Performance of resolving a concrete type with WithConcreteTypeDynamicRegistrations

[appel1]

I'm looking into switching to DryIoc and it looks like performance resolving a concrete type that hasn't been registered using WithConcreteTypeDynamicRegistrations is quite a bit slower that resolving a registered service. It's very possible I've done something stupid in my benchmark? Is there a better way to not have to register everything?

Job=.NET Framework 4.8.1  Runtime=.NET Framework 4.8.1

| Method                               | type     | Mean        | Error     | StdDev     | Median      | Gen0   | Allocated |
|------------------------------------- |--------- |------------:|----------:|-----------:|------------:|-------:|----------:|
| DryIoc_WithConcreteTypeRegistrations | INoDep   |    26.48 ns |  0.140 ns |   0.117 ns |    26.51 ns | 0.0038 |      24 B |
| DryIoc_WithConcreteTypeRegistrations | ICtorDep |    31.85 ns |  0.635 ns |   1.593 ns |    31.29 ns | 0.0076 |      48 B |
| DryIoc_WithConcreteTypeRegistrations | NoDep    | 1,168.51 ns | 23.129 ns |  58.026 ns | 1,137.47 ns | 0.1411 |     899 B |
| DryIoc_WithConcreteTypeRegistrations | CtorDep  | 2,424.66 ns | 48.017 ns | 109.359 ns | 2,371.38 ns | 0.2060 |    1308 B |

    [MemoryDiagnoser]
    [SimpleJob(RuntimeMoniker.Net481)]
    // [SimpleJob(RuntimeMoniker.Net80)]
    [Orderer(SummaryOrderPolicy.FastestToSlowest)]
    public class ContainerBenchmarks
    {
        private IContainer m_dryioc_withConcreteTypeRegistrations;

        [GlobalSetup]
        public void Setup()
        {
            m_dryioc_withConcreteTypeRegistrations = RegisterDryIoc_WithConcreteTypeRegistrations();
        }

        public Container RegisterDryIoc_WithConcreteTypeRegistrations()
        {
            var container = new Container(rules => rules.WithConcreteTypeDynamicRegistrations());
            container.Register<IServiceWithoutDependencies, ServiceWithoutDependencies>();
            container.Register<IDependencyViaProperty, DependencyViaProperty>();
            container.Register<IDependencyViaCtor, DependencyViaCtor>();

            return container;
        }

        public static Value[] ResolveDryIoc_WithConcreteTypeRegistrations_Types() => [
            new(typeof(IServiceWithoutDependencies), "INoDep"),
            new(typeof(ServiceWithoutDependencies),"NoDep"),
            new(typeof(DependencyViaCtor),"CtorDep"),
            new(typeof(IDependencyViaCtor),"ICtorDep"),
        ];

        [Benchmark]
        [ArgumentsSource(nameof(ResolveDryIoc_WithConcreteTypeRegistrations_Types))]
        public object DryIoc_WithConcreteTypeRegistrations(Value type) => m_dryioc_withConcreteTypeRegistrations.Resolve(type.Type);
  }

    public interface IDependencyViaCtor { }

    public class DependencyViaCtorBase([OptionalDependency] IServiceWithoutDependencies? bar = null)
    {
        public IServiceWithoutDependencies? Bar { get; } = bar;
    }

    public class DependencyViaCtor(IServiceWithoutDependencies? bar = null) : DependencyViaCtorBase(bar), IDependencyViaCtor
    {
    }

    public interface IServiceWithoutDependencies
    { }

    public class ServiceWithoutDependencies : IServiceWithoutDependencies
    { 
    }

    public record Value(Type Type, string Name)
    {
        public override string ToString() => Name;
    }

[dadhi]

Hi, @appel1

I have tried your benchmark and got similar results:

BenchmarkDotNet v0.13.11, Windows 11 (10.0.22631.3672/23H2/2023Update/SunValley3)
Intel Core i9-8950HK CPU 2.90GHz (Coffee Lake), 1 CPU, 12 logical and 6 physical cores
.NET SDK 8.0.204
  [Host]   : .NET 8.0.4 (8.0.424.16909), X64 RyuJIT AVX2
  .NET 8.0 : .NET 8.0.4 (8.0.424.16909), X64 RyuJIT AVX2

Job=.NET 8.0  Runtime=.NET 8.0

| Method                               | type     | Mean        | Error     | StdDev     | Gen0   | Allocated |
|------------------------------------- |--------- |------------:|----------:|-----------:|-------:|----------:|
| DryIoc_WithConcreteTypeRegistrations | INoDep   |    24.97 ns |  0.644 ns |   1.889 ns | 0.0038 |      24 B |
| DryIoc_WithConcreteTypeRegistrations | ICtorDep |    29.35 ns |  0.862 ns |   2.542 ns | 0.0076 |      48 B |
| DryIoc_WithConcreteTypeRegistrations | NoDep    |   983.30 ns | 28.969 ns |  85.416 ns | 0.1144 |     728 B |
| DryIoc_WithConcreteTypeRegistrations | CtorDep  | 2,024.36 ns | 54.419 ns | 160.457 ns | 0.2060 |    1296 B |

Sure, the results are slower, because the automatic concrete types resolution is far from the happy path.
But anyway, I did not expect such difference.
I will check the reason for it.