The LAPACK forum has moved to https://github.com/Reference-LAPACK/lapack/discussions.

ScaLAPACK test case failure

Open discussion regarding features, bugs, issues, vendors, etc.

ScaLAPACK test case failure

Postby user_avg » Fri Aug 17, 2007 9:44 am

I am trying to test ScaLAPACK. However, one of the executable 'xzsep' is failing. I am getting the following error:

job aborted:
rank: node: exit code: message
0: optron: terminated
1: optron: fatal error: Fatal error in MPI_Bcast: Message truncated, error stack
:
MPI_Bcast(791)......................: MPI_Bcast(buf=0x00000000005FE170, count=1,
dtype=USER<vector>, root=0, comm=0x84000005) failed
MPIR_Bcast(192).....................:
MPIC_Recv(80).......................:
MPIDI_CH3U_Request_unpack_uebuf(529): Message truncated; 44 bytes received but buffer size is 24
2: optron: terminated
3: optron: terminated

It says the buffer size is 24, but 44 bytes received. It seems some buffer related thing but I cant get what the exact problem is and how to handle that? Appreciate any help...
user_avg
 
Posts: 3
Joined: Mon Jun 11, 2007 9:16 am

Postby Julien Langou » Fri Aug 17, 2007 10:23 am

Oups that does look not good.

1- Is it the only executable that fails? Can you try others? What about the xdsep, zcsep and
zssep?

2- what MPI are you using?

3- can we get the output file to check which routine/example in xzsep makes the code fail.

Julien.
Julien Langou
 
Posts: 835
Joined: Thu Dec 09, 2004 12:32 pm
Location: Denver, CO, USA

Postby user_avg » Mon Aug 20, 2007 2:21 am

Hi Julien,

1) Yes, 'xzsep' is the only executable which is failing. Tried the ones mentioned by you, they pass all the tests.

2) I am using MSMPI with PGI 64 bit compilers on Windows. Currently I am testing on a single machine and not on a cluster

3) The output that I posted in the previous post is what I get after I run the following command: 'mpiexec -n 4 xzsep.exe'

thanks again,
Abhishek
user_avg
 
Posts: 3
Joined: Mon Jun 11, 2007 9:16 am

Postby Julien Langou » Mon Aug 20, 2007 9:39 am

Hello,
thanks for the information. Does the test leave any output?
We have access to one Windows cluster so we might want to retry
the test. It was working fine on our machine as far as I remember.
We take you inform. A priori you can use ScaLAPACK for a bunch of
things also there is something weird going the Symmetric Eigenvalue
Problem.
Julien.
Julien Langou
 
Posts: 835
Joined: Thu Dec 09, 2004 12:32 pm
Location: Denver, CO, USA

Postby user_avg » Mon Aug 20, 2007 9:58 am

Hi julien

No, the xzsep.exe doesnt leave any output except the above MPI fatal error message. Please note that I am building using PGI C and Fortran 64 bit compilers and on native Windows. As it was working on your side will it be possible for you to share the makefiles?

thanks,
Abhishek
user_avg
 
Posts: 3
Joined: Mon Jun 11, 2007 9:16 am


Return to User Discussion

Who is online

Users browsing this forum: No registered users and 4 guests