FEAP User Forum

FEAP => Parallel FEAP => Topic started by: william.lnx on October 04, 2017, 12:38:14 PM

Title: Header field was corrupted in graph partitioning
Post by: william.lnx on October 04, 2017, 12:38:14 PM

Hello everyone,

Recently I am having the next error when I run the $FEAPHOME8_4/parfeap/feap from a server:

Mesh output:  Filename = Ifile_0001
  Number partition nodes =        66902  Number total nodes =      133803
  Start COORD: TIME =   153.539993   
  Start COORD: TIME =   153.539993   
  End   COORD: TIME =   153.963989   
  Start ELEMT: TIME =   153.963989               0
  End   ELEMT: TIME =   153.972000               0        1223
  Ghost  Stress Elements =         1101  Partn Stress Elements =          122
  Matrix Store: Diagonal =         707  Off-Diagonal =         889
 
  Mesh output:  Filename = Ifile_0002
  Number partition nodes =            0  Number total nodes =      133803
  Start COORD: TIME =   154.355988   
  Start COORD: TIME =   154.355988   
  End   COORD: TIME =   154.355988   
  Start ELEMT: TIME =   154.355988               0
  End   ELEMT: TIME =   154.360001               0           0
  Ghost  Stress Elements =            0  Partn Stress Elements =            0
  Matrix Store: Diagonal =           0  Off-Diagonal =           0
 
  Total Ghost Stress Elements =        1101
  Total Partn Stress Elements =         122
  Total all   Stress Elements =       87178
Header field was corrupted

I realized that the last partition is always having problems (no matter how many processors are used) .  I have checked the input files and they are OK since I have run them from another computer. Have any of you have the same problem?

Any suggestion will be appreciated.

William,

Title: Re: Header field was corrupted in graph partitioning
Post by: Prof. S. Govindjee on October 04, 2017, 02:04:13 PM

Seems to be a memory error. 
Is the graph unusual?  We did recently make one fix to the graph partitioning routine.  I wonder if you have the same problem.

Can you post the unpartitioned mesh?

Title: Re: Header field was corrupted in graph partitioning
Post by: william.lnx on October 04, 2017, 03:14:26 PM

Hello,

I can't share the mesh, but the same happens when using a simple mesh generated with the BLOCk command. I am using the command below to made the partitioning

BATCh
 GRAPh,,nproc
 OUTDomains
END

I am guessing that there might be some problems with my Metis installation. Usually, the "header field was corrupted" message appears when I made changes in the makefile.in file.

Regards.

Title: Re: Header field was corrupted in graph partitioning
Post by: FEAP_Admin on October 04, 2017, 07:47:29 PM

Ok. Then the best advice that can be offered is to use a debugger or put debugging prints into pmacr7.F to see where things go wrong.

If it happens with a basic block, then you have broken something.

We, btw, always use METIS from our PETSc installation.