[Perf] Regression in Xml_ToStream Serializers

[DrewScoggins]

Run Information

Architecture	x64
OS	Windows 10.0.18362
Baseline	71d0d2b275efa091f235c569f01869f39bfb7638
Compare	e983168f871dcba090ff6ec08c0d2c7d4353fb8e
Diff	Diff

Regressions in MicroBenchmarks.Serializers.Xml_ToStream<MyEventsListerViewModel>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
XmlSerializer_ - Duration of single invocation	392.35 μs	491.84 μs	1.25	0.46	False

Historical Data in Reporting System

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'MicroBenchmarks.Serializers.Xml_ToStream<MyEventsListerViewModel>*'

Payloads

Baseline
Compare

Histogram

MicroBenchmarks.Serializers.Xml_ToStream<MyEventsListerViewModel>.XmlSerializer_

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

### Run Information

Architecture	x64
OS	Windows 10.0.18362
Baseline	71d0d2b275efa091f235c569f01869f39bfb7638
Compare	e983168f871dcba090ff6ec08c0d2c7d4353fb8e
Diff	Diff

Regressions in MicroBenchmarks.Serializers.Xml_ToStream<IndexViewModel>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
XmlSerializer_ - Duration of single invocation	40.90 μs	49.20 μs	1.20	0.13	False

Historical Data in Reporting System

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'MicroBenchmarks.Serializers.Xml_ToStream<IndexViewModel>*'

Payloads

Baseline
Compare

Histogram

MicroBenchmarks.Serializers.Xml_ToStream<IndexViewModel>.XmlSerializer_

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[DrewScoggins]

Seems related to #55101

[DrewScoggins]

Improvement:

Run Information

Architecture	x64
OS	Windows 10.0.18362
Baseline	71d0d2b275efa091f235c569f01869f39bfb7638
Compare	e983168f871dcba090ff6ec08c0d2c7d4353fb8e
Diff	Diff

Improvemnts in MicroBenchmarks.Serializers.Xml_FromStream<MyEventsListerViewModel>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
XmlSerializer_ - Duration of single invocation	1.17 ms	895.25 μs	0.77	0.01	True

Historical Data in Reporting System

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'MicroBenchmarks.Serializers.Xml_FromStream<MyEventsListerViewModel>*'

Payloads

Baseline
Compare

Histogram

MicroBenchmarks.Serializers.Xml_FromStream<MyEventsListerViewModel>.XmlSerializer_

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

### Run Information

Architecture	x64
OS	Windows 10.0.18362
Baseline	71d0d2b275efa091f235c569f01869f39bfb7638
Compare	e983168f871dcba090ff6ec08c0d2c7d4353fb8e
Diff	Diff

Improvemnts in MicroBenchmarks.Serializers.Xml_FromStream<IndexViewModel>

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
XmlSerializer_ - Duration of single invocation	102.70 μs	88.16 μs	0.86	0.01	False

Historical Data in Reporting System

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'MicroBenchmarks.Serializers.Xml_FromStream<IndexViewModel>*'

Payloads

Baseline
Compare

Histogram

MicroBenchmarks.Serializers.Xml_FromStream<IndexViewModel>.XmlSerializer_

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

[mconnew]

@DrewScoggins, #55101 wouldn't have caused a regression unless the types being serialized contain DateTimeOffset properties. If that's the case, then the regression is okay as we're doing more work as we're now actually serializing DateTimeOffset now. If there's no DateTimeOffset properties involved, then the problem was somewhere else.

[DrewScoggins]

OK, those types do have DateTimeOffset. Before #55101 we were not doing the serialization at all, and now we are? If that is the case then seems like a regression would be expected.

[mconnew]

Yes, that is exactly the case. Previously it would serialize as an empty XML element. Now it will output an XML format date time, and will deserialize it too.

[stephentoub]

Closing as expected / by-design. Thanks.