Brian Ward

> At that point we might as well deprecate both _lp and _jacobian functions in favor of using _target functions for everything Maybe? It still seems useful from some perspective...

The FFT functions we use from Eigen are also in their unsupported module, so that seems like an okay place to start

Assigning to a variable in a Stan program will always evaluate the expression, so using these will be very difficult to do efficiently if just implemented as a normal function....

> I think that's true even if the variable being assigned to is a function argument. This is a bit more complicated. Built in functions can accept unevaluated templates and...

What would be the least intrusive way to enable this in a way that still lets users disable it? Most of our makefile logic just checks if STAN_THREADS is _defined_,...

One possible way I thought of is giving the opt-out a new name, and then we can do something like ```makefile ifndef STAN_SINGLETHREADED STAN_THREADS = 1 endif ``` near the...

I noticed RStan appears to unconditionally enable threading on Windows: https://github.com/stan-dev/rstan/blob/develop/StanHeaders/src/Makevars.win#L3

Here is a summary of what I know on Windows: Naively enabling STAN_THREADS while building with mingw's GCC leads to about a 2.5x slowdown on the [`schools-4` model](https://github.com/stan-dev/example-models/blob/master/bugs_examples/vol2/schools/schools-4.stan) (which has...

I believe we still need signatures in the serializer/deserializer in the Stan repo, no?

Need to think about the implementation a bit. Might be worth starting in stanc3 and seeing which is more natural: 1. The naming structure you suggest. The backend will need...