Start using the best resilience expansion in SIL #23170

slavapestov · 2019-03-07T22:13:48Z

Follow-up to #23263.

Finally start using the best resilience expansion in SIL type lowering.

Completes rdar://problem/24057844, https://bugs.swift.org/browse/SR-261.

slavapestov · 2019-03-13T06:40:18Z

@swift-ci Please benchmark

slavapestov · 2019-03-13T06:40:28Z

@swift-ci Please test source compatibility

slavapestov · 2019-03-13T07:04:15Z

@swift-ci Please test

swift-ci · 2019-03-13T07:34:07Z

Performance: -O

TEST	OLD	NEW	DELTA	RATIO
Regression
IterateData	1462	4088	+179.6%	0.36x
Improvement
ObjectiveCBridgeStubDateMutation	394	255	-35.3%	1.55x
ObjectiveCBridgeStubFromNSDate	6560	5230	-20.3%	1.25x (?)
FlattenListLoop	4337	3974	-8.4%	1.09x (?)
MapReduce	397	368	-7.3%	1.08x (?)
FlattenListFlatMap	6541	6072	-7.2%	1.08x (?)
MapReduceAnyCollection	397	369	-7.1%	1.08x (?)

Performance: -Osize

TEST	OLD	NEW	DELTA	RATIO
Regression
IterateData	1462	4042	+176.5%	0.36x
CharacterPropertiesStashed	1870	2050	+9.6%	0.91x (?)
CharacterPropertiesStashedMemo	1840	2010	+9.2%	0.92x (?)
ObjectiveCBridgeToNSArray	14000	15100	+7.9%	0.93x (?)
Improvement
ObjectiveCBridgeStubDateMutation	400	257	-35.7%	1.56x
ObjectiveCBridgeStubFromNSDate	6600	5150	-22.0%	1.28x (?)
DictionaryLiteral	7710	6940	-10.0%	1.11x (?)
FlattenListLoop	4432	4069	-8.2%	1.09x (?)
Array2D	7520	6912	-8.1%	1.09x (?)
MapReduce	433	404	-6.7%	1.07x (?)

Performance: -Onone

TEST	OLD	NEW	DELTA	RATIO
Regression
IterateData	5782	7974	+37.9%	0.73x
FatCompactMap	456050	532290	+16.7%	0.86x (?)
RemoveWhereFilterInts	3296	3798	+15.2%	0.87x (?)
RemoveWhereFilterStrings	4117	4706	+14.3%	0.87x (?)
ArrayAppendLazyMap	289930	325270	+12.2%	0.89x (?)
FloatingPointPrinting_Float_interpolated	54600	60600	+11.0%	0.90x (?)
FloatingPointPrinting_Float80_interpolated	95600	105600	+10.5%	0.91x (?)
Breadcrumbs.IdxToUTF16Range.longASCII	364	402	+10.4%	0.91x (?)
Breadcrumbs.UTF16ToIdxRange.longASCII	356	393	+10.4%	0.91x (?)
DataCreateEmptyArray	148700	164150	+10.4%	0.91x (?)
Breadcrumbs.CopyUTF16CodeUnits.ASCII	190	209	+10.0%	0.91x (?)
FloatingPointPrinting_Double_interpolated	72200	79200	+9.7%	0.91x (?)
Breadcrumbs.CopyUTF16CodeUnits.Mixed	222	243	+9.5%	0.91x (?)
LazilyFilteredArrays2	127100	138800	+9.2%	0.92x (?)
NormalizedIterator_ascii	491	536	+9.2%	0.92x (?)
Breadcrumbs.UTF16ToIdxRange.longMixed	443	483	+9.0%	0.92x (?)
ObjectiveCBridgeFromNSArrayAnyObjectToStringForced	46600	50400	+8.2%	0.92x (?)
WordCountUniqueASCII	12220	13200	+8.0%	0.93x (?)
NormalizedIterator_latin1	790	852	+7.8%	0.93x (?)
CharIteration_tweet_unicodeScalars_Backwards	732800	789080	+7.7%	0.93x (?)
Improvement
ObjectiveCBridgeStubFromNSDate	6230	4910	-21.2%	1.27x (?)
ObjectiveCBridgeStubDateMutation	771	628	-18.5%	1.23x
DictionaryCompactMapValuesOfCastValue	73764	63990	-13.3%	1.15x (?)
LuhnAlgoEager	5002	4460	-10.8%	1.12x (?)
UTF8Decode_InitDecoding_ascii	452	405	-10.4%	1.12x (?)
DictionaryGroup	8646	7782	-10.0%	1.11x (?)
DropWhileAnySeqCntRange	55216	50097	-9.3%	1.10x (?)
DropWhileAnySeqCRangeIter	55112	50070	-9.1%	1.10x (?)
DictionaryLiteral	8030	7360	-8.3%	1.09x (?)
RangeReplaceableCollectionPlusDefault	9732	9024	-7.3%	1.08x (?)
DataReplaceMediumBuffer	11800	11000	-6.8%	1.07x (?)

Code size: -swiftlibs

TEST	OLD	NEW	DELTA	RATIO
Regression
libswiftNetwork.dylib	155648	172032	+10.5%	0.90x
libswiftFoundation.dylib	1581056	1642496	+3.9%	0.96x
libswiftCore.dylib	3612672	3690496	+2.2%	0.98x
Improvement
libswiftDispatch.dylib	90112	81920	-9.1%	1.10x
libswiftCloudKit.dylib	77824	73728	-5.3%	1.06x
libswiftStdlibUnittest.dylib	352256	348160	-1.2%	1.01x

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

swift-ci · 2019-03-13T08:19:24Z

Build failed
Swift Test Linux Platform
Git Sha - 13f76bd14406c6058b802426d2116b240598be1c

swift-ci · 2019-03-13T08:55:45Z

Build failed
Swift Test OS X Platform
Git Sha - 13f76bd14406c6058b802426d2116b240598be1c

atrick · 2019-04-26T18:26:48Z

test/SILOptimizer/globalopt_resilience.swift

+// CHECK: return [[RESULT]] : $()
+
+// CHECK-LABEL: sil @$s4test26cannotConvertToValueUseAltyyF : $@convention(thin) () -> ()
+// CHECK: [[TMP:%.*]] = alloc_stack $ResilientStruct


I would like at least a FIXME here.

FIXME: The optimizer should easily eliminate this copy. The problem is that copy forwarding assumes that address-only types are always explicitly destroyed.

This is a large patch; I couldn't split it up further while still keeping things working. There are four things being changed at once here: - Places that call SILType::isAddressOnly()/isLoadable() now call the SILFunction overload and not the SILModule one. - SILFunction's overloads of getTypeLowering() and getLoweredType() now pass the function's resilience expansion down, instead of hardcoding ResilienceExpansion::Minimal. - Various other places with '// FIXME: Expansion' now use a better resilience expansion. - A few tests were updated to reflect SILGen's improved code generation, and some new tests are added to cover more code paths that previously were uncovered and only manifested themselves as standard library build failures while I was working on this change.

… a SILModule Finally we can remove the old overloads.

slavapestov · 2019-04-27T02:48:33Z

@swift-ci Please smoke test

slavapestov · 2019-04-27T02:48:45Z

@swift-ci Please test source compatibility

slavapestov force-pushed the enable-sil-resilience-expansion branch 3 times, most recently from 4152571 to 13f76bd Compare March 13, 2019 06:39

slavapestov changed the title ~~Start using the correct resilience expansion in SIL~~ Start using the best resilience expansion in SIL Mar 13, 2019

This was referenced Mar 14, 2019

Start using best resilience expansion in SIL #23286

Merged

IRGen: Use maximal resilience expansion to lower SIL types #23288

Merged

slavapestov force-pushed the enable-sil-resilience-expansion branch from 13f76bd to b69c6b9 Compare March 19, 2019 04:05

slavapestov force-pushed the enable-sil-resilience-expansion branch from b69c6b9 to 89f5287 Compare April 5, 2019 16:06

slavapestov force-pushed the enable-sil-resilience-expansion branch from 89f5287 to d7c7ead Compare April 17, 2019 12:00

atrick reviewed Apr 26, 2019

View reviewed changes

slavapestov force-pushed the enable-sil-resilience-expansion branch from d7c7ead to f168c00 Compare April 26, 2019 22:12

slavapestov marked this pull request as ready for review April 26, 2019 22:13

slavapestov force-pushed the enable-sil-resilience-expansion branch from f168c00 to eef3b22 Compare April 27, 2019 01:45

slavapestov added 2 commits April 26, 2019 22:47

SIL: Remove SILType::isAddressOnly()/isLoadable() overloads that take…

d069cc4

… a SILModule Finally we can remove the old overloads.

slavapestov force-pushed the enable-sil-resilience-expansion branch from eef3b22 to d069cc4 Compare April 27, 2019 02:48

slavapestov merged commit 9587a3a into swiftlang:master Apr 27, 2019

slavapestov mentioned this pull request Apr 27, 2019

[SR-261] Plumb ResilienceExpansion through SIL type lowering #42883

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Start using the best resilience expansion in SIL #23170

Start using the best resilience expansion in SIL #23170

Uh oh!

slavapestov commented Mar 7, 2019 •

edited

Loading

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

swift-ci commented Mar 13, 2019

Uh oh!

swift-ci commented Mar 13, 2019

Uh oh!

swift-ci commented Mar 13, 2019

Uh oh!

atrick Apr 26, 2019

Uh oh!

slavapestov commented Apr 27, 2019

Uh oh!

slavapestov commented Apr 27, 2019

Uh oh!

Uh oh!

Start using the best resilience expansion in SIL #23170

Start using the best resilience expansion in SIL #23170

Uh oh!

Conversation

slavapestov commented Mar 7, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

slavapestov commented Mar 13, 2019

Uh oh!

swift-ci commented Mar 13, 2019

Performance: -O

Performance: -Osize

Performance: -Onone

Code size: -swiftlibs

Uh oh!

swift-ci commented Mar 13, 2019

Uh oh!

swift-ci commented Mar 13, 2019

Uh oh!

atrick Apr 26, 2019

Choose a reason for hiding this comment

Uh oh!

slavapestov commented Apr 27, 2019

Uh oh!

slavapestov commented Apr 27, 2019

Uh oh!

Uh oh!

slavapestov commented Mar 7, 2019 •

edited

Loading