Perf regression on StringBuilder.Append

[Symbai]

There is a small performance regression comparing 2.2 vs 3.0 (10%), but a bigger regression comparing .NET Framework vs 3.0 (18%).

BenchmarkDotNet=v0.11.5, OS=Windows 10.0.18362
Intel Core i7-4960X CPU 3.60GHz (Haswell), 1 CPU, 12 logical and 6 physical cores
.NET Core SDK=3.0.100-preview8-013015
  [Host]     : .NET Core 3.0.0-preview8-27911-03 (CoreCLR 4.700.19.36002, CoreFX 4.700.19.36101), 64bit RyuJIT
  Job-EGUGMC : .NET Core 2.2.3 (CoreCLR 4.6.27414.05, CoreFX 4.6.27414.05), 64bit RyuJIT
  Clr        : .NET Framework 4.7.2 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.8.3815.0
  Core       : .NET Core 3.0.0-preview8-27911-03 (CoreCLR 4.700.19.36002, CoreFX 4.700.19.36101), 64bit RyuJIT

Method	Runtime	Toolchain	Mean	Error	StdDev
Small	Core	.NET Core 2.2	179.1 ns	0.7665 ns	0.5984 ns
Medium	Core	.NET Core 2.2	8,932.6 ns	25.0438 ns	22.2007 ns
Large	Core	.NET Core 2.2	876,877.0 ns	4,255.3692 ns	3,553.4260 ns
Small	Clr	Default	171.3 ns	0.3084 ns	0.2734 ns
Medium	Clr	Default	8,121.2 ns	17.6129 ns	14.7076 ns
Large	Clr	Default	805,701.2 ns	2,628.9582 ns	2,459.1291 ns
Small	Core	Default	188.3 ns	1.1118 ns	0.9284 ns
Medium	Core	Default	10,112.1 ns	40.8397 ns	38.2015 ns
Large	Core	Default	982,194.7 ns	3,529.5412 ns	3,128.8463 ns

[ClrJob]
[CoreJob]
[RPlotExporter]
public  class ObjectTest
{
    const string testString = "blablabla";
      
    [Benchmark]
    public bool Small()
    {
        var sb = new StringBuilder();
        for (int i = 0; i < 10; i++)
        {
            sb.Append(testString);
        }
        return sb.Length > 0;
    }
    [Benchmark]
    public bool Medium()
    {
        var sb = new StringBuilder();
        for (int i = 0; i < 1000; i++)
        {
            sb.Append(testString);
        }
        return sb.Length > 0;
    }
    [Benchmark]
    public bool Large()
    {
        var sb = new StringBuilder();
        for (int i = 0; i < 100000; i++)
        {
            sb.Append(testString);
        }
        return sb.Length > 0;
    }
}

[danmoseley]

@adamsitnik in case he's interested

[danmoseley]

cc @billwert

[danmoseley]

This seems well covered by existing benchmark
https://github.com/dotnet/performance/blob/6b31a4513a9ab70a45b86ee188c1e7275c820644/src/benchmarks/micro/corefx/System.Runtime/Perf.StringBuilder.cs#L29-L39

[adamsitnik]

The source of the problem has been identified and now it's being tracked in dotnet/coreclr#27511 so I am closing this issue.

Moreover, I've also improved the performance of Append(char) so when dotnet/coreclr#27511 is implemented the .NET Core implementation is going to be actually faster (not just on par)