MultipleReturnTerminators + SingleReturnTerminator

[saethlin]

MultipleReturnTerminators was added with the thinking that we would do better MIR optimizations with that form. But LLVM does better when there is a single return terminator. So adding a new pass which turns back into the single terminator form should give us the best of both worlds.

Locally, it looks like this change just rearranges/renames some things. If this doesn't look like an improvement in perf, I'm going to suggest that we remove MultipleReturnTerminators.

[rustbot]

r? @davidtwco

(rustbot has picked a reviewer for you, use r? to override)

[saethlin]

Ah crap sorry I forgot to ask for review from ghost

@bors try @rust-timer queue

[bors]

⌛ Trying commit ca521b0e1fb89c0813cd1d75f1fa8d0112d818c4 with merge ebf0d2d3df7ddf24af902c6cb1401231a299dc2d...

[bors]

☀️ Try build successful - checks-actions
Build commit: ebf0d2d3df7ddf24af902c6cb1401231a299dc2d (ebf0d2d3df7ddf24af902c6cb1401231a299dc2d)

[rust-timer]

Finished benchmarking commit (ebf0d2d3df7ddf24af902c6cb1401231a299dc2d): 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)	0.5%	[0.2%, 1.9%]	6
Regressions ❌ (secondary)	1.1%	[1.1%, 1.1%]	1
Improvements ✅ (primary)	-0.7%	[-1.0%, -0.2%]	7
Improvements ✅ (secondary)	-0.5%	[-0.6%, -0.4%]	4
All ❌✅ (primary)	-0.1%	[-1.0%, 1.9%]	13

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)	6.4%	[0.0%, 10.0%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.6%	[-2.6%, -2.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	4.1%	[-2.6%, 10.0%]	4

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)	2.1%	[2.1%, 2.1%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.1%	[2.1%, 2.1%]	1

[cjgillot]

Can the SingleReturnTerminator be done on-the-fly during codegen?
MultipleReturnTerminator can have a secondary effect of shrinking MIR, which makes encoding/decoding faster.
Having SingleReturnTerminator before encoding nullifies this second effect.

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit b878a1c88dad18f70c45c3753469a49ce1e1bca6 with merge 0c68d2596c4617e0a308139af6ad56d4ac858bf1...

[bors]

☀️ Try build successful - checks-actions
Build commit: 0c68d2596c4617e0a308139af6ad56d4ac858bf1 (0c68d2596c4617e0a308139af6ad56d4ac858bf1)

[bors]

☀️ Try build successful - checks-actions
Build commit: 0c68d2596c4617e0a308139af6ad56d4ac858bf1 (0c68d2596c4617e0a308139af6ad56d4ac858bf1)

[rust-timer]

Finished benchmarking commit (0c68d2596c4617e0a308139af6ad56d4ac858bf1): 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%	[0.3%, 1.0%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.9%	[-1.4%, -0.6%]	3
Improvements ✅ (secondary)	-1.3%	[-1.4%, -1.1%]	3
All ❌✅ (primary)	-0.2%	[-1.4%, 1.0%]	6

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)	2.4%	[1.7%, 3.1%]	2
Regressions ❌ (secondary)	1.7%	[1.7%, 1.7%]	1
Improvements ✅ (primary)	-1.2%	[-2.4%, -0.1%]	2
Improvements ✅ (secondary)	-2.7%	[-3.0%, -2.4%]	2
All ❌✅ (primary)	0.6%	[-2.4%, 3.1%]	4

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)	4.9%	[4.5%, 5.3%]	3
Improvements ✅ (primary)	-1.5%	[-1.5%, -1.5%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.5%	[-1.5%, -1.5%]	1

[rust-timer]

Finished benchmarking commit (597042ea7995e0681a2e2fbdcbaa41ff8bcd6c3b): 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%	[0.3%, 1.1%]	15
Regressions ❌ (secondary)	1.3%	[0.2%, 2.5%]	13
Improvements ✅ (primary)	-0.4%	[-0.8%, -0.3%]	6
Improvements ✅ (secondary)	-1.0%	[-1.4%, -0.4%]	3
All ❌✅ (primary)	0.3%	[-0.8%, 1.1%]	21

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)	2.4%	[0.2%, 3.9%]	5
Regressions ❌ (secondary)	3.1%	[3.1%, 3.1%]	1
Improvements ✅ (primary)	-3.9%	[-4.9%, -3.0%]	2
Improvements ✅ (secondary)	-1.6%	[-2.2%, -1.3%]	3
All ❌✅ (primary)	0.6%	[-4.9%, 3.9%]	7

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)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-18.6%	[-18.6%, -18.6%]	1
All ❌✅ (primary)	-	-	0

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit b9e06f1 with merge cdbe287ec197b4df04fec87680c5779610f69a85...

[bors]

☀️ Try build successful - checks-actions
Build commit: cdbe287ec197b4df04fec87680c5779610f69a85 (cdbe287ec197b4df04fec87680c5779610f69a85)

[saethlin]

@rust-timer queue

[saethlin]

@rust-timer build cdbe287ec197b4df04fec87680c5779610f69a85

[rust-timer]

Finished benchmarking commit (cdbe287ec197b4df04fec87680c5779610f69a85): 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)	0.8%	[0.4%, 1.4%]	6
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.6%	[-0.7%, -0.5%]	2
Improvements ✅ (secondary)	-1.4%	[-1.4%, -1.4%]	2
All ❌✅ (primary)	0.4%	[-0.7%, 1.4%]	8

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)	3.1%	[0.2%, 5.5%]	4
Regressions ❌ (secondary)	2.3%	[2.3%, 2.3%]	1
Improvements ✅ (primary)	-2.6%	[-2.7%, -2.5%]	2
Improvements ✅ (secondary)	-4.6%	[-5.0%, -4.3%]	2
All ❌✅ (primary)	1.2%	[-2.7%, 5.5%]	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)	1.5%	[1.5%, 1.5%]	1
Regressions ❌ (secondary)	2.3%	[2.3%, 2.3%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.5%	[1.5%, 1.5%]	1

[Noratrieb]

Marking as waiting on author given that it's a draft and has conflicts.

[saethlin]

The theoretical benefit of this pass is that it could reduce the size of our MIR on disk. But based on the experimentation in this PR, it looks like this pass has a very marginal impact on the amount of MIR we produce. In addition, it looks to me like encoding/decoding MIR isn't a large enough fraction of our compile time to be concerned about marginal tweaks like this.