COWOpts: handle debug_value instructions

[eeckstein]

Don't let debug_value instructions bail the optimization.

This fixes a couple of performance regressions, which were introduced by adding more debug_value instructions (#38736).

rdar://82327743

[eeckstein]

@swift-ci smoke test

[eeckstein]

@swift-ci benchmark

[swift-ci]

Performance (x86_64): -O

Regression	OLD	NEW	DELTA	RATIO
DictionaryOfAnyHashableStrings_insert	2744	5572	+103.1%	0.49x
Set.isDisjoint.Int25	268	345	+28.7%	0.78x (?)
Set.isDisjoint.Int50	268	335	+25.0%	0.80x (?)
DictionaryKeysContainsNative	21	25	+19.0%	0.84x (?)
FloatingPointPrinting_Float_description_uniform	4700	5100	+8.5%	0.92x (?)
Improvement	OLD	NEW	DELTA	RATIO
XorLoop	2033	381	-81.3%	5.34x
ArrayAppendReserved	1600	460	-71.2%	3.48x
RC4	422	149	-64.7%	2.83x
ArrayPlusEqualFiveElementCollection	9065	3367	-62.9%	2.69x
ArrayPlusEqualSingleElementCollection	2068	799	-61.4%	2.59x
ArrayAppend	1920	780	-59.4%	2.46x
BucketSort	276	123	-55.4%	2.24x
FlattenListLoop	5499	2548	-53.7%	2.16x
StringFromLongWholeSubstring	5	4	-20.0%	1.25x
AngryPhonebook.Armenian	259	208	-19.7%	1.25x (?)
AngryPhonebook.Cyrillic	276	223	-19.2%	1.24x
AngryPhonebook.Strasse	217	183	-15.7%	1.19x
Breadcrumbs.MutatedIdxToUTF16.Mixed	310	273	-11.9%	1.14x (?)

Code size: -O

Improvement	OLD	NEW	DELTA	RATIO
XorLoop.o	2099	1602	-23.7%	1.31x
ArraySubscript.o	3108	2382	-23.4%	1.30x
BinaryFloatingPointProperties.o	5264	5104	-3.0%	1.03x
Hash.o	22407	21783	-2.8%	1.03x
FlattenList.o	4106	3994	-2.7%	1.03x
BucketSort.o	9015	8799	-2.4%	1.02x
ArrayAppend.o	25337	24777	-2.2%	1.02x
StringEdits.o	10229	10005	-2.2%	1.02x
MonteCarloE.o	2937	2889	-1.6%	1.02x
ProtocolConformance.o	4031	3983	-1.2%	1.01x
DiffingMyers.o	6909	6829	-1.2%	1.01x
DataBenchmarks.o	59879	59231	-1.1%	1.01x
Array2D.o	2981	2949	-1.1%	1.01x

Performance (x86_64): -Osize

Regression	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeStubFromNSDate	6270	7560	+20.6%	0.83x (?)
DictionaryKeysContainsNative	26	30	+15.4%	0.87x (?)
Improvement	OLD	NEW	DELTA	RATIO
XorLoop	1890	394	-79.2%	4.80x
ArrayAppendReserved	1480	460	-68.9%	3.22x
FlattenListLoop	4345	1693	-61.0%	2.57x
ArrayPlusEqualFiveElementCollection	8325	3367	-59.6%	2.47x
ArrayPlusEqualSingleElementCollection	1927	799	-58.5%	2.41x
RC4	401	172	-57.1%	2.33x
ArrayAppend	1700	780	-54.1%	2.18x
BucketSort	330	196	-40.6%	1.68x
AngryPhonebook.Armenian	259	207	-20.1%	1.25x
StringFromLongWholeSubstring	5	4	-20.0%	1.25x
AngryPhonebook.Cyrillic	276	223	-19.2%	1.24x
AngryPhonebook.Strasse	217	183	-15.7%	1.19x
Breadcrumbs.MutatedUTF16ToIdx.Mixed	304	265	-12.8%	1.15x (?)
NSStringConversion.MutableCopy.UTF8	1045	924	-11.6%	1.13x (?)
AngryPhonebook.Cyrillic.Small	587	523	-10.9%	1.12x (?)
MonteCarloE	10820	9740	-10.0%	1.11x
AngryPhonebook.Armenian.Small	569	517	-9.1%	1.10x (?)
NSStringConversion.MutableCopy.Rebridge.UTF8	846	774	-8.5%	1.09x (?)
AngryPhonebook.Strasse.Small	646	595	-7.9%	1.09x (?)
ArraySubscript	1208	1120	-7.3%	1.08x (?)

Code size: -Osize

Improvement	OLD	NEW	DELTA	RATIO
XorLoop.o	1999	1368	-31.6%	1.46x
ArraySubscript.o	3085	2382	-22.8%	1.30x
RC4.o	3387	3085	-8.9%	1.10x
BinaryFloatingPointProperties.o	5271	5111	-3.0%	1.03x
NopDeinit.o	4412	4288	-2.8%	1.03x
PrimsNonStrongRef.o	86729	84381	-2.7%	1.03x
BucketSort.o	8608	8398	-2.4%	1.03x
Hash.o	19001	18557	-2.3%	1.02x
ArrayAppend.o	22817	22309	-2.2%	1.02x
MonteCarloE.o	2922	2857	-2.2%	1.02x
FlattenList.o	3930	3846	-2.1%	1.02x
Array2D.o	2717	2666	-1.9%	1.02x
ProtocolConformance.o	3786	3742	-1.2%	1.01x
StringEdits.o	10163	10048	-1.1%	1.01x
RemoveWhere.o	13615	13473	-1.0%	1.01x

Performance (x86_64): -Onone

Regression	OLD	NEW	DELTA	RATIO
StringBuilderWithLongSubstring	4350	4920	+13.1%	0.88x (?)
SIMDReduce.Int8x16.Initializer	617480	680102	+10.1%	0.91x (?)
String.data.Small	81	89	+9.9%	0.91x (?)
Improvement	OLD	NEW	DELTA	RATIO
AngryPhonebook.Armenian	259	208	-19.7%	1.25x
AngryPhonebook.Cyrillic	276	223	-19.2%	1.24x
AngryPhonebook.Strasse	217	184	-15.2%	1.18x
Breadcrumbs.MutatedUTF16ToIdx.Mixed	317	272	-14.2%	1.17x (?)
Breadcrumbs.MutatedIdxToUTF16.Mixed	337	292	-13.4%	1.15x (?)
DictionaryBridgeToObjC_BulkAccess	175	163	-6.9%	1.07x (?)

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]

@swift-ci smoke test linux

[meg-gupta]

Thanks!