[Experiment] Do not emit noalias for inline functions

[HKalbasi]

r? @saethlin

[saethlin]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: b1487a9 (b1487a9dc062157b115f8459b4c4d22d2570fc21, parent: dd7fda570040e8a736f7d8bc28ddd1b444aabc82)

[rust-timer]

Finished benchmarking commit (b1487a9): 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)	0.2%	[0.1%, 0.5%]	54
Regressions ❌ (secondary)	0.5%	[0.1%, 1.1%]	47
Improvements ✅ (primary)	-0.6%	[-2.2%, -0.1%]	40
Improvements ✅ (secondary)	-0.6%	[-3.6%, -0.1%]	33
All ❌✅ (primary)	-0.1%	[-2.2%, 0.5%]	94

Max RSS (memory usage)

Results (primary -0.2%, secondary -2.0%)

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

	mean	range	count
Regressions ❌ (primary)	3.7%	[3.7%, 3.7%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.2%	[-2.3%, -2.1%]	2
Improvements ✅ (secondary)	-2.0%	[-2.6%, -1.4%]	2
All ❌✅ (primary)	-0.2%	[-2.3%, 3.7%]	3

Cycles

Results (primary 0.9%, secondary 1.8%)

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

	mean	range	count
Regressions ❌ (primary)	2.8%	[2.3%, 4.6%]	6
Regressions ❌ (secondary)	3.6%	[2.1%, 4.6%]	19
Improvements ✅ (primary)	-2.0%	[-2.4%, -1.7%]	4
Improvements ✅ (secondary)	-2.4%	[-3.0%, -2.1%]	8
All ❌✅ (primary)	0.9%	[-2.4%, 4.6%]	10

Binary size

Results (primary 0.1%, 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)	0.1%	[0.0%, 0.2%]	11
Regressions ❌ (secondary)	0.1%	[0.1%, 0.6%]	38
Improvements ✅ (primary)	-0.1%	[-0.1%, -0.1%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[-0.1%, 0.2%]	15

Bootstrap: 472.936s -> 470.008s (-0.62%)
Artifact size: 389.95 MiB -> 390.01 MiB (0.02%)

[saethlin]

That result definitely indicates some kind of missed optimization in LLVM, noalias shouldn't be regressing codegen. Though isolating why might be fiendishly difficult.

Overall, the above report reminds me this perf experiment I did a little while ago: #121068 (comment) (basically the same result, check builds are slower which means less optimization of the compiler, but the hashmap runtime benchmarks benefit)

[HKalbasi]

Sigh. So we can't attach that meaning to the #[inline] considering how it is currently used. I think we can still use it for experiments, but we need something else for the long term plan.

Now that we are here, let's test it with #[inline(always)] as well. Maybe the missed optimizations were due to functions that requested #[inline] but didn't get inlined. If it regresses this time as well, there is something strange happening...

[HKalbasi]

@bors try @rust-timer queue

[rust-timer]

Insufficient permissions to issue commands to rust-timer.

[rust-bors]

☀️ Try build successful (CI)
Build commit: f1987df (f1987df82d2c47926125142f58fb0d7f7a85f06a, parent: dd7fda570040e8a736f7d8bc28ddd1b444aabc82)