Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Run InstSimplify before UnreachablePropagation #123256

Closed
wants to merge 1 commit into from

Conversation

saethlin
Copy link
Member

@saethlin saethlin commented Mar 31, 2024

This is the pass ordering change I was musing about: #122975 (comment)

r? @ghost

@rustbot rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Mar 31, 2024
@saethlin
Copy link
Member Author

@bors try @rust-timer queue

@rust-timer

This comment has been minimized.

@rustbot rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 31, 2024
bors added a commit to rust-lang-ci/rust that referenced this pull request Mar 31, 2024
Run InstSimplify before UnreachablePropagation

This is the pass ordering change I was musing about: rust-lang#122975 (comment)

r? `@ghost`
@bors
Copy link
Contributor

bors commented Mar 31, 2024

⌛ Trying commit 0b77e30 with merge 49d1f0a...

@rust-log-analyzer

This comment has been minimized.

@bors
Copy link
Contributor

bors commented Mar 31, 2024

☀️ Try build successful - checks-actions
Build commit: 49d1f0a (49d1f0a7f4d467843f95684ca0eb51bca6cce28d)

@rust-timer

This comment has been minimized.

@rust-timer
Copy link
Collaborator

Finished benchmarking commit (49d1f0a): 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)
1.0% [0.5%, 1.6%] 4
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
-0.5% [-1.0%, -0.3%] 16
Improvements ✅
(secondary)
-0.5% [-1.0%, -0.2%] 9
All ❌✅ (primary) -0.2% [-1.0%, 1.6%] 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)
8.5% [2.2%, 19.1%] 3
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
-2.8% [-5.0%, -0.0%] 4
Improvements ✅
(secondary)
- - 0
All ❌✅ (primary) 2.1% [-5.0%, 19.1%] 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)
1.2% [1.0%, 1.4%] 2
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
- - 0
Improvements ✅
(secondary)
-3.7% [-4.5%, -2.9%] 2
All ❌✅ (primary) 1.2% [1.0%, 1.4%] 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)
0.1% [0.0%, 0.7%] 35
Regressions ❌
(secondary)
0.0% [0.0%, 0.0%] 4
Improvements ✅
(primary)
-0.4% [-1.1%, -0.0%] 50
Improvements ✅
(secondary)
-0.6% [-1.3%, -0.1%] 25
All ❌✅ (primary) -0.2% [-1.1%, 0.7%] 85

Bootstrap: 670.332s -> 671.609s (0.19%)
Artifact size: 315.71 MiB -> 315.67 MiB (-0.01%)

@rustbot rustbot added perf-regression Performance regression. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Mar 31, 2024
@saethlin
Copy link
Member Author

The optimized MIR for Result::clone changes from this:

    bb2: {
        StorageLive(_6);
        _6 = &(((*_1) as Err).0: E);
        StorageLive(_7);
        StorageLive(_8);
        _8 = &(((*_1) as Err).0: E);
        _7 = <E as clone::Clone>::clone(move _8) -> [return: bb5, unwind continue];
    }

To this:

    bb2: {
        _5 = &(((*_1) as Err).0: E);
        StorageLive(_6); 
        _6 = <E as clone::Clone>::clone(move _5) -> [return: bb5, unwind continue];
    }   

That is very surprising to me.

@saethlin
Copy link
Member Author

Oh I see the problem. ReferencePropagation does this:

         _6 = &(((*_1) as Err).0: E);
         StorageLive(_7);
         StorageLive(_8);
-        _8 = &(*_6);
+        _8 = &(((*_1) as Err).0: E);
         _7 = <E as Clone>::clone(move _8) -> [return: bb5, unwind continue];
     }

Then _6 is unused, but it has debuginfo, so SimplifyLocals won't remove it.

But if we move InstCombine before ReferencePropagation, InstCombine gets to see the reborrow and turns it into an assignment. Then SimplifyLocals considers optimizing out _8 instead of _6, which it can, because _8 has not debuginfo.

@cjgillot I only see this pattern crop up twice in core, but it seems worth doing something about. Do you think it's possible to get ReferencePropagation to behave the way we want here?

@saethlin
Copy link
Member Author

@bors try @rust-timer queue

@rust-timer

This comment has been minimized.

@rustbot rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 31, 2024
@bors
Copy link
Contributor

bors commented Mar 31, 2024

⌛ Trying commit 0da8676 with merge 66e8eb9...

bors added a commit to rust-lang-ci/rust that referenced this pull request Mar 31, 2024
Run InstSimplify before UnreachablePropagation

This is the pass ordering change I was musing about: rust-lang#122975 (comment)

r? `@ghost`
@rust-log-analyzer

This comment has been minimized.

@bors
Copy link
Contributor

bors commented Mar 31, 2024

☀️ Try build successful - checks-actions
Build commit: 66e8eb9 (66e8eb98489e088d62233c3026e6ae02f3ad6f04)

@rust-timer

This comment has been minimized.

@rust-timer
Copy link
Collaborator

Finished benchmarking commit (66e8eb9): 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% [0.4%, 1.6%] 5
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
-0.5% [-0.8%, -0.2%] 15
Improvements ✅
(secondary)
-0.4% [-0.6%, -0.2%] 9
All ❌✅ (primary) -0.1% [-0.8%, 1.6%] 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)
2.6% [0.7%, 4.4%] 4
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
-3.3% [-3.9%, -2.7%] 2
Improvements ✅
(secondary)
- - 0
All ❌✅ (primary) 0.6% [-3.9%, 4.4%] 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)
1.0% [0.7%, 1.4%] 3
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
- - 0
Improvements ✅
(secondary)
-2.4% [-2.4%, -2.4%] 1
All ❌✅ (primary) 1.0% [0.7%, 1.4%] 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%, 0.7%] 39
Regressions ❌
(secondary)
0.0% [0.0%, 0.0%] 4
Improvements ✅
(primary)
-0.4% [-1.1%, -0.0%] 48
Improvements ✅
(secondary)
-0.6% [-1.4%, -0.1%] 25
All ❌✅ (primary) -0.2% [-1.1%, 0.7%] 87

Bootstrap: 668.134s -> 670.326s (0.33%)
Artifact size: 315.67 MiB -> 315.67 MiB (0.00%)

@rustbot rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 31, 2024
@cjgillot
Copy link
Contributor

cjgillot commented Apr 1, 2024

Oh I see the problem. ReferencePropagation does this:

         _6 = &(((*_1) as Err).0: E);
         StorageLive(_7);
         StorageLive(_8);
-        _8 = &(*_6);
+        _8 = &(((*_1) as Err).0: E);
         _7 = <E as Clone>::clone(move _8) -> [return: bb5, unwind continue];
     }

Then _6 is unused, but it has debuginfo, so SimplifyLocals won't remove it.

But if we move InstCombine before ReferencePropagation, InstCombine gets to see the reborrow and turns it into an assignment. Then SimplifyLocals considers optimizing out _8 instead of _6, which it can, because _8 has not debuginfo.

@cjgillot I only see this pattern crop up twice in core, but it seems worth doing something about. Do you think it's possible to get ReferencePropagation to behave the way we want here?

Reference propagation could absolutely do that transform. We'd lose some information about reborrows, but none we actually use, so that's OK.

@bors
Copy link
Contributor

bors commented Apr 3, 2024

☔ The latest upstream changes (presumably #122225) made this pull request unmergeable. Please resolve the merge conflicts.

@saethlin
Copy link
Member Author

saethlin commented Apr 5, 2024

I'm going to close this because I think the right action here is to not do what I was envisioning, but to improve ReferencePropagation.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
perf-regression Performance regression. S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants