Stream: t-compiler

Topic: creating MIR before type check


nikomatsakis (Nov 16 2018 at 19:41, on Zulip):

you can encode uninferred HIR into MIR without much effort

certainly not the MIR we have today; I suppose some kind of extended MIR, sure

nikomatsakis (Nov 16 2018 at 19:41, on Zulip):

I'm not sure the advantage of this

nikomatsakis (Nov 16 2018 at 19:43, on Zulip):

@eddyb so some of the things that MIR (today) needs that you don't know before type-check:

I've thought about having type check take as input HIR and generate as output HAIR, but this sounds like another route altogether. I'm not sure why it'd be better to this "MIR+" first? Is having the CFG buying you something?

nikomatsakis (Nov 16 2018 at 19:43, on Zulip):

(Context is this message from @eddyb during the steering meeting)

oli (Nov 16 2018 at 19:55, on Zulip):

I think one positive effect would be that inference runs more naturally on a dataflow graph

eddyb (Nov 17 2018 at 08:28, on Zulip):

you can do NLL inference at the same time, I would think

Last update: Nov 22 2019 at 04:30UTC