Account for immutably borrowed locals in MIR copy-prop and GVN

[cjgillot]

For the most part, we consider that immutably borrowed Freeze locals still fulfill SSA conditions. As the borrow is immutable, any use of the local will have the value given by the single assignment, and there can be no surprise.

This allows copy-prop to merge a non-borrowed local with a borrowed local. We chose to keep copy-classes heads unborrowed, as those may be easier to optimize in later passes.

This also allows to GVN the value behind an immutable borrow. If a SSA local is borrowed, dereferencing that borrow is equivalent to copying the local's value: re-executing the assignment between the borrow and the dereference would be UB.

r? @ghost for perf

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit fd856c7 with merge 5d2ef83...

[bors]

☀️ Try build successful - checks-actions
Build commit: 5d2ef83 (5d2ef83d9fbd186559a2618781c405a364bce787)

[rust-timer]

Finished benchmarking commit (5d2ef83): 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)	0.6%	[0.2%, 2.2%]	8
Regressions ❌ (secondary)	0.3%	[0.2%, 0.4%]	8
Improvements ✅ (primary)	-0.6%	[-1.2%, -0.4%]	7
Improvements ✅ (secondary)	-0.6%	[-1.3%, -0.2%]	7
All ❌✅ (primary)	0.0%	[-1.2%, 2.2%]	15

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)	4.7%	[2.7%, 8.1%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.6%	[-5.4%, -2.1%]	12
Improvements ✅ (secondary)	-3.2%	[-7.1%, -0.8%]	86
All ❌✅ (primary)	-1.9%	[-5.4%, 8.1%]	15

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)	2.2%	[2.2%, 2.2%]	1
Regressions ❌ (secondary)	1.9%	[1.9%, 1.9%]	1
Improvements ✅ (primary)	-1.3%	[-1.3%, -1.3%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[-1.3%, 2.2%]	2

Binary size

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)	0.3%	[0.0%, 1.5%]	22
Regressions ❌ (secondary)	0.1%	[0.1%, 0.1%]	4
Improvements ✅ (primary)	-0.3%	[-1.5%, -0.0%]	40
Improvements ✅ (secondary)	-0.8%	[-1.9%, -0.2%]	12
All ❌✅ (primary)	-0.1%	[-1.5%, 1.5%]	62

Bootstrap: 667.569s -> 666.983s (-0.09%)
Artifact size: 318.50 MiB -> 318.41 MiB (-0.03%)

[cjgillot]

Perf report: weak effect on the benchmark suite in average. Crate metadata size is reduced up to 2 %. Binary size are slightly reduced too, up to 1.9%, but with some increases.

r? @oli-obk
cc @rust-lang/wg-mir-opt

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

Some changes occurred in coverage tests.

cc @Zalathar

[oli-obk]

@bors r+

[bors]

📌 Commit 8ce8fbf has been approved by oli-obk

It is now in the queue for this repository.

[bors]

⌛ Testing commit 8ce8fbf with merge 30b6c13...

[bors]

⌛ Testing commit 8ce8fbf with merge 92b015f...

[bors]

💔 Test failed - checks-actions

[bors]

💔 Test failed - checks-actions

[Zalathar]

According to #111883 (comment) that ICE test maybe shouldn't exist either way?

It was only recently added by #122997.

[oli-obk]

@bors r+

[bors]

📌 Commit c19866f has been approved by oli-obk

It is now in the queue for this repository.

[bors]

⌛ Testing commit c19866f with merge 460356f...

[fmease]

⌛ Testing commit c19866f with merge 460356f...

That's a zombie.

@bors retry for good measure

[RalfJung]

How does this treat addr_of!(local)? This should be treated as allowing mutation of the local via the pointer.

[bors]

⌛ Testing commit c19866f with merge d2d24e3...

[bors]

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing d2d24e3 to master...

[rust-timer]

Finished benchmarking commit (d2d24e3): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Next Steps: If you can justify the regressions found in this perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please open an issue or create a new PR that fixes the regressions, add a comment linking to the newly created issue or PR, and then add the perf-regression-triaged label to this PR.

@rustbot label: +perf-regression
cc @rust-lang/wg-compiler-performance

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)	0.3%	[0.2%, 0.9%]	10
Regressions ❌ (secondary)	0.8%	[0.2%, 2.6%]	4
Improvements ✅ (primary)	-0.5%	[-1.1%, -0.2%]	6
Improvements ✅ (secondary)	-0.5%	[-1.0%, -0.3%]	8
All ❌✅ (primary)	0.0%	[-1.1%, 0.9%]	16

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)	2.9%	[1.7%, 4.7%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-5.3%	[-6.8%, -3.6%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.8%	[-6.8%, 4.7%]	7

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)	0.8%	[0.8%, 0.9%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.7%	[-0.7%, -0.7%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.3%	[-0.7%, 0.9%]	3

Binary size

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)	0.1%	[0.0%, 0.2%]	23
Regressions ❌ (secondary)	0.1%	[0.1%, 0.1%]	4
Improvements ✅ (primary)	-0.3%	[-1.2%, -0.0%]	44
Improvements ✅ (secondary)	-0.6%	[-1.9%, -0.1%]	14
All ❌✅ (primary)	-0.2%	[-1.2%, 0.2%]	67

Bootstrap: 674.658s -> 674.126s (-0.08%)
Artifact size: 315.94 MiB -> 315.84 MiB (-0.03%)

[pnkfelix]

How does this treat addr_of!(local)? This should be treated as allowing mutation of the local via the pointer.

Did this Q from @RalfJung get addressed, @cjgillot ?

[pnkfelix]

Visiting for perf triage

• html5ever opt-full regressed by 0.92%; libc in various incremental scenarios regressed by 0.30% to 0.39%.
• the libc changes were anticipated in the perf build prior to merge; html5ever opt-full was not predicted there.
• pnkfelix hypothesizes that this just reflects some extra-work from the compiler attempting to do the copy-propagation and global-value-numbering mir-optimizations on a larger set of immutably-borrowed locals, and is acceptable given the expected benefits.
• marking as triaged

@rustbot label: +perf-regression-triaged