SALOME 9.5.0 - Geom + Mesh -> 2+ GB file, reopen error bad:alloc

Dear All,

as it stands, I am currently "fiddling" with SALOME Platform for a work related topic, trying to reproduce some literature CFD results at present.

SALOME Platform 9.5.0 - universal binaries on both Linux and Windows (unzip and run):

-> I create a geometry using the geometry module.

-> I create a mesh using the meshing module. (using netgen to be specific)

So far so good. I then save the file using the hdf format (and potentially export my mesh in unv if I wish to use it deirectly).

I discovered today that two of my files, each around 2.3GBlarge, do not permit me to re-open the mesh. I receive an error "std::bad_alloc" and loose any and all associated mesh data. This does NOT happen with smaller files. A 480MB file of the same geometry with a coarser mesh works without issues. This is also NOT a RAM issue as I encounter the error on Windows (less RAM) and on Linux (more RAM) where the mesh was originally created.

As it stands, I can recreate the mesh as it is a simple test case, but once we consider the real case, calculating the mesh will require more effort and time and always saving a mesh-free geometry with just  the configuration does not seem lie an enticing option, though may be a necessity...

The only clue I have is file size...

Any ideas? Comments? Suggestions?

Hello!

Could you attach a Python dump file of your study to check the issue. Invoke menu File > Dump Study to produce the file.

Best regards

Edward

Previously SMESH expert wrote:

Hello!

Could you attach a Python dump file of your study to check the issue. Invoke menu File > Dump Study to produce the file.

Best regards

Edward

Given that I am trying to reproduce some literature data with a basic geometry, I guess this should be fine. I removed my username in the file path, however it does not contain any special characters and I have the same result between Linux and Windows with very different file paths.

v001 -> open file, dump study -> empty file, thus not included

v002 -> open file, go to geometry module (no issues so far), dump study

v003 -> carry on to mesh module (mesh has disappeared and does not show...), dump study

v004 -> open file, go to mesh module directly, see error, OK it then dump study

On the plus side, I learned how I can extract the parametrisation should this ever happen in the future - good to know.

OK, I encountered this issue again and colleagues confirmed they have equally seen it in the past.

The solution is to enable the multi-file save in SALOME Platform from the settings.
This then saves a mesh as a dedicated .med file and it is possible to store for example very large CFD meshes and avoid the limitations related to a single hdf file.