Stream: t-compiler/wg-mir-opt

Topic: mir validation enhancements


oli (Jul 14 2020 at 08:02, on Zulip):

choco said:

Hello everyone, I also came from the tweet that oli made, so if there's something to gets my hands dirty I'm down!

@choco I have something that isn't actually an optimization, but validates that our optimizations don't generate broken code. The work for this is exactly the same as one does for optimizations, so I think it's a good starter. https://github.com/rust-lang/rust/issues/73832

oli (Jul 14 2020 at 08:03, on Zulip):

I wrote some quickstart instructions in https://rust-lang.zulipchat.com/#narrow/stream/189540-t-compiler.2Fwg-mir-opt/topic/multiple.20return.20terminators/near/203699347

oli (Jul 14 2020 at 08:04, on Zulip):

the part of creating a test does not apply to you, you just rerun the command after your changes

Jonas Schievink (Jul 14 2020 at 08:15, on Zulip):

Is libcore building with validation? I thought it was broken.

oli (Jul 14 2020 at 08:18, on Zulip):

oh... maybe then that's not a good issue to start with

oli (Jul 14 2020 at 08:22, on Zulip):

Ok, different thing. I have something: https://github.com/rust-lang/rust/issues/65693 . If you want to take that issue, you'd be optimizing the MIR datastructures, making MIR optimizations less complex, instead of optimizing instances of MIR.

oli (Jul 14 2020 at 08:22, on Zulip):

Again, no custom issue is requires, and no change in test output expected

oli (Jul 14 2020 at 08:23, on Zulip):

I'm gonna throw some more documentation on that issue

oli (Jul 14 2020 at 08:31, on Zulip):

let me know if you run into any situations that require non-trivial code changes

choco (Jul 14 2020 at 11:26, on Zulip):

Thanks! I will start looking at this today after work :)

Last update: Sep 28 2020 at 15:15UTC