[SILGen] Use temporary initializer for switch address subject.

[zoecarver]

If the subject of a switch statement is a tuple address, using an initializer to initialize it gets rid of one stack allocation for each tuple element.

Based on #31779.

Right now we also have to create a temporary allocation but, I have another patch that removes it in SILGen.

Update: removed example that is no longer correct.

[zoecarver]

@swift-ci please benchmark.

[zoecarver]

@swift-ci please benchmark

[swift-ci]

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
EqualSubstringSubstring	22	30	+36.4%	0.73x
LessSubstringSubstring	22	30	+36.4%	0.73x
EqualSubstringSubstringGenericEquatable	22	30	+36.4%	0.73x
LessSubstringSubstringGenericComparable	22	30	+36.4%	0.73x
EqualStringSubstring	23	31	+34.8%	0.74x
EqualSubstringString	23	30	+30.4%	0.77x
ObjectiveCBridgeStubFromNSDate	3790	4630	+22.2%	0.82x (?)
ObjectiveCBridgeStringHash	72	84	+16.7%	0.86x
ArrayAppendLatin1Substring	19332	22536	+16.6%	0.86x (?)
ArrayAppendAsciiSubstring	18936	22068	+16.5%	0.86x (?)
ArrayAppendUTF16Substring	18972	22068	+16.3%	0.86x (?)
UTF8Decode_InitDecoding	141	164	+16.3%	0.86x
UTF8Decode_InitFromCustom_contiguous	142	164	+15.5%	0.87x (?)
Breadcrumbs.MutatedIdxToUTF16.Mixed	185	209	+13.0%	0.89x
DataAccessBytesMedium	45	50	+11.1%	0.90x (?)
Data.hash.Empty	45	50	+11.1%	0.90x
StringComparison_longSharedPrefix	321	356	+10.9%	0.90x (?)
ObjectiveCBridgeStringGetASCIIContents	287	317	+10.5%	0.91x (?)
AngryPhonebook.Strasse	136	149	+9.6%	0.91x (?)
UTF8Decode_InitFromCustom_noncontiguous	302	329	+8.9%	0.92x (?)
IterateData	901	980	+8.8%	0.92x (?)
SubstringEqualString	276	297	+7.6%	0.93x (?)
Improvement	OLD	NEW	DELTA	RATIO
ChainedFilterMap	621	504	-18.8%	1.23x (?)
ArrayAppendLazyMap	850	690	-18.8%	1.23x (?)

Code size: -O

Regression	OLD	NEW	DELTA	RATIO
DictionaryKeysContains.o	8299	8620	+3.9%	0.96x
DictionaryCompactMapValues.o	13093	13509	+3.2%	0.97x
Improvement	OLD	NEW	DELTA	RATIO
ChainedFilterMap.o	3429	3365	-1.9%	1.02x

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
ArrayAppendRepeatCol	420	850	+102.4%	0.49x
EqualSubstringSubstring	22	30	+36.4%	0.73x
EqualStringSubstring	22	30	+36.4%	0.73x
EqualSubstringSubstringGenericEquatable	22	30	+36.4%	0.73x
EqualSubstringString	22	30	+36.4%	0.73x
LessSubstringSubstring	23	31	+34.8%	0.74x
LessSubstringSubstringGenericComparable	23	30	+30.4%	0.77x
ObjectiveCBridgeStringHash	71	84	+18.3%	0.85x (?)
UTF8Decode_InitDecoding	140	165	+17.9%	0.85x
UTF8Decode_InitFromCustom_contiguous	144	168	+16.7%	0.86x
Breadcrumbs.MutatedUTF16ToIdx.Mixed	181	204	+12.7%	0.89x
Breadcrumbs.MutatedIdxToUTF16.Mixed	185	208	+12.4%	0.89x
StringComparison_longSharedPrefix	324	361	+11.4%	0.90x (?)
ArrayAppendLatin1Substring	19368	21492	+11.0%	0.90x (?)
ArrayAppendUTF16Substring	19008	21060	+10.8%	0.90x (?)
ArrayAppendAsciiSubstring	19008	21060	+10.8%	0.90x (?)
AngryPhonebook.Strasse	136	150	+10.3%	0.91x (?)
ObjectiveCBridgeStringGetASCIIContents	288	317	+10.1%	0.91x
UTF8Decode_InitFromData	148	162	+9.5%	0.91x (?)
FlattenListFlatMap	4164	4540	+9.0%	0.92x (?)
AngryPhonebook	252	273	+8.3%	0.92x
Calculator	156	168	+7.7%	0.93x (?)
Improvement	OLD	NEW	DELTA	RATIO
Dictionary4OfObjects	328	276	-15.9%	1.19x
Dictionary4	240	207	-13.7%	1.16x
DropWhileAnySeqCntRange	105	96	-8.6%	1.09x (?)
StrToInt	1210	1120	-7.4%	1.08x (?)
DropWhileAnyCollectionLazy	157	146	-7.0%	1.08x (?)
RGBHistogram	1640	1530	-6.7%	1.07x (?)

Code size: -Osize

Regression	OLD	NEW	DELTA	RATIO
DictionaryKeysContains.o	8108	8508	+4.9%	0.95x
DictionaryCompactMapValues.o	12330	12730	+3.2%	0.97x
Improvement	OLD	NEW	DELTA	RATIO
SortLettersInPlace.o	8173	8005	-2.1%	1.02x
ChainedFilterMap.o	3334	3286	-1.4%	1.01x

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
CaptureProp	416020	567160	+36.3%	0.73x
LessSubstringSubstring	27	35	+29.6%	0.77x
EqualSubstringSubstringGenericEquatable	27	35	+29.6%	0.77x
EqualSubstringSubstring	27	34	+25.9%	0.79x
LessSubstringSubstringGenericComparable	27	34	+25.9%	0.79x
EqualStringSubstring	28	35	+25.0%	0.80x
EqualSubstringString	28	35	+25.0%	0.80x
ObjectiveCBridgeStringHash	72	86	+19.4%	0.84x (?)
UTF8Decode_InitDecoding	146	173	+18.5%	0.84x
UTF8Decode_InitFromCustom_contiguous	149	174	+16.8%	0.86x
Breadcrumbs.MutatedIdxToUTF16.Mixed	198	223	+12.6%	0.89x (?)
Breadcrumbs.MutatedUTF16ToIdx.Mixed	196	220	+12.2%	0.89x (?)
StringRemoveDupes	490	547	+11.6%	0.90x (?)
ObjectiveCBridgeStringGetASCIIContents	287	317	+10.5%	0.91x (?)
AngryPhonebook.Strasse	136	149	+9.6%	0.91x (?)
Improvement	OLD	NEW	DELTA	RATIO
StringWalk	2920	2480	-15.1%	1.18x (?)
ArrayOfPOD	723	643	-11.1%	1.12x (?)
PrefixSequenceLazy	3543	3219	-9.1%	1.10x (?)
ObjectiveCBridgeStubToNSDateRef	2720	2480	-8.8%	1.10x (?)
PrefixWhileSequenceLazy	4696	4306	-8.3%	1.09x (?)
PrefixSequence	3513	3228	-8.1%	1.09x (?)

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 mini
  Model Identifier: Macmini8,1
  Processor Name: 6-Core Intel Core i7
  Processor Speed: 3.2 GHz
  Number of Processors: 1
  Total Number of Cores: 6
  L2 Cache (per Core): 256 KB
  L3 Cache: 12 MB
  Memory: 64 GB

[shahmishal]

Please update the base branch to main by Oct 5th otherwise the pull request will be closed automatically.

• How to change the base branch: (Link)
• More detail about the branch update: (Link)