try to make rustc-dev work with and without LTO by RalfJung · Pull Request #161535 · rust-lang/rust

RalfJung · 2026-08-22T14:57:01Z

View all comments

This might fix the issues introduced by #161260.

RalfJung · 2026-08-22T14:57:21Z

@bors try jobs=dist*linux

try to make rustc-dev work with and without LTO try-job: dist*linux

weihanglo · 2026-08-22T15:08:11Z

try jobs=dist*linux

TIL glob is supported.

Zoxc · 2026-08-22T15:39:50Z

-            // Compiling C deps, like jemalloc and llvm-wrapper, should be with
-            // the same LTO mode as the Rust code they are linked into.
+            // Compiling C deps, like jemalloc and llvm-wrapper, should be with the same LTO mode as
+            // the Rust code they are linked into. We don't pass the checks cc-rs uses to


should be with the same LTO mode as the Rust code they are linked into
Just a note that this doesn't seem very true.

View changes since the review

FWIW, the discussion happens in #t-compiler > Analysing broken jemalloc rlib

Just a note that this doesn't seem very true.

I just reformatted that comment, I don't know if it is correct.
What would be more correct?

weihanglo · 2026-08-22T16:30:29Z

Should we do something like this after the let lto_flags = ...?

            if lto_cflag.is_some()
                && builder.config.bootstrap_override_lld.is_used()
                && !target.is_msvc()
            {
                self.rustflags.arg("-Clink-args=-Wl,--fat-lto-objects");
            }

I am testing dist build on my machine still though.

rust-bors · 2026-08-22T17:17:39Z

☀️ Try build successful (CI)
Build commit: a4edfb9 (a4edfb99457ab3be0979b69702c5a12bb7d12269)
Base parent: b0ca9c7 (b0ca9c712455d865c3bf0bf5325a3e79677a2d97)

weihanglo · 2026-08-22T17:44:37Z

I guess we want a perf build

@rust-timer build a4edfb9

RalfJung · 2026-08-22T18:11:14Z

Should we do something like this after the let lto_flags = ...?

I have no idea.

RalfJung · 2026-08-22T21:30:05Z

This did fix Miri but I guess it also broke LTO. I'll try the magic linker incantation you mentioned above then.

RalfJung · 2026-08-22T21:32:42Z

@bors try jobs=dist*linux
@rust-timer queue

try to make rustc-dev work with and without LTO try-job: dist*linux

RalfJung · 2026-08-22T21:33:31Z

+            // Make sure the linker actually uses the fat LTO part.
+            if lto_cflag.is_some()
+                && builder.config.bootstrap_override_lld.is_used()
+                && !target.is_msvc()


FWIW I don't understand why we need to check for MSVC here, given that above we already check for clang.

Also, when does it ever make sense to set -ffat-lto-objects without setting the link-args as well?

View changes since the review

IIRC clang-cl.exe still produces COFF object, right? So even it is clang-like we should not pass that link arg, as it is not supported yet.

https://llvm.org/docs/FatLTO.html#supported-file-formats

Or alternativelu we should just apply to x86_64 Linux as we previously discussed.

https://releases.llvm.org/22.1.0/tools/lld/docs/ReleaseNotes.html#coff-improvements

Oh! In LLVM 22 COFF gains the support of fat LTO.
Though according to bjorn3 we are still on 21 right now.

Also, when does it ever make sense to set -ffat-lto-objects without setting the link-args as well?

And yeah it probably makes little sense. We may be able to just figure out the minimal condition and coerce them in one of block

Maybe the entire code can be simplified to this?

let lto_cflag = if matches!(self.mode, Mode::Rustc | Mode::ToolRustcPrivate) && is_lto_stage(&self.compiler) && builder.cc_tool(target).is_like_clang() && builder.config.bootstrap_override_lld.is_used() // unsure if we should take care of non-dist config for downstream packagers { let lto_cflag = match builder.config.rust_lto { RustcLto::Thin => Some("-flto=thin -ffat-lto-objects"), RustcLto::Fat => Some("-flto=full -ffat-lto-objects"), RustcLto::ThinLocal | RustcLto::Off => None, }; if lto_cflag.is_some() { self.rustflags.arg("-Clink-args=-Wl,--fat-lto-objects"); } lto_cflag } else { None };

I copied that, and it seems to work. 🤷

weihanglo · 2026-08-22T23:25:40Z

So I got a result of libtikv_jemalloc_sys.rlib size on aarch64

native (pre PR 161260, c656540): 4.81 MB
bitcode only (PR 161260, 124c16e): 4.84 MB
fat LTO object (this PR, d2077da): 9.54 MB

All above are uncompressed size.

rust-bors · 2026-08-23T00:00:28Z

☀️ Try build successful (CI)
Build commit: 5db500f (5db500f538a1f4727c942992cbc3d05b437d5930)
Base parent: 78c04b6 (78c04b6a348438fb9396b8864cd34f92fb99cd53)

rust-timer · 2026-08-23T01:25:00Z

Finished benchmarking commit (5db500f): comparison URL.

Overall result: ❌ regressions - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

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

Max RSS (memory usage)

This perf run didn't have relevant results for this metric.

Cycles

Results (secondary -8.0%)

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)	1.3%	[1.3%, 1.3%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-12.7%	[-13.0%, -12.4%]	2
All ❌✅ (primary)	-	-	0

Binary size

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

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)	-	-	0
Improvements ✅ (primary)	-0.0%	[-0.0%, -0.0%]	5
Improvements ✅ (secondary)	-0.0%	[-0.0%, -0.0%]	1
All ❌✅ (primary)	-0.0%	[-0.0%, -0.0%]	5

Bootstrap: 468.574s -> 466.803s (-0.38%)
Artifact size: 400.20 MiB -> 400.15 MiB (-0.01%)

weihanglo · 2026-08-23T03:48:27Z

And for x86_64 uncompressed libtikv_jemalloc_sys.rlib:

no native state
bitcode only (PR 161260, 124c16e): 4.82 MB
fat LTO (this PR trybuild, 5db500f): 9.41 MB

RalfJung · 2026-08-23T06:31:09Z

On aarch64-linux, the rustc-dev component went from 125 MiB to 128 MiB.
On x86-64-linux, it went from 142 MiB to 143 MiB.

That seems fine? That's compressed size obviously though, so this may consume a lot more disk space if it compresses very well. If I check locally the whole folder lib/rustlib/x86_64-unknown-linux-gnu/lib it went from 773 MiB to 777 MiB... looks like that tikv file is the only file that got bigger.^^

RalfJung · 2026-08-23T11:51:35Z

Yeah I think this should be ready now. I don't actually understand why the conditions are how they are, I just copied that from @weihanglo. ;) I kicked off another perf run to be sure but that last reorganization shouldn't change anything.

r? @Kobzol

rustbot · 2026-08-23T11:51:39Z

Kobzol is not on the review rotation at the moment.
They may take a while to respond.

weihanglo

Looks good to me as I introduced the regression and proposed the change :D

View changes since this review

rust-timer · 2026-08-23T13:36:00Z

Finished benchmarking commit (a259cfd): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

This perf run didn't have relevant results for this metric.

Max RSS (memory usage)

Results (secondary 3.2%)

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)	3.2%	[3.2%, 3.2%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -2.1%, secondary -2.8%)

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)	-	-	0
Improvements ✅ (primary)	-2.1%	[-2.1%, -2.1%]	1
Improvements ✅ (secondary)	-2.8%	[-2.8%, -2.8%]	1
All ❌✅ (primary)	-2.1%	[-2.1%, -2.1%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 469.531s -> 468.171s (-0.29%)
Artifact size: 400.26 MiB -> 400.30 MiB (0.01%)

Kobzol · 2026-08-24T06:24:32Z

                && is_lto_stage(&self.compiler)
                && builder.cc_tool(target).is_like_clang()
+                // unsure if we should take care of non-dist config for downstream packagers
+                && builder.config.bootstrap_override_lld.is_used()


Hmm, this config is only used for speeding up the build process itself, and it should in fact be completely unnecessary on x64 now, to the point where we should remove it from its CI config.

Why is this condition needed? Isn't the whole point of the config below that the resulting code will be compilable even without LLD?

View changes since the review

I'm totally fine with removing it, all these conditions are a mystery to me. ;)

Kobzol · 2026-08-24T06:35:26Z

@bors try @rust-timer queue

try to make rustc-dev work with and without LTO

RalfJung · 2026-08-24T06:50:48Z

+                };
+                // Make sure the linker actually uses the fat LTO part.
+                if lto_cflag.is_some() {
+                    self.rustflags.arg("-Clink-args=-Wl,--fat-lto-objects");


IIUC, this flag is needed to make the linker even recognize the bitcode in fat LTO objects. It is strange that this is not the default (unlike LTO for "thin" objects which apparently is the default). Does this mean we're also not LTO'ing jemalloc for the distributed Miri, Clippy etc any more?

View changes since the review

I'll do a Clippy benchmark after the current one finishes to check.

rust-bors · 2026-08-24T07:34:34Z

☀️ Try build successful (CI)
Build commit: 0ff57e8 (0ff57e8a1e520723d1162f33201a36e14c9db339)
Base parent: da51146 (da5114692c9ebe46b869488c5f34f92eb10b98c1)

rust-timer · 2026-08-24T08:51:41Z

Finished benchmarking commit (0ff57e8): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

This perf run didn't have relevant results for this metric.

Max RSS (memory usage)

Results (secondary 7.5%)

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)	7.5%	[7.5%, 7.5%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -2.4%, secondary -0.9%)

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)	6.5%	[6.5%, 6.5%]	1
Improvements ✅ (primary)	-2.4%	[-2.4%, -2.4%]	1
Improvements ✅ (secondary)	-3.4%	[-4.6%, -2.3%]	3
All ❌✅ (primary)	-2.4%	[-2.4%, -2.4%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 469.244s -> 471.258s (0.43%)
Artifact size: 400.19 MiB -> 400.31 MiB (0.03%)

Kobzol · 2026-08-24T08:54:41Z

@bors try jobs=dist-x86_64-linux profiles=clippy

Kobzol · 2026-08-24T08:54:53Z

@bors try jobs=dist-x86_64-linux @rust-timer queue profiles=clippy

rust-timer · 2026-08-24T08:54:56Z

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

try to make rustc-dev work with and without LTO try-job: dist-x86_64-linux

rust-bors · 2026-08-24T10:34:17Z

☀️ Try build successful (CI)
Build commit: 85c799e (85c799e5714eeb9d5b29f2e950896cf34d26f2c3)
Base parent: 04a3cad (04a3cad16c522be4fd15ca76d3c119c14be68954)

rust-timer · 2026-08-24T10:34:21Z

Queued 85c799e with parent 04a3cad, future comparison URL.
There is currently 1 preceding artifact in the queue.
It will probably take at least ~1.4 hours until the benchmark run finishes.

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. T-bootstrap Relevant to the bootstrap subteam: Rust's build system (x.py and src/bootstrap) labels Aug 22, 2026

rust-bors Bot pushed a commit that referenced this pull request Aug 22, 2026

Auto merge of #161535 - RalfJung:aarch64-lto-compat, r=<try>

a4edfb9

try to make rustc-dev work with and without LTO try-job: dist*linux

This comment has been minimized.

Sign in to view

Zoxc reviewed Aug 22, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

This comment was marked as outdated.

Sign in to view

rustbot added the perf-regression Performance regression. label Aug 22, 2026

RalfJung force-pushed the aarch64-lto-compat branch from 8975695 to d2077da Compare August 22, 2026 21:32

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 Aug 22, 2026

rust-bors Bot pushed a commit that referenced this pull request Aug 22, 2026

Auto merge of #161535 - RalfJung:aarch64-lto-compat, r=<try>

5db500f

try to make rustc-dev work with and without LTO try-job: dist*linux

RalfJung commented Aug 22, 2026

View reviewed changes

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 Aug 23, 2026

rustbot assigned Kobzol and unassigned jieyouxu Aug 23, 2026

RalfJung added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 23, 2026

weihanglo approved these changes Aug 23, 2026

View reviewed changes

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 23, 2026

Kobzol reviewed Aug 24, 2026

View reviewed changes

make rustc-dev work with and without LTO

85cc84f

RalfJung force-pushed the aarch64-lto-compat branch from 60ea0f4 to 85cc84f Compare August 24, 2026 06:34

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 Aug 24, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Aug 24, 2026

Auto merge of #161535 - RalfJung:aarch64-lto-compat, r=<try>

0ff57e8

try to make rustc-dev work with and without LTO

RalfJung commented Aug 24, 2026

View reviewed changes

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 24, 2026

This comment was marked as outdated.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 24, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Aug 24, 2026

Auto merge of #161535 - RalfJung:aarch64-lto-compat, r=<try>

85c799e

try to make rustc-dev work with and without LTO try-job: dist-x86_64-linux

Uh oh!

Uh oh!

Conversation

RalfJung commented Aug 22, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RalfJung commented Aug 22, 2026

Uh oh!

This comment has been minimized.

weihanglo commented Aug 22, 2026

Uh oh!

Zoxc Aug 22, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

weihanglo commented Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rust-bors Bot commented Aug 22, 2026

Uh oh!

weihanglo commented Aug 22, 2026

Uh oh!

This comment has been minimized.

RalfJung commented Aug 22, 2026

Uh oh!

This comment was marked as outdated.

RalfJung commented Aug 22, 2026

Uh oh!

RalfJung commented Aug 22, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

RalfJung Aug 22, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

weihanglo Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

weihanglo Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

weihanglo commented Aug 22, 2026

Uh oh!

rust-bors Bot commented Aug 23, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Aug 23, 2026

Overall result: ❌ regressions - no action needed

Uh oh!

weihanglo commented Aug 23, 2026

Uh oh!

RalfJung commented Aug 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RalfJung commented Aug 23, 2026

Uh oh!

rustbot commented Aug 23, 2026

Uh oh!

weihanglo left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rust-timer commented Aug 23, 2026

Overall result: no relevant changes - no action needed

Uh oh!

Kobzol Aug 24, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

RalfJung commented Aug 22, 2026 •

edited by rustbot

Loading

Zoxc Aug 22, 2026 •

edited by rustbot

Loading

weihanglo commented Aug 22, 2026 •

edited

Loading

RalfJung Aug 22, 2026 •

edited by rustbot

Loading

weihanglo Aug 22, 2026 •

edited

Loading

weihanglo Aug 22, 2026 •

edited

Loading

RalfJung commented Aug 23, 2026 •

edited

Loading

weihanglo left a comment •

edited by rustbot

Loading

Kobzol Aug 24, 2026 •

edited by rustbot

Loading

RalfJung Aug 24, 2026 •

edited by rustbot

Loading