[perf experiment] Enable DeduplicateBlocks

[saethlin]

It looks to me like all this really does is delete duplicated empty unreachable blocks, and the occasional StorageDead + drop block. Perhaps something unexpected turns up in perf.

r? @ghost

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 098af8fbffa06926b3d72b2e136cc2e3c224431f with merge 11d6ba903e6762c93f69ccf2ece4c445cfc575ae...

[bors]

☀️ Try build successful - checks-actions
Build commit: 11d6ba903e6762c93f69ccf2ece4c445cfc575ae (11d6ba903e6762c93f69ccf2ece4c445cfc575ae)

[rust-timer]

Finished benchmarking commit (11d6ba903e6762c93f69ccf2ece4c445cfc575ae): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.9%	[0.3%, 2.1%]	7
Regressions ❌ (secondary)	1.4%	[0.2%, 2.6%]	11
Improvements ✅ (primary)	-0.6%	[-1.0%, -0.3%]	15
Improvements ✅ (secondary)	-0.5%	[-0.6%, -0.3%]	4
All ❌✅ (primary)	-0.2%	[-1.0%, 2.1%]	22

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	2.7%	[1.2%, 4.7%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.7%	[-3.0%, -0.0%]	3
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.5%	[-3.0%, 4.7%]	6

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	2.2%	[2.2%, 2.2%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.2%	[-1.2%, -1.2%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.5%	[-1.2%, 2.2%]	2

[saethlin]

Cachegrind diffs look like this:

--------------------------------------------------------------------------------
Ir          
--------------------------------------------------------------------------------
173,178,186  PROGRAM TOTALS

--------------------------------------------------------------------------------
Ir           file:function
--------------------------------------------------------------------------------
176,073,369  ???:<rustc_data_structures::obligation_forest::ObligationForest<rustc_trait_selection::traits::fulfill::PendingPredicateObligation>>::process_obligations::<rustc_trait_selection::traits::fulfill::FulfillProcessor>

rust/compiler/rustc_data_structures/src/obligation_forest/mod.rs

Line 413 in a2112fc

pub fn process_obligations<P>(&mut self, processor: &mut P) -> P::OUT

Which is interesting. I have never seen a case like this where the whole regression is accounted for by a codegen change in one function.

[saethlin]

Compiling rustc, here is the size of the blocks that get deduplicated:

(  1)     9855 (91.4%, 91.4%): 0
(  2)      597 ( 5.5%, 96.9%): 2
(  3)      260 ( 2.4%, 99.3%): 1
(  4)       55 ( 0.5%, 99.8%): 3
(  5)        7 ( 0.1%, 99.9%): 4
(  6)        1 ( 0.0%, 99.9%): 5

And with DestinationPropagation enabled, this distribution slants much more towards 0.

The limit of this pass was previously set at 10. Things must have been very different back in the day to justify setting it that high. I'm setting the limit down to 3.

@bors try @rust-timer queue

[bors]

⌛ Trying commit ee29825 with merge 303c0a1f5a0e2a8be2e7b8e1eac1978c0a63f3e5...

[bors]

☀️ Try build successful - checks-actions
Build commit: 303c0a1f5a0e2a8be2e7b8e1eac1978c0a63f3e5 (303c0a1f5a0e2a8be2e7b8e1eac1978c0a63f3e5)

[rust-timer]

Finished benchmarking commit (303c0a1f5a0e2a8be2e7b8e1eac1978c0a63f3e5): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.3%, 1.1%]	6
Regressions ❌ (secondary)	1.4%	[0.1%, 2.5%]	10
Improvements ✅ (primary)	-0.6%	[-0.8%, -0.4%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.2%	[-0.8%, 1.1%]	10

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.7%	[1.7%, 1.7%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.4%	[-4.7%, -0.0%]	2
Improvements ✅ (secondary)	-3.4%	[-3.7%, -3.1%]	2
All ❌✅ (primary)	-1.0%	[-4.7%, 1.7%]	3

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.5%	[2.5%, 2.5%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

[saethlin]

Adding a tighter limit to this seems to have removed our improvements but hardly altered the regressions. I would not have expected that. So if I remove the limit entirely, I should see huge regression due to making optimized_mir slow, but what else will show up?

@bors try @rust-timer queue

[bors]

⌛ Trying commit 3b77094 with merge 8992fea7489ae2ab19a68d91257bc3ea778fcf79...

[bors]

☀️ Try build successful - checks-actions
Build commit: 8992fea7489ae2ab19a68d91257bc3ea778fcf79 (8992fea7489ae2ab19a68d91257bc3ea778fcf79)

[rust-timer]

Finished benchmarking commit (8992fea7489ae2ab19a68d91257bc3ea778fcf79): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.6%	[0.2%, 0.9%]	5
Regressions ❌ (secondary)	1.3%	[0.4%, 2.6%]	17
Improvements ✅ (primary)	-0.5%	[-0.9%, -0.2%]	13
Improvements ✅ (secondary)	-0.5%	[-0.7%, -0.3%]	3
All ❌✅ (primary)	-0.2%	[-0.9%, 0.9%]	18

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	2.3%	[1.8%, 3.0%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.3%	[-4.5%, -0.0%]	2
Improvements ✅ (secondary)	-2.9%	[-2.9%, -2.9%]	1
All ❌✅ (primary)	0.5%	[-4.5%, 3.0%]	5

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.6%	[1.6%, 1.6%]	1
Regressions ❌ (secondary)	2.6%	[2.2%, 3.1%]	4
Improvements ✅ (primary)	-0.6%	[-0.6%, -0.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.5%	[-0.6%, 1.6%]	2

[Noratrieb]

Marking as waiting on author because it's still a draft.

[saethlin]

#106428 enabled deduplicating unreachable blocks. The distribution of block sizes now looks like this (it used to be dominated by 0):

(  1)     5006 (63.2%, 63.2%): 1
(  2)     1469 (18.5%, 81.7%): 0
(  3)     1301 (16.4%, 98.1%): 2
(  4)       62 ( 0.8%, 98.9%): 3
(  5)       54 ( 0.7%, 99.6%): 4
(  6)        9 ( 0.1%, 99.7%): 6
(  7)        6 ( 0.1%, 99.8%): 5
(  8)        3 ( 0.0%, 99.8%): 7
(  9)        2 ( 0.0%, 99.9%): 8
( 10)        1 ( 0.0%, 99.9%): 9

It looks like most of the 0-statement blocks are drop terminators. the 1-statement blocks are split between _0 = + goto and StorageDead + goto.

I suspect that this distribution will be significantly reshaped by jump threading or whatever other passes land, and since this pass isn't deleting enough MIR to produce a benefit, I'm closing this.