[metadata prespecialization] On by default. #29488

nate-chandler · 2020-01-27T23:54:25Z

Previously, -Xfrontend -prespecialize-generic-metadata had to be passed in order for generic metadata to be prespecialized. Now it is prespecialized unless -Xfrontend -disable-generic-metadata-prespecialization is passed.

nate-chandler · 2020-01-27T23:54:37Z

@swift-ci please test

nate-chandler · 2020-01-27T23:54:46Z

@swift-ci please test windows platform

nate-chandler · 2020-01-28T00:57:04Z

@swift-ci please clean test windows platform

swift-ci · 2020-01-28T01:00:06Z

Build failed
Swift Test Linux Platform
Git Sha - 35242caac6ed5659f210a9333b81d139f1f66851

swift-ci · 2020-01-28T01:26:26Z

Build failed
Swift Test OS X Platform
Git Sha - 35242caac6ed5659f210a9333b81d139f1f66851

nate-chandler · 2020-01-28T19:25:31Z

@swift-ci please test macos platform

swift-ci · 2020-01-28T19:26:23Z

Build failed
Swift Test OS X Platform
Git Sha - 35242caac6ed5659f210a9333b81d139f1f66851

nate-chandler · 2020-01-28T21:21:40Z

@swift-ci please test macos platform

nate-chandler · 2020-01-28T21:31:55Z

@swift-ci please clean test windows platform

swift-ci · 2020-01-28T22:25:30Z

Build failed
Swift Test OS X Platform
Git Sha - 8185fe5161337dd81a3bd6042bc8ed2299e15e96

nate-chandler · 2020-01-29T02:19:53Z

@swift-ci please test

swift-ci · 2020-01-29T02:22:20Z

Build failed
Swift Test Linux Platform
Git Sha - 8185fe5161337dd81a3bd6042bc8ed2299e15e96

swift-ci · 2020-01-29T02:24:12Z

Build failed
Swift Test OS X Platform
Git Sha - 8185fe5161337dd81a3bd6042bc8ed2299e15e96

nate-chandler · 2020-01-30T03:02:12Z

@swift-ci please test linux platform

swift-ci · 2020-01-30T03:04:42Z

Build failed
Swift Test Linux Platform
Git Sha - 5c93f07b5301685b93c1abbb922351a915773e62

Previously, when emitting the metadata accessor, the generic arguments of the type were enumerated one after the next. That was fine in most cases but is incorrect in cases where the actual number of generic arguments is less than apparent number as when two arguments are required to be equal. Now, the arguments are enumerated according to the requirements vended by the GenericTypeRequirements struct.

nate-chandler · 2020-01-30T19:31:03Z

@swift-ci please test linux platform

swift-ci · 2020-01-30T19:33:34Z

Build failed
Swift Test Linux Platform
Git Sha - 9bcd84275668fad4c0899d9e48a896412a359f91

nate-chandler · 2020-01-31T00:20:58Z

@swift-ci please test macos platform

nate-chandler · 2020-01-31T02:23:12Z

@swift-ci please test linux platform

swift-ci · 2020-02-04T19:48:02Z

Build failed
Swift Test Linux Platform
Git Sha - fbd509460f60959656bc7ff9d240cfa1fdd28072

Previously, some ad hoc checks were done in order to determine whether the metadata access for a generic type was trivial. Now, standard predicates are used, specifically IRGenModule's member functions isDependentConformance and isResilientConformance.

nate-chandler · 2020-02-04T19:59:52Z

@swift-ci please test

swift-ci · 2020-02-04T20:02:19Z

Build failed
Swift Test Linux Platform
Git Sha - 8c79ba5d979110a5e87f2034a87dbb019dbda7e7

swift-ci · 2020-02-04T20:04:21Z

Build failed
Swift Test OS X Platform
Git Sha - 8c79ba5d979110a5e87f2034a87dbb019dbda7e7

nate-chandler · 2020-02-04T22:52:39Z

@swift-ci please test windows platform

nate-chandler · 2020-02-04T23:23:38Z

@swift-ci please benchmark

swift-ci · 2020-02-05T00:08:34Z

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
PrefixAnySeqCntRange	33	37	+12.1%	0.89x (?)

Improvement	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeFromNSSetAnyObjectToString	72500	65000	-10.3%	1.12x (?)

Code size: -O

Performance: -Osize

Code size: -Osize

Performance: -Onone

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: 8-Core Intel Xeon E5
  Processor Speed: 3 GHz
  Number of Processors: 1
  Total Number of Cores: 8
  L2 Cache (per Core): 256 KB
  L3 Cache: 25 MB
  Memory: 64 GB

nate-chandler · 2020-02-05T20:40:08Z

@swift-ci please clean test windows platform

shahmishal · 2020-02-05T21:35:24Z

@swift-ci please clean test windows platform

nate-chandler · 2020-02-06T00:09:40Z

@swift-ci please test

nate-chandler · 2020-02-06T00:09:55Z

@swift-ci please test windows platform

swift-ci · 2020-02-06T00:10:03Z

Build failed
Swift Test Linux Platform
Git Sha - fca1669d154ade308a0efdc964867faa146f3bec

swift-ci · 2020-02-06T00:10:18Z

Build failed
Swift Test OS X Platform
Git Sha - fca1669d154ade308a0efdc964867faa146f3bec

nate-chandler · 2020-02-06T01:05:42Z

@swift-ci please test

nate-chandler · 2020-02-06T01:05:54Z

@swift-ci please test windows platform

swift-ci · 2020-02-06T01:08:15Z

Build failed
Swift Test Linux Platform
Git Sha - 536571d583044ab25755d4f5424059fe3853c210

swift-ci · 2020-02-06T01:10:16Z

Build failed
Swift Test OS X Platform
Git Sha - 536571d583044ab25755d4f5424059fe3853c210

Previously, -Xfrontend -prespecialize-generic-metadata had to be passed in order for generic metadata to be prespecialized. Now it is prespecialized unless -Xfrontend -disable-generic-metadata-prespecialization is passed.

Temporarily disable metadata prespecialization on platforms other than MacOS, iOS, tvOS, watchOS, or Linux. At the moment, tests are failing on Windows with linker errors such as demangleToMetadata-558ea9.o : error LNK2001: unresolved external symbol $ss5Int64VN demangleToMetadata-558ea9.o : error LNK2001: unresolved external symbol $sSSSHsWP Once the issue leading to those linker errors has been resolved, the feature will be enabled on Windows.

nate-chandler · 2020-02-06T02:08:21Z

@swift-ci please test

nate-chandler · 2020-02-06T02:08:28Z

@swift-ci please test windows platform

swift-ci · 2020-02-06T02:08:42Z

Build failed
Swift Test Linux Platform
Git Sha - 07fa6a67208ce09f84b4cfd19141d87e071b42a1

swift-ci · 2020-02-06T02:10:30Z

Build failed
Swift Test OS X Platform
Git Sha - 07fa6a67208ce09f84b4cfd19141d87e071b42a1

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 35242ca to e499d8b Compare January 28, 2020 19:24

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch 2 times, most recently from 414e834 to 8185fe5 Compare January 28, 2020 21:21

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch 2 times, most recently from cca094a to 5c93f07 Compare January 29, 2020 02:19

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch 2 times, most recently from 1f50c8e to 9bcd842 Compare January 30, 2020 03:01

nate-chandler changed the title ~~[frontend] Default on metadata prespecialization.~~ [metadata prespecialization] On by default. Jan 30, 2020

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 9bcd842 to dc77f0a Compare January 30, 2020 19:30

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from dc77f0a to 9a03cd7 Compare January 31, 2020 00:20

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 8c79ba5 to fca1669 Compare February 4, 2020 19:59

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 536571d to 07fa6a6 Compare February 6, 2020 01:01

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 07fa6a6 to 99ec30a Compare February 6, 2020 01:19

nate-chandler added 2 commits February 5, 2020 17:53

[metadata prespecialization] On by default.

0ae9e68

Previously, -Xfrontend -prespecialize-generic-metadata had to be passed in order for generic metadata to be prespecialized. Now it is prespecialized unless -Xfrontend -disable-generic-metadata-prespecialization is passed.

nate-chandler force-pushed the generic-metadata-prespecialization-components/default-on branch from 99ec30a to ae1f41e Compare February 6, 2020 02:08

nate-chandler merged commit 4a97914 into swiftlang:master Feb 6, 2020

nate-chandler deleted the generic-metadata-prespecialization-components/default-on branch February 6, 2020 17:10

[metadata prespecialization] On by default. #29488

[metadata prespecialization] On by default. #29488

Uh oh!

Conversation

nate-chandler commented Jan 27, 2020

Uh oh!

nate-chandler commented Jan 27, 2020

Uh oh!

nate-chandler commented Jan 27, 2020

Uh oh!

nate-chandler commented Jan 28, 2020

Uh oh!

swift-ci commented Jan 28, 2020

Uh oh!

swift-ci commented Jan 28, 2020

Uh oh!

nate-chandler commented Jan 28, 2020

Uh oh!

swift-ci commented Jan 28, 2020

Uh oh!

nate-chandler commented Jan 28, 2020

Uh oh!

nate-chandler commented Jan 28, 2020

Uh oh!

swift-ci commented Jan 28, 2020

Uh oh!

nate-chandler commented Jan 29, 2020

Uh oh!

swift-ci commented Jan 29, 2020

Uh oh!

swift-ci commented Jan 29, 2020

Uh oh!

nate-chandler commented Jan 30, 2020

Uh oh!

swift-ci commented Jan 30, 2020

Uh oh!

nate-chandler commented Jan 30, 2020

Uh oh!

swift-ci commented Jan 30, 2020

Uh oh!

nate-chandler commented Jan 31, 2020

Uh oh!

nate-chandler commented Jan 31, 2020

Uh oh!

swift-ci commented Feb 4, 2020

Uh oh!

nate-chandler commented Feb 4, 2020

Uh oh!

swift-ci commented Feb 4, 2020

Uh oh!

swift-ci commented Feb 4, 2020

Uh oh!

nate-chandler commented Feb 4, 2020

Uh oh!

nate-chandler commented Feb 4, 2020

Uh oh!

swift-ci commented Feb 5, 2020

Performance: -O

Code size: -O

Performance: -Osize

Code size: -Osize

Performance: -Onone

Code size: -swiftlibs

Uh oh!

nate-chandler commented Feb 5, 2020

Uh oh!

shahmishal commented Feb 5, 2020

Uh oh!

nate-chandler commented Feb 6, 2020

Uh oh!

nate-chandler commented Feb 6, 2020

Uh oh!

swift-ci commented Feb 6, 2020

Uh oh!

swift-ci commented Feb 6, 2020

Uh oh!

nate-chandler commented Feb 6, 2020

Uh oh!

nate-chandler commented Feb 6, 2020

Uh oh!