code.cs

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using System;
using System.Threading.Tasks;

#if DEBUG
var obj = new AsyncMachineryBenchmarks();
// we should have done 100 things (just yields) either way
Console.WriteLine(await obj.Recursive()); // 100
Console.WriteLine(await obj.Loop()); // 100
#else
BenchmarkRunner.Run(typeof(AsyncMachineryBenchmarks).Assembly, args: args);
#endif

[MemoryDiagnoser]
public class AsyncMachineryBenchmarks
{
    [Params(100)]
    public int Count { get; set; } = 100;

    [Benchmark(Baseline = true)]
    public Task<int> Recursive() => RecurseAsync(100);

    static async Task<int> RecurseAsync(int count)
    {
        await Task.Yield();
        return count > 1 ? (1 + await RecurseAsync(count - 1)) : 1;
    }

    [Benchmark]
    public async Task<int> Loop()
    {
        int total = 0, count = Count;
        while (count-- > 0)
        {
            await Task.Yield();
            total++;
        }
        return total;
    }
}

results.txt

note we can ignore the Mean here - that's going to be hugely limited by randomness of `Yield` and the thread-pool performance,
which is fast but not very deterministic; what we're interested in is the allocations, which is the machinery overheads

| Method    | Count | Mean     | Error    | StdDev   | Ratio | RatioSD | Gen0   | Allocated | Alloc Ratio |
|---------- |------ |---------:|---------:|---------:|------:|--------:|-------:|----------:|------------:|
| Recursive | 100   | 32.96 us | 0.312 us | 0.292 us |  1.00 |    0.00 | 0.6104 |   10460 B |        1.00 |
| Loop      | 100   | 41.63 us | 0.429 us | 0.381 us |  1.26 |    0.02 |      - |     171 B |        0.02 |

BenchmarkDotNet v0.13.12, Windows 11 (10.0.26100.1)
AMD Ryzen 9 7900X, 1 CPU, 24 logical and 12 physical cores
.NET SDK 8.0.203
  [Host]     : .NET 8.0.3 (8.0.324.11423), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  DefaultJob : .NET 8.0.3 (8.0.324.11423), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI