Interesting but probably out-of-scope idea: Use copy-and-patch compilation about microhs HOT 2 OPEN

I read the paper you linked to. That's a very interesting approach. I will keep that in mind if I ever decide to do a JIT. Thanks!

from microhs.

I have indeed thought of some kind of JIT, which would treat a function body as a super-combinator and generate code for it. But you can do substantially better than copying a body for each of the combinators, and instead generate code that avoids a lot of the overhead associated with combinators.
Indeed, the very first functional language compiler I wrote (in 1981) used template copying, where there was a template for each super-combinator. The code did a memcpy of the template and then patched the wholes. So I'm familiar with the idea. :)
We will see where things go. I have many things on my TODO list.

Nitpick: MicroHs compiles to a fixed set of combinators, so it's not really what traditionally is called supercombinators.

from microhs.