[DO NOT MERGE][stdlib] Keep Dictionary storage at least 25% full when possible

[lorentey]

This is an experimental implementation of shrinking Dictionary storage after removals.

It isn't clear this is worth doing, and the bulk of the actual work towards landing it would be in benchmarking and adding test coverage, which is not at all done. However, the implementation is an interesting exercise in library evolution.

It's almost possible to make this feature deploy back to earlier OSes -- except Set/Dictionary doesn't record the reserved capacity in 5.0 & 5.1, and that arguably rules out shrinking. (We'd also have to make the minimum load factor part of the ABI, but that's less of an issue.)

rdar://problem/18114559

[lorentey]

Just to see if we have adequate benchmarks for Dictionary removals:

@swift-ci benchmark

[lorentey]

@swift-ci test

[swift-ci]

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
DictionaryRemove	3150	7460	+136.8%	0.42x
DictionaryRemoveOfObjects	21700	27900	+28.6%	0.78x
ObjectiveCBridgeStubFromNSDateRef	3760	4190	+11.4%	0.90x (?)
Improvement	OLD	NEW	DELTA	RATIO
String.data.Medium	98	86	-12.2%	1.14x (?)
CharacterLiteralsLarge	108	97	-10.2%	1.11x (?)
String.data.LargeUnicode	106	96	-9.4%	1.10x (?)
CharacterLiteralsSmall	340	317	-6.8%	1.07x (?)

Code size: -O

Regression	OLD	NEW	DELTA	RATIO
DictionaryRemove.o	15985	16337	+2.2%	0.98x

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
DictionaryRemove	5640	15100	+167.7%	0.37x
DictionaryRemoveOfObjects	29900	48800	+63.2%	0.61x
Improvement	OLD	NEW	DELTA	RATIO
ObjectiveCBridgeStubFromNSStringRef	207	175	-15.5%	1.18x (?)
ObjectiveCBridgeFromNSSetAnyObjectToString	78000	71500	-8.3%	1.09x (?)
NSStringConversion.UTF8	831	776	-6.6%	1.07x (?)

Code size: -Osize

Improvement	OLD	NEW	DELTA	RATIO
DictionaryRemove.o	10833	10657	-1.6%	1.02x

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
DictionaryRemove	13230	24070	+81.9%	0.55x
DictionaryRemoveOfObjects	44400	69300	+56.1%	0.64x
Improvement	OLD	NEW	DELTA	RATIO
ArrayOfPOD	1146	1064	-7.2%	1.08x (?)

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: 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

[lorentey]

A 2x slowdown matches how insertion behaves when capacity was reserved vs when it wasn't.

Some of it is probably spent on runtime availability checks that we could perhaps eliminate.

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - a2e2f9b

[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)