Add mir-opt tests to track MIR quality.

[cjgillot]

cc @scottmcm @saethlin

If you have other ideas, please say so.

[rustbot]

r? @Mark-Simulacrum

(rustbot has picked a reviewer for you, use r? to override)

[bors]

☔ The latest upstream changes (presumably #107404) made this pull request unmergeable. Please resolve the merge conflicts.

[saethlin]

👍 I think this is a much better way to factor the MIR tests. We should clearly delineate the point of the various MIR tests we have, this has been a frequent point of confusion for me personally, and this is improving that.

---

-O makes me uncomfortable, because I strongly suspect most people do not use that build configuration, and just opt for Cargo's --release which by default is -Copt-level=3 not -Copt-level=2. Currently this has no impact on MIR opts, thought Inline does mention the enum:

rust/compiler/rustc_mir_transform/src/inline.rs

Lines 53 to 54 in 3462f79

	(sess.opts.optimize == OptLevel::Default
	|| sess.opts.optimize == OptLevel::Aggressive)

The fact that all the codegen tests use -O bothers me for the same reason, so on balance I think there is some space for saying let's use -O for now and if we want to use -Copt-level=3 swap all our optimization tests to that at once.

---

These tests are uncommonly fragile because they contain line and column numbers from the standard library implementation. If we start to accumulate tests along these lines and don't normalize or erase the line/column information from the tests we can expect any change to library/ to break these tests.

I'm actually not sure if the fact that we currently have line/column info is a bug, because looking carefully over the diff here, I see this:

        scope 4 (inlined iter::range::<impl Iterator for std::ops::Range<u32>>::next) { // at $DIR/range_iter.rs:21:14: 21:24
            debug self => _8;            // in scope 4 at $SRC_DIR/core/src/iter/range.rs:LL:COL

so it looks like something is trying to help out here, but mostly it isn't.

[scottmcm]

This looks good to me.

It would be nice to have an option to also scrub the inlined locations -- especially for things from other crates -- but that's an existing thing that doesn't need to block adding and moving tests here. (And if they're too painful, we can always remove tests.)

r? @scottmcm
@bors r+ rollup=iffy (moves enough tests that conflicts seem likely)

[bors]

📌 Commit 332b7f5 has been approved by scottmcm

It is now in the queue for this repository.

[bors]

⌛ Testing commit 332b7f5 with merge 253b727...

[bors]

☀️ Test successful - checks-actions
Approved by: scottmcm
Pushing 253b727 to master...

[rust-timer]

Finished benchmarking commit (253b727): comparison URL.

Overall result: ❌ regressions - no action needed

@rustbot label: -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)	0.6%	[0.6%, 0.6%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.6%	[0.6%, 0.6%]	1

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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.6%	[-3.6%, -3.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-3.6%	[-3.6%, -3.6%]	1

Cycles

This benchmark run did not return any relevant results for this metric.