[CSBindings] Detect situations when type variable bindings could be i… #20205

xedin · 2018-11-01T00:44:36Z

…ncomplete

When bind param constraint is associated with a given type
variable a set of bindings collected for it is potentially
incomplete, because binding gathering doesn't look through related
type variables. Which means that such type variable has to be
de-prioritized until bind param constraint is resolved
or there is just nothing else to try, otherwise there is a
risk that solver would skip some of the valid solutions.

Resolves: rdar://problem/45659733

xedin · 2018-11-01T00:46:05Z

/cc @rudkx

xedin · 2018-11-01T00:47:06Z

In the future we'd be better off merging fully bound and other flags with this "incomplete" flag.

xedin · 2018-11-01T00:48:23Z

@swift-ci please test

xedin · 2018-11-01T00:48:42Z

@swift-ci please test source compatibility

test/Constraints/generics.swift

…ncomplete When `bind param` constraint is associated with a given type variable a set of bindings collected for it is potentially incomplete, because binding gathering doesn't look through related type variables. Which means that such type variable has to be de-prioritized until `bind param` constraint is resolved or there is just nothing else to try, otherwise there is a risk that solver would skip some of the valid solutions. Resolves: rdar://problem/45659733

xedin · 2018-11-01T00:59:40Z

@swift-ci please test

xedin · 2018-11-01T00:59:49Z

@swift-ci please test source compatibility

xedin · 2018-11-01T03:38:12Z

@swift-ci please smoke test compiler performance

xedin · 2018-11-01T03:42:36Z

@swift-ci please smoke test compiler performance

DougGregor · 2018-11-01T04:27:01Z

lib/Sema/ConstraintSystem.h

+
+      // As a last resort, let's check if the bindings are
+      // potentially incomplete, and if so, let's de-prioritize them.
+      return x.PotentiallyIncomplete < y.PotentiallyIncomplete;


It's Yet Another ranking step for potential bindings, and I agree with you that this should be merged back with FullyBound (or something like it) at some point. If you need this PR to make progress... okay...

I actually tried to make it so FullyBound is set when there is a bind param constraint but it breaks ordering even worse (tried a bunch of different combinations), I’m hoping to re-visit fully bound soon. Trying to unblock some code with this PR...

Maybe the best possible answer here is to re-work bind param itself, so we don’t get into situations like that, or make some of the choices retry-able

swift-ci · 2018-11-01T05:05:07Z

Build comment file:

Summary for master smoketest

Unexpected test results, excluded stats for Kingfisher, ReactiveCocoa, ReactiveSwift

No regressions above thresholds

Debug
- debug brief
- debug detailed
Release
- release brief
- release detailed

Debug

debug brief

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (3)

name	old	new	delta	delta_pct
Frontend.NumInstructionsExecuted	80,482,651,641	80,534,212,473	51,560,832	0.06%
LLVM.NumLLVMBytesOutput	2,898,548	2,898,548	0	0.0%
time.swift-driver.wall	11.0s	11.0s	10.7ms	0.1%

debug detailed

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (23)

name	old	new	delta	delta_pct
AST.NumImportedExternalDefinitions	3,577	3,577	0	0.0%
AST.NumLoadedModules	654	654	0	0.0%
AST.NumTotalClangImportedEntities	13,508	13,508	0	0.0%
AST.NumUsedConformances	508	508	0	0.0%
IRModule.NumIRBasicBlocks	12,092	12,092	0	0.0%
IRModule.NumIRFunctions	5,240	5,240	0	0.0%
IRModule.NumIRGlobals	4,708	4,708	0	0.0%
IRModule.NumIRInsts	133,974	133,974	0	0.0%
IRModule.NumIRValueSymbols	9,677	9,677	0	0.0%
LLVM.NumLLVMBytesOutput	2,898,548	2,898,548	0	0.0%
SILModule.NumSILGenFunctions	2,482	2,482	0	0.0%
SILModule.NumSILOptFunctions	3,541	3,541	0	0.0%
Sema.NumConformancesDeserialized	8,942	8,942	0	0.0%
Sema.NumConstraintScopes	43,096	43,076	-20	-0.05%
Sema.NumDeclsDeserialized	97,842	97,842	0	0.0%
Sema.NumDeclsValidated	5,358	5,358	0	0.0%
Sema.NumFunctionsTypechecked	3,274	3,274	0	0.0%
Sema.NumGenericSignatureBuilders	2,738	2,738	0	0.0%
Sema.NumLazyGenericEnvironments	20,626	20,626	0	0.0%
Sema.NumLazyGenericEnvironmentsLoaded	432	432	0	0.0%
Sema.NumLazyIterableDeclContexts	18,938	18,938	0	0.0%
Sema.NumTypesDeserialized	39,229	39,229	0	0.0%
Sema.NumTypesValidated	3,385	3,385	0	0.0%

Release

release brief

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (3)

name	old	new	delta	delta_pct
Frontend.NumInstructionsExecuted	96,093,480,692	96,059,692,258	-33,788,434	-0.04%
LLVM.NumLLVMBytesOutput	3,076,532	3,076,516	-16	-0.0%
time.swift-driver.wall	18.1s	18.1s	-1.4ms	-0.01%

release detailed

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (23)

name	old	new	delta	delta_pct
AST.NumImportedExternalDefinitions	829	829	0	0.0%
AST.NumLoadedModules	65	65	0	0.0%
AST.NumTotalClangImportedEntities	3,124	3,124	0	0.0%
AST.NumUsedConformances	508	508	0	0.0%
IRModule.NumIRBasicBlocks	13,533	13,533	0	0.0%
IRModule.NumIRFunctions	4,221	4,221	0	0.0%
IRModule.NumIRGlobals	4,218	4,218	0	0.0%
IRModule.NumIRInsts	116,051	116,051	0	0.0%
IRModule.NumIRValueSymbols	8,318	8,318	0	0.0%
LLVM.NumLLVMBytesOutput	3,076,532	3,076,516	-16	-0.0%
SILModule.NumSILGenFunctions	1,959	1,959	0	0.0%
SILModule.NumSILOptFunctions	2,720	2,720	0	0.0%
Sema.NumConformancesDeserialized	5,676	5,676	0	0.0%
Sema.NumConstraintScopes	37,838	37,818	-20	-0.05%
Sema.NumDeclsDeserialized	21,982	21,982	0	0.0%
Sema.NumDeclsValidated	3,000	3,000	0	0.0%
Sema.NumFunctionsTypechecked	1,676	1,676	0	0.0%
Sema.NumGenericSignatureBuilders	632	632	0	0.0%
Sema.NumLazyGenericEnvironments	4,473	4,473	0	0.0%
Sema.NumLazyGenericEnvironmentsLoaded	82	82	0	0.0%
Sema.NumLazyIterableDeclContexts	3,224	3,224	0	0.0%
Sema.NumTypesDeserialized	11,538	11,538	0	0.0%
Sema.NumTypesValidated	1,517	1,517	0	0.0%

xedin · 2018-11-01T20:36:25Z

@swift-ci please smoke test

xedin · 2018-11-01T20:36:38Z

@swift-ci please smoke test

xedin · 2018-11-01T20:36:55Z

I'll run smoke test one more time before merging, just in case...

xedin · 2018-11-01T22:11:11Z

@swift-ci please smoke test Linux platform

xedin · 2018-11-02T00:00:37Z

@swift-ci please smoke test Linux platform

xedin · 2018-11-02T08:11:50Z

@swift-ci please smoke test Linux platform

xedin requested a review from slavapestov November 1, 2018 00:44

xedin force-pushed the rdar-45659733 branch from 723d6ca to 3384ca0 Compare November 1, 2018 00:48

swiftlang deleted a comment from swift-ci Nov 1, 2018

slavapestov reviewed Nov 1, 2018

View reviewed changes

test/Constraints/generics.swift Outdated Show resolved Hide resolved

slavapestov approved these changes Nov 1, 2018

View reviewed changes

xedin requested a review from DougGregor November 1, 2018 00:56

xedin force-pushed the rdar-45659733 branch from 3384ca0 to 4f70889 Compare November 1, 2018 00:59

swiftlang deleted a comment from swift-ci Nov 1, 2018

DougGregor approved these changes Nov 1, 2018

View reviewed changes

xedin merged commit a3ce27c into swiftlang:master Nov 2, 2018

[CSBindings] Detect situations when type variable bindings could be i… #20205

[CSBindings] Detect situations when type variable bindings could be i… #20205

Uh oh!

Conversation

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

DougGregor Nov 1, 2018

Choose a reason for hiding this comment

Uh oh!

xedin Nov 1, 2018

Choose a reason for hiding this comment

Uh oh!

xedin Nov 1, 2018

Choose a reason for hiding this comment

Uh oh!

swift-ci commented Nov 1, 2018

Build comment file:

Summary for master smoketest

Debug

debug brief

debug detailed

Release

release brief

release detailed

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 1, 2018

Uh oh!

xedin commented Nov 2, 2018

Uh oh!

xedin commented Nov 2, 2018

Uh oh!

Uh oh!