Benchmarks for #113857 (davidwrighton)

[EgorBot]

Processing dotnet/runtime#113857 (comment) command:

Command

-amd -arm -windows_intel

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using System.Numerics;

BenchmarkSwitcher.FromAssembly(typeof(Program).Assembly).Run(args);

public class Tests
{
    private object _lockObj = new object();
    private int _field = 0;

    [Benchmark(Baseline = true)]
    public void Lock_OnLockObj()
    {
        _field++;
        lock(_lockObj)
        {
            _field++;
        }
    }

    [Benchmark]
    public void LockOnThis_ThisKnownToBeNonNull()
    {
        _field++;
        lock(this)
        {
            _field++;
        }
    }

    [Benchmark]
    public void LockOnThis()
    {
        lock(this)
        {
            _field++;
        }
    }
}

(EgorBot will reply in this issue)

[EgorBot]

❌ Failed on AzureGenoa: Job failed, see logs.

cc @davidwrighton (logs). EgorBot manual: link.

[EgorBot]

❌ Failed on AzureCobalt100: Job failed, see logs.

cc @davidwrighton (logs). EgorBot manual: link.

[EgorBo]

@davidwrighton it's a random build error that I don't understand why it happens sometimes:

EXEC : error : Unable to load shared library 'jitinterface_x64' or one of its dependencies. In order to help diagnose loading problems, consider using a tool like strace. If you're using glibc, consider setting the LD_DEBUG environment variable: [/home/egorbot/runtime/src/coreclr/crossgen-corelib.proj]
  /home/egorbot/runtime/artifacts/bin/coreclr/linux.x64.Release/x64/crossgen2/jitinterface_x64.so: cannot open shared object file: No such file or directory
  /home/egorbot/runtime/artifacts/bin/coreclr/linux.x64.Release/x64/crossgen2/libjitinterface_x64.so: cannot open shared object file: No such file or directory
  /home/egorbot/runtime/artifacts/bin/coreclr/linux.x64.Release/x64/crossgen2/jitinterface_x64: cannot open shared object file: No such file or directory
  /home/egorbot/runtime/artifacts/bin/coreclr/linux.x64.Release/x64/crossgen2/libjitinterface_x64: cannot open shared object file: No such file or directory

let me debug it again.. must be some LIBC mismatch

[EgorBo]

Windows job should finish just fine

[EgorBo]

it's dotnet/runtime#111731

[EgorBot]

Benchmark results on windows_azure_cascadelake

BenchmarkDotNet v0.14.0, Windows 11 (10.0.26100.2314) (Hyper-V)
azure_cascadelake
  Job-YOAKDR : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  Job-VPJOMB : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI

Method	Job	Toolchain	Mean	Error	Ratio
Lock_OnLockObj	Job-YOAKDR	\core_root_base\corerun.exe	19.27 ns	0.020 ns	1.00
LockOnThis_ThisKnownToBeNonNull	Job-YOAKDR	\core_root_base\corerun.exe	19.58 ns	0.016 ns	1.02
LockOnThis	Job-YOAKDR	\core_root_base\corerun.exe	19.37 ns	0.032 ns	1.01
Lock_OnLockObj	Job-VPJOMB	\core_root_diff\corerun.exe	19.25 ns	0.017 ns	1.00
LockOnThis_ThisKnownToBeNonNull	Job-VPJOMB	\core_root_diff\corerun.exe	19.60 ns	0.045 ns	1.02
LockOnThis	Job-VPJOMB	\core_root_diff\corerun.exe	19.33 ns	0.030 ns	1.00

BDN_Artifacts.zip

cc @davidwrighton (agent_logs.txt). EgorBot manual: link.

[EgorBo]

Ok, seems that I've fixed the problem with a workaround

[EgorBo]

@EgorBot -amd -arm -pr 113857

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using System.Numerics;

BenchmarkSwitcher.FromAssembly(typeof(Program).Assembly).Run(args);

public class Tests
{
    private object _lockObj = new object();
    private int _field = 0;

    [Benchmark(Baseline = true)]
    public void Lock_OnLockObj()
    {
        _field++;
        lock(_lockObj)
        {
            _field++;
        }
    }

    [Benchmark]
    public void LockOnThis_ThisKnownToBeNonNull()
    {
        _field++;
        lock(this)
        {
            _field++;
        }
    }

    [Benchmark]
    public void LockOnThis()
    {
        lock(this)
        {
            _field++;
        }
    }
}

[davidwrighton]

@EgorBo thank you. I was really hoping not to burden you with more work.

[EgorBot]

Benchmark results on linux_azure_cobalt100

BenchmarkDotNet v0.14.0, Ubuntu 24.04.2 LTS (Noble Numbat)
azure_cobalt100
  Job-QKYKGR : .NET 10.0.0 (42.42.42.42424), Arm64 RyuJIT AdvSIMD
  Job-VDGVUF : .NET 10.0.0 (42.42.42.42424), Arm64 RyuJIT AdvSIMD

Method	Toolchain	Mean	Error	Ratio
Lock_OnLockObj	Main	20.05 ns	0.002 ns	1.00
LockOnThis_ThisKnownToBeNonNull	Main	20.18 ns	0.002 ns	1.01
LockOnThis	Main	20.41 ns	0.009 ns	1.02
Lock_OnLockObj	PR	19.91 ns	0.004 ns	0.99
LockOnThis_ThisKnownToBeNonNull	PR	20.38 ns	0.003 ns	1.02
LockOnThis	PR	20.45 ns	0.006 ns	1.02

BDN_Artifacts.zip

cc @EgorBo (agent_logs.txt). EgorBot manual: link.

[EgorBot]

Benchmark results on linux_azure_genoa

BenchmarkDotNet v0.14.0, Ubuntu 24.04.2 LTS (Noble Numbat)
AMD EPYC 9V74, 1 CPU, 16 logical and 8 physical cores
  Job-YNEKDB : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  Job-WYCSJD : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI

Method	Toolchain	Mean	Error	Ratio
Lock_OnLockObj	Main	8.494 ns	0.0092 ns	1.00
LockOnThis_ThisKnownToBeNonNull	Main	8.159 ns	0.0073 ns	0.96
LockOnThis	Main	8.173 ns	0.0012 ns	0.96
Lock_OnLockObj	PR	8.283 ns	0.0183 ns	0.98
LockOnThis_ThisKnownToBeNonNull	PR	8.177 ns	0.0052 ns	0.96
LockOnThis	PR	8.174 ns	0.0008 ns	0.96

BDN_Artifacts.zip

cc @EgorBo (agent_logs.txt). EgorBot manual: link.

[EgorBo]

@EgorBo thank you. I was really hoping not to burden you with more work.

No problem! I wanted to fix this issue for quite some time

[am11]

Do we want to run comparisons with https://github.com/dotnet/runtime/commits/ac13edc2df258639b106edfe88147520d0ff3247, one commit in main before dotnet/runtime#113242 was merged?

or only dotnet/runtime#113857 vs. main?

[EgorBo]

Do we want to run comparisons with https://github.com/dotnet/runtime/commits/ac13edc2df258639b106edfe88147520d0ff3247, one commit in main before dotnet/runtime#113242 was merged?

or only dotnet/runtime#113857 vs. main?

The bot compares PR vs current tip of Main

[am11]

Yup, but I was asking if David wants to compare the previous PR because this one is just a follow up and previous PR had few commits before the merge which weren't benchmarked.