cg_llvm: stop identifying ADTs in LLVM IR

[erikdesjardins]

This is an extension of #94107. It may be a minor perf win.

Fixes #96242.

Now that we use opaque pointers, ADTs can no longer be recursive, so we
do not need to name them. Previously, this would be necessary if you had
a struct like

struct Foo(Box<Foo>, u64, u64);

which would be represented with something like

%Foo = type { %Foo*, i64, i64 }

which is now just

{ ptr, i64, i64 }

r? @tmiasko

[workingjubilee]

@bors try @rust-timer queue

[workingjubilee]

whoops, uhh
@bors try

[bors]

⌛ Trying commit 1d7f728 with merge 0c56080e7cdb00fc640662bd2954f5a58254a3f9...

[bors]

☀️ Try build successful - checks-actions
Build commit: 0c56080e7cdb00fc640662bd2954f5a58254a3f9 (0c56080e7cdb00fc640662bd2954f5a58254a3f9)

[rust-timer]

Finished benchmarking commit (0c56080e7cdb00fc640662bd2954f5a58254a3f9): comparison URL.

Overall result: ❌✅ regressions and improvements - 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.

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 a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.4%, 0.4%]	1
Regressions ❌ (secondary)	0.4%	[0.3%, 0.5%]	2
Improvements ✅ (primary)	-0.8%	[-1.2%, -0.3%]	5
Improvements ✅ (secondary)	-1.2%	[-1.6%, -0.8%]	2
All ❌✅ (primary)	-0.6%	[-1.2%, 0.4%]	6

Max RSS (memory usage)

Results

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%	[-1.3%, -1.3%]	1
Improvements ✅ (secondary)	-3.8%	[-3.8%, -3.8%]	1
All ❌✅ (primary)	-1.3%	[-1.3%, -1.3%]	1

Cycles

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

Binary size

Results

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)	-	-	0
Improvements ✅ (secondary)	-1.4%	[-2.6%, -0.2%]	2
All ❌✅ (primary)	-	-	0

Bootstrap: missing data

[erikdesjardins]

Perf shows a ~1% improvement to a few opt builds, which is about what I expected.

But then also a 0.5% regression to 3 doc builds. This doesn't make sense, rustdoc shouldn't be doing any codegen (I hope), and this change should not be able to impact optimizations. I suspect it is something like: changing some struct types from being named to unnamed changes module hashes during LTO, which slightly perturbs some code in rustdoc, making it slower if we're unlucky.

Cachegrind diff:

Files compared:   results/cgfilt-d12c6e947ceacf3b22c154caf9532b390d8dc88a-ucd-Doc-Full; results/cgfilt-0c56080e7cdb00fc640662bd2954f5a58254a3f9-ucd-Doc-Full
...

--------------------------------------------------------------------------------
Ir         
--------------------------------------------------------------------------------
39,529,196  PROGRAM TOTALS

--------------------------------------------------------------------------------
Ir          file:function
--------------------------------------------------------------------------------
49,843,901  ???:<rustc_infer::infer::freshen::TypeFreshener as rustc_type_ir::fold::FallibleTypeFolder<rustc_middle::ty::context::TyCtxt>>::try_fold_ty
-6,447,963  ???:<rustc_middle::ty::Ty as rustc_type_ir::fold::TypeSuperFoldable<rustc_middle::ty::context::TyCtxt>>::try_super_fold_with::<rustc_infer::infer::freshen::TypeFreshener>
 1,377,686  ???:<rustc_middle::ty::Ty as rustc_serialize::serialize::Decodable<rustc_metadata::rmeta::decoder::DecodeContext>>::decode
-1,214,742  ???:<rustc_middle::hir::map::ItemCollector as rustc_hir::intravisit::Visitor>::visit_expr
 1,209,392  ???:<rustc_span::hygiene::SyntaxContext as rustc_serialize::serialize::Decodable<rustc_metadata::rmeta::decoder::DecodeContext>>::decode
-1,133,433  ???:<rustc_middle::ty::generics::GenericParamDef as rustc_serialize::serialize::Decodable<rustc_metadata::rmeta::decoder::DecodeContext>>::decode
-1,115,806  ???:<rustc_span::hygiene::SyntaxContextData as rustc_serialize::serialize::Decodable<rustc_metadata::rmeta::decoder::DecodeContext>>::decode
-1,051,024  ???:<rustc_arena::TypedArena<rustc_span::source_map::Spanned<rustc_ast::ast::LitKind>> as core::ops::drop::Drop>::drop
...

Looks like some inlining decision changes near TypeFreshener. I could try adding some more inline annotations on try_fold_ty (TypeFreshener::fold_ty is #[inline] already) or other nearby functions, but given how sensitive the type folding code is, and the small number and magnitude of regressions, I am not inclined to do so.

[workingjubilee]

Rustdoc has been known to evince severe regressions merely from adding library API before, so.

[erikdesjardins]

Hmm, yeah, looking at the three doc builds that regressed, they've been jumping up and down by the same magnitude seen in this PR:

webrender-2022-doc
This PR: +0.35%
Recent oscillations: +/-0.3%

ucd-doc
This PR: +0.45%
Recent oscillations: +/-0.5%

match-stress-doc
This PR: +0.33%
Recent oscillations: +/-0.3%

I think, then, I'm justified in doing
@rustbot label +perf-regression-triaged

[tmiasko]

@bors r+

[bors]

📌 Commit 1d7f728 has been approved by tmiasko

It is now in the queue for this repository.

[bors]

⌛ Testing commit 1d7f728 with merge 73dc6f0...

[bors]

☀️ Test successful - checks-actions
Approved by: tmiasko
Pushing 73dc6f0 to master...

[rust-timer]

Finished benchmarking commit (73dc6f0): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.8%	[-1.2%, -0.6%]	5
Improvements ✅ (secondary)	-1.1%	[-1.5%, -0.6%]	6
All ❌✅ (primary)	-0.8%	[-1.2%, -0.6%]	5

Max RSS (memory usage)

Results

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)	4.4%	[4.4%, 4.4%]	1
Regressions ❌ (secondary)	2.5%	[1.7%, 2.9%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.0%	[-3.0%, -3.0%]	1
All ❌✅ (primary)	4.4%	[4.4%, 4.4%]	1

Cycles

Results

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)	4.0%	[2.9%, 5.3%]	6
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	4.0%	[2.9%, 5.3%]	6

Binary size

Results

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)	-0.0%	[-0.0%, -0.0%]	3
Improvements ✅ (secondary)	-1.4%	[-2.6%, -0.2%]	2
All ❌✅ (primary)	-0.0%	[-0.0%, -0.0%]	3

Bootstrap: 649.336s -> 647.034s (-0.35%)