Fix logic related to isTriviallyDuplicatable.

[atrick]

In SILInstruction::isTriviallyDuplicatable():

• Make deallocating instructions trivially duplicatable. They are by
  any useful definition--duplicating an instruction does not imply
  reordering it. Tail duplication was already treating deallocations
  as duplicatable, but doing it inconsistently. Sometimes it checks
  isTriviallyDuplicatable, and sometimes it doesn't, which appears to
  have been an accident. Disallowing duplication of deallocations will
  cause severe performance regressions. Instead, consistently allow
  them to be duplicated, making tail duplication more powerful, which
  could expose other bugs.

• Do not duplicate on-stack AllocRefInst (without special
  consideration). This is a correctness fix that apparently was never
  exposed.

Fix SILLoop::canDuplicate():

• Handle isDeallocatingStack. It's not clear how we were avoiding an
  assertion before when a stack allocatable reference was confined to
  a loop--probably just by luck.

• Handle begin/end_access inside a loop. This is extremely important
  and probably prevented many loop optimizations from working with
  exclusivity.

Update LoopRotate canDuplicateOrMoveToPreheader(). This is NFC.

[atrick]

I figured out this is what was been blocking another correctness fix:
#27444

That fix was regressing performance because it was correctly checking isTriviallyDuplicatable before cloning blocks.

[atrick]

@swift-ci test

[atrick]

@swift-ci benchmark

[atrick]

@swift-ci test source compatibility

[swift-ci]

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
Set.isDisjoint.Seq.Box.Empty	145	160	+10.3%	0.91x (?)
Improvement	OLD	NEW	DELTA	RATIO
ClassArrayGetter2	1830	120	-93.4%	15.25x
MapReduceClass2	210	42	-80.0%	5.00x
MapReduceClassShort2	376	203	-46.0%	1.85x
Set.isDisjoint.Seq.Box25	484	296	-38.8%	1.64x
Set.isDisjoint.Seq.Box0	658	408	-38.0%	1.61x
Set.isSuperset.Seq.Box25	153	96	-37.3%	1.59x
Set.isSuperset.Seq.Box0	242	184	-24.0%	1.32x
ArrayAppendGenericStructs	1760	1340	-23.9%	1.31x (?)
MapReduceLazyCollectionShort	37	34	-8.1%	1.09x (?)

Code size: -O

Regression	OLD	NEW	DELTA	RATIO
DictionaryGroup.o	16197	17573	+8.5%	0.92x
MapReduce.o	31909	32261	+1.1%	0.99x
SetTests.o	141957	143437	+1.0%	0.99x

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeStubFromNSDateRef	3780	4200	+11.1%	0.90x (?)
Improvement	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeFromNSStringForced	2735	2545	-6.9%	1.07x (?)

Code size: -Osize

Regression	OLD	NEW	DELTA	RATIO
Hash.o	18997	19445	+2.4%	0.98x

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeStubToNSDate2	1320	1470	+11.4%	0.90x (?)
Improvement	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeStubFromNSDateRef	4850	4370	-9.9%	1.11x (?)

Code size: -swiftlibs

How to read the data

The tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.

If you see any unexpected regressions, you should consider fixing the
regressions before you merge the PR.

Noise: Sometimes the performance results (not code size!) contain false
alarms. Unexpected regressions which are marked with '(?)' are probably noise.
If you see regressions which you cannot explain you can try to run the
benchmarks again. If regressions still show up, please consult with the
performance team (@eeckstein).

Hardware Overview

  Model Name: Mac Pro
  Model Identifier: MacPro6,1
  Processor Name: 12-Core Intel Xeon E5
  Processor Speed: 2.7 GHz
  Number of Processors: 1
  Total Number of Cores: 12
  L2 Cache (per Core): 256 KB
  L3 Cache: 30 MB
  Memory: 64 GB

[eeckstein]

LGTM. Tough I don't know why @aschwaighofer excluded partial_apply on stack. He should approve this specific change.

[aschwaighofer]

The case of partial_apply [stack] is handled earlier by querying the isAllocatingStack() predicate so this part of the change is NFC.

[aschwaighofer]

You could add partial_apply [stack] here.

[atrick]

I don't think the previous support in Loop canDuplicate for on-stack PartialApply actually worked because it only handled the PartialApply, not the corresponding DeallocStack.

I added the full support here
#28261