Greetings to all!

Tobi sent me an email earlier today related to this and I might as well leave a public note as well, to share with everyone the suggestions I had... so this is a quick copy-paste-adapt for future reference, until I or anyone else bothers with writing this at openfoamwiki.net

I have no idea yet for the current generation of Ryzen CPUs (Ryzen 3000 series), but I do know of this report for EPYC: http://www.prace-ri.eu/best-practice-guide-amd-epyc
If you look for Table 5, you will see the options they suggest for GCC/G++.

However, the "znver1" architecture is possibly not the best for this generation of Ryzen/Threadripper... there is an alternative, which is to use:
It will only work properly with a recent GCC version for the more recent CPUs.

Beyond this, it might take some trial and error. Some guidelines are given here: https://wiki.gentoo.org/wiki/GCC_optimization

You can use the following strategy to test various builds with different optimization flags:

1. Go into the folder "wmake/rules/linux64Gcc"
2. Copy the files "cOpt" and "c++Opt" to another name, for example: "cOptNative" and "c++OptNative"
3. "cOPT" and "c++OPT" are the lines that need to be updated.
4. Create a new alias in your ".bashrc" file for this, for example:
   
   Code:

   alias ofdevNative='source $HOME/OpenFOAM/OpenFOAM-dev/etc/bashrc WM_COMPILE_OPTION=OptNative'

5. Start a new terminal and activate this alias ofdevNative.
6. Then run ./Allwmake inside "OpenFOAM-dev".
7. Repeat the same strategy for other names and therefore you can do several builds with small changes to the optimization flags.

Warning: Last time I checked, AVX and AVX2 are not used by OpenFOAM, so don't bother with them.

Best regards,
Bruno