Tracking issue for generator code quality

[tmandry]

So far we haven't done many deep dives into the code LLVM produces for generators. I think there's a lot of room for improvement here, both in the compiler and in LLVM.

Issues (please add more as they come up):

• [codegen] unnecessary panicking branch in foo().await (vs equivalent FutureImpl.await) #71093 unnecessary panicking branch in foo().await (vs equivalent FutureImpl.await)

Some more ideas are in #71093 (comment)

Related tracking issues:

• Tracking Issue for Coroutine Memory Usage #69826 Generator Memory Usage
• Tracking issue for more precise coroutine captures #69663 More precise generator captures

[Mark-Simulacrum]

@tmandry It'd be good to have a handle on when we expect to close this (e.g., is this an 'active work' type issue or not)

[tmandry]

@tmandry It'd be good to have a handle on when we expect to close this (e.g., is this an 'active work' type issue or not)

It's more of a place to group issues that are related so we can better track work. The other two tracking issues I linked above are more examples of this.

I don't know if we can ever expect to be "done" with improving generator memory usage, for instance, but if all known problems are fixed (or not going to be) then we can close the issue.

EDIT: I have a few code quality-related issues I could add here, as things to investigate or future projects, I just haven't written them up yet.

[pnkfelix]

Visiting during T-compiler backlog bonanza. This isn't the same kind of "tracking issue" that we associated with the C-tracking-issue label (in terms of a specific feature being designed)

@rustbot label: +metabug -C-tracking-issue