Do partial SsaLocals analysis in unoptimized builds

[saethlin]

I've been musing about the perf implications of all the queries that is_freeze results in.

(it's not like this makes GVN lightweight or light on queries overall, because evaluating consts is still quite a lot of queries)

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 696799b with merge 46ae8eb...

[bors]

☀️ Try build successful - checks-actions
Build commit: 46ae8eb (46ae8eb67ec584d3f0bbf4a2618c0d3b6e562473)

[rust-timer]

Finished benchmarking commit (46ae8eb): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

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 the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.2%, 1.2%]	17
Regressions ❌ (secondary)	0.7%	[0.1%, 1.1%]	8
Improvements ✅ (primary)	-0.5%	[-0.9%, -0.2%]	21
Improvements ✅ (secondary)	-10.2%	[-27.7%, -0.2%]	13
All ❌✅ (primary)	0.1%	[-0.9%, 1.2%]	38

Max RSS (memory usage)

Results (primary 0.1%, secondary -14.2%)

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)	3.9%	[2.0%, 6.8%]	3
Regressions ❌ (secondary)	2.7%	[2.6%, 3.0%]	3
Improvements ✅ (primary)	-2.1%	[-3.7%, -1.0%]	5
Improvements ✅ (secondary)	-26.8%	[-29.2%, -25.2%]	4
All ❌✅ (primary)	0.1%	[-3.7%, 6.8%]	8

Cycles

Results (primary 1.0%, secondary -8.2%)

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)	1.0%	[1.0%, 1.1%]	2
Regressions ❌ (secondary)	8.1%	[2.3%, 10.3%]	7
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-31.0%	[-36.5%, -10.6%]	5
All ❌✅ (primary)	1.0%	[1.0%, 1.1%]	2

Binary size

Results (primary -0.5%, secondary -4.7%)

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)	0.2%	[0.0%, 0.9%]	39
Regressions ❌ (secondary)	0.1%	[0.1%, 0.2%]	8
Improvements ✅ (primary)	-1.2%	[-6.2%, -0.0%]	33
Improvements ✅ (secondary)	-6.4%	[-25.3%, -0.0%]	24
All ❌✅ (primary)	-0.5%	[-6.2%, 0.9%]	72

Bootstrap: 766.815s -> 768.247s (0.19%)
Artifact size: 330.84 MiB -> 330.83 MiB (-0.00%)

[saethlin]

Interesting. Looks like this removes ~98% of the LLVM time for deep-vector.

Also it looks for incr-unchanged and incr-patched builds we have a handful of improvements and no regressions, but for full and incr-full builds we're almost all regressions. The self-profile breakdown points to GVN as the culprit (of course) except for the image debug-full build where the regression is... in LLVM??? image debug-full is also the largest binary size regression which corroborates that. How strange.

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit a013042 with merge ef3fe2f...

[saethlin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 1580046 with merge b14fa64...

[bors]

☀️ Try build successful - checks-actions
Build commit: b14fa64 (b14fa6478d5f801a9257f06e5ad8527530b7cfee)

[rust-timer]

Finished benchmarking commit (b14fa64): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

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 the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.2%, 1.2%]	23
Regressions ❌ (secondary)	0.5%	[0.1%, 1.0%]	8
Improvements ✅ (primary)	-0.4%	[-1.0%, -0.2%]	26
Improvements ✅ (secondary)	-0.8%	[-1.4%, -0.2%]	9
All ❌✅ (primary)	0.1%	[-1.0%, 1.2%]	49

Max RSS (memory usage)

Results (primary 2.7%, secondary -3.2%)

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)	5.7%	[3.1%, 10.5%]	5
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.4%	[-2.9%, -2.1%]	3
Improvements ✅ (secondary)	-3.2%	[-3.2%, -3.2%]	1
All ❌✅ (primary)	2.7%	[-2.9%, 10.5%]	8

Cycles

Results (primary 1.0%)

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)	1.0%	[1.0%, 1.0%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.0%	[1.0%, 1.0%]	1

Binary size

Results (primary -0.4%, secondary -1.1%)

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)	0.2%	[0.0%, 0.9%]	30
Regressions ❌ (secondary)	0.1%	[0.1%, 0.2%]	8
Improvements ✅ (primary)	-0.7%	[-3.5%, -0.0%]	50
Improvements ✅ (secondary)	-1.7%	[-6.1%, -0.0%]	18
All ❌✅ (primary)	-0.4%	[-3.5%, 0.9%]	80

Bootstrap: 790.529s -> 790.638s (0.01%)
Artifact size: 347.33 MiB -> 347.26 MiB (-0.02%)