[stdlib] Implement withContiguousStorageIfAvailable for raw buffer types

[glessard]

Until now, UnsafeRawBufferPointer and UnsafeMutableRawBufferPointer have not had an implementation of the withContiguousStorageIfAvailable and withContiguousMutableStorageIfAvailable functions, making a frequently-used fast path unavailable to those types.

The recently-landed SE-0333 (apple/swift#39529) means implementing the withContiguous{Mutable}StorageIfAvailable functions for the raw buffer types is now possible.

Resolves SR-15433 (rdar://84980367)

[glessard]

@swift-ci please smoke test linux platform

[glessard]

@swift-ci please test macOS platform

[glessard]

@swift-ci apple silicon benchmark

[glessard]

@swift-ci apple silicon benchmark

[swift-ci]

Performance (arm64): -O

Regression	OLD	NEW	DELTA	RATIO
RawBuffer.copyContents	10	12	+20.0%	0.83x (?)
NSArray.bridged.objectAtIndex	193	218	+13.0%	0.89x (?)
ArrayPlusEqualArrayOfInt	170	190	+11.8%	0.89x (?)
ObjectiveCBridgeFromNSStringForced	1435	1560	+8.7%	0.92x (?)
Improvement	OLD	NEW	DELTA	RATIO
Data.init.Sequence.64kB.Count.I	3	2	-33.3%	1.50x (?)
Data.init.Sequence.2047B.Count.I	8	7	-12.5%	1.14x (?)
DataCreateMedium	1000	900	-10.0%	1.11x (?)
BufferFillFromSlice	13	12	-7.7%	1.08x (?)

Code size: -O

Performance (arm64): -Osize

Regression	OLD	NEW	DELTA	RATIO
RawBuffer.copyContents	10	11	+10.0%	0.91x (?)
RawBufferCopyBytes	11	12	+9.1%	0.92x (?)
NSArray.bridged.objectAtIndex	202	220	+8.9%	0.92x (?)
Improvement	OLD	NEW	DELTA	RATIO
ArrayAppendOptionals	430	400	-7.0%	1.07x (?)
ObjectiveCBridgeStubNSDataAppend	1180	1100	-6.8%	1.07x (?)

Code size: -Osize

Performance (arm64): -Onone

Regression	OLD	NEW	DELTA	RATIO
ArrayAppendToGeneric	170	190	+11.8%	0.89x (?)
RawBufferCopyBytes	11	12	+9.1%	0.92x (?)
Improvement	OLD	NEW	DELTA	RATIO
ParseFloat.Float.Exp	8	7	-12.5%	1.14x (?)

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 mini
  Model Identifier: Macmini9,1
  Total Number of Cores: 8 (4 performance and 4 efficiency)
  Memory: 16 GB

[glessard]

I don't think any of these benchmark (performance) results are relevant. Thoughts, @lorentey?

[eeckstein]

yes, it looks like noise.

[lorentey]

This looks good!

Incidentally: I keep having a hard time grokking the nature of withMemoryRebound -- is rebinding memory semantically mutating the global state of a Swift program? Or is its effect localized somehow, to, say, a single thread? (I.e., is it really safe to call it from a nonmutating context, while some other parts of the program may be accessing the same memory, bound differently?)

[lorentey]

RawBuffer.copyContents is the one result that may be worth a look. Afaict it goes through URBP._copyContents instead of this path, though, so...

[glessard]

I looked at RawBuffer.copyContents, and it doesn't go through this path. Seems like a fluke.