[Perf] Linux/x64: Regressions in System.IO.Compression

[performanceautofiler]

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	ViperUbuntu
Baseline	2b3088b99e97f0d6d3dd24dcd41ab330757f5aa2
Compare	f9016e3507a9669894c500c1cdd4a45aaef24354
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.IO.Compression.ZLib

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Compress - Duration of single invocation ADX - Test Multi Config Graph	258.85 μs	1.16 ms	4.50	0.04	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	281.64 μs	326.49 μs	1.16	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	364.03 μs	511.00 μs	1.40	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	574.52 μs	1.61 ms	2.81	0.02	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	2.63 ms	7.14 ms	2.72	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	111.58 μs	139.90 μs	1.25	0.42	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	1.78 ms	2.36 ms	1.32	0.02	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	261.50 μs	323.78 μs	1.24	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	45.79 μs	63.59 μs	1.39	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	47.45 μs	70.89 μs	1.49	0.01	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
python3 .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.IO.Compression.ZLib*'

System.IO.Compression.ZLib.Compress(level: Optimal, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Decompress(level: Fastest, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Decompress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Compress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Compress(level: Optimal, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Compress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Compress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Decompress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Decompress(level: Optimal, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.ZLib.Decompress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	ViperUbuntu
Baseline	2b3088b99e97f0d6d3dd24dcd41ab330757f5aa2
Compare	f9016e3507a9669894c500c1cdd4a45aaef24354
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.IO.Compression.Gzip

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Decompress - Duration of single invocation ADX - Test Multi Config Graph	359.47 μs	570.68 μs	1.59	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	2.70 ms	7.20 ms	2.66	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	47.40 μs	86.22 μs	1.82	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	915.09 μs	2.20 ms	2.40	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	112.26 μs	150.80 μs	1.34	0.19	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	285.22 μs	377.61 μs	1.32	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	252.01 μs	547.55 μs	2.17	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	1.79 ms	2.36 ms	1.32	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	262.29 μs	375.85 μs	1.43	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	581.96 μs	1.67 ms	2.86	0.05	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
python3 .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.IO.Compression.Gzip*'

System.IO.Compression.Gzip.Decompress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Compress(level: Optimal, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Decompress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Compress(level: Fastest, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Compress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Decompress(level: Fastest, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Decompress(level: Optimal, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Compress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Decompress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Gzip.Compress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	ViperUbuntu
Baseline	2b3088b99e97f0d6d3dd24dcd41ab330757f5aa2
Compare	f9016e3507a9669894c500c1cdd4a45aaef24354
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.IO.Compression.Deflate

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Decompress - Duration of single invocation ADX - Test Multi Config Graph	45.64 μs	51.87 μs	1.14	0.01	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	251.08 μs	438.63 μs	1.75	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	936.74 μs	2.09 ms	2.23	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	1.78 ms	2.27 ms	1.28	0.03	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	575.91 μs	1.56 ms	2.71	0.04	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	363.33 μs	460.75 μs	1.27	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	257.88 μs	1.16 ms	4.51	0.05	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	260.70 μs	287.18 μs	1.10	0.01	False
Compress - Duration of single invocation ADX - Test Multi Config Graph	111.34 μs	124.19 μs	1.12	0.30	False
Decompress - Duration of single invocation ADX - Test Multi Config Graph	47.43 μs	57.84 μs	1.22	0.01	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
python3 .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.IO.Compression.Deflate*'

System.IO.Compression.Deflate.Decompress(level: Optimal, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Decompress(level: Optimal, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Compress(level: Fastest, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Compress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Compress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Decompress(level: Fastest, file: "alice29.txt")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Compress(level: Optimal, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Decompress(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Compress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Deflate.Decompress(level: Fastest, file: "sum")

ETL Files

Histogram

JIT Disasms

Docs

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

---

Run Information

Name	Value
Architecture	x64
OS	ubuntu 22.04
Queue	ViperUbuntu
Baseline	2b3088b99e97f0d6d3dd24dcd41ab330757f5aa2
Compare	f9016e3507a9669894c500c1cdd4a45aaef24354
Diff	Diff
Configs	CompilationMode:tiered, RunKind:micro

Regressions in System.IO.Compression.Brotli

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio
Compress_WithoutState - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	478.41 ms	513.99 ms	1.07	0.01	False
Compress_WithState - Duration of single invocation 📝 - Benchmark Source ADX - Test Multi Config Graph	478.30 ms	512.86 ms	1.07	0.01	False

Test Report

Repro

General Docs link: https://github.com/dotnet/performance/blob/main/docs/benchmarking-workflow-dotnet-runtime.md

git clone https://github.com/dotnet/performance.git
python3 .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.IO.Compression.Brotli*'

System.IO.Compression.Brotli.Compress_WithoutState(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

System.IO.Compression.Brotli.Compress_WithState(level: Optimal, file: "TestDocument.pdf")

ETL Files

Histogram

JIT Disasms

Docs

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

[dotnet-policy-service]

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

[LoopedBard3]

Other related regressions:
Linux/x64: dotnet/perf-autofiling-issues#38694

[LoopedBard3]

Seems to be caused by: #104904. @carlossanlop. We are only seeing this regression on Linux.

[carlossanlop]

I don't know what changed. I also saw the results for arm64 Linux, they're really bad and conflict with what we got in this previous report with a ton of improvements: dotnet/perf-autofiling-issues#38280

cc @jkotas @jkoritzinsky

[LoopedBard3]

I don't know what changed. I also saw the results for arm64 Linux, they're really bad and conflict with what we got in this previous report with a ton of improvements: dotnet/perf-autofiling-issues#38280

cc @jkotas @jkoritzinsky

The diff for the set of these regressions, which seems to be the same time as the arm64 ones, is: 0f2e63f...bbabbc9. The commit that jumps out the most is #104904 where it seems something is causing the behavior to match pre-zlib changes behavior.

[AndyAyersMS]

If you look at one of these charts you can see two improvements and two regressions back to old levels. The first regression was when the initial zlib change was reverted. The second reversion seems like an unexpected outcome from #104904

[carlossanlop]

Yep, thanks. We confirmed it and we are working on a fix: #105352

[ericstj]

@carlossanlop can this be resolved now that #105352 is merged?

[carlossanlop]

I think so, yes. If the issue is still happening, we will get a new issue opened by the perf lab.