Experiment with interpreted glue

[pcwalton]

We spend lots of time generating glue, even for stuff like cmp glue that is cold. We could emit a declarative description of the type and have a generic "cmp" function that uses that description.

[msullivan]

That seems like it will trade compile time for run time, which seems like not something we want to do? On the other hand, it might be a win because of cache effects... Hm.

[pcwalton]

Yeah, I think it's worth trying. I think this is what OCaml does at least. At the very least it should help -O0.

[graydon]

I believe this 'experiment' is successful, yes? Closing.