Set dead_on_return attribute for indirect arguments #145093

nikic · 2025-08-08T12:25:09Z

Set the dead_on_return attribute (added in LLVM 21) for arguments that are passed indirectly, but not byval.

This indicates that the value of the argument on return does not matter, enabling additional dead store elimination.

From LangRef:

This attribute indicates that the memory pointed to by the argument is dead upon function return, both upon normal return and if the calls unwinds, meaning that the caller will not depend on its contents. Stores that would be observable either on the return path or on the unwind path may be elided.

Specifically, the behavior is as-if any memory written through the pointer during the execution of the function is overwritten with a poison value upon function return. The caller may access the memory, but any load not preceded by a store will return poison.

This attribute does not imply aliasing properties. For pointer arguments that do not alias other memory locations, noalias attribute may be used in conjunction. Conversely, this attribute always implies dead_on_unwind.

This attribute cannot be applied to return values.

This fixes parts of #96497.

rustbot · 2025-08-08T12:25:13Z

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

nikic · 2025-08-08T12:26:22Z

I don't really expect this to have any appreciate impact in practice, but let's check...

@bors try @rust-timer queue

Set dead_on_return attribute for indirect arguments

rust-bors · 2025-08-08T12:26:27Z

⌛ Trying commit 9c1bbfa with merge 7642ceb…

To cancel the try build, run the command @bors try cancel.

rust-bors · 2025-08-08T14:37:52Z

☀️ Try build successful (CI)
Build commit: 7642ceb (7642cebd8079c4acdbbae34e69cfc8d7855e1086, parent: 2886b36df4a646dd8d82fb65bf0c9d8d96c1f71a)

rust-timer · 2025-08-08T15:48:51Z

Finished benchmarking commit (7642ceb): comparison URL.

Overall result: no relevant changes - no action needed

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.

@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 (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)	-	-	0
Regressions ❌ (secondary)	2.6%	[2.6%, 2.6%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results (secondary 2.9%)

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)	2.9%	[2.9%, 2.9%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 465.538s -> 465.32s (-0.05%)
Artifact size: 377.38 MiB -> 377.36 MiB (-0.00%)

erikdesjardins · 2025-08-08T19:19:55Z

tests/codegen-llvm/dead_on_return.rs

+pub fn test(mut s: String) {
+    // CHECK-LABEL: @test
+    // CHECK: __rust_dealloc
+    // CHECK-NOT: store
+    s = String::new();
+}


This test passing indicates that this fixes (part of) #96497, specifically #96497 (comment)

Ah yes, I wasn't aware we had an issue for this.

nnethercote · 2025-08-08T23:16:59Z

r=me. The only question is does this fix all of #96497, or just the comment that @erikdesjardins linked to?

Set the dead_on_return attribute (added in LLVM 21) for arguments that are passed indirectly, but not byval. This indicates that the value of the argument on return does not matter, enabling additional dead store elimination.

nikic · 2025-08-11T10:42:18Z

r=me. The only question is does this fix all of #96497, or just the comment that @erikdesjardins linked to?

I've added two more cases from the issue as tests. One case that's not fixed is the Vec::fill one (https://rust.godbolt.org/z/zPj3nrfbh). So I've adjusted the PR description to not close the issue yet.

nnethercote · 2025-08-11T10:46:04Z

Great, thank you. r=me, merge when you are ready.

nikic · 2025-08-11T12:20:59Z

@bors r=nnethercote

bors · 2025-08-11T12:21:02Z

📌 Commit ebef9d7 has been approved by nnethercote

It is now in the queue for this repository.

bors · 2025-08-12T06:20:58Z

⌛ Testing commit ebef9d7 with merge 21f6a33...

Set dead_on_return attribute for indirect arguments Set the dead_on_return attribute (added in LLVM 21) for arguments that are passed indirectly, but not byval. This indicates that the value of the argument on return does not matter, enabling additional dead store elimination. From LangRef: > This attribute indicates that the memory pointed to by the argument is dead upon function return, both upon normal return and if the calls unwinds, meaning that the caller will not depend on its contents. Stores that would be observable either on the return path or on the unwind path may be elided. > > Specifically, the behavior is as-if any memory written through the pointer during the execution of the function is overwritten with a poison value upon function return. The caller may access the memory, but any load not preceded by a store will return poison. > > This attribute does not imply aliasing properties. For pointer arguments that do not alias other memory locations, noalias attribute may be used in conjunction. Conversely, this attribute always implies dead_on_unwind. > > This attribute cannot be applied to return values. This fixes parts of #96497.

rust-log-analyzer · 2025-08-12T08:26:05Z

The job x86_64-apple-1 failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)

Launching test 'f32 small integer values'
Launching test 'f64 small integer values'
Launching test 'f32 subnormal'
Launching test 'f64 subnormal'
Command `/Users/runner/work/rust/rust/build/x86_64-apple-darwin/stage0/bin/cargo run --target x86_64-apple-darwin -Zbinary-dep-depinfo -j 4 -Zroot-dir=/Users/runner/work/rust/rust --locked --color always --release --manifest-path /Users/runner/work/rust/rust/src/tools/test-float-parse/Cargo.toml -- --skip-huge [workdir=/Users/runner/work/rust/rust]` failed with exit code 1
Created at: src/bootstrap/src/core/builder/cargo.rs:423:33
Executed at: src/bootstrap/src/core/build_steps/test.rs:3763:30

Command has failed. Rerun with -v to see more details.
Build completed unsuccessfully in 1:54:14
  local time: Tue Aug 12 08:25:47 UTC 2025
  network time: Tue, 12 Aug 2025 08:25:47 GMT
##[error]Process completed with exit code 1.
Post job cleanup.

bors · 2025-08-12T08:26:35Z

💔 Test failed - checks-actions

nikic · 2025-08-12T09:19:38Z

@bors retry Not super clear what actually failed there, but it looks spurious.

bors · 2025-08-12T10:21:43Z

⌛ Testing commit ebef9d7 with merge fec132b...

Set dead_on_return attribute for indirect arguments Set the dead_on_return attribute (added in LLVM 21) for arguments that are passed indirectly, but not byval. This indicates that the value of the argument on return does not matter, enabling additional dead store elimination. From LangRef: > This attribute indicates that the memory pointed to by the argument is dead upon function return, both upon normal return and if the calls unwinds, meaning that the caller will not depend on its contents. Stores that would be observable either on the return path or on the unwind path may be elided. > > Specifically, the behavior is as-if any memory written through the pointer during the execution of the function is overwritten with a poison value upon function return. The caller may access the memory, but any load not preceded by a store will return poison. > > This attribute does not imply aliasing properties. For pointer arguments that do not alias other memory locations, noalias attribute may be used in conjunction. Conversely, this attribute always implies dead_on_unwind. > > This attribute cannot be applied to return values. This fixes parts of #96497.

rust-log-analyzer · 2025-08-12T12:35:40Z

The job x86_64-msvc-ext1 failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)

   Verifying foo v0.0.1 (C:\a\rust\rust\build\x86_64-pc-windows-msvc\stage1-tools\x86_64-pc-windows-msvc\tmp\cit\t2372\foo)
 Downloading crates ...
  Downloaded bar v1.0.0 (registry `dummy-registry`)
   Compiling bar v1.0.0
   Compiling foo v0.0.1 (C:\a\rust\rust\build\x86_64-pc-windows-msvc\stage1-tools\x86_64-pc-windows-msvc\tmp\cit\t2372\foo\target\package\foo-0.0.1)
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.69s

stack backtrace:
   0: __rustc::rust_begin_unwind
   1: core::panicking::panic_fmt
   2: cargo_test_support::panic_error::pe
   3: <cargo_test_support::Execs>::run
   4: testsuite::package::reserved_windows_name
   5: <<std::hash::random::RandomState>::new::KEYS::{closure#1} as core::ops::function::FnOnce<(core::option::Option<&mut core::option::Option<core::cell::Cell<(u64, u64)>>>,)>>::call_once
   6: test::run_test
   7: test::run_test
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

bors · 2025-08-12T12:36:18Z

💔 Test failed - checks-actions

nikic · 2025-08-12T12:50:32Z

@bors retry Same reserved_windows_name failure also happened on #145161 recently.

rustbot assigned nnethercote Aug 8, 2025

This comment has been minimized.

Sign in to view

rust-bors bot added a commit that referenced this pull request Aug 8, 2025

Auto merge of #145093 - nikic:dead-on-return, r=<try>

7642ceb

Set dead_on_return attribute for indirect arguments

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 8, 2025

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 Aug 8, 2025

erikdesjardins reviewed Aug 8, 2025

View reviewed changes

nnethercote 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 Aug 8, 2025

Set dead_on_return attribute for indirect arguments

ebef9d7

Set the dead_on_return attribute (added in LLVM 21) for arguments that are passed indirectly, but not byval. This indicates that the value of the argument on return does not matter, enabling additional dead store elimination.

nikic force-pushed the dead-on-return branch from 9c1bbfa to ebef9d7 Compare August 11, 2025 10:39

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Aug 11, 2025

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Aug 12, 2025

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 12, 2025

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Aug 12, 2025

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 12, 2025

Set dead_on_return attribute for indirect arguments #145093

Are you sure you want to change the base?

Set dead_on_return attribute for indirect arguments #145093

Conversation

nikic commented Aug 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Aug 8, 2025

Uh oh!

nikic commented Aug 8, 2025

Uh oh!

This comment has been minimized.

rust-bors bot commented Aug 8, 2025

Uh oh!

rust-bors bot commented Aug 8, 2025

Uh oh!

This comment has been minimized.

rust-timer commented Aug 8, 2025

Overall result: no relevant changes - no action needed

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

erikdesjardins Aug 8, 2025

Choose a reason for hiding this comment

Uh oh!

nikic Aug 8, 2025

Choose a reason for hiding this comment

Uh oh!

nnethercote commented Aug 8, 2025

Uh oh!

nikic commented Aug 11, 2025

Uh oh!

nnethercote commented Aug 11, 2025

Uh oh!

nikic commented Aug 11, 2025

Uh oh!

bors commented Aug 11, 2025

Uh oh!

bors commented Aug 12, 2025

Uh oh!

rust-log-analyzer commented Aug 12, 2025

Uh oh!

bors commented Aug 12, 2025

Uh oh!

nikic commented Aug 12, 2025

Uh oh!

bors commented Aug 12, 2025

Uh oh!

rust-log-analyzer commented Aug 12, 2025

Uh oh!

bors commented Aug 12, 2025

Uh oh!

nikic commented Aug 12, 2025

Uh oh!

Uh oh!

nikic commented Aug 8, 2025 •

edited

Loading