3 ms·
1.2 is great (the nascent MSVC support is especially promising), but I'm most excited by the ongoing work in nightly to completely overhaul the compiler middle-
by kibwen 11y ago
1.2 is great (the nascent MSVC support is especially promising), but I'm most excited by the ongoing work in nightly to completely overhaul the compiler middle-end. The high-level overview is that the middle-end is being refactored to operate on its own intermediate representation instead of the raw AST, which should open the door to a whole host of improvements in compiler speed and code generation. It's technical debt whose bill is long overdue. :)
- ewillbefull 11y agoWill the MIR/HIR stuff also help with adding features? I'm thinking of things like non-lexical scoping/SEME regions/etc. which might be easier to implement with it.
- steveklabnik 11y agoWhen I started reading your comment, I was ready to respond with "yes, one example is non-lexical lifetimes", sooo :) Another is compiler plugins: it's much easier to stabilize an MIR than it is to stabilize the compiler's internal functions.
- pcwalton 11y agoYes, MIR will help with non-lexical borrows.
- dbaupp 11y agoYep, the MIR allows the compiler to understand control flow more accurately and more precisely. Non-lexical lifetimes in particular are one of the big features that the MIR should help enable.
- grayrest 11y agoDoes it open the door to being able to incrementally update the representation and have the compiler used as a library for IDE support?
- steveklabnik 11y agoYup!