WORLD): Call stack within LAM:
Rank (0, MPI_COMM_WORLD):  - MPI_Comm_accept()
Rank (0, MPI_COMM_WORLD):  - main()
                                          
---------------------------------------------------
After a long time, the output is still like this and the
usage of CPU is 0.
I use ctrl-C to cancel it and then do command
"lamnodes", but this time
lamnodes also hangs up, no output appears. Only after I do
lamboot again,
lamnodes becomes all right.
   I don't know what's the problem. Can someone help me?

                                   Yours sincerely
                                   Guibin Liu


====================================================
 laminfo
             LAM/MPI: 7.1.2
              Prefix: /cluster/lammpi-7.1.2
        Architecture: x86_64-unknown-linux-gnu
       Configured by: root
       Configured on: Wed Dec 20 00:54:19 CST 2006
      Configure host: ClusterServer
      Memory manager: ptmalloc2
          C bindings: yes
        C++ bindings: yes
    Fortran bindings: yes
          C compiler: gcc
        C++ compiler: g++
    Fortran compiler: ifort
     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/