GVN: Only propagate borrows from SSA locals

[dianqk]

Fixes #141313. This is a more principled fix than #147886.

Using a reference that is not a borrowing of an SSA local at a new location may be UB.

The PR has two major changes.

The first one, when introducing a new dereference at a new location, is that the reference must point to an SSA local or be an immutable argument. dereference_address has handled SSA locals.

The second one, if we cannot guard to the reference point to an SSA local in visit_assign, we have to rewrite the value to opaque. This avoids unifying the following dereferences that also are references:

let b: &T = *a;
// ... `a` is allowed to be modified. `c` and `b` have different borrowing lifetime.
// Unifying them will extend the lifetime of `b`.
let c: &T = *a; 

See also #130853.

This still allows unifying non-reference dereferences:

let a: &T = ...;
let b: T = *a;
// ... a is NOT allowed to be modified.
let c: T = *a; 

r? @cjgillot

[dianqk]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: f23e5de (f23e5de34a174f8bd281d69038b68bf6332b1af6, parent: fcd630976c460c819c4bbcaf107d0c94501205d8)

[rust-timer]

Finished benchmarking commit (f23e5de): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.3%	[0.1%, 0.9%]	20
Regressions ❌ (secondary)	0.4%	[0.1%, 2.0%]	27
Improvements ✅ (primary)	-0.6%	[-1.2%, -0.2%]	4
Improvements ✅ (secondary)	-0.2%	[-0.4%, -0.1%]	7
All ❌✅ (primary)	0.1%	[-1.2%, 0.9%]	24

Max RSS (memory usage)

Results (primary -1.1%, secondary -4.7%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.4%	[1.9%, 2.8%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.4%	[-6.8%, -0.6%]	3
Improvements ✅ (secondary)	-4.7%	[-6.5%, -3.7%]	3
All ❌✅ (primary)	-1.1%	[-6.8%, 2.8%]	5

Cycles

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

Binary size

Results (primary 0.1%, secondary 0.0%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.0%, 6.9%]	11
Regressions ❌ (secondary)	0.1%	[0.0%, 0.2%]	14
Improvements ✅ (primary)	-0.1%	[-1.0%, -0.0%]	27
Improvements ✅ (secondary)	-0.1%	[-0.1%, -0.0%]	5
All ❌✅ (primary)	0.1%	[-1.0%, 6.9%]	38

Bootstrap: 475.254s -> 472.456s (-0.59%)
Artifact size: 390.85 MiB -> 390.82 MiB (-0.01%)

[dianqk]

@bors try parent=fcd630976c460c819c4bbcaf107d0c94501205d8 @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 2183340 (218334063dc5d70b7f9d69202bf22750167c6cbf, parent: fcd630976c460c819c4bbcaf107d0c94501205d8)

[rust-timer]

Finished benchmarking commit (2183340): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.3%	[0.1%, 0.9%]	17
Regressions ❌ (secondary)	0.3%	[0.1%, 0.8%]	23
Improvements ✅ (primary)	-0.6%	[-1.2%, -0.2%]	4
Improvements ✅ (secondary)	-0.2%	[-0.4%, -0.1%]	8
All ❌✅ (primary)	0.1%	[-1.2%, 0.9%]	21

Max RSS (memory usage)

Results (primary -1.9%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.7%	[1.7%, 1.7%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.0%	[-4.6%, -0.6%]	3
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.9%	[-4.6%, 1.7%]	4

Cycles

Results (secondary -2.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.3%	[-2.3%, -2.3%]	1
All ❌✅ (primary)	-	-	0

Binary size

Results (primary 0.2%, secondary 0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.0%, 6.9%]	11
Regressions ❌ (secondary)	0.1%	[0.0%, 0.2%]	14
Improvements ✅ (primary)	-0.2%	[-0.9%, -0.0%]	11
Improvements ✅ (secondary)	-0.1%	[-0.1%, -0.0%]	5
All ❌✅ (primary)	0.2%	[-0.9%, 6.9%]	22

Bootstrap: 475.254s -> 472.783s (-0.52%)
Artifact size: 390.85 MiB -> 390.84 MiB (-0.00%)

[dianqk]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 0e5b1a1 (0e5b1a1b4144d4e37e094f7d74087be1e0077bfb, parent: e96bb7e44fbcc23c1e6009e8d0ee8ab208668fb4)

[rust-timer]

Finished benchmarking commit (0e5b1a1): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.2%	[0.2%, 0.3%]	33
Regressions ❌ (secondary)	0.2%	[0.1%, 0.3%]	25
Improvements ✅ (primary)	-0.6%	[-1.2%, -0.2%]	3
Improvements ✅ (secondary)	-0.1%	[-0.3%, -0.1%]	9
All ❌✅ (primary)	0.2%	[-1.2%, 0.3%]	36

Max RSS (memory usage)

Results (primary -0.9%, secondary -2.6%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	3.0%	[2.3%, 3.6%]	2
Regressions ❌ (secondary)	2.1%	[1.5%, 3.1%]	3
Improvements ✅ (primary)	-2.2%	[-3.7%, -1.4%]	6
Improvements ✅ (secondary)	-4.4%	[-5.4%, -0.5%]	8
All ❌✅ (primary)	-0.9%	[-3.7%, 3.6%]	8

Cycles

Results (primary 2.9%, secondary 3.8%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.9%	[1.6%, 4.5%]	22
Regressions ❌ (secondary)	4.2%	[1.9%, 8.8%]	41
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.2%	[-4.2%, -4.1%]	2
All ❌✅ (primary)	2.9%	[1.6%, 4.5%]	22

Binary size

Results (primary 0.0%, secondary -0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.0%, 0.3%]	8
Regressions ❌ (secondary)	0.0%	[0.0%, 0.0%]	1
Improvements ✅ (primary)	-0.1%	[-0.1%, -0.0%]	15
Improvements ✅ (secondary)	-0.1%	[-0.2%, -0.0%]	9
All ❌✅ (primary)	0.0%	[-0.1%, 0.3%]	23

Bootstrap: 470.891s -> 474.803s (0.83%)
Artifact size: 383.44 MiB -> 385.80 MiB (0.62%)

[dianqk]

It's weird.
@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: f1385c1 (f1385c1681402a00f99d1eb766520242f9fde309, parent: 466ea4e6c39f8a43727edcc726ca86b499e14d83)

[rust-timer]

Finished benchmarking commit (f1385c1): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.3%	[0.2%, 0.3%]	6
Regressions ❌ (secondary)	0.6%	[0.1%, 1.0%]	10
Improvements ✅ (primary)	-0.6%	[-1.1%, -0.1%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.0%	[-1.1%, 0.3%]	8

Max RSS (memory usage)

Results (primary -1.9%, secondary -2.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	3.1%	[3.1%, 3.1%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.7%	[-5.3%, -1.7%]	6
Improvements ✅ (secondary)	-2.2%	[-2.2%, -2.2%]	1
All ❌✅ (primary)	-1.9%	[-5.3%, 3.1%]	7

Cycles

Results (primary -1.9%, secondary -0.0%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.8%	[2.8%, 2.8%]	1
Improvements ✅ (primary)	-1.9%	[-1.9%, -1.9%]	1
Improvements ✅ (secondary)	-2.8%	[-2.8%, -2.8%]	1
All ❌✅ (primary)	-1.9%	[-1.9%, -1.9%]	1

Binary size

Results (primary 0.0%, secondary -0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.0%, 0.4%]	7
Regressions ❌ (secondary)	0.0%	[0.0%, 0.0%]	1
Improvements ✅ (primary)	-0.1%	[-0.1%, -0.0%]	15
Improvements ✅ (secondary)	-0.1%	[-0.2%, -0.0%]	8
All ❌✅ (primary)	0.0%	[-0.1%, 0.4%]	22

Bootstrap: 476.499s -> 476.268s (-0.05%)
Artifact size: 397.91 MiB -> 397.83 MiB (-0.02%)

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[dianqk]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: f2def88 (f2def887c3d08e9845884a5677473348acc2e23f, parent: db3e99bbab28c6ca778b13222becdea54533d908)

[rust-timer]

Finished benchmarking commit (f2def88): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	1.5%	[0.9%, 2.1%]	2
Regressions ❌ (secondary)	0.7%	[0.3%, 2.0%]	5
Improvements ✅ (primary)	-1.5%	[-1.9%, -1.0%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.0%	[-1.9%, 2.1%]	4

Max RSS (memory usage)

Results (primary -0.6%, secondary 0.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.5%	[1.7%, 3.5%]	3
Regressions ❌ (secondary)	1.5%	[0.5%, 2.5%]	2
Improvements ✅ (primary)	-5.3%	[-7.6%, -3.1%]	2
Improvements ✅ (secondary)	-2.0%	[-2.0%, -2.0%]	1
All ❌✅ (primary)	-0.6%	[-7.6%, 3.5%]	5

Cycles

Results (primary 2.8%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.8%	[2.8%, 2.8%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.8%	[2.8%, 2.8%]	1

Binary size

Results (primary 0.0%, secondary 0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.0%, 1.2%]	5
Regressions ❌ (secondary)	1.2%	[1.2%, 1.2%]	1
Improvements ✅ (primary)	-0.3%	[-1.4%, -0.0%]	6
Improvements ✅ (secondary)	-0.1%	[-0.2%, -0.0%]	8
All ❌✅ (primary)	0.0%	[-1.4%, 1.2%]	11

Bootstrap: 472.94s -> 474.175s (0.26%)
Artifact size: 398.12 MiB -> 398.12 MiB (-0.00%)

[dianqk]

The two latest perf results that were built with LLVM 22 are different. I am not sure why, but the latest result seems acceptable to me.