Matthew Knepley

https://journals.sagepub.com/doi/abs/10.1177/1094342012444795

I agree with you that a good empirical way to get at this is allocating a bunch of vectors. My thought was that slowdown could happen if memory filled up...

I do not know of anyone reporting such as thing. I think many people do this, but I do not have a specific example. I used to do this _forever_...

@amneetb The solver itself is a tiny amount of memory, so if something is going on it must be about work vectors or a matrix. Can you make the call...

@amneetb It is giving you the "resident set size" which is the best measurement we have: https://unix.stackexchange.com/questions/30940/getrusage-system-call-what-is-maximum-resident-set-size You are doing the same thing my tiny code is doing, so I...

The easiest way to get inf is dividing by zero. We have PetscIsInfOrNanReal() which can be used anywhere in the code to track down the first instance. We call it...

Did you valgrind this? Another easy way to get inf is from uninitialized data.

Note that PETSc can manage Triangle, TetGen, MMG, and ParMMG meshing as well.

This sounds very much like something we added in PETSc that has been used all over the place: https://petsc.org/main/docs/manualpages/DM/DMProjectFunctionLocal.html https://petsc.org/main/docs/manualpages/DM/DMProjectFieldLocal.html The idea is that you project something, like an analytic...

@amneetb Do you have a true low Mach test case? I have not been able to find one. Lots of people run the methane burner from Madja and Sethian, but...