G'day all.
Quoting Simon Peyton-Jones
Interesting thought. I think you're describing a possible extension to the SpecConstr transformation described in "Call pattern specialisation for Haskell"
Without reading the paper, that sounds right to me.
Here 'x' is free in 'g', but x's value is known at g's call sites. It's not enough just to know "x is a cons" inside g; we must also have access to z,zs.
I would have thought that GHC already optimised this in a separate pass. Doesn't it do some kind of redundant case test elimination?
And it's arguably a bad shortcoming that currently the mere act of lambda lifting can affect how effective SpecConstr is.
Indeed. It also suggests to me that perhaps lambda lifting the pattern matching "retries" might _always_ be a good idea, given that the programmer has no way of controlling inlining for this kind of intermediately-generated function. Cheers, Andrew Bromage