FEAP User Forum
FEAP => Parallel FEAP => Topic started by: george.kissas on March 02, 2017, 07:37:48 AM
-
Dear All,
I need help with the following problem. I am using the parallel FEAP v8.4 to solve a diffusion equation. I run the executable serially with directives
INIT DISPl
OUTMesh
GRAPh PARTiton 4
OUTDomains
in order to partition the mesh and everything works fine( i get Ithermal_0001,Ithermal_0002,Ithermal_0003,Ithermal_0004). Then I'm running parallel feap using "mpirun -n 4 $FEAPHOME8_4/parfeap/feap -ksp_type gmres -pc_type jacobi " but i get
F I N I T E E L E M E N T A N A L Y S I S P R O G R A M
FEAP (C) Regents of the University of California
All Rights Reserved.
VERSION: Release 8.4.1g
DATE: 07 March 2014
Files are set as: Status Filename
Input (read ) : Exists Ithermal
Output (write) : Exists Othermal
Restart (read ) : New Rthermal
Restart (write) : New Rthermal
Plots (write) : New Pthermal
Caution, existing write files will be overwritten.
Are filenames correct?( y or n; r = redefine all, s = stop) :
F I N I T E E L E M E N T A N A L Y S I S P R O G R A M
FEAP (C) Regents of the University of California
All Rights Reserved.
VERSION: Release 8.4.1g
DATE: 07 March 2014
Files are set as: Status Filename
Input (read ) : Exists Ithermal
Output (write) : Exists Othermal
Restart (read ) : New Rthermal
Restart (write) : New Rthermal
Plots (write) : New Pthermal
Caution, existing write files will be overwritten.
Are filenames correct?( y or n; r = redefine all, s = stop) :
*ERROR* FILNAM: Reinput data
F I N I T E E L E M E N T A N A L Y S I S P R O G R A M
FEAP (C) Regents of the University of California
All Rights Reserved.
VERSION: Release 8.4.1g
DATE: 07 March 2014
Files are set as: Status Filename
Input (read ) : Exists Ithermal
Output (write) : Exists Othermal
Restart (read ) : New Rthermal
Restart (write) : New Rthermal
Plots (write) : New Pthermal
Caution, existing write files will be overwritten.
Are filenames correct?( y or n; r = redefine all, s = stop) :
*ERROR* FILNAM: Reinput data
F I N I T E E L E M E N T A N A L Y S I S P R O G R A M
FEAP (C) Regents of the University of California
All Rights Reserved.
VERSION: Release 8.4.1g
DATE: 07 March 2014
Files are set as: Status Filename
Input (read ) : Exists Ithermal
Output (write) : Exists Othermal
Restart (read ) : New Rthermal
Restart (write) : New Rthermal
Plots (write) : New Pthermal
Caution, existing write files will be overwritten.
Are filenames correct?( y or n; r = redefine all, s = stop) :
*ERROR* FILNAM: Reinput data
I tried to redefine the input as Ithermal_0001 but the same problem occurs. I reinstalled FEAP and PETSc, but the problem remains.
Thank you,
GK
-
Running a parallel job requires the following steps.
1. You must have a 'flat file' to start with (a file with no extra nodes or elements, so no tie commands). If you have an input file with things like tie commands. You can generate a flat file using the OUTMesh command. I think the name will be Ithermal.rev. If you input file is already flat, then you can skip this step.
2. Run (serially) the version of FEAP in the parfeap directory using your flat file as input. Execute the GRAPh,NODE,4 command followed by OUTDomains. This will create for you 4 parallel input files Ithermal_0001, Itermal_0002, etc.
3. Now run your problem in parallel using the version of FEAP in the parfeap directory. For example mpirun -np 4 feap -ksp_type cg -pc_type jacobi -ksp_monitor -options_left . When prompted for the input file use Ithermal_0001. Note only the first process will ask for the input file name, the other processes get their file name from the first process.
-
Dear Admin,
Thank you for your reply. I have done everything you mentioned and the problem remains. The problem is that FEAP solves only one partition serially( i get only the output of one partition).I post the content of Ithermal and solve.thermal batch command to take a look.
Ithermal
BATCh
INIT DISPl
GRAPh NODE 4
OUTDomains
END
solve.thermal
BATCh
PETSc ON
INIT DISPl
DT,,0.010
TRANs,BACKward
LOOP,,100
TIME
ITER TOL 1.D-07 1.D-08 1.D+20
TANG,,1
DISPL ALL
NEXT TIME
END
Can it be a PETSc problem?
Thank you,
GK
-
It is always a possibility that it is a PETSC problem. Did you run the installation tests with PETSc?
The other thing to do is to look at the Ithermal_000X files to make sure they look reasonable.
If the unpartitioned and partitioned files are not too large, you could try posting them.
-
Dear Professor Govindjee,
Thank you for your reply. Unfortunately the files are too large, so i cannot upload them. I will try to verify that PETSc is working fine and if the problem remains, I will get back to you.
Thank you,
GK
-
Dear Professor Govindjee,
One more question.Which version of PETSc is compatible with FEAP v8.4?
Thank you,
GK
-
According to the comments in parfeap/usolve.F the 8.4 release version works with Petsc 3.3.x. I believe it also works with version 3.4.x.
When 3.5.x came out they made changes to the API which require editing the call to KSPSetOperators. With the release of 3.7.x additional changes are required like changing finclude to petsc/finclude and updating the calls to PETSCOptionsXXX.