[Runtime] Have ConcurrentReadableHashMap store indices inline when the table is sufficiently small.

[mikeash]

We use the lower two bits of the indices value to indicate the index size and inline-ness. When stored inline, we can pack 7 or 15 (on 32-bit and 64-bit respectively) values inline, meaning we don't need a heap allocation until we exceed that size, and avoiding the cost of an additional read in those cases.

[mikeash]

@swift-ci please benchmark

[mikeash]

@swift-ci please test

[swift-ci]

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
FlattenListLoop	1639	2503	+52.7%	0.65x (?)
FlattenListFlatMap	4774	6556	+37.3%	0.73x (?)
ObjectiveCBridgeFromNSArrayAnyObjectForced	6600	7200	+9.1%	0.92x (?)
SuffixAnyCollection	11	12	+9.1%	0.92x (?)
InsertCharacterEndIndex	165	178	+7.9%	0.93x (?)
Improvement	OLD	NEW	DELTA	RATIO
NSStringConversion.MutableCopy.Rebridge	998	852	-14.6%	1.17x (?)
EqualSubstringString	64	59	-7.8%	1.08x (?)
EqualStringSubstring	66	61	-7.6%	1.08x (?)

Code size: -O

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
StringFromLongWholeSubstring	4	5	+25.0%	0.80x
Improvement	OLD	NEW	DELTA	RATIO
FlattenListFlatMap	6801	5815	-14.5%	1.17x (?)
EqualStringSubstring	66	61	-7.6%	1.08x (?)

Code size: -Osize

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
SevenBoom	1818	2189	+20.4%	0.83x (?)
EqualSubstringSubstringGenericEquatable	144	159	+10.4%	0.91x (?)
LessSubstringSubstringGenericComparable	146	160	+9.6%	0.91x (?)
NSStringConversion.Rebridge.Mutable	1823	1991	+9.2%	0.92x (?)
StringBuilder	4540	4940	+8.8%	0.92x (?)
DictionaryLiteral	11720	12640	+7.8%	0.93x (?)
ArrayOfPOD	1051	1133	+7.8%	0.93x (?)
StringUTF16Builder	4510	4860	+7.8%	0.93x (?)
Improvement	OLD	NEW	DELTA	RATIO
DictionaryCompactMapValuesOfCastValue	72468	65988	-8.9%	1.10x (?)

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

[xedin]

Make sense to me!

[rjmccall]

LGTM

[mikeash]

Thanks for checking it over! This doesn't produce the performance win I was hoping for (fewer dependent memory accesses resulting in a minor boost) but as long as it's not a significant slowdown, it's worth it for the memory savings alone. I'm going to re-test and then get it in.

[mikeash]

@swift-ci please test

[mikeash]

@swift-ci please test windows platform

[mikeash]

@swift-ci please test windows platform

[mikeash]

@swift-ci please test windows platform