When HIR auto-refs a comparison operator, clean it up in MIR

[scottmcm]

Today, if you're comparing &&Ts, it ends up auto-reffing in HIR. So the MIR ends up calling PartialEq/Cmp with &&&T, and the MIR inliner can only get that down to &T: https://rust.godbolt.org/z/hje6jd4Yf.

So this adds an always-run pass to look at Calls in MIR with from_hir_call: false to just call the correct Partial{Eq,Cmp} implementation directly, even if it's debug and we're not running the inliner, to avoid needing to ever monomorphize a bunch of useless forwarding impls.

This hopes to avoid ever needing something like #108372 where we'd tell people to manually dereference the sides of their comparisons.

r? @ghost

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[scottmcm]

This could go either way, so let's see:
@bors try @rust-timer queue

[bors]

⌛ Trying commit 7b74759 with merge d8ac1e4bc289879ff692ef5472e1cebd5c3cc42b...

[JakobDegen]

This doesn't seem correct to me. Could you add something like this as a test:

struct S;

struct T;

impl PartialEq<&&&T> for &&S {
    fn eq(&self, _t: &&&&T) -> bool {
        todo!()
    }
}

fn main() {
    let s = S;
    let t = T;
    dbg!(&&s == &&&t);
}

That having been said, do we know why inlining doesn't cover this? I would very strongly prefer not having to add a special case for this. (I have more review comments but will leave those for later)

[bors]

☀️ Try build successful - checks-actions
Build commit: d8ac1e4bc289879ff692ef5472e1cebd5c3cc42b (d8ac1e4bc289879ff692ef5472e1cebd5c3cc42b)

[scottmcm]

Oh, right, that's coherent despite the blanket :(

Guess that means this approach just can't work.

[rust-timer]

Finished benchmarking commit (d8ac1e4bc289879ff692ef5472e1cebd5c3cc42b): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.4%, 0.4%]	1
Regressions ❌ (secondary)	1.3%	[1.3%, 1.3%]	1
Improvements ✅ (primary)	-1.2%	[-1.7%, -0.7%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.7%	[-1.7%, 0.4%]	3

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.1%	[1.1%, 1.1%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.8%	[-2.1%, -1.5%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.8%	[-2.1%, 1.1%]	3

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.6%	[-0.6%, -0.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.6%	[-0.6%, -0.6%]	1

[scottmcm]

Well, I like 4% on ripgrep opt-full, but this isn't nearly worth it. I'll try something else instead.