We don't need `NonNull::as_ptr` debuginfo #133899

scottmcm · 2024-12-05T08:48:14Z

In order to stop pessimizing the use of local variables in core, skip debug info for MIR temporaries in tiny (single-BB) functions.

For functions as simple as this -- Pin::new, etc -- nobody every actually wants debuginfo for them in the first place. They're more like intrinsics than real functions, and stepping over them is good.

scottmcm · 2024-12-05T08:48:35Z

@bors try @rust-timer queue

[EXPERIMENT] We don't need `NonNull::as_ptr` debuginfo Stop pessimizing the use of local variables in core by skipping debug info for MIR temporaries in tiny (single-BB) functions. For functions as simple as this -- `Pin::new`, etc -- nobody every actually wants debuginfo for them in the first place. They're more like intrinsics than real functions, and stepping over them is good. r? ghost

bors · 2024-12-05T08:49:47Z

⌛ Trying commit 49b1b09 with merge fe36381...

fmease · 2024-12-05T10:20:05Z

sync @bors r-

bors · 2024-12-05T10:35:06Z

☀️ Try build successful - checks-actions
Build commit: fe36381 (fe36381f279b8d83a2a75289c291c8962f09421f)

rust-timer · 2024-12-05T15:39:56Z

Finished benchmarking commit (fe36381): comparison URL.

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

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 the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	0.9%	[0.9%, 0.9%]	1
Regressions ❌ (secondary)	0.4%	[0.2%, 1.2%]	6
Improvements ✅ (primary)	-1.0%	[-2.2%, -0.1%]	35
Improvements ✅ (secondary)	-3.0%	[-3.2%, -2.7%]	2
All ❌✅ (primary)	-0.9%	[-2.2%, 0.9%]	36

Max RSS (memory usage)

Results (primary -2.3%, secondary -3.4%)

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%	[1.2%, 2.6%]	3
Regressions ❌ (secondary)	2.9%	[2.9%, 2.9%]	1
Improvements ✅ (primary)	-3.5%	[-7.2%, -0.7%]	10
Improvements ✅ (secondary)	-6.5%	[-8.0%, -4.9%]	2
All ❌✅ (primary)	-2.3%	[-7.2%, 2.6%]	13

Cycles

Results (primary -1.3%, secondary -1.3%)

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)	1.1%	[1.1%, 1.1%]	1
Regressions ❌ (secondary)	1.8%	[1.8%, 1.8%]	1
Improvements ✅ (primary)	-1.6%	[-2.4%, -1.0%]	9
Improvements ✅ (secondary)	-2.8%	[-3.5%, -2.1%]	2
All ❌✅ (primary)	-1.3%	[-2.4%, 1.1%]	10

Binary size

Results (primary -1.4%, secondary -0.8%)

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)	-1.4%	[-6.0%, -0.0%]	90
Improvements ✅ (secondary)	-0.8%	[-6.6%, -0.0%]	44
All ❌✅ (primary)	-1.4%	[-6.0%, -0.0%]	90

Bootstrap: 768.023s -> 767.749s (-0.04%)
Artifact size: 330.84 MiB -> 330.73 MiB (-0.03%)

rustbot · 2024-12-05T16:19:26Z

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

scottmcm · 2024-12-05T16:21:11Z

Oh, I'm quite happy with those perf results :) 560,934 bytes off hyper-debug.

r? mir

Kobzol · 2024-12-05T20:57:12Z

The (especially binary size) wins look awesome! Just thinking out loud, I wonder if there's a way to evaluate the debugging impact of these changes though? I noticed that in recent Rust versions debugging stepping regressed for me, with stepping over (a line without await) suddenly starting to jump into some assembly instructions or going inside functions unnecessarily.

scottmcm · 2024-12-05T22:21:04Z

My saying "strip" here might have made it sound like it's doing more than it is.

The following are unchanged, and should continue to work exactly as before:

Anything in your own code, as this behaviour is only enabled while building the sysroot (core+alloc+std)
Anything related to line/column information about the execution, because this still preserves all the spans of the statements in the function
When you step into a core function that loops, you can still debug that loop, since any function with more than one basic block will be unchanged by this PR.
When you step into a core function from your own code, looking at the values of the parameters to the function will still work, as this always leaves that debuginfo intact as well

What doesn't work:

When you've stepped into a core function, looking at the values of local variables that are not parameters
If you step into a core function from another core function (like stepping into Pin::new_unchecked from Pin::new), you might not be able to see the value of the sub-function's parameters.

For a specific example, my argument is that while it's true that there's a local in NonNull::dangling

rust/library/core/src/ptr/non_null.rs

Lines 111 to 117 in c94848c

    
           pub const fn dangling() -> Self { 
        
               // SAFETY: ptr::dangling_mut() returns a non-null well-aligned pointer. 
        
               unsafe { 
        
                   let ptr = crate::ptr::dangling_mut::<T>(); 
        
                   NonNull::new_unchecked(ptr) 
        
               } 
        
           }

I don't think anyone's ever cared about that in the debuginfo, since it's a trivial straight-line function (even though we do have six inlined function call scopes in it today).

Kobzol · 2024-12-05T22:27:07Z

I see, thanks for clarifying! That makes sense, I agree that this kind of local variable debuginfo is not very useful. In fact, I think that it would be nice if we had a way to skip generating such information for all libraries that a crate depends on, to improve compile times and reduce binary size of debug builds, as these are relatively rarely needed for actual stepping (as opposed to the user's own code). But that's a separate discussion.

oli-obk · 2024-12-06T08:08:43Z

tests/mir-opt/inline/inline_coroutine.main.Inline.panic-abort.diff

@@ -16,7 +16,6 @@
 +     scope 3 (inlined Pin::<&mut {coroutine@$DIR/inline_coroutine.rs:20:5: 20:8}>::new) {
 +         debug pointer => _3;


Shouldn't this one also be gone? Pin::new (after inlining the function call in its body and stripping debug info) is just the same as Pin::new_unchecked, but its debug info seems to stay around.

pointer is the parameter to Pin::new, so it's not removed

rust/library/core/src/pin.rs

Line 1191 in cdeddae

pub const fn new(pointer: Ptr) -> Pin<Ptr> {

(Rather than if it were a local in the function)

But isn't the same true for the pointer arg of new_unchecked?

Ah, if the currently-being-optimized function were in libcore, the arg of new would also be gone...

Makes sense, but subtle

Yeah, definitely subtle. We could be less subtle by just removing all the debuginfo from trivial functions, but I thought it would be worth allowing at least that outer bit so that if you're unsure what you're calling into core with, that first step in having those arguments visible seemed potentially useful.

Basically I'm trying to make as small a change as I can while still avoiding the bad incentive of local variables in core making the compiler slower.

oli-obk · 2024-12-09T18:47:48Z

@bors r+

bors · 2024-12-09T18:47:50Z

📌 Commit 49b1b09 has been approved by oli-obk

It is now in the queue for this repository.

scottmcm · 2024-12-11T16:14:38Z

Ok, failure was spurious, so back into the queue.
@bors r=oli-obk

bors · 2024-12-11T16:14:42Z

📌 Commit a7fc76a has been approved by oli-obk

It is now in the queue for this repository.

jyn514 · 2024-12-12T09:59:08Z

In fact, I think that it would be nice if we had a way to skip generating such information for all libraries that a crate depends on, to improve compile times and reduce binary size of debug builds, as these are relatively rarely needed for actual stepping (as opposed to the user's own code)

FWIW i use a debugger on rust binaries somewhat frequently and almost always i’m doing so because one of my dependencies does something i don’t understand or expect. so removing that debuginfo would make my life harder.

if you intend this to be the default for profile = dev that seems ok. but i really would want it to have an opt-out.

jyn514 · 2024-12-12T10:00:57Z

i think you could solve the problem you’re talking about more directly if gdb/lldb had a command for “never step into functions defined in this crate”. maybe that already exists today, i’m not sure.

Kobzol · 2024-12-12T10:00:58Z

I meant it as the opposite, i.e. by default build all debuginfo, but have an opt-in mechanism to skip debuginfo of dependencies.

jyn514 · 2024-12-12T10:03:45Z

I meant it as the opposite, i.e. by default build all debuginfo, but have an opt-in mechanism to skip debuginfo of dependencies.

you can already do this today, no?

[profile.release]
debug = true

// disable debug symbols for all packages except this one
[profile.release.package."*"]
debug = false

jyn514 · 2024-12-12T10:11:28Z

i think you could solve the problem you’re talking about more directly if gdb/lldb had a command for “never step into functions defined in this crate”. maybe that already exists today, i’m not sure.

yeah here we go: https://sourceware.org/gdb/current/onlinedocs/gdb.html/Skipping-Over-Functions-and-Files.html
we could set this by default in rust-gdb so it skips "uninteresting" frames.

Kobzol · 2024-12-12T10:54:20Z

I meant it as the opposite, i.e. by default build all debuginfo, but have an opt-in mechanism to skip debuginfo of dependencies.

you can already do this today, no?
[profile.release]
debug = true

// disable debug symbols for all packages except this one
[profile.release.package."*"]
debug = false

Ah, I forgot about this, good point!

yeah here we go: https://sourceware.org/gdb/current/onlinedocs/gdb.html/Skipping-Over-Functions-and-Files.html
we could set this by default in rust-gdb so it skips "uninteresting" frames.

Yeah, that's useful, I think I used this in the IJ Rust plugin when implementing an option to skip stepping through stdlib sources. Hard to say which functions to mark as being auto skippable though.

We don't need `NonNull::as_ptr` debuginfo In order to stop pessimizing the use of local variables in core, skip debug info for MIR temporaries in tiny (single-BB) functions. For functions as simple as this -- `Pin::new`, etc -- nobody every actually wants debuginfo for them in the first place. They're more like intrinsics than real functions, and stepping over them is good.

bors · 2024-12-13T03:50:08Z

⌛ Testing commit a7fc76a with merge 7a5dab9...

rust-log-analyzer · 2024-12-13T05:19:03Z

The job i686-mingw failed! Check out the build log: (web) (plain)

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

bors · 2024-12-13T05:20:04Z

💔 Test failed - checks-actions

scottmcm · 2024-12-13T06:08:07Z

That job just passed in test-with-try above #133899 (comment) so sounds spurious again :/

@bors retry

bors · 2024-12-13T08:32:25Z

⌛ Testing commit a7fc76a with merge dd436ae...

bors · 2024-12-13T11:31:19Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing dd436ae to master...

rust-timer · 2024-12-13T12:48:28Z

Finished benchmarking commit (dd436ae): comparison URL.

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

Our benchmarks found a performance regression caused by this PR.
This might be an actual regression, but it can also be just noise.

Next Steps:

If the regression was expected or you think it can be justified,
please write a comment with sufficient written justification, and add
@rustbot label: +perf-regression-triaged to it, to mark the regression as triaged.
If you think that you know of a way to resolve the regression, try to create
a new PR with a fix for the regression.
If you do not understand the regression or you think that it is just noise,
you can ask the @rust-lang/wg-compiler-performance working group for help (members of this group
were already notified of this PR).

@rustbot label: +perf-regression
cc @rust-lang/wg-compiler-performance

Instruction count

This is the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	0.6%	[0.2%, 0.9%]	2
Regressions ❌ (secondary)	0.5%	[0.2%, 1.1%]	4
Improvements ✅ (primary)	-0.9%	[-2.1%, -0.3%]	34
Improvements ✅ (secondary)	-2.7%	[-3.0%, -2.5%]	2
All ❌✅ (primary)	-0.9%	[-2.1%, 0.9%]	36

Max RSS (memory usage)

Results (primary -2.3%, secondary -4.7%)

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%	[1.0%, 3.2%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.3%	[-7.0%, -0.9%]	9
Improvements ✅ (secondary)	-4.7%	[-7.9%, -2.3%]	3
All ❌✅ (primary)	-2.3%	[-7.0%, 3.2%]	11

Cycles

Results (primary -1.2%, secondary -2.4%)

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)	1.1%	[1.1%, 1.1%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.4%	[-2.3%, -0.8%]	11
Improvements ✅ (secondary)	-2.4%	[-2.9%, -1.9%]	2
All ❌✅ (primary)	-1.2%	[-2.3%, 1.1%]	12

Binary size

Results (primary -1.3%, secondary -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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.3%	[-6.1%, -0.0%]	93
Improvements ✅ (secondary)	-0.9%	[-6.6%, -0.0%]	36
All ❌✅ (primary)	-1.3%	[-6.1%, -0.0%]	93

Bootstrap: 769.59s -> 770.497s (0.12%)
Artifact size: 331.01 MiB -> 330.36 MiB (-0.20%)

Kobzol · 2024-12-23T14:02:38Z

More improvements than regressions, and in particular very nice binary size wins.

@rustbot label: +perf-regression-triaged

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 Dec 5, 2024

bors 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 Dec 5, 2024

This comment has been minimized.

Sign in to view

fmease 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 Dec 5, 2024

rustbot added perf-regression Performance regression. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Dec 5, 2024

scottmcm marked this pull request as ready for review December 5, 2024 16:19

rustbot assigned matthewjasper Dec 5, 2024

scottmcm changed the title ~~[EXPERIMENT] We don't need NonNull::as_ptr debuginfo~~ We don't need NonNull::as_ptr debuginfo Dec 5, 2024

oli-obk reviewed Dec 6, 2024

View reviewed changes

scottmcm added the perf-regression-triaged The performance regression has been triaged. label Dec 9, 2024

ChrisDenton added CI-spurious-fail-mingw CI spurious failure: target env mingw and removed CI-spurious-fail-msvc CI spurious failure: target env msvc labels Dec 11, 2024

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 Dec 11, 2024

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 Dec 13, 2024

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 Dec 13, 2024

bors added the merged-by-bors This PR was explicitly merged by bors. label Dec 13, 2024

bors merged commit dd436ae into rust-lang:master Dec 13, 2024
7 checks passed

rustbot added this to the 1.85.0 milestone Dec 13, 2024

bors mentioned this pull request Dec 13, 2024

[mir-opt] GVN some more transmute cases #133324

Merged

scottmcm deleted the strip-mir-debuginfo branch December 14, 2024 02:10

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

We don't need `NonNull::as_ptr` debuginfo #133899

We don't need `NonNull::as_ptr` debuginfo #133899

scottmcm commented Dec 5, 2024 •

edited

Loading

scottmcm commented Dec 5, 2024

This comment has been minimized.

bors commented Dec 5, 2024

fmease commented Dec 5, 2024

bors commented Dec 5, 2024

This comment has been minimized.

rust-timer commented Dec 5, 2024

rustbot commented Dec 5, 2024

scottmcm commented Dec 5, 2024 •

edited

Loading

Kobzol commented Dec 5, 2024

scottmcm commented Dec 5, 2024 •

edited

Loading

Kobzol commented Dec 5, 2024

oli-obk Dec 6, 2024

scottmcm Dec 6, 2024

oli-obk Dec 7, 2024

scottmcm Dec 8, 2024

oli-obk commented Dec 9, 2024

bors commented Dec 9, 2024

scottmcm commented Dec 11, 2024

bors commented Dec 11, 2024

jyn514 commented Dec 12, 2024

jyn514 commented Dec 12, 2024

Kobzol commented Dec 12, 2024

jyn514 commented Dec 12, 2024

jyn514 commented Dec 12, 2024

Kobzol commented Dec 12, 2024

bors commented Dec 13, 2024

rust-log-analyzer commented Dec 13, 2024

bors commented Dec 13, 2024

scottmcm commented Dec 13, 2024

bors commented Dec 13, 2024

bors commented Dec 13, 2024

rust-timer commented Dec 13, 2024

Kobzol commented Dec 23, 2024

		@@ -16,7 +16,6 @@
		+ scope 3 (inlined Pin::<&mut {coroutine@$DIR/inline_coroutine.rs:20:5: 20:8}>::new) {
		+ debug pointer => _3;

We don't need NonNull::as_ptr debuginfo #133899

We don't need NonNull::as_ptr debuginfo #133899

Conversation

scottmcm commented Dec 5, 2024 • edited Loading

scottmcm commented Dec 5, 2024

This comment has been minimized.

bors commented Dec 5, 2024

fmease commented Dec 5, 2024

bors commented Dec 5, 2024

This comment has been minimized.

rust-timer commented Dec 5, 2024

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

rustbot commented Dec 5, 2024

scottmcm commented Dec 5, 2024 • edited Loading

Kobzol commented Dec 5, 2024

scottmcm commented Dec 5, 2024 • edited Loading

Kobzol commented Dec 5, 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

oli-obk commented Dec 9, 2024

bors commented Dec 9, 2024

scottmcm commented Dec 11, 2024

bors commented Dec 11, 2024

jyn514 commented Dec 12, 2024

jyn514 commented Dec 12, 2024

Kobzol commented Dec 12, 2024

jyn514 commented Dec 12, 2024

jyn514 commented Dec 12, 2024

Kobzol commented Dec 12, 2024

bors commented Dec 13, 2024

rust-log-analyzer commented Dec 13, 2024

bors commented Dec 13, 2024

scottmcm commented Dec 13, 2024

bors commented Dec 13, 2024

bors commented Dec 13, 2024

rust-timer commented Dec 13, 2024

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

Kobzol commented Dec 23, 2024

We don't need `NonNull::as_ptr` debuginfo #133899

We don't need `NonNull::as_ptr` debuginfo #133899

scottmcm commented Dec 5, 2024 •

edited

Loading

scottmcm commented Dec 5, 2024 •

edited

Loading

scottmcm commented Dec 5, 2024 •

edited

Loading