Speed up KeyAnalyzer for substring based frozen collections

[IDisposable]

Building on PR #88516, I tried making everything as explicitly inline-able and sealed to get the JIT to eliminate any conditional jumps in the hot loop by teasing out the various comparator options into trait-like classes to mix in to get sealed classes where all the logic in conditional-free (down to the internals of the String.AsSpan() methods themselves. This required making some potentially hard-to-read changes, but the essence is that now we construct a comparer and hash set for the from-the-left attempt, and then lazily construct another comparer and matching hash set only if we need one for from-the-right attempt.

Since in the vast number of cases, we're going to spend the bulk of the execution time iterating through the uniqueStrings we're presented with, the overhead of having another comparer/collection is pretty low in comparison.

To ensure we don't bloat memory. I reversed when we clean the set to right before we return false from HasSufficientUniquenessFactor since we know that we're going to get called again for a different index/count... this means that both HashSet<string>s spend their lifetimes empty except when being tested (we could clear them on success, as well, but the old code didn't bother).

The resulting SubstringComparer classes could be useful elsewhere, but are currently just internal to the System.Collections.Immutable assembly. I suspect we could/should pass them out in the AnalysisResults and use the chosen substring extractor as it already has the offset, length, and ignore/honor case encapsulated, but that would be another pass. If we do that, I would like to add some tests for those classes first.

I also added some more tests to the KeyAnalyser tests to verify newly exposed behavior.

The other changes imported from PR #88516 here are:

• Eliminate the the inner TryUseSubstring because we can just early return the calculated results as we build them
• Hoist the calculation of acceptableNonUniqueCount out to the top level since it never changes (which means we pass that into the HasSufficientUniquenessFactor method for it to "use up" internally (passed by value, so unchanged at call-site)
• Eliminated the delegate ReadOnlySpan<char> GetSpan and use, which helps reduce dynamic dispatch overhead in the CreateAnalysisResults method
• Eliminated the IsLeft field of the SubstringComparer since we can tell by the Index being negative that we're doing right-justified slicing (and documented that on the class)
• Changed the logic managing the Index and Count on the comparer for right-justified substrings.
• Added [MethodImpl(MethodImplOptions.AggressiveInlining)] to the Equals and GetHashCode overrides.

[ghost]

Tagging subscribers to this area: @dotnet/area-system-collections
See info in area-owners.md if you want to be subscribed.

Issue Details

---

null

Author:	IDisposable
Assignees:	-
Labels:	area-System.Collections, community-contribution
Milestone:	-

[IDisposable]

Performance tests.

• Baseline is the main as of d5c4a4e
• Faster Freeze is the code in PR Move the IsLeft/IsRight decision out of the loop and use computed substring set #88516 which shows between a 1%-3% performance improvement and unchanged allocation (see that PR for discussion)
• Trait Based is this code (which picks up where faster freeze left off), and shows identical allocation and a 1%-3% performance gain on my test machine.

BenchmarkDotNet=v0.13.2.2052-nightly, OS=Windows 11 (10.0.22631.2115)
Intel Core i7-10875H CPU 2.30GHz, 1 CPU, 16 logical and 8 physical cores
.NET SDK=8.0.100-preview.7.23322.33
[Host] : .NET 8.0.0 (8.0.23.32106), X64 RyuJIT AVX2
Job-SVXVNR : .NET 8.0.0 (42.42.42.42424), X64 RyuJIT AVX2
Job-CESATQ : .NET 8.0.0 (42.42.42.42424), X64 RyuJIT AVX2
Job-YVAYLV : .NET 8.0.0 (42.42.42.42424), X64 RyuJIT AVX2
PowerPlanMode=00000000-0000-0000-0000-000000000000 Arguments=/p:EnableUnsafeBinaryFormatterSerialization=true IterationTime=250.0000 ms MaxIterationCount=20 MinIterationCount=15 WarmupCount=1 

Method	Job	Toolchain	Count	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(2ms)	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
ToFrozenDictionary	Job-SVXVNR	Trait Based	10	1.233 μs	0.0240 μs	0.0247 μs	1.241 μs	1.183 μs	1.279 μs	0.99	Same	0.02	0.2038	-	-	1.68 KB	1.00
ToFrozenDictionary	Job-CESATQ	Baseline	10	1.250 μs	0.0246 μs	0.0242 μs	1.257 μs	1.210 μs	1.281 μs	1.00	Base	0.00	0.2056	-	-	1.68 KB	1.00
ToFrozenDictionary	Job-YVAYLV	Faster Freeze	10	1.225 μs	0.0244 μs	0.0280 μs	1.215 μs	1.190 μs	1.260 μs	0.98	Same	0.03	0.2031	-	-	1.67 KB	1.00
ToFrozenDictionary	Job-SVXVNR	Trait Based	100	7.263 μs	0.1453 μs	0.1554 μs	7.286 μs	7.035 μs	7.590 μs	0.99	Same	0.03	1.4245	0.0297	-	11.83 KB	1.00
ToFrozenDictionary	Job-CESATQ	Baseline	100	7.307 μs	0.1418 μs	0.1393 μs	7.366 μs	7.068 μs	7.471 μs	1.00	Base	0.00	1.4462	0.0284	-	11.83 KB	1.00
ToFrozenDictionary	Job-YVAYLV	Faster Freeze	100	7.461 μs	0.1212 μs	0.1134 μs	7.504 μs	7.253 μs	7.637 μs	1.02	Same	0.02	1.4284	0.0292	-	11.82 KB	1.00
ToFrozenDictionary	Job-SVXVNR	Trait Based	1000	68.360 μs	1.0448 μs	0.9773 μs	68.455 μs	66.576 μs	69.596 μs	0.98	Same	0.02	11.1702	2.6596	-	92.64 KB	1.00
ToFrozenDictionary	Job-CESATQ	Baseline	1000	69.919 μs	0.8861 μs	0.7399 μs	70.078 μs	68.216 μs	70.834 μs	1.00	Base	0.00	11.1607	2.5112	-	92.64 KB	1.00
ToFrozenDictionary	Job-YVAYLV	Faster Freeze	1000	67.473 μs	1.1315 μs	1.1620 μs	67.719 μs	65.659 μs	69.484 μs	0.97	Same	0.02	11.0759	2.6371	-	92.63 KB	1.00
ToFrozenDictionary	Job-SVXVNR	Trait Based	10000	1,104.651 μs	4.3610 μs	3.6416 μs	1,104.166 μs	1,098.515 μs	1,112.397 μs	1.00	Same	0.01	148.4375	148.4375	148.4375	904.43 KB	1.00
ToFrozenDictionary	Job-CESATQ	Baseline	10000	1,108.193 μs	10.2913 μs	8.0347 μs	1,109.215 μs	1,087.775 μs	1,123.516 μs	1.00	Base	0.00	145.8333	145.8333	145.8333	904.43 KB	1.00
ToFrozenDictionary	Job-YVAYLV	Faster Freeze	10000	1,093.550 μs	12.1788 μs	10.1698 μs	1,094.617 μs	1,064.160 μs	1,107.952 μs	0.99	Same	0.01	145.8333	145.8333	145.8333	904.42 KB	1.00

[IDisposable]

This was a fantastic learning experience... This gained, at most, 3% performance without adversely impacting the allocation patterns and helped me learn the GSW pattern's application other places we want to eliminate virtual and delete overheads.

[danmoseley]

What is GSW pattern in this context? Maybe i missed it.

[IDisposable]

What is GSW pattern in this context?

Using a GenericSpecializedWrapper to "pass in" the explicit implementation details at template declaration time as a struct that allows the compile to devirtualize all the interface methods. I saw it in the various frozen set implementations.

[IDisposable]

I've done another pass to clean up the documentation and make it a little cleaner.

[IDisposable]

Superseded by PR #89863