Ablity to use .pvtu files for parallel postprocessing with ParaView #210
Conversation
plugin/mpi/common_hpddm.hpp
Outdated
| pvd << " </Collection>\n"; | ||
| pvd << "</VTKFile>\n"; | ||
| } | ||
| *name = file_without_extension + "_" + (size > 1 ? str[2].str() + "_" : "") + str[0].str() + (size > 1 ? "_" + /*str[1].str()*/to_string(rank) : "") + "." + extension; |
There was a problem hiding this comment.
Is there a specific reason why you replaced str[1].str() by to_string(rank)?
I understand this makes the ProcZero definition in plugin/seq/iovtk.cpp easier when mpiSize > 1, but it makes it harder IMHO to browse through files. Is this a ParaView requirement? Otherwise, I'd be happy to revert just this change, and "fix" PvtuWriter and PvdWriter to handle a rank 0 whose tailing string is, e.g., with 16 processes, _00.xyz (instead of _0.xyz).
There was a problem hiding this comment.
Thank you for reviewing the code.
The reason to use to_string(rank) over str[1].str() was indeed to make it easier to identify the rank 0 in plugin/seq/iovtk.cpp. And also to avoid the reoccurring string assembly via ostringstream within PvtuWriter, since this has to be performed in a for loop of the size mpiSize by a single proc (mpirank=0 which is incharge of writing the pvtu file). There is no ParaView requirement attached to this choice.
If you prefer the str[1].str(), which I agree makes the file browsing easier, we can definitely make changes in PvtuWriter and PvdWriter to make it work. Do you want me to make these changes, and push them (I already tried something locally and they seem to work)?
There was a problem hiding this comment.
Sure, feel free to push whatever you want. Be careful though, I force-pushed in your branch to split your initial commit in two. If it's too much trouble for your, you can just send me the patch w.r.t. your local branch.
There was a problem hiding this comment.
I have made the necessary changes here.
No description provided.