[Gmsh] gmsh & hdf5 files: just a reflexion

[paul.carrico at free.fr]

Hi All, 

I'm using gmsh to postprocess finite element results, but for huge
models (or if I want to postprocess several physical quantities), I've
ever been confronted to limitations (mainly memory issues). 

Recently I started to work with hdf5 libraries (or netcdf ones which is
based on the previous Hdf5 ones), and I'm wondering if gmsh would not
take advantages of such ones, either for the nodes list, elements
connectivity or nodedata and elementdata processing, through typically a
kind of "include/input" instruction"? another advantages are: 

- hdf5 is implicitly parallized, 

- hdf5 is a binary format (compressed or not), optimized for I/O access,


- only the data we want to process are loaded ... in other word, not
only the full data need to be placed first in memory 

-etc. 

Just a thought I wanted to share, while I've not idea about the
difficulties and the amount of work (and some people probably have had
such idea) ... gmsh remains for me a great tool 

Bye 

Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/gmsh/attachments/20180509/bd3fc83e/attachment.html>