perf experiment for #147185

[RalfJung]

Trying to figure out why #147185 regresses syn builds so much.

[rustbot]

r? @BoxyUwU

rustbot has assigned @BoxyUwU.
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

[RalfJung]

@bors try
@rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: f3ba3f5 (f3ba3f55843086cd3e7e91cfccec0c492804392d, parent: 907705abea355952d84ed39cea256ed4570bdda6)

[rust-timer]

Finished benchmarking commit (f3ba3f5): comparison URL.

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

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.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

@bors rollup=never
@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)	0.4%	[0.4%, 0.5%]	4
Improvements ✅ (primary)	-1.7%	[-2.9%, -0.3%]	9
Improvements ✅ (secondary)	-1.0%	[-1.0%, -1.0%]	1
All ❌✅ (primary)	-1.7%	[-2.9%, -0.3%]	9

Max RSS (memory usage)

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

Cycles

Results (primary 2.0%, secondary 0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.0%	[2.0%, 2.0%]	1
Regressions ❌ (secondary)	4.1%	[2.6%, 5.6%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.8%	[-4.9%, -2.7%]	2
All ❌✅ (primary)	2.0%	[2.0%, 2.0%]	1

Binary size

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

Bootstrap: 474.692s -> 475.969s (0.27%)
Artifact size: 390.07 MiB -> 390.07 MiB (-0.00%)

[RalfJung]

I have no explanation how this can make such a big difference for syn... even if syn has thousands of repr(transparent) types, this should hardly change anything as it just checks one more flag?

[RalfJung]

Cc @rust-lang/wg-compiler-performance in case someone has any idea how this could be happening.

[RalfJung]

@bors try
@rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 9fcd3c8 (9fcd3c810e6e2bb555bfca0af2cc081141a7a911, parent: c6d42d774d1edfb270b8faaefacc67c213b0260b)

[rust-timer]

Finished benchmarking commit (9fcd3c8): 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 (primary 2.1%, secondary 1.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.1%	[2.1%, 2.1%]	1
Regressions ❌ (secondary)	3.1%	[2.2%, 4.0%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.0%	[-3.0%, -3.0%]	1
All ❌✅ (primary)	2.1%	[2.1%, 2.1%]	1

Cycles

Results (primary 2.1%, secondary -0.4%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.1%	[2.1%, 2.1%]	1
Regressions ❌ (secondary)	2.3%	[2.0%, 2.6%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.1%	[-3.4%, -2.7%]	2
All ❌✅ (primary)	2.1%	[2.1%, 2.1%]	1

Binary size

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

Bootstrap: 475.179s -> 475.205s (0.01%)
Artifact size: 390.26 MiB -> 390.27 MiB (0.00%)

[RalfJung]

@bors try
@rust-timer queue

[rust-timer]

Awaiting bors try build completion.

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

[rust-bors]

💔 Test for 3a5a811 failed: CI

[Kobzol]

https://perf.rust-lang.org/detailed-query.html?commit=f3ba3f55843086cd3e7e91cfccec0c492804392d&benchmark=syn-2.0.101-check&scenario=incr-unchanged&base_commit=907705abea355952d84ed39cea256ed4570bdda6 there were some query changes, maybe that could be it? Cachegrind diff looks kinda interesting:

< -39,350,149 ???:
-6,659,324 <&mut serde_json::ser::Serializer<&mut alloc::boxed::Box<dyn std::io::Write + core::marker::Send>> as serde_core::ser::Serializer>::serialize_str
-4,215,955 <std::io::buffered::bufwriter::BufWriter<std::io::stdio::Stderr> as std::io::Write>::write_all
-2,730,341 <rustc_errors::styled_buffer::StyledBuffer>::putc
-2,580,516 <rustc_span::source_map::SourceMap>::lookup_char_pos
-2,511,566 rustc_errors::emitter::normalize_whitespace
-2,492,928 <rustc_errors::styled_buffer::StyledBuffer>::render
-2,307,220 <rustc_span::SourceFile>::get_line
-1,407,502 <rustc_errors::emitter::HumanEmitter>::draw_line
-1,249,210 _rjem_je_arena_ralloc
-942,782 <rustc_errors::styled_buffer::StyledBuffer>::ensure_lines
-868,556 realloc
-818,220 core::str::count::do_count_chars
-803,060 <serde_json::ser::Compound<&mut alloc::boxed::Box<dyn std::io::Write + core::marker::Send>, serde_json::ser::CompactFormatter> as serde_core::ser::SerializeStruct>::serialize_field::<usize>
-733,739 free
-691,479 <rustc_errors::emitter::HumanEmitter>::emit_messages_default_inner::{closure#0}
-567,521 core::fmt::write
-544,679 rustc_incremental::persist::load::setup_dep_graph
-535,045 <<rustc_errors::json::Diagnostic>::from_errors_diagnostic::BufWriter as std::io::Write>::write
-510,903 malloc
-464,106 <rustc_errors::json::DiagnosticSpan as serde_core::ser::Serialize>::serialize::<&mut serde_json::ser::Serializer<&mut alloc::boxed::Box<dyn std::io::Write + core::marker::Send>>>
-422,919 <anstyle::style::Style>::fmt_to
-326,780 <hashbrown::raw::RawTable<((usize, rustc_data_structures::stable_hasher::HashingControls), rustc_data_structures::fingerprint::Fingerprint)>>::reserve_rehash::<hashbrown::map::make_hasher<(usize, rustc_data_structures::stable_hasher::HashingControls), rustc_data_structures::fingerprint::Fingerprint, rustc_hash::FxBuildHasher>::{closure#0}>
326,776 <hashbrown::raw::RawTable<((*const (), rustc_data_structures::stable_hasher::HashingControls), rustc_data_structures::fingerprint::Fingerprint)>>::reserve_rehash::<hashbrown::map::make_hasher<(*const (), rustc_data_structures::stable_hasher::HashingControls), rustc_data_structures::fingerprint::Fingerprint, rustc_hash::FxBuildHasher>::{closure#0}>
-303,372 <std::io::default_write_fmt::Adapter<<rustc_errors::json::Diagnostic>::from_errors_diagnostic::BufWriter> as core::fmt::Write>::write_str
-261,576 <rustc_errors::json::DiagnosticSpanLine>::from_span
-258,560 <rustc_errors::styled_buffer::StyledBuffer>::append
-229,844 <rustc_span::source_map::SourceMap>::span_to_lines
-217,550 core::fmt::Formatter::pad
-196,577 _rjem_je_arena_cache_bin_fill_small
-186,318 rustc_span::char_width
-165,353 <rustc_span::span_encoding::Span>::to
-154,852 rustc_errors::emitter::emit_to_destination
-129,414 _rjem_je_malloc_default
-104,344 <rustc_errors::json::DiagnosticSpan>::from_span_full::<core::iter::sources::from_fn::FromFn<<rustc_span::span_encoding::Span>::macro_backtrace::{closure#0}>>
-101,388 <anstyle::style::Style as core::fmt::Display>::fmt
-92,904 <<rustc_errors::json::Diagnostic>::from_errors_diagnostic::BufWriter as std::io::Write>::write_fmt
-90,530 <anstyle::color::DisplayBuffer>::write_str
-87,516 <rustc_span::source_map::SourceMap>::span_to_snippet
-85,554 rustc_span::str_width
-83,035 <alloc::string::String as core::fmt::Write>::write_str
-80,730 <serde_json::ser::Compound<&mut alloc::boxed::Box<dyn std::io::Write + core::marker::Send>, serde_json::ser::CompactFormatter> as serde_core::ser::SerializeStruct>::serialize_field::<u32>
-79,817 <alloc::raw_vec::RawVec<rustc_expand::mbe::TokenTree>>::grow_one
79,817 <alloc::raw_vec::RawVec<object::write::Symbol>>::grow_one
-66,234 <rustc_errors::snippet::Style>::anstyle
-55,212 <core::str::lossy::Utf8Chunks as core::iter::traits::iterator::Iterator>::next
-54,930 core::ptr::drop_in_place::<rustc_errors::json::DiagnosticSpan>
-54,173 <serde_json::ser::Compound<&mut alloc::boxed::Box<dyn std::io::Write + core::marker::Send>, serde_json::ser::CompactFormatter> as serde_core::ser::SerializeStruct>::serialize_field::<core::option::Option<alloc::string::String>>
-53,844 rustc_middle::ty::codec::encode_with_shorthand::<rustc_middle::query::on_disk_cache::CacheEncoder, rustc_middle::ty::Ty, <rustc_middle::query::on_disk_cache::CacheEncoder as rustc_middle::ty::codec::TyEncoder>::type_shorthands>
-53,628 <rustc_errors::emitter::HumanEmitter>::draw_line_num
-52,752 rustc_middle::lint::explain_lint_level_source
-50,079 <rustc_errors::emitter::HumanEmitter as rustc_errors::emitter::Emitter>::emit_diagnostic
-49,110 _rjem_je_tcache_bin_flush_small
-46,860 <core::iter::sources::from_fn::FromFn<<rustc_span::span_encoding::Span>::macro_backtrace::{closure#0}> as core::iter::traits::iterator::Iterator>::next
-45,249 <rustc_errors::emitter::FileWithAnnotatedLines>::collect_annotations::add_annotation_to_file
-44,424 <rustc_span::source_map::SourceMap>::span_until_char
-42,623 <core::str::iter::SplitInternal<char>>::next
< -1,618,401 ./string/../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:__memcpy_avx_unaligned_erms
< -47,730 ./nptl/./nptl/pthread_mutex_trylock.c:pthread_mutex_trylock@@GLIBC_2.34

Maybe there are a bunch of new warnings emitted after your PR?

warning: `syn` (lib) generated 19 warnings (run `cargo fix --lib -p syn` to apply 19 suggestions)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 2.09s

vs

warning: `syn` (lib) generated 67 warnings (45 duplicates) (run `cargo fix --lib -p syn` to apply 22 suggestions)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 2.28s

[RalfJung]

The query changes looked entirely insignificant to me... or maybe I just don't know how to read that page.

Maybe there are a bunch of new warnings emitted after your PR?

Where did you get that output from?

We did crater this, but maybe this is an ancient version of syn that still triggers the lint. The PR also changed repr_transparent_external_private_fields to report-in-deps, after all.

[Kobzol]

I just ran cargo build with the previous/new compiler version on https://github.com/rust-lang/rustc-perf/tree/master/collector/compile-benchmarks/syn-2.0.101.

[Kobzol]

The syn version is 6 months old, for reference, so not exactly ancient.

[RalfJung]

Ah, on the latest nightly we are getting

warning: lint `repr_transparent_external_private_fields` has been renamed to `repr_transparent_non_zst_fields`
--> src/token.rs:146:28
|
146 | #[allow(unknown_lints, repr_transparent_external_private_fields)] // False positive: https://github.com/rust-lang/rust/issues/78586#issuecomment-1722680482
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: use the new name: `repr_transparent_non_zst_fields`
|
= note: `#[warn(renamed_and_removed_lints)]` on by default
warning: lint `repr_transparent_external_private_fields` has been renamed to `repr_transparent_non_zst_fields`
--> src/token.rs:327:36
|
327 | #[allow(unknown_lints, repr_transparent_external_private_fields)] // False positive: https://github.com/rust-lang/rust/issues/78586#issuecomment-1722680482
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: use the new name: `repr_transparent_non_zst_fields`
...
521 | / define_punctuation_structs! {
522 | | "_" pub struct Underscore/1 /// wildcard patterns, inferred types, unnamed items in constants, extern crates, use declarations, and destructuring assignment
523 | | }
| |_- in this macro invocation
|
= note: this warning originates in the macro `define_punctuation_structs` (in Nightly builds, run with -Z macro-backtrace for more info)
warning: lint `repr_transparent_external_private_fields` has been renamed to `repr_transparent_non_zst_fields`
--> src/token.rs:327:36
|
327 | #[allow(unknown_lints, repr_transparent_external_private_fields)] // False positive: https://github.com/rust-lang/rust/issues/78586#issuecomment-1722680482
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: use the new name: `repr_transparent_non_zst_fields`

And I guess what we are seeing is the effort of rendering those lints.

Good catch, thanks!

[Kobzol]

Yeah, the diagnostics emitting code is ludicrously slow.