VOD: https://www.youtube.com/watch?v=9EN1jA2Q57c
Agenda. After the previous stream I realised that
optimisations should work as functions over an AST (not a Lean
Expr!). Doing optimisations functions over an AST allows
you to prove things a lot easier, but requires a lot more work in the
form of denotational semantics.
I promised myself I would never go anywhere near those things, but it absolutely makes sense if you want to transpile these verified optimisations to other programs (for example, export optimisations to C for someone’s optimiser).
reassociate.