Do not consider &mut *x as mutating x in CopyProp

[cjgillot]

This PR removes an unfortunate overly cautious case from the current implementation.

Found by #105274 cc @saethlin

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[saethlin]

Nobody got assigned automatically 🤔 should someone be?

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 69a2d71e17b2821a445730d2360331e38229af2e with merge 0e92e4b96312299b07d11647a466dc737e5207c3...

[bors]

☀️ Try build successful - checks-actions
Build commit: 0e92e4b96312299b07d11647a466dc737e5207c3 (0e92e4b96312299b07d11647a466dc737e5207c3)

[saethlin]

FYI this perf run does not include #105274 which might be relevant.

[rust-timer]

Finished benchmarking commit (0e92e4b96312299b07d11647a466dc737e5207c3): comparison URL.

Overall result: no relevant changes - no 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.

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

Instruction count

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

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)	3.8%	[3.8%, 3.8%]	1
Regressions ❌ (secondary)	3.6%	[3.6%, 3.6%]	1
Improvements ✅ (primary)	-3.1%	[-3.1%, -3.1%]	2
Improvements ✅ (secondary)	-2.6%	[-3.8%, -1.0%]	19
All ❌✅ (primary)	-0.8%	[-3.1%, 3.8%]	3

Cycles

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

[cjgillot]

Now that #105274 is merged.
@bors try @rust-timer queue

[bors]

⌛ Trying commit 69a2d71e17b2821a445730d2360331e38229af2e with merge 2894f9ae226ad439e75cb76ad98f6811eb326247...

[bors]

☀️ Try build successful - checks-actions
Build commit: 2894f9ae226ad439e75cb76ad98f6811eb326247 (2894f9ae226ad439e75cb76ad98f6811eb326247)

[rust-timer]

Finished benchmarking commit (2894f9ae226ad439e75cb76ad98f6811eb326247): 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)	0.9%	[0.8%, 1.0%]	2
Regressions ❌ (secondary)	1.3%	[1.3%, 1.3%]	1
Improvements ✅ (primary)	-0.3%	[-0.4%, -0.3%]	3
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[-0.4%, 1.0%]	5

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)	5.0%	[3.4%, 6.6%]	2
Regressions ❌ (secondary)	4.2%	[4.2%, 4.2%]	1
Improvements ✅ (primary)	-2.2%	[-3.3%, -0.1%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.2%	[-3.3%, 6.6%]	6

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
Improvements ✅ (secondary)	-6.3%	[-7.1%, -5.6%]	6
All ❌✅ (primary)	-	-	0

[cjgillot]

r? compiler

[saethlin]

Does &raw mut *x deserve the same treatment?

[cjgillot]

This patch handles both cases by considering that a mutation of '*x' is not a mutation of 'x'.

[eholk]

@RalfJung - Does this change have any stacked borrows implications?

@bors r? @JakobDegen

[rustbot]

Failed to set assignee to JakobDegen: invalid assignee

Note: Only org members, users with write permissions, or people who have commented on the PR may be assigned.

[eholk]

I took a quick look over the PR, and it seems reasonable to me, but I'd rather someone who knows this code better sign off.

@bors r? @wesleywiser

[RalfJung]

Does this change have any stacked borrows implications?

That's hard to say without reverse engineering CopyProp. @cjgillot could you describe with some examples which new transformations this entails?

Stacked Borrows currently does consider &mut *x to be a write to x. Tree Borrows does not. But this is a pretty fundamental trade-off; Tree Borrows loses some optimizations due to this choice. However it's hard for me to say whether that is even relevant for CopyProp.

[cjgillot]

Basically, this allows to transform

_2 = _1
_3 = &mut *_1
_4 = _1
anything(_2, _4)

Into

_3 = &mut *_1
anything(_1, _1)

This would be allowed whatever the type of _1, either a reference or a raw pointer.
I don't see how SB considers &mut *x to write to x. I understand it changes the borrow stack global state, but I don't see where it writes to x, neither address nor provenance.

If we consider &mut *x to write to x, ie by modifying provenance, this transformation is incorrect.

[JakobDegen]

Yeah, I feel pretty strongly that we need to allow this opt. The scenario in which it risks being wrong is if the _2 = _1 actually modifies the value - that can be the result of implicit retags, but we should just not do those.

[oli-obk]

@bors r+

r? @oli-obk

[bors]

📌 Commit 209eb8a has been approved by oli-obk

It is now in the queue for this repository.

[bors]

⌛ Testing commit 209eb8a with merge 66a2d62...

[bors]

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 66a2d62 to master...

[rust-timer]

Finished benchmarking commit (66a2d62): 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)	0.4%	[0.2%, 1.0%]	18
Regressions ❌ (secondary)	0.4%	[0.3%, 0.5%]	11
Improvements ✅ (primary)	-0.3%	[-0.3%, -0.3%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.3%	[-0.3%, 1.0%]	20

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)	2.1%	[1.9%, 2.2%]	2
Regressions ❌ (secondary)	1.5%	[1.5%, 1.5%]	2
Improvements ✅ (primary)	-4.0%	[-4.8%, -3.3%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.0%	[-4.8%, 2.2%]	4

Cycles

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

[cjgillot]

The regression in opt builds comes from LLVM doing more work, which is expected.

The regression for serde_derive check is a change in how btree's search_tree over region constraints is codegened, and in the stable hasher.

[RalfJung]

Yeah, I feel pretty strongly that we need to allow this opt. The scenario in which it risks being wrong is if the _2 = _1 actually modifies the value - that can be the result of implicit retags, but we should just not do those.

Implicit retags on assignments specifically. I agree we shouldn't have those.

Retags on Rvalue::Ref are a different matter IMO. But those should be fine here.