I had a go at developing a converter that creates ascii or binary vtu files from ascii frd files. The converter uses vtu11 (https://github.com/phmkopp/vtu11) for writing the vtu file. The frd file is mapped (https://github.com/mandreyel/mio) to allow fast access to its contents.
Please feel free to test the converter with your output files and I would appreciate any feedback or notes on any errors with the conversion. For a 3GB frd file the conversion time is 125seconds (approx 2 minutes) on my machine (Intel(R) Xeon(R) Silver 4210 CPU @ 2.20GHz 2.19 GHz (2 processors) - 128GB RAM)
Many thanks @SergioP . I have recompiled using Visual Studio only without the Intel Compiler; it is a bit slower but still works.
I did think about sharing the code but I need to spend some time cleaning it up…many test routines and other bits of code yet to be deleted.
Hi Jason,
Your converter could be a killer for my ccx2paraview. I hate you
Please, share the code for Linux users. Let me know, if you would like to host your project here: CALCULIX · GitHub
ccx2paraview is actually one of the main tools in my numerical analysis workflow/toolkit along with PrePomax; so thank you for developing and sharing ccx2paraview.
The base for the converter is a rough C++ conversion of the frd reader from PrePomax and I would need some time to clean up the code before sharing. I got it up and running last weekend for my test cases and wanted some feedback from others with there models before trimming it down.
May I ask why you need the mapping? As far as I understand memory mapping is useful for random access on large files. Converting a file should be possible in a “read once from beginning to end while processing the data on-the-fly” fashion. Or do I miss something? I am particularly unfamiliar with VTU, but I do write converters of FE I/O files with focus on performance, so I am really curious about your decision to map the file.
I have a bad habit of trying to peek at my frd files while CalculiX is still running and have at times received a message that a write operation to the frd file could not be performed. With that in mind, I wanted something to grab and create a quick snapshot of the frd file so that the converter can work with it just in case CalculiX may need write access to the file; this was preferred rather than using std::getline to read the files contents line by line, particularly for very large files.
After the mapping, I use std::strtok_r to get the lines of text from the map; this is very similar to using std::getline. It is still a work in progress so I can compile another version using std::getline to see if there is any changes in the total conversion time.
