Avoid depending on forever-red DepNode when encoding metadata.

[cjgillot]

Split from #114669 for perf

r? @petrochenkov

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 1584b0e with merge 22d2c51...

[bors]

☀️ Try build successful - checks-actions
Build commit: 22d2c51 (22d2c5105e640237ee0a534b4a7d84635a5b4480)

[rust-timer]

Finished benchmarking commit (22d2c51): 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)	0.7%	[0.1%, 1.5%]	33
Regressions ❌ (secondary)	0.5%	[0.2%, 1.1%]	10
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.7%	[0.1%, 1.5%]	33

Max RSS (memory usage)

Results (primary 2.0%, secondary 0.7%)

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

	mean	range	count
Regressions ❌ (primary)	2.0%	[0.6%, 7.0%]	7
Regressions ❌ (secondary)	0.7%	[0.7%, 0.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.0%	[0.6%, 7.0%]	7

Cycles

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

Binary size

Results (primary 0.4%, secondary 0.6%)

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

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.0%, 2.2%]	76
Regressions ❌ (secondary)	0.6%	[0.0%, 3.3%]	36
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[0.0%, 2.2%]	76

Bootstrap: 694.099s -> 694.846s (0.11%)
Artifact size: 372.14 MiB -> 372.24 MiB (0.03%)

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit f12e837 with merge 3797379...

[bors]

☀️ Try build successful - checks-actions
Build commit: 3797379 (3797379f7091833992e5116135662ecd38c057d2)

[rust-timer]

Finished benchmarking commit (3797379): 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)	0.5%	[0.1%, 2.1%]	84
Regressions ❌ (secondary)	1.2%	[0.2%, 2.0%]	20
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-21.9%	[-42.6%, -0.3%]	7
All ❌✅ (primary)	0.5%	[0.1%, 2.1%]	84

Max RSS (memory usage)

Results (primary 0.4%, secondary -11.4%)

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

	mean	range	count
Regressions ❌ (primary)	1.6%	[1.0%, 2.3%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.4%	[-3.4%, -3.4%]	1
Improvements ✅ (secondary)	-11.4%	[-15.0%, -8.4%]	6
All ❌✅ (primary)	0.4%	[-3.4%, 2.3%]	4

Cycles

Results (primary -1.1%, secondary -5.8%)

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

	mean	range	count
Regressions ❌ (primary)	1.7%	[1.1%, 2.7%]	5
Regressions ❌ (secondary)	3.3%	[1.5%, 10.2%]	9
Improvements ✅ (primary)	-2.0%	[-3.0%, -1.4%]	15
Improvements ✅ (secondary)	-17.5%	[-37.9%, -2.1%]	7
All ❌✅ (primary)	-1.1%	[-3.0%, 2.7%]	20

Binary size

Results (primary 0.4%, secondary 0.6%)

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

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.0%, 2.2%]	76
Regressions ❌ (secondary)	0.6%	[0.0%, 3.3%]	36
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[0.0%, 2.2%]	76

Bootstrap: 460.931s -> 461.575s (0.14%)
Artifact size: 372.21 MiB -> 372.24 MiB (0.01%)

[petrochenkov]

This is actually a smaller regression than I was expecting.
And the cases in which the resolutions(()) stay green (I assume) give a huge speedup.

And this can likely be sped up further

• by not hashing visibilities_for_hashing and maybe some other fields twice.
• by un-querifying other cheap accesses to tables in ResolverGlobalCtxt

I think the doc_link_resolutions change can also be reverted now.

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 992fa62 with merge a46eba4...

[bors]

☀️ Try build successful - checks-actions
Build commit: a46eba4 (a46eba429ac4af042b119d009c71fc20c27de559)

[rust-timer]

Finished benchmarking commit (a46eba4): 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)	0.5%	[0.2%, 2.0%]	77
Regressions ❌ (secondary)	1.1%	[0.2%, 2.1%]	22
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-25.5%	[-42.6%, -16.4%]	6
All ❌✅ (primary)	0.5%	[0.2%, 2.0%]	77

Max RSS (memory usage)

Results (primary -2.5%, secondary -6.7%)

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)	3.2%	[2.9%, 3.6%]	2
Improvements ✅ (primary)	-2.5%	[-2.8%, -2.2%]	2
Improvements ✅ (secondary)	-9.5%	[-14.7%, -0.4%]	7
All ❌✅ (primary)	-2.5%	[-2.8%, -2.2%]	2

Cycles

Results (primary 2.1%, secondary -7.7%)

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

	mean	range	count
Regressions ❌ (primary)	2.1%	[1.3%, 4.1%]	9
Regressions ❌ (secondary)	2.3%	[1.7%, 3.0%]	7
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-17.7%	[-38.2%, -1.4%]	7
All ❌✅ (primary)	2.1%	[1.3%, 4.1%]	9

Binary size

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

Bootstrap: 462.08s -> 461.207s (-0.19%)
Artifact size: 372.26 MiB -> 372.20 MiB (-0.02%)

[cjgillot]

@petrochenkov this latest version only marks resolutions to be hashed. The perf has not terribly changed compared to the previous run. Cachegrind says the regression comes from hashing more spans and hashing many ModChildren. Do you want me to keep looking for perf improvements in this PR, or can it be done as follow-ups?

[petrochenkov]

What is the difference between the previous version (#143247 (comment)) and the current one? I only see the doc_link_resolutions change.

I'm still interested in checking whether this

And this can likely be sped up further

• by not hashing visibilities_for_hashing and maybe some other fields twice.
• by un-querifying other cheap accesses to tables in ResolverGlobalCtxt

improved things, but the current results are good enough to merge and unblock #114669.

This investigation can be done in a follow up, just need to not forget about it.

[cjgillot]

What is the difference between the previous version (#143247 (comment)) and the current one? I only see the doc_link_resolutions change.

Yes, I only did that change. This makes the PR the minimal change for what #114669 needs.

[petrochenkov]

@bors r+

[bors]

📌 Commit 992fa62 has been approved by petrochenkov

It is now in the queue for this repository.