[semantic-arc] Eliminate simple recursive copy without borrows involved. #34812

gottesmm · 2020-11-18T21:31:23Z

This doesn't happen as much as the guaranteed + owned value optimization but can
occur once we start eliminating borrow scopes. I noticed that I needed this when
looking at the arbitrary RAUW for ownership API.

gottesmm · 2020-11-18T21:31:30Z

@swift-ci test

gottesmm · 2020-11-18T21:31:36Z

@swift-ci test source compatibility

gottesmm · 2020-11-18T21:31:45Z

@swift-ci smoke benchmark

swift-ci · 2020-11-18T21:49:28Z

Build failed
Swift Test OS X Platform
Git Sha - 37628233612dfef8505bef2496ad08a99ea38240

atrick · 2020-11-18T21:50:59Z

@gottesmm is this another workaround for not enabling CopyPropagation? Can we tag these somehow so we can remove all the ad-hoc complexity later?

swift-ci · 2020-11-18T22:24:55Z

Build failed
Swift Test Linux Platform
Git Sha - 37628233612dfef8505bef2496ad08a99ea38240

swift-ci · 2020-11-18T22:35:49Z

Build failed before running benchmark.

gottesmm · 2020-11-18T22:37:39Z

This needs #34809

gottesmm · 2020-11-19T03:02:38Z

I just added the commit from 34809 on top.

gottesmm · 2020-11-19T03:02:43Z

@swift-ci test

gottesmm · 2020-11-19T03:03:05Z

@swift-ci smoke benchmark

gottesmm · 2020-11-19T03:03:18Z

@swift-ci test source compatibility

swift-ci · 2020-11-19T04:10:24Z

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
StringFromLongWholeSubstring	4	5	+25.0%	0.80x
InsertCharacterEndIndexNonASCII	59	65	+10.2%	0.91x (?)
NSStringConversion.MutableCopy.Rebridge.Long	1081	1169	+8.1%	0.92x (?)
FloatingPointPrinting_Float_description_small	5400	5832	+8.0%	0.93x (?)
InsertCharacterEndIndex	165	178	+7.9%	0.93x (?)

Improvement	OLD	NEW	DELTA	RATIO
PrefixAnySequence	2316	1372	-40.8%	1.69x
DropWhileAnySequence	2811	1830	-34.9%	1.54x
DropFirstAnySequence	2797	1833	-34.5%	1.53x
NSStringConversion.UTF8	1033	934	-9.6%	1.11x (?)
FlattenListFlatMap	6659	6117	-8.1%	1.09x (?)
AngryPhonebook.Strasse.Small	1022	949	-7.1%	1.08x (?)
ObjectiveCBridgeStringHash	114	106	-7.0%	1.08x (?)
DataToStringLargeUnicode	7400	6900	-6.8%	1.07x (?)
NSStringConversion.MutableCopy.Rebridge.Medium	640	598	-6.6%	1.07x (?)

Code size: -O

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
StringFromLongWholeSubstring	4	5	+25.0%	0.80x
DropFirstAnySeqCRangeIter	164	199	+21.3%	0.82x
DropFirstAnySeqCntRange	164	199	+21.3%	0.82x
DropWhileAnyCollection	182	217	+19.2%	0.84x
DropWhileAnySeqCntRange	178	201	+12.9%	0.89x
DropWhileAnySeqCRangeIter	195	219	+12.3%	0.89x (?)
DropFirstAnySeqCRangeIterLazy	225	252	+12.0%	0.89x (?)
DropFirstAnySeqCntRangeLazy	225	251	+11.6%	0.90x (?)
PrefixAnyCollection	165	183	+10.9%	0.90x
DropFirstAnyCollection	165	182	+10.3%	0.91x (?)
DropWhileAnyCollectionLazy	252	276	+9.5%	0.91x (?)
DropWhileAnySeqCRangeIterLazy	247	270	+9.3%	0.91x
DropWhileAnySeqCntRangeLazy	247	270	+9.3%	0.91x
SuffixAnyCollection	65	71	+9.2%	0.92x (?)
DropLastAnyCollection	65	71	+9.2%	0.92x (?)
PrefixWhileAnyCollection	218	235	+7.8%	0.93x (?)

Improvement	OLD	NEW	DELTA	RATIO
DropWhileAnySequence	3191	2074	-35.0%	1.54x
DropFirstAnySequence	5712	4583	-19.8%	1.25x
DropWhileSequenceLazy	123	105	-14.6%	1.17x
PrefixAnySequence	6112	5356	-12.4%	1.14x
PrefixAnySeqCRangeIter	193	175	-9.3%	1.10x
UTF8Decode_InitFromData_ascii_as_ascii	673	615	-8.6%	1.09x (?)

Code size: -Osize

Improvement	OLD	NEW	DELTA	RATIO
DropWhile.o	16226	15773	-2.8%	1.03x

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
RandomDoubleDef	93400	105300	+12.7%	0.89x (?)
RandomDoubleLCG	57750	64284	+11.3%	0.90x (?)

Improvement	OLD	NEW	DELTA	RATIO
StringToDataEmpty	4750	3900	-17.9%	1.22x (?)
StringToDataSmall	4850	4000	-17.5%	1.21x (?)
DataToStringMedium	11650	9950	-14.6%	1.17x (?)
StringInterpolation	17600	16000	-9.1%	1.10x (?)
DataAppendDataMediumToSmall	6440	5940	-7.8%	1.08x (?)
DictionaryRemoveOfObjects	61000	56400	-7.5%	1.08x (?)
ArrayOfPOD	1131	1047	-7.4%	1.08x (?)

Code size: -swiftlibs

Improvement	OLD	NEW	DELTA	RATIO
libswiftFoundation.dylib	1359872	1343488	-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

gottesmm · 2020-11-19T09:22:58Z

@swift-ci test windows platform

gottesmm · 2020-11-19T09:23:40Z

@swift-ci benchmark

gottesmm · 2020-11-19T09:29:27Z

@atrick interesting... this one seemed to hit code-size on DropWhile.o. Thats real ARC traffic being eliminated. Also some stuff in Foundation.

gottesmm · 2020-11-19T09:34:01Z

lib/SILOptimizer/SemanticARC/CopyValueOpts.cpp

+  SmallVector<Operand *, 8> parentLifetimeEndingUses;
+  for (auto *origValueUse : originalValue->getUses())
+    if (origValueUse->isLifetimeEnding() &&
+        !isa<OwnershipForwardingInst>(origValueUse->getUser()))


Hmm... reading this... I think what I wrote here could be made better. I think at this point, the optimization is not dealing with LiveRanges, so the right thing to do is just gather up all consuming uses of the value, ignoring forwarding.

*ignoring forwarding I mean just treating forwarding insts as normal consuming instructions.

swift-ci · 2020-11-19T10:34:31Z

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
FlattenListLoop	1004	1545	+53.9%	0.65x (?)
FlattenListFlatMap	3204	4440	+38.6%	0.72x (?)
PrefixSequenceLazy	22	26	+18.2%	0.85x
PrefixSequence	22	26	+18.2%	0.85x

Improvement	OLD	NEW	DELTA	RATIO
PrefixAnySequence	1835	1046	-43.0%	1.75x
DropWhileAnySequence	2211	1394	-37.0%	1.59x
DropFirstAnySequence	2171	1390	-36.0%	1.56x
DropWhileArrayLazy	58	49	-15.5%	1.18x (?)
IterateData	944	803	-14.9%	1.18x (?)
ObjectiveCBridgeStringHash	70	64	-8.6%	1.09x (?)
ObjectiveCBridgeFromNSDictionaryAnyObject	25800	23900	-7.4%	1.08x (?)
DropWhileArray	28	26	-7.1%	1.08x (?)

Code size: -O

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
PrefixArray	13	26	+100.0%	0.50x
DropWhileArrayLazy	49	71	+44.9%	0.69x
PrefixCountableRangeLazy	18	26	+44.4%	0.69x
PrefixSequenceLazy	40	53	+32.5%	0.75x
PrefixSequence	40	53	+32.5%	0.75x
DropFirstSequence	40	50	+25.0%	0.80x
DropFirstSequenceLazy	40	50	+25.0%	0.80x
PrefixWhileAnyCollection	113	139	+23.0%	0.81x
DropLastAnyCollection	35	41	+17.1%	0.85x
DropFirstAnyCollection	86	99	+15.1%	0.87x (?)

Improvement	OLD	NEW	DELTA	RATIO
DropFirstArray	26	13	-50.0%	2.00x
DropFirstArrayLazy	26	14	-46.2%	1.86x
DropFirstCountableRangeLazy	26	16	-38.5%	1.62x
DropWhileAnySequence	2423	1650	-31.9%	1.47x
PrefixWhileAnyCollectionLazy	108	81	-25.0%	1.33x (?)
PrefixWhileAnySeqCRangeIterLazy	107	81	-24.3%	1.32x (?)
PrefixWhileAnySeqCntRangeLazy	107	81	-24.3%	1.32x (?)
DropWhileSequenceLazy	81	62	-23.5%	1.31x
DropFirstAnySequence	4121	3295	-20.0%	1.25x
PrefixAnySeqCRangeIterLazy	101	81	-19.8%	1.25x (?)
PrefixAnySeqCntRangeLazy	101	81	-19.8%	1.25x (?)
PrefixAnyCollection	109	89	-18.3%	1.22x
DropWhileCountableRangeLazy	71	58	-18.3%	1.22x
FlattenListLoop	1207	999	-17.2%	1.21x (?)
DropFirstAnySeqCRangeIterLazy	143	119	-16.8%	1.20x
DropFirstAnySeqCntRangeLazy	143	119	-16.8%	1.20x (?)
PrefixAnySequence	4565	3805	-16.6%	1.20x
IterateData	940	803	-14.6%	1.17x
SuffixAnyCollection	41	37	-9.8%	1.11x (?)
DropWhileAnySeqCRangeIter	109	100	-8.3%	1.09x (?)
ObjectiveCBridgeStringHash	69	64	-7.2%	1.08x (?)
DropLastAnySeqCRangeIterLazy	49800	46220	-7.2%	1.08x (?)

Code size: -Osize

Improvement	OLD	NEW	DELTA	RATIO
DropWhile.o	16226	15773	-2.8%	1.03x

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
ArrayAppendLatin1Substring	28692	31284	+9.0%	0.92x (?)

Improvement	OLD	NEW	DELTA	RATIO
DropWhileAnySeqCntRangeLazy	53299	48439	-9.1%	1.10x (?)
DropLastAnySeqCRangeIterLazy	51874	47524	-8.4%	1.09x (?)
AngryPhonebook.Cyrillic	168	154	-8.3%	1.09x (?)
AngryPhonebook.Armenian	156	143	-8.3%	1.09x (?)
Data.hash.Empty	145	133	-8.3%	1.09x (?)
AngryPhonebook.Cyrillic.Small	622	574	-7.7%	1.08x (?)
AngryPhonebook.Armenian.Small	622	575	-7.6%	1.08x (?)
DropLastAnyCollection	12605	11683	-7.3%	1.08x (?)
DropWhileAnySeqCntRange	54409	50453	-7.3%	1.08x (?)
AngryPhonebook.Strasse.Small	727	675	-7.2%	1.08x (?)
DropLastAnySeqCRangeIter	50366	46773	-7.1%	1.08x (?)
AngryPhonebook.Strasse	136	127	-6.6%	1.07x (?)
DropFirstAnySeqCRangeIter	50235	46913	-6.6%	1.07x (?)

Code size: -swiftlibs

Improvement	OLD	NEW	DELTA	RATIO
libswiftFoundation.dylib	1359872	1343488	-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 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

gottesmm · 2020-11-20T02:00:53Z

@swift-ci smoke test

This doesn't happen as much as the guaranteed + owned value optimization but can occur once we start eliminating borrow scopes. I noticed that I needed this when looking at the arbitrary RAUW for ownership API.

gottesmm · 2020-11-20T06:45:49Z

Just had to mixup a merge conflict due to me landing a different PR that touched the same test.

gottesmm · 2020-11-20T06:45:59Z

@swift-ci smoke test

gottesmm requested review from atrick and meg-gupta November 18, 2020 21:31

gottesmm commented Nov 19, 2020

View reviewed changes

gottesmm force-pushed the pr-b1ab03067e8684c69df8ad52319746f5b9d145e0 branch from 4603e72 to b9b6eea Compare November 20, 2020 02:00

[semantic-arc] Eliminate simple recursive copy without borrows involved.

1fd4003

This doesn't happen as much as the guaranteed + owned value optimization but can occur once we start eliminating borrow scopes. I noticed that I needed this when looking at the arbitrary RAUW for ownership API.

gottesmm force-pushed the pr-b1ab03067e8684c69df8ad52319746f5b9d145e0 branch from b9b6eea to 1fd4003 Compare November 20, 2020 06:44

gottesmm merged commit 7bb9aa7 into swiftlang:main Nov 20, 2020

gottesmm deleted the pr-b1ab03067e8684c69df8ad52319746f5b9d145e0 branch November 20, 2020 09:06

atrick mentioned this pull request Nov 22, 2020

[semantic-arc] Optimize more lifetime joining when a copy_value, destroy_value are in the same block. #34844

Merged

[semantic-arc] Eliminate simple recursive copy without borrows involved. #34812

[semantic-arc] Eliminate simple recursive copy without borrows involved. #34812

Uh oh!

Conversation

gottesmm commented Nov 18, 2020

Uh oh!

gottesmm commented Nov 18, 2020

Uh oh!

gottesmm commented Nov 18, 2020

Uh oh!

gottesmm commented Nov 18, 2020

Uh oh!

swift-ci commented Nov 18, 2020

Uh oh!

atrick commented Nov 18, 2020

Uh oh!

swift-ci commented Nov 18, 2020

Uh oh!

swift-ci commented Nov 18, 2020

Uh oh!

gottesmm commented Nov 18, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

swift-ci commented Nov 19, 2020

Performance: -O

Code size: -O

Performance: -Osize

Code size: -Osize

Performance: -Onone

Code size: -swiftlibs

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm commented Nov 19, 2020

Uh oh!

gottesmm Nov 19, 2020

Choose a reason for hiding this comment

Uh oh!

gottesmm Nov 19, 2020

Choose a reason for hiding this comment

Uh oh!

swift-ci commented Nov 19, 2020

Performance: -O

Code size: -O

Performance: -Osize

Code size: -Osize

Performance: -Onone

Code size: -swiftlibs

Uh oh!

gottesmm commented Nov 20, 2020

Uh oh!

gottesmm commented Nov 20, 2020

Uh oh!

gottesmm commented Nov 20, 2020

Uh oh!

Uh oh!