Converting mesh format with --mesh-only from .inp to .e on supercomputer

[Tim Hsu]

I'm wondering if anybody has had experience with Abaqus .inp meshes, which I use for my simulations.

The .inp meshes actually work fine for my purpose. But I've begun to look into splitting .inp meshes in order to avoid memory issues from large meshes. And it seems splitter-opt only works for .e meshes, so I want to convert .inp meshes to .e format using --mesh-only.

On my linux workstation the conversion works. But somehow on the supercomputer that I use, I get the following message:

ERROR: Unrecognized file extension: Mesh/file=meshes/micro/msri_cat/russiandoll/s1/MSRI_CAT_box1_shrink1.inp
   I understand the following:

     *.dat   -- Tecplot ASCII file
     *.e     -- Sandia's ExodusII format
     *.exd   -- Sandia's ExodusII format
     *.fro   -- ACDL's surface triangulation file
     *.gmv   -- LANL's GMV (General Mesh Viewer) format
     *.mesh  -- MEdit mesh format
     *.mgf   -- MGF binary mesh format
     *.msh   -- GMSH ASCII file
     *.n     -- Sandia's Nemesis format
     *.nem   -- Sandia's Nemesis format
     *.plt   -- Tecplot binary file
     *.poly  -- TetGen ASCII file
     *.ucd   -- AVS's ASCII UCD format
     *.ugrid -- Kelly's DIVA ASCII format
     *.unv   -- I-deas Universal format
     *.vtu   -- VTK (paraview-readable) format
     *.xda   -- libMesh ASCII format
     *.xdr   -- libMesh binary format,


 Exiting without writing output

 -----------------------------------------------------------------------------------------------------
| Time:           Fri Nov 10 15:26:40 2017                                                            |
| OS:             Linux                                                                               |
| HostName:       n256                                                                                |
| OS Release:     3.18.44                                                                             |
| OS Version:     #1 SMP Sun Nov 13 05:15:33 EST 2016                                                 |
| Machine:        x86_64                                                                              |
| Username:       thsu                                                                                |
| Configuration:  ../configure  'INSTALL=/nfs/home/6/thsu/myprojects/moose/scripts/../libmesh/build-aux/install-sh -C'|
|  '--with-methods=opt oprof dbg'                                                                     |
|  '--prefix=/nfs/home/6/thsu/myprojects/moose/scripts/../libmesh/installed'                          |
|  '--enable-silent-rules'                                                                            |
|  '--enable-unique-id'                                                                               |
|  '--disable-warnings'                                                                               |
|  '--enable-unique-ptr'                                                                              |
|  '--enable-openmp'                                                                                  |
|  '--disable-maintainer-mode'                                                                        |
|  '--enable-petsc-required'                                                                          |
|  'METHODS=opt oprof dbg'                                                                            |
|  'PETSC_DIR=/nfs/apps/PETSC/3.6.4/gnu/6.1.0/openmpi/1.10.2'                                         |
 -----------------------------------------------------------------------------------------------------
 ------------------------------------------------------------------------------------------------------------
| Ermine Performance: Alive time=344.209, Active time=338.629                                                |
 ------------------------------------------------------------------------------------------------------------
| Event                         nCalls     Total Time  Avg Time    Total Time  Avg Time    % of Active Time  |
|                                          w/o Sub     w/o Sub     With Sub    With Sub    w/o S    With S   |
|------------------------------------------------------------------------------------------------------------|
|                                                                                                            |
|                                                                                                            |
| Application                                                                                                |
|   Full Runtime                1          0.0000      0.000009    338.6285    338.628543  0.00     100.00   |
|                                                                                                            |
| Setup                                                                                                      |
|   Application Setup           1          164.7091    164.709076  338.6285    338.628527  48.64    100.00   |
|   Read Mesh                   1          173.9195    173.919451  173.9195    173.919451  51.36    51.36    |
 ------------------------------------------------------------------------------------------------------------
| Totals:                       3          338.6285                                        100.00            |
 ------------------------------------------------------------------------------------------------------------

This is quite weird, especially because without converting, my simulations work fine directly on top of .inp meshes. Has anybody run into similar problems before?

[Peterson, JW]

On Fri, Nov 10, 2017 at 2:18 PM, Tim Hsu <tim...@gmail.com> wrote:

I'm wondering if anybody has had experience with Abaqus .inp meshes, which I use for my simulations.

The .inp meshes actually work fine for my purpose. But I've begun to look into splitting .inp meshes in order to avoid memory issues from large meshes. And it seems splitter-opt only works for .e meshes, so I want to convert .inp meshes to .e format using --mesh-only.

On my linux workstation the conversion works. But somehow on the supercomputer that I use, I get the following message:

ERROR: Unrecognized file extension: Mesh/file=meshes/micro/msri_cat/russiandoll/s1/MSRI_CAT_box1_shrink1.inp
   I understand the following:

     *.dat   -- Tecplot ASCII file
     *.e     -- Sandia's ExodusII format
     *.exd   -- Sandia's ExodusII format
     *.fro   -- ACDL's surface triangulation file
     *.gmv   -- LANL's GMV (General Mesh Viewer) format
     *.mesh  -- MEdit mesh format
     *.mgf   -- MGF binary mesh format
     *.msh   -- GMSH ASCII file
     *.n     -- Sandia's Nemesis format
     *.nem   -- Sandia's Nemesis format
     *.plt   -- Tecplot binary file
     *.poly  -- TetGen ASCII file
     *.ucd   -- AVS's ASCII UCD format
     *.ugrid -- Kelly's DIVA ASCII format
     *.unv   -- I-deas Universal format
     *.vtu   -- VTK (paraview-readable) format
     *.xda   -- libMesh ASCII format
     *.xdr   -- libMesh binary format,

Are you using a much older version of MOOSE on your cluster?

The "I understand the following" message here does not look like the one in the version of libmesh used by MOOSE today, specifically a line about supporting the .inp extension should be there.

Your MOOSE would have to be very old, though, since Abaqus support was added to libmesh by commit 7960b9114, back in 2011...

--

John

[Tim Hsu]

The moose repository and libmesh are up-to-date (via git clone and update_and_rebuild_libmesh.sh). The moose installation on the cluster was handled by the admin around last year.

Hmm...

[Tim Hsu]

I figured out my problem with "--mesh-only". Perhaps I don't know how the syntax is supposed to work. But the conversion only works if I put the "--mesh-only" phrase at the very end of command line. An example would be:

./myApp-opt -i inputFile.i \

Mesh/file=someMesh.inp \

Outputs/exodus=true \

--mesh-only (this has to be at the very the end)

But the error message (mentioned earlier) not showing .inp support is a bit weird. Maybe the text part of the codes isn't updated?

[Cody Permann]

Interesting... So there is at least a few small problem here with our command line help: "--mesh-only" takes an optional output filename which isn't listed when you run your executable with the help flag. What's happening is your CLI override is being interpreted as the filename that would be used to write the output mesh. Thus, moving the --mesh-only command to the end of the line fixed your issue. 

I'm interested in improving the user experience and the error message you receive so can you send us the exact incorrect command line you attempted to run? I'm trying to replicate your exact failure mode but haven't been able to get that same message (with the .inp problem) so far. We did verify that the mesh formats you see on your screen are the OUTPUT formats (not input formats which is why we thought your libMesh was outdated). We do NOT write Abaqus format.

Things to improve:

1) libMesh's error message (it should say whether it's trying to read or write when it prints out that message)

2) MOOSE's CLI help message (needs to list the optional mesh filename)

3) Maybe better diagnostics (when we can help it) with complex command lines

Thanks for your assistance in helping us improve the software.

Cody

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/5a8938a4-c431-4609-8116-3e98a156f30d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Tim Hsu]

Sure thing. My command was something like this:

(under csh script for qsub)

set fname = "msri_cat_s2_1"

set fpath = "micro/msri_cat/russiandoll/s2"

## Run ERMINE 

#mpirun ermine-opt -i inputs/micro_cat.i --mesh-only \

#Mesh/file=meshes/${fpath}/${fname}.inp \

#Outputs/file_base=outputs/${fpath}/${fname} \

#Outputs/exodus=true

[Tim Hsu]

Uh. Didn't mean to comment out the command.

(under csh script for qsub)

set fname = "msri_cat_s2_1"

set fpath = "micro/msri_cat/russiandoll/s2"

## Run ERMINE 

mpirun ermine-opt -i inputs/micro_cat.i --mesh-only \

Mesh/file=meshes/${fpath}/${fname}.inp \

Outputs/file_base=outputs/${fpath}/${fname} \

Outputs/exodus=true