Denis-Gabriel Caprace

I have come to a better understanding of what happens here. It seems that the NaNs appear in the computation of the preconditioner of the (A)NK solver. Iteration 0 performs...

Thanks @sseraj for the quick reply. Among the tests I've run, I've come back to the tutorial wing based on the `test_solve.py`script in the regression test set. This runs without...

Thank you all for the comments. I noticed the absence of a call to other turbulence models in `blockResCore`: for the preconditioner, the routine is called in frozen turbulence mode...

I've made the changes to compute the residuals for the other turbulence models in the blockResCore, as you suggested @anilyil, plus a couple other places where it seemed to be...

We recommend the use of LLVM Clang to compile the binaries of ExaFMM. The manual procedure explained [here](https://nbviewer.jupyter.org/url/edoalvar2.groups.et.byu.net/LabNotebook/202008/FLOWVPMSetupFinal.ipynb) can be adapted as follows. An automated procedure is also under development...

Thanks for documenting this Eric! I made a commit on FLOWExaFMM so that the change at line 585-592 is built in. I also edited my message above so that the...

## What currently works The modified Makefile works and enables building Snopt.jl. Note that, in the case of dynamic linking, some dependencies may or may not be found at execution...

FYI: @tylercritchfield and I tried this on his MAC and we observed the same behavior.

The code in this branch is backward compatible, i.e. the default behavior is still to compile from the user-provided source code. So yes it could be merged, unless you want...