SIL: Don't drop generic signatures where all parameters are concrete

[slavapestov]

No description provided.

[slavapestov]

@jrose-apple I realized this changes optimizer behavior; if the benchmark results are affected we'll have to tweak this a little bit.

[slavapestov]

@swift-ci Please smoke test

[slavapestov]

@swift-ci Please test source compatibility

[slavapestov]

@swift-ci Please benchmark

[swift-ci]

Performance: -O

TEST	OLD	NEW	DELTA	RATIO
Regression
LessSubstringSubstring	39	44	+12.8%	0.89x
EqualStringSubstring	39	43	+10.3%	0.91x
EqualSubstringString	39	43	+10.3%	0.91x
LessSubstringSubstringGenericComparable	39	43	+10.3%	0.91x
EqualSubstringSubstring	39	42	+7.7%	0.93x (?)
EqualSubstringSubstringGenericEquatable	39	42	+7.7%	0.93x
ObjectiveCBridgeStubFromNSDateRef	4060	4370	+7.6%	0.93x (?)
Improvement
DictionaryBridgeToObjC_Access	1014	905	-10.7%	1.12x (?)
FlattenListLoop	4336	3976	-8.3%	1.09x
Array2D	7520	6912	-8.1%	1.09x (?)
MapReduce	397	368	-7.3%	1.08x (?)
FlattenListFlatMap	6540	6074	-7.1%	1.08x (?)
MapReduceAnyCollection	397	369	-7.1%	1.08x

Performance: -Osize

TEST	OLD	NEW	DELTA	RATIO
Regression
LessSubstringSubstringGenericComparable	39	44	+12.8%	0.89x
EqualSubstringSubstring	39	43	+10.3%	0.91x
EqualStringSubstring	39	43	+10.3%	0.91x (?)
LessSubstringSubstring	40	44	+10.0%	0.91x
EqualSubstringSubstringGenericEquatable	40	44	+10.0%	0.91x
EqualSubstringString	40	44	+10.0%	0.91x
Improvement
FlattenListLoop	4433	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
ArrayOfPOD	776	855	+10.2%	0.91x (?)
EqualStringSubstring	46	50	+8.7%	0.92x (?)
Improvement
DictionaryGroup	8663	7832	-9.6%	1.11x (?)
DropWhileAnySeqCRangeIter	54913	49966	-9.0%	1.10x (?)
DropWhileAnySeqCntRange	55173	50292	-8.8%	1.10x (?)
StrComplexWalk	7280	6640	-8.8%	1.10x

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

[slavapestov]

Ok, this should now fix the optimizer crash. If it doesn't do it, I'll have to do a spot fix and postpone this for now since it's already more painful than I imagined.

[slavapestov]

@swift-ci Please smoke test

[slavapestov]

@swift-ci Please test source compatibility

[slavapestov]

@swift-ci Please test source compatibility

[slavapestov]

@swift-ci Please test source compatibility

[slavapestov]

@swift-ci Please benchmark

[swift-ci]

Performance: -O

TEST	OLD	NEW	DELTA	RATIO
Regression
LessSubstringSubstring	39	44	+12.8%	0.89x
EqualStringSubstring	39	43	+10.3%	0.91x (?)
EqualSubstringString	39	43	+10.3%	0.91x
LessSubstringSubstringGenericComparable	39	43	+10.3%	0.91x
FlattenListLoop	3976	4336	+9.1%	0.92x
Array2D	6912	7520	+8.8%	0.92x
DictionaryOfAnyHashableStrings_lookup	3912	4224	+8.0%	0.93x (?)
MapReduce	368	397	+7.9%	0.93x
EqualSubstringSubstring	39	42	+7.7%	0.93x
EqualSubstringSubstringGenericEquatable	39	42	+7.7%	0.93x (?)
FlattenListFlatMap	6083	6549	+7.7%	0.93x (?)
SubstringEqualString	405	436	+7.7%	0.93x (?)
MapReduceAnyCollection	369	397	+7.6%	0.93x (?)

Performance: -Osize

TEST	OLD	NEW	DELTA	RATIO
Regression
LessSubstringSubstringGenericComparable	39	44	+12.8%	0.89x
EqualSubstringSubstring	39	43	+10.3%	0.91x (?)
EqualStringSubstring	39	43	+10.3%	0.91x (?)
LessSubstringSubstring	40	44	+10.0%	0.91x
EqualSubstringSubstringGenericEquatable	40	44	+10.0%	0.91x
EqualSubstringString	40	44	+10.0%	0.91x
FlattenListLoop	4070	4431	+8.9%	0.92x (?)
Array2D	6912	7520	+8.8%	0.92x (?)
SubstringEqualString	404	438	+8.4%	0.92x (?)
Set.isDisjoint.Box.Empty	190	205	+7.9%	0.93x

Performance: -Onone

TEST	OLD	NEW	DELTA	RATIO
Regression
ArrayOfPOD	776	857	+10.4%	0.91x (?)
Improvement
StrComplexWalk	7280	6560	-9.9%	1.11x
WordCountUniqueASCII	13040	11980	-8.1%	1.09x (?)
WordCountUniqueUTF16	13390	12340	-7.8%	1.09x (?)

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

[jrose-apple]

Is it just that we're considering these less profitable to inline or something?

[slavapestov]

@jrose-apple We don't inline generic functions at all actually

[gottesmm]

@slavapestov this is not true. I think we do it in the standard library. This is a piece of work that we did not have enough manpower to finish when @swiftix left. I think the problem was that we did not find a good heuristic yet that didn't create too much code-size increase. But for the stdlib it wasn't so bad and we saw nice perf increase.

[atrick]

@slavapestov are you picking this for 5.1? If so, I don't need to pick my fix.

[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)