Only apply LTO to rustdoc at stage 2

[Kobzol]

It doesn't make much sense at stage 1, and it was broken anyway. This was implemented in #135832. The issue with LTO and stage 1 rustdoc was reported here.

r? @onur-ozkan

[Kobzol]

@bors try @rust-timer queue

Just to make sure that it doesn't regress perf.

[bors]

⌛ Trying commit 046385b with merge f02aecb...

[jieyouxu]

If LTO applied to rustdoc is broken at stage 1, won't it still be broken if applied at stage 2...? 🤔

[Kobzol]

It is only "broken" because we also skip LTO in stage 1 for librustc_driver.so and friends, upon which rustdoc depends. So we need to do the same for rustdoc too, otherwise it will miss the LTO bitcode.

[saethlin]

because we also skip LTO in stage 1 for librustc_driver.so and friends

So does this mean my lto = "thin" setting in config.toml is ineffectual for x build library and I need to do a stage2 to build to see optimizations that are remotely similar to a distributed build?

[jieyouxu]

That seems strange in general, if that is indeed the case... 🤔

[Kobzol]

because we also skip LTO in stage 1 for librustc_driver.so and friends

So does this mean my lto = "thin" setting in config.toml is ineffectual for x build library and I need to do a stage2 to build to see optimizations that are remotely similar to a distributed build?

If you're talking about LTO optimizing librustc_driver.so, then yes, it is currently only performed for stage 2, not lower stages. The CI performance is much more affected by PGO + BOLT (~30%) than LTO (5-10%) though.

[bors]

☀️ Try build successful - checks-actions
Build commit: f02aecb (f02aecb6948aa43ec48735055369961182d0a915)

[onur-ozkan]

LGTM other than a non-blocker nit

[onur-ozkan]

@bors r+ rollup

[bors]

📌 Commit cfa3518 has been approved by onur-ozkan

It is now in the queue for this repository.

[lqd]

perf results will be ready in 20 mins

[rust-timer]

Finished benchmarking commit (f02aecb): 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 -1.1%, secondary 0.3%)

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)	-	-	0
Regressions ❌ (secondary)	6.4%	[6.4%, 6.4%]	1
Improvements ✅ (primary)	-1.1%	[-1.1%, -1.1%]	1
Improvements ✅ (secondary)	-2.7%	[-2.8%, -2.7%]	2
All ❌✅ (primary)	-1.1%	[-1.1%, -1.1%]	1

Cycles

Results (primary 2.7%, secondary 3.2%)

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)	2.7%	[2.1%, 3.3%]	5
Regressions ❌ (secondary)	3.2%	[2.5%, 3.9%]	15
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.7%	[2.1%, 3.3%]	5

Binary size

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

Bootstrap: 778.541s -> 779.409s (0.11%)
Artifact size: 328.82 MiB -> 328.83 MiB (0.00%)

[lqd]

@Kobzol: it's likely unrelated because they're not doc benchmarks, but look at these cycles results what the eff

[Kobzol]

Yeah it's most probably unrelated. Probably solar flares.

[bors]

⌛ Testing commit cfa3518 with merge 433a03d...

[rust-log-analyzer]

A job failed! Check out the build log: (web) (plain)

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

[bors]

💔 Test failed - checks-actions

[onur-ozkan]

A job failed! Check out the build log: (web) (plain)

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

What the hack?

@bors retry

[jieyouxu]

There was I think a github actions outage at that time or sth. Or sometimes a job just doesn't get picked up for whatever reason.