rustdoc: split build_impl into build_{local,external}_impl

[lolbinarycat]

splitting these out should let us avoid quite a few more branches, and hopefully also improve cache locality of the code.

followup to #145851

[rustbot]

r? @GuillaumeGomez

rustbot has assigned @GuillaumeGomez.
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

[lolbinarycat]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: c00bc1f (c00bc1f8148ea75996f48ef92478ecb1e7e7c9ed, parent: 91ee6a4057ce4bf1ab6d2f932cae497488d67c81)

[rust-timer]

Finished benchmarking commit (c00bc1f): comparison URL.

Overall result: ❌ regressions - 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)	1.0%	[0.2%, 3.0%]	20
Regressions ❌ (secondary)	1.9%	[0.4%, 3.2%]	25
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.0%	[0.2%, 3.0%]	20

Max RSS (memory usage)

Results (primary -1.4%, 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)	2.7%	[2.7%, 2.7%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.7%	[-3.8%, -0.8%]	14
Improvements ✅ (secondary)	-2.0%	[-4.2%, -1.0%]	17
All ❌✅ (primary)	-1.4%	[-3.8%, 2.7%]	15

Cycles

Results (primary -2.2%, 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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.2%	[-2.2%, -2.2%]	1
Improvements ✅ (secondary)	-2.0%	[-2.9%, -1.2%]	2
All ❌✅ (primary)	-2.2%	[-2.2%, -2.2%]	1

Binary size

Results (secondary 0.0%)

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)	0.0%	[0.0%, 0.0%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Bootstrap: 469.036s -> 466.669s (-0.50%)
Artifact size: 391.17 MiB -> 391.16 MiB (-0.00%)

[lolbinarycat]

because this is built off the previous performance PR, here's the comparison URL that should actually be used: https://perf.rust-lang.org/compare.html?start=05c0f818fe424341111ec6f6ccc79df099d0a142&end=c00bc1f8148ea75996f48ef92478ecb1e7e7c9ed&stat=instructions%3Au&opt=false&debug=false&check=false

quite discouraging actually, and i'm not sure how to intuit this. you would think removing branches would speed things up, at least a little bit.

[GuillaumeGomez]

Hard to compare times. The main comparison is how many instructions were run, so I guess improving branch prediction doesn't allow to do that.

[lolbinarycat]

I mean, removing a branching instruction should save an instruction, in theory.

but looking at branch-misses, that metric also regressed.

i think what's happening here is the tiny bit of duplicated code at the start is getting inlined into actually a fairly significant amount of instructions, including a branch, and that's canceling out any improvements (i believe those old removed match statements actually branch predict very well, since they get executed a bunch with locals, then a bunch with external impls).

there might be something i can do to make this worth it, so i'll keep this open for now, but i'm not confident.

[bors]

☔ The latest upstream changes (presumably #138907) made this pull request unmergeable. Please resolve the merge conflicts.

[apiraino]

I'm going to pull out T-compiler from the review as I think we need to review here (but in case LMK)

@rustbot label -T-compiler