rmeta: avoid embedding StabilityLevel::Unstable reason multiple times into .rmeta\.rlib files

[klensy]

Avoids bloating size of some rmeta\rlib files by not placing default string for StabilityLevel::Unstable reason multiple times, affects only stdlib\rustc artifacts. For stdlib cuts about 3% (diff of total size for patched\unpatched *.rmeta files of stage1-std) of file size, depending on crates.

fixes #88180

[rust-highfive]

r? @lcnr

(rust-highfive has picked a reviewer for you, use r? to override)

[klensy]

Or maybe encode with something like this, but values deduped?

rust/compiler/rustc_metadata/src/rmeta/mod.rs

Lines 126 to 135 in ca1e68b

	/// A list of lazily-decoded values, with the added capability of random access.
	///
	/// Random-access table (i.e. offering constant-time `get`/`set`), similar to
	/// `LazyArray<T>`, but without requiring encoding or decoding all the values
	/// eagerly and in-order.
	struct LazyTable<I, T> {
	position: NonZeroUsize,
	encoded_size: usize,
	_marker: PhantomData<fn(I) -> T>,
	}

[cjgillot]

There is precedent for such encoding, for instance type_shorthands or ExpnIndex.
Do you have an estimate of the amount of duplicated strings?

[klensy]

There is precedent for such encoding, for instance type_shorthands or ExpnIndex. Do you have an estimate of the amount of duplicated strings?

from #88180 when issue was created:

found 172 times in libaddr2line.rmeta
found 911 times in libhashbrown.rmeta
found 9590 times in libobject.rmeta

[klensy]

Checked now, for stage0 compiler artifact libobject.rmeta: 11237 times, for example.

[cjgillot]

Why not the following:

• encode the integer value of Symbol;
• encode somewhere else a FxHashMap<u32, String> of only actually encoded Symbols.

This would maximize the deduplication while being self-contained.
If we don't care about being self-contained, we can skip encoding the strings for static Symbols (ie. those listed in rustc_span::symbol).

What do you think?

[klensy]

rust/compiler/rustc_span/src/symbol.rs

Lines 1822 to 1833 in aedf78e

	impl<S: Encoder> Encodable<S> for Symbol {
	fn encode(&self, s: &mut S) {
	s.emit_str(self.as_str());
	}
	}
	impl<D: Decoder> Decodable<D> for Symbol {
	#[inline]
	fn decode(d: &mut D) -> Symbol {
	Symbol::intern(&d.read_str())
	}
	}

Something like replacing s.emit_str(self.as_str()); with with s.emit_u32(self.as_u32()); + saving in hash table, and in decode reading from that hash table? Only left to find the place where to save that hash table.

[cjgillot]

To generate those hash-tables, you can use specialization like we do for impl {De,En}codable for Span. Then, the hash-tables can be stored as any other in CrateRoot. We may want to only decode it once.

[lcnr]

r? @cjgillot

[cjgillot]

@klensy do you still want to merge this, or should it be closed in favour of #98851?

[bors]

☔ The latest upstream changes (presumably #98975) made this pull request unmergeable. Please resolve the merge conflicts.

[klensy]

I think this should be merged (and #98851 tested after that), as current default error message looks more like lint warn\error (that can possibly be translatable in future) and shouldn't be embed into compiled artifacts, will rebase.

[klensy]

@rustbot ready

[cjgillot]

@bors r+ rollup=never
^ may have perf implications

[bors]

📌 Commit b38c948 has been approved by cjgillot

It is now in the queue for this repository.

[bors]

⌛ Testing commit b38c948 with merge 7f93d4a...

[bors]

☀️ Test successful - checks-actions
Approved by: cjgillot
Pushing 7f93d4a to master...

[rust-timer]

Finished benchmarking commit (7f93d4a): comparison url.

Instruction count

• Primary benchmarks: no relevant changes found
• Secondary benchmarks: 😿 relevant regression found

	mean1	max	count2
Regressions 😿 (primary)	N/A	N/A	0
Regressions 😿 (secondary)	0.3%	0.3%	1
Improvements 🎉 (primary)	N/A	N/A	0
Improvements 🎉 (secondary)	N/A	N/A	0
All 😿🎉 (primary)	N/A	N/A	0

Max RSS (memory usage)

Results

• Primary benchmarks: 🎉 relevant improvements found
• Secondary benchmarks: mixed results

	mean1	max	count2
Regressions 😿 (primary)	N/A	N/A	0
Regressions 😿 (secondary)	2.4%	2.4%	1
Improvements 🎉 (primary)	-1.3%	-1.6%	6
Improvements 🎉 (secondary)	-2.0%	-3.8%	7
All 😿🎉 (primary)	-1.3%	-1.6%	6

Cycles

Results

• Primary benchmarks: 😿 relevant regression found
• Secondary benchmarks: 😿 relevant regressions found

	mean1	max	count2
Regressions 😿 (primary)	2.2%	2.2%	1
Regressions 😿 (secondary)	2.9%	3.6%	5
Improvements 🎉 (primary)	N/A	N/A	0
Improvements 🎉 (secondary)	N/A	N/A	0
All 😿🎉 (primary)	2.2%	2.2%	1

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

@rustbot label: -perf-regression

Footnotes

1. the arithmetic mean of the percent change ↩ ↩² ↩³

2. number of relevant changes ↩ ↩² ↩³