Yes. I was the one who originally filed the issue, I believe :-)
> ? Hopefully, they are implementing everything we need.
The initial plan was to only do a limited set of operators (like
scalar * matrix). If they cover everything, there are probably
some places we can reduce promotion. But for most of the matrix
operations, we do our own derivatives now and only call Eigen with
double values. We want to keep it that way if we ever need to
send things off to GPU or other parallelization through Eigen that
wouldn't work with var inputs.
- Bob