mir-opt: Merge all branch BBs into a single copy statement #129931

DianQK · 2024-09-03T13:44:29Z

#128299 simplified

match a {
    Foo::A(x) => Foo::A(*x),
    Foo::B => Foo::B
}

to

match a {
    Foo::A(x) => a, // copy a
    Foo::B => Foo::B
}

The switch branch can be simplified into a single copy statement. This PR implements a relatively general simplification.

rustbot · 2024-09-03T13:44:37Z

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

DianQK · 2024-09-03T13:45:05Z

r? ghost

rustbot · 2024-09-03T13:45:08Z

Failed to set assignee to ghost: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

DianQK · 2024-09-03T13:46:55Z

r? cjgillot

bors · 2024-09-14T17:07:18Z

☔ The latest upstream changes (presumably #128299) made this pull request unmergeable. Please resolve the merge conflicts.

DianQK · 2024-09-14T22:02:59Z

@bors try @rust-timer queue

bors · 2024-09-14T22:04:09Z

⌛ Trying commit 622247a with merge c324112...

Merge these copy statements that simplified the canonical enum clone method by GVN This is blocked by rust-lang#128299.

bors · 2024-09-14T23:56:31Z

☀️ Try build successful - checks-actions
Build commit: c324112 (c3241126e94d7a53a1805ad854686e0bd061b5df)

rust-timer · 2024-09-15T02:22:20Z

Finished benchmarking commit (c324112): 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 (primary 0.9%)

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)	4.0%	[3.0%, 5.1%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.8%	[-5.9%, -1.7%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.9%	[-5.9%, 5.1%]	5

Cycles

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

Binary size

Results (primary 0.0%, secondary -0.0%)

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%]	11
Regressions ❌ (secondary)	0.0%	[0.0%, 0.1%]	2
Improvements ✅ (primary)	-0.1%	[-0.3%, -0.0%]	6
Improvements ✅ (secondary)	-0.0%	[-0.0%, -0.0%]	3
All ❌✅ (primary)	0.0%	[-0.3%, 0.2%]	17

Bootstrap: 760.71s -> 756.595s (-0.54%)
Artifact size: 341.13 MiB -> 341.17 MiB (0.01%)

cjgillot

My high-level reaction to this PR is that it's both too specific in what it's trying to match (a clone impl) and too general in its implementation (handles storage statements...).

I suggest having two heuristics:

for clone impls, use some kind of StructuralClone trait so that we fully replace the derived impls Clone with simpler MIR;
for general MIR, lift the restrictions on this pass (bb0 and assignment to _0 namely).

What do you think?

cjgillot · 2024-09-15T13:07:59Z

compiler/rustc_mir_transform/src/lib.rs

@@ -604,6 +602,7 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
            &dead_store_elimination::DeadStoreElimination::Initial,
            &gvn::GVN,
            &simplify::SimplifyLocals::AfterGVN,
+            &match_branches::MatchBranchSimplification,


Should this run on clone shims too? Or do we want a trait-based solution to detect trivial clone impls?

Should this run on clone shims too?

Could you explain more?

Or do we want a trait-based solution to detect trivial clone impls?

It makes sense to me for clone impls.

compiler/rustc_mir_transform/src/match_branches.rs

DianQK · 2024-09-18T13:08:45Z

* for clone impls, use some kind of `StructuralClone` trait so that we fully replace the derived impls `Clone` with simpler MIR;

* for general MIR, lift the restrictions on this pass (bb0 and assignment to _0 namely).

What do you think?

It looks like makes sense.
IIUC, the StructuralClone is exact matching clone shims? I expect this can improve compile-time.

@rustbot author

DianQK

I didn't submit all the comments through "Submit review"; there are still some pending comments. o

I am currently modifying the code to move it into a separate pass.

compiler/rustc_mir_transform/src/match_branches.rs

DianQK · 2024-09-18T12:44:37Z

compiler/rustc_mir_transform/src/lib.rs

@@ -604,6 +602,7 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
            &dead_store_elimination::DeadStoreElimination::Initial,
            &gvn::GVN,
            &simplify::SimplifyLocals::AfterGVN,
+            &match_branches::MatchBranchSimplification,


Should this run on clone shims too?

Could you explain more?

Or do we want a trait-based solution to detect trivial clone impls?

It makes sense to me for clone impls.

bors · 2024-10-06T12:09:06Z

⌛ Trying commit 7846743 with merge 9637884...

bors · 2024-10-06T13:52:45Z

☀️ Try build successful - checks-actions
Build commit: 9637884 (963788449d01d434fa9a69b331e4f5c41f10eab4)

rust-timer · 2024-10-06T15:23:18Z

Finished benchmarking commit (9637884): comparison URL.

Overall result: ❌✅ regressions and improvements - 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 is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.5%	[0.3%, 0.8%]	2
Regressions ❌ (secondary)	0.5%	[0.5%, 0.5%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.5%	[-0.5%, -0.5%]	1
All ❌✅ (primary)	0.5%	[0.3%, 0.8%]	2

Max RSS (memory usage)

Results (primary 3.2%, secondary 2.5%)

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)	3.2%	[2.2%, 4.7%]	4
Regressions ❌ (secondary)	2.5%	[2.5%, 2.5%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.2%	[2.2%, 4.7%]	4

Cycles

Results (primary 1.9%)

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)	1.9%	[1.0%, 2.8%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.9%	[1.0%, 2.8%]	2

Binary size

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

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%	[0.1%, 0.4%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.1%	[-0.4%, -0.0%]	8
Improvements ✅ (secondary)	-0.0%	[-0.0%, -0.0%]	1
All ❌✅ (primary)	-0.0%	[-0.4%, 0.4%]	11

Bootstrap: 775.25s -> 775.189s (-0.01%)
Artifact size: 329.52 MiB -> 329.45 MiB (-0.02%)

DianQK · 2024-10-07T02:13:54Z

@rustbot review

compiler/rustc_mir_transform/src/merge_branches.rs

cjgillot · 2024-10-12T22:55:01Z

compiler/rustc_mir_transform/src/merge_branches.rs

+    Some(dest_place)
+}
+
+fn find_copy_assign<'tcx>(


Could you doc-comment this? I'm still not sure what pattern you are looking for.

cjgillot · 2024-10-12T22:58:41Z

compiler/rustc_mir_transform/src/merge_branches.rs

+                        // There is only one statement that cannot be ignored
+                        // that can be used as an expected copy statement.
+                        expected_assign_stmt = Some(statement_index);
+                        lived_stmts.remove(statement_index);


I know we already discussed this, but I'd rather have DSE run multiple times in the pipeline, than having re-coded here. This complicates the code a lot, where we could get rid of the statements once.

Since debuginfo handles it differently, I'll try to modify the DSE then get rid of these.

compiler/rustc_mir_transform/src/merge_branches.rs

cjgillot · 2024-10-12T23:05:25Z

compiler/rustc_mir_transform/src/lib.rs

@@ -609,6 +610,7 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
            &dead_store_elimination::DeadStoreElimination::Initial,
            &gvn::GVN,
            &simplify::SimplifyLocals::AfterGVN,
+            &merge_branches::MergeBranchSimplification,


Why here in the pipeline?

This needs to run after GVN. I expect SimplifyLocals::AfterGVN can simplify the analysis in MergeBranchSimplification. Additionally, after MergeBranchSimplification simplifies the CFG, it will benefit subsequent passes that are related to the CFG.

DianQK · 2024-10-16T01:24:52Z

I will probably update in two days.

@rustbot author

This reverts commit 9692513.

This reverts commit 25d434b.

DianQK · 2024-10-17T14:47:00Z

Just a few trivial modifications and rebase.

DianQK · 2024-10-20T13:25:19Z

I have tried to update the document and DSE.

@rustbot review

bors · 2024-10-30T11:02:59Z

☔ The latest upstream changes (presumably #132349) made this pull request unmergeable. Please resolve the merge conflicts.

DianQK · 2024-10-30T14:43:16Z

Blocked by #132353.

@rustbot author
@rustbot label +S-blocked

rustbot assigned nnethercote Sep 3, 2024

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 Sep 3, 2024

rustbot assigned cjgillot and unassigned nnethercote Sep 3, 2024

This comment has been minimized.

Sign in to view

DianQK force-pushed the match-br-copy branch from efce2e8 to adb6166 Compare September 3, 2024 14:44

DianQK force-pushed the match-br-copy branch from adb6166 to 622247a Compare September 14, 2024 22:02

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Sep 14, 2024

bors added a commit to rust-lang-ci/rust that referenced this pull request Sep 14, 2024

Auto merge of rust-lang#129931 - DianQK:match-br-copy, r=<try>

c324112

Merge these copy statements that simplified the canonical enum clone method by GVN This is blocked by rust-lang#128299.

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Sep 15, 2024

cjgillot reviewed Sep 15, 2024

View reviewed changes

DianQK mentioned this pull request Sep 17, 2024

tests: allow trunc/select instructions to be missing #130454

Merged

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Sep 18, 2024

DianQK commented Oct 5, 2024

View reviewed changes

DianQK force-pushed the match-br-copy branch from 622247a to dddf046 Compare October 5, 2024 13:18

DianQK changed the title ~~Merge these copy statements that simplified the canonical enum clone method by GVN~~ mir-opt: Merge all branch BBs into a single copy statement Oct 5, 2024

This comment has been minimized.

Sign in to view

rustbot removed S-waiting-on-perf Status: Waiting on a perf run to be completed. perf-regression Performance regression. labels Oct 6, 2024

DianQK force-pushed the match-br-copy branch from 7846743 to e1d9550 Compare October 7, 2024 02:13

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 7, 2024

cjgillot reviewed Oct 12, 2024

View reviewed changes

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 16, 2024

DianQK force-pushed the match-br-copy branch from e1d9550 to c25e375 Compare October 17, 2024 14:44

DianQK added 2 commits October 17, 2024 22:45

Revert "tests: allow trunc/select instructions to be missing"

0ccf8ba

This reverts commit 9692513.

Revert "Update try_question_mark_nop.rs test"

9bb8cb8

This reverts commit 25d434b.

DianQK force-pushed the match-br-copy branch from c25e375 to 3882859 Compare October 17, 2024 14:45

DianQK added 2 commits October 20, 2024 20:59

mir-opt: Allow other passes to use DSE

70392ec

mir-opt: Merge all branch BBs into a single copy statement for enum

99eed4b

DianQK force-pushed the match-br-copy branch from 3882859 to 99eed4b Compare October 20, 2024 13:11

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 20, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

mir-opt: Merge all branch BBs into a single copy statement #129931

mir-opt: Merge all branch BBs into a single copy statement #129931

DianQK commented Sep 3, 2024 •

edited

Loading

rustbot commented Sep 3, 2024

DianQK commented Sep 3, 2024

rustbot commented Sep 3, 2024

DianQK commented Sep 3, 2024

This comment has been minimized.

bors commented Sep 14, 2024

DianQK commented Sep 14, 2024

This comment has been minimized.

bors commented Sep 14, 2024

bors commented Sep 14, 2024

This comment has been minimized.

rust-timer commented Sep 15, 2024

cjgillot left a comment

cjgillot Sep 15, 2024

DianQK Sep 18, 2024

DianQK commented Sep 18, 2024

DianQK left a comment

DianQK Sep 18, 2024

bors commented Oct 6, 2024

bors commented Oct 6, 2024

This comment has been minimized.

rust-timer commented Oct 6, 2024

DianQK commented Oct 7, 2024

cjgillot Oct 12, 2024

cjgillot Oct 12, 2024

DianQK Oct 16, 2024

cjgillot Oct 12, 2024

DianQK Oct 20, 2024

DianQK commented Oct 16, 2024

DianQK commented Oct 17, 2024

DianQK commented Oct 20, 2024

bors commented Oct 30, 2024

DianQK commented Oct 30, 2024

mir-opt: Merge all branch BBs into a single copy statement #129931

Are you sure you want to change the base?

mir-opt: Merge all branch BBs into a single copy statement #129931

Conversation

DianQK commented Sep 3, 2024 • edited Loading

rustbot commented Sep 3, 2024

DianQK commented Sep 3, 2024

rustbot commented Sep 3, 2024

DianQK commented Sep 3, 2024

This comment has been minimized.

bors commented Sep 14, 2024

DianQK commented Sep 14, 2024

This comment has been minimized.

bors commented Sep 14, 2024

bors commented Sep 14, 2024

This comment has been minimized.

rust-timer commented Sep 15, 2024

Overall result: no relevant changes - no action needed

cjgillot left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

DianQK commented Sep 18, 2024

DianQK left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bors commented Oct 6, 2024

bors commented Oct 6, 2024

This comment has been minimized.

rust-timer commented Oct 6, 2024

Overall result: ❌✅ regressions and improvements - no action needed

DianQK commented Oct 7, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

DianQK commented Oct 16, 2024

DianQK commented Oct 17, 2024

DianQK commented Oct 20, 2024

bors commented Oct 30, 2024

DianQK commented Oct 30, 2024

DianQK commented Sep 3, 2024 •

edited

Loading