roystgnr

> so are we saying we can move this to fe.C and call that same method for all families (except for any specific checks like element type)? - hopefully, by...

That's not a huge difference but it is much more than I'd expected. Init might be 1% slower or that might be my imagination, but compute looks 15% slower? IMHO...

What does this do? `_new_resize` gets initialized to 0, then I don't see it ever being used again. `vec_dim`, `n_vector_vars`, and `n_scalar_vars` get calculated (somewhat confusingly, but that's an existing...

We default to v8.11 when we have HDF5 available, but I think it's important to keep v5.22 around for users without HDF5. The diff is much larger, but it looks...

IIRC you can use h5dump to get ASCII out of the ExodusII-HDF5 file and pipe that to ncgen to put it into an ExodusII-NetCDF3 file.

Double-checked with John; he's also happy to merge. Sorry I let this get buried before at least looking at the diff...

I'm not very familiar with that script (my apologies: libMesh builds on Windows aren't in CI or really supported, we just don't turn down third-party patches from people who've suffered...

PBR and RI are implemented as of 1254b554 - the combination of which appears to do exactly jack for performance. Going to take a crack at RV next, but only...

Honestly? IMHO the way to avoid doing derivative computations with DualNumber is "don't pass in a DualNumber". If someone's code is sufficiently data type independent that you set T=DualNumber in...

We're still spending 52% of runtime in new/delete. More informatively, we're seeing 6 new/delete calls per computeQpResidual() call in Meta. The dot product `x*x+y*y+z*z` would naively force construction of temporaries...