JIT: test case showing inexact type for this in constructors

[AndyAyersMS]

Add test case showing that in constructors the exact type of this is not generally known. To devirtualize we need to rely on type propagation from the newobj.

Related discussion in #16198.

[AndyAyersMS]

@briansull PTAL
cc @dotnet/jit-contrib

[Edit: results below are for a now-abandoned invalid opt, so please disregard]

jit-diffs results:

Total bytes of diff: -101 (0.00% of base)
    diff is an improvement.
Total byte diff includes 0 bytes from reconciling methods
        Base had    0 unique methods,        0 unique bytes
        Diff had    0 unique methods,        0 unique bytes
Top file regressions by size (bytes):
         277 : System.Text.Encoding.CodePages.dasm (0.39% of base)
          10 : System.Private.CoreLib.dasm (0.00% of base)
Top file improvements by size (bytes):
        -102 : Microsoft.CodeAnalysis.VisualBasic.dasm (0.00% of base)
         -76 : Microsoft.CodeAnalysis.CSharp.dasm (0.00% of base)
         -32 : Newtonsoft.Json.dasm (-0.01% of base)
         -27 : System.Private.Xml.dasm (0.00% of base)
         -21 : System.Collections.NonGeneric.dasm (-0.08% of base)
18 total files with size differences (16 improved, 2 regressed), 112 unchanged.
Top method regessions by size (bytes):
          78 : System.Text.Encoding.CodePages.dasm - ISCIIDecoder:.ctor(ref):this
          57 : System.Text.Encoding.CodePages.dasm - ISO2022Encoder:.ctor(ref):this
          40 : System.Text.Encoding.CodePages.dasm - ISCIIEncoder:.ctor(ref):this
          30 : System.Private.CoreLib.dasm - System.Text.ASCIIEncoding:.ctor():this
          28 : System.Text.Encoding.CodePages.dasm - System.Text.EncoderNLS:.ctor(ref):this
Top method improvements by size (bytes):
         -41 : Microsoft.CodeAnalysis.CSharp.dasm - Microsoft.CodeAnalysis.CSharp.Symbols.SourceUserDefinedOperatorSymbolBase:.ctor(int,ref,ref,ref,ref,ref,struct,ref,bool):this
         -31 : Microsoft.CodeAnalysis.VisualBasic.dasm - AnonymousDelegateTemplateSymbol:.ctor(ref,struct):this
         -27 : Microsoft.CodeAnalysis.VisualBasic.dasm - Microsoft.CodeAnalysis.VisualBasic.Symbols.SourceEventSymbol:.ctor(ref,ref,ref,ref,ref):this
         -26 : Microsoft.CodeAnalysis.VisualBasic.dasm - Microsoft.CodeAnalysis.VisualBasic.Symbols.SourceNamedTypeSymbol:.ctor(ref,ref,ref):this
         -18 : System.Private.Xml.dasm - System.Xml.XmlEncodedRawTextWriter:.ctor(ref,ref):this (2 methods)
60 total methods with size differences (44 improved, 16 regressed), 154780 unchanged.

[erozenfeld]

LGTM

[jkotas]

Is the following still going to print "Child" with this optimization?

using System;
using System.Threading;
using System.Threading.Tasks;

class Test {

public Test() {
    Console.WriteLine(ToString());
}

static void Main() {
     new Child();
}

}

class Child : Test {

public override string ToString() {
    return "Child";
}

}

[AndyAyersMS]

Yes, it does.

That particular call was already being devirtualized because of inlining into Main and type propagation from newobj.

In cases where we don't inline the constructor, this change kicks in.

[edited to fix a few typos...]

[AndyAyersMS]

@jkotas thought about it some more and you were right.

We also need to check that the constructor type is final. Relevant example.

using System;
using System.Runtime.CompilerServices;

class Test {

    public override string ToString() {
        return "Test";
    }

    [MethodImpl(MethodImplOptions.NoInlining)]
    public Test() {
        Console.WriteLine(ToString());     // we don't know type of "this" exactly here
    }
    
    static void Main() {
        new Child();
    }
}

class Child : Test {

    [MethodImpl(MethodImplOptions.NoInlining)]
    public Child() { }

    public override string ToString() {
        return "Child";
    }
}

Looks like I should add a test in addition to a fix.

[AndyAyersMS]

And since we already know about final classes the upshot is this now doesn't tell us anything we don't already know.

So I'll update this to just add a test case as we don't seem to have any basic tests that would catch this.

[AndyAyersMS]

Tizen failure unrelated, will ignore.