Regressions in System.Threading.Tests.Perf_Lock

[performanceautofiler]

Run Information

Architecture	arm64
OS	Windows 10.0.19041
Baseline	3f51e1449f78041bd4f213602f2bbfa4af321dde
Compare	4ce23f46ed20683a1b6649e9a43e09459007c9d8
Diff	Diff

Regressions in System.Threading.Tests.Perf_Lock

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
ReaderWriterLockSlimPerf - Duration of single invocation	27.52 ns	31.28 ns	1.14	0.29	False

Test Report

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Tests.Perf_Lock*'

Payloads

Baseline
Compare

Histogram

System.Threading.Tests.Perf_Lock.ReaderWriterLockSlimPerf

Description of detection logic

IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsRegressionBase: Marked as regression because the compare was 5% greater than the baseline, and the value was not too small.
IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsRegressionWindowed: Marked as regression because 31.277895484004436 > 28.89098714960821.
IsChangePoint: Marked as a change because one of 12/2/2022 9:06:56 PM, 12/8/2022 3:49:24 AM falls between 11/29/2022 2:55:01 PM and 12/8/2022 3:49:24 AM.
IsRegressionStdDev: Marked as regression because -3.8699101464857404 (T) = (0 -32.26726895312294) / Math.Sqrt((5.026255417921273 / (22)) + (11.34885565459455 / (13))) is less than -2.034515297446192 = MathNet.Numerics.Distributions.StudentT.InvCDF(0, 1, (22) + (13) - 2, .025) and -0.14399453580674726 = (28.205789401229957 - 32.26726895312294) / 28.205789401229957 is less than -0.05.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsChangeEdgeDetector: Marked not as a regression because Edge Detector said so.

Docs

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

[dotnet-issue-labeler]

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.

[cincuranet]

Probably caused by #79096. Is that expected @stephentoub?

[ghost]

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

Issue Details

---

Run Information

Architecture	arm64
OS	Windows 10.0.19041
Baseline	3f51e1449f78041bd4f213602f2bbfa4af321dde
Compare	4ce23f46ed20683a1b6649e9a43e09459007c9d8
Diff	Diff

Regressions in System.Threading.Tests.Perf_Lock

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
ReaderWriterLockSlimPerf - Duration of single invocation	27.52 ns	31.28 ns	1.14	0.29	False

Test Report

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Tests.Perf_Lock*'

Payloads

Baseline
Compare

Histogram

System.Threading.Tests.Perf_Lock.ReaderWriterLockSlimPerf

Description of detection logic

IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsRegressionBase: Marked as regression because the compare was 5% greater than the baseline, and the value was not too small.
IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsRegressionWindowed: Marked as regression because 31.277895484004436 > 28.89098714960821.
IsChangePoint: Marked as a change because one of 12/2/2022 9:06:56 PM, 12/8/2022 3:49:24 AM falls between 11/29/2022 2:55:01 PM and 12/8/2022 3:49:24 AM.
IsRegressionStdDev: Marked as regression because -3.8699101464857404 (T) = (0 -32.26726895312294) / Math.Sqrt((5.026255417921273 / (22)) + (11.34885565459455 / (13))) is less than -2.034515297446192 = MathNet.Numerics.Distributions.StudentT.InvCDF(0, 1, (22) + (13) - 2, .025) and -0.14399453580674726 = (28.205789401229957 - 32.26726895312294) / 28.205789401229957 is less than -0.05.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsChangeEdgeDetector: Marked not as a regression because Edge Detector said so.

Docs

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

Author:	performanceautofiler[bot]
Assignees:	EgorBo
Labels:	area-System.Threading.Tasks, tenet-performance, tenet-performance-benchmarks, untriaged
Milestone:	-

[stephentoub]

@EgorBo, in this example:
SharpLab
is it expected that in the manually-inlined case we get:

L0004: cmp byte ptr [rcx+8], 0

but in the JIT-inlined case we get:

L0004: movzx eax, byte ptr [rcx+8]
L0008: test eax, eax

?

using System;

public class C
{
    private bool _disposed;
    
    public void M1()
    {
        if (_disposed)
        {
            Throw();
        }
    }
    
    public void M2()
    {
        ThrowIf(_disposed);
    }
    
    private static void ThrowIf(bool condition)
    {
        if (condition)
        {
            Throw();
        }
    }
    
    private static void Throw() => throw new ObjectDisposedException(null);
}

C.M1()
    L0000: sub rsp, 0x28
    L0004: cmp byte ptr [rcx+8], 0
    L0008: jne short L000f
    L000a: add rsp, 0x28
    L000e: ret
    L000f: call 0x00007fff66550078
    L0014: int3

C.M2()
    L0000: sub rsp, 0x28
    L0004: movzx eax, byte ptr [rcx+8]
    L0008: test eax, eax
    L000a: jne short L0011
    L000c: add rsp, 0x28
    L0010: ret
    L0011: call 0x00007fff66550078
    L0016: int3

[ghost]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Run Information

Architecture	arm64
OS	Windows 10.0.19041
Baseline	3f51e1449f78041bd4f213602f2bbfa4af321dde
Compare	4ce23f46ed20683a1b6649e9a43e09459007c9d8
Diff	Diff

Regressions in System.Threading.Tests.Perf_Lock

Benchmark	Baseline	Test	Test/Base	Test Quality	Edge Detector	Baseline IR	Compare IR	IR Ratio	Baseline ETL	Compare ETL
ReaderWriterLockSlimPerf - Duration of single invocation	27.52 ns	31.28 ns	1.14	0.29	False

Test Report

Repro

git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net8.0 --filter 'System.Threading.Tests.Perf_Lock*'

Payloads

Baseline
Compare

Histogram

System.Threading.Tests.Perf_Lock.ReaderWriterLockSlimPerf

Description of detection logic

IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsRegressionBase: Marked as regression because the compare was 5% greater than the baseline, and the value was not too small.
IsRegressionChecked: Marked as regression because the three check build points were 0.05 greater than the baseline.
IsRegressionWindowed: Marked as regression because 31.277895484004436 > 28.89098714960821.
IsChangePoint: Marked as a change because one of 12/2/2022 9:06:56 PM, 12/8/2022 3:49:24 AM falls between 11/29/2022 2:55:01 PM and 12/8/2022 3:49:24 AM.
IsRegressionStdDev: Marked as regression because -3.8699101464857404 (T) = (0 -32.26726895312294) / Math.Sqrt((5.026255417921273 / (22)) + (11.34885565459455 / (13))) is less than -2.034515297446192 = MathNet.Numerics.Distributions.StudentT.InvCDF(0, 1, (22) + (13) - 2, .025) and -0.14399453580674726 = (28.205789401229957 - 32.26726895312294) / 28.205789401229957 is less than -0.05.
IsImprovementBase: Marked as not an improvement because the compare was not 5% less than the baseline, or the value was too small.
IsChangeEdgeDetector: Marked not as a regression because Edge Detector said so.

Docs

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

Author:	performanceautofiler[bot]
Assignees:	EgorBo
Labels:	tenet-performance, tenet-performance-benchmarks, area-CodeGen-coreclr, untriaged
Milestone:	-

[EgorBo]

It seems like a missing oportunity for Forward Sub:

STMT00004
               [000010] -A-XG------                         *  ASG       bool  
               [000009] D------N---                         +--*  LCL_VAR   bool   V02 tmp1         
               [000001] ---XG------                         \--*  FIELD     bool   C:_disposed
               [000000] -----------                            \--*  LCL_VAR   ref    V00 this

STMT00002
               [000007] -----------                         *  JTRUE     void  
               [000006] -----------                         \--*  EQ        int   
               [000004] -----------                            +--*  LCL_VAR   int    V02 tmp1
               [000005] -----------                            \--*  CNS_INT   int    0

cc @AndyAyersMS @jakobbotsch

can we substitute bool to int?

[jakobbotsch]

Currently there is an early out for it:

runtime/src/coreclr/jit/forwardsub.cpp

Lines 675 to 685 in 8d5f520

	// Next statement seems suitable.
	// See if we can forward sub without changing semantics.
	//
	// Bail if types disagree.
	// Might be able to tolerate these by retyping.
	//
	if (fsv.GetNode()->TypeGet() != fwdSubNode->TypeGet())
	{
	JITDUMP(" mismatched types (substitution)\n");
	return false;
	}

I'm not sure what issues were hit without the check, I think we can allow it by potentially inserting a cast (and there is already some logic to handle that, though only for normalize-on-store -- may need to be handled for normalize-on-load too).

[jakobbotsch]

I opened #83969 for it.

[EgorBo]

Dynamic PGO "fixed" it