[TEST] Use global Rust jemalloc

[Zoxc]

No description provided.

[Zoxc]

@bors try @rust-timer queue

[rust-bors]

💔 Test for 64c00df failed: CI. Failed job:

• try - dist-x86_64-linux-quick (web logs, enhanced plaintext logs)

[Zoxc]

On macOS 11:

Benchmark	Before	Before		After
	Time	Time	%	Time	%
🟣 regex:check	1.1744s	1.1759s	0.13%	1.0958s	💚 -6.69%
🟣 syn:check	1.9310s	1.9309s	-0.01%	1.7920s	💚 -7.20%
🟣 clap:check	2.0073s	2.0093s	0.10%	1.8822s	💚 -6.23%
🟣 hyper:check	0.3262s	0.3260s	-0.06%	0.3143s	💚 -3.63%
Total	5.4389s	5.4421s	0.06%	5.0844s	💚 -6.52%
Summary	1.0000s	1.0004s	0.04%	0.9406s	💚 -5.94%

[bjorn3]

This only works when libstd is statically linked into librustc_driver which we don't do on MinGW:

rust/src/bootstrap/src/core/builder/mod.rs

Lines 1125 to 1129 in 92c7010

	/// Returns if `std` should be statically linked into `rustc_driver`.
	/// It's currently not done on `windows-gnu` due to linker bugs.
	pub fn link_std_into_rustc_driver(&self, target: TargetSelection) -> bool {
	!target.triple.ends_with("-windows-gnu")
	}

If libstd is dynamically linked, this is UB as libstd and rustc may use a different allocator depending on the target and if protected visibility is used (PE and Mach-O effectively behave as if protected visibility is always used, while on ELF we may want to make -Zdefault-visibility=protected the default in the future). Rustc should probably emit an error for this case, but does not do so currently.

View changes since the review

[Zoxc]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: ca3fc89 (ca3fc89a2aeeda43050fdebc0d92c676d4063482, parent: 2f43fe430374fce8c75d477bdb78715d5c0349c4)

[rust-timer]

Finished benchmarking commit (ca3fc89): comparison URL.

Overall result: ✅ improvements - no action needed

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

@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)	0.0%	[0.0%, 0.0%]	3
Improvements ✅ (primary)	-0.3%	[-0.8%, -0.2%]	75
Improvements ✅ (secondary)	-0.4%	[-0.8%, -0.0%]	126
All ❌✅ (primary)	-0.3%	[-0.8%, -0.2%]	75

Max RSS (memory usage)

Results (primary 0.1%, 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)	1.4%	[1.4%, 1.4%]	1
Regressions ❌ (secondary)	2.5%	[1.1%, 3.8%]	2
Improvements ✅ (primary)	-1.3%	[-1.3%, -1.3%]	1
Improvements ✅ (secondary)	-2.2%	[-2.2%, -2.2%]	1
All ❌✅ (primary)	0.1%	[-1.3%, 1.4%]	2

Cycles

Results (primary 1.7%, secondary 0.5%)

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

	mean	range	count
Regressions ❌ (primary)	1.7%	[1.7%, 1.7%]	1
Regressions ❌ (secondary)	3.8%	[2.2%, 6.1%]	5
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.7%	[-4.0%, -1.2%]	5
All ❌✅ (primary)	1.7%	[1.7%, 1.7%]	1

Binary size

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

Bootstrap: 487.764s -> 489.254s (0.31%)
Artifact size: 393.48 MiB -> 394.61 MiB (0.29%)

[rust-bors]

☔ The latest upstream changes (presumably #160517) made this pull request unmergeable. Please resolve the merge conflicts by rebasing.