MonoInterpreter doesn't correctly handle overshifting of a vector

[tannergooding]

As found in #89339, monointerpreter does not currently correctly mask the shiftAmount and produces an incorrect result when the shift amount is larger than the number of bits available in the underlying T.

The simplest repro is:

Vector128<byte> vector = Vector128.Create(1);
Console.WriteLine(vector << 8);

Where the result should be 1, 1, ..., 1.

The same should be true for Vector128<short> using 16, Vector128<int> using 32, and Vector128<long> using 64.

The shiftAmount in all cases should be functionally vector << (shiftAmount & ((sizeof(T) * 8) - 1))

Tagging subscribers to this area: @dotnet/area-system-numerics
See info in area-owners.md if you want to be subscribed.

Issue Details

---

As found in #89339, monointerpreter does not currently correctly mask the shiftAmount and produces an incorrect result when the shift amount is larger than the number of bits available in the underlying T.

The simplest repro is:

Vector128<byte> vector = Vector128.Create(1);
Console.WriteLine(vector << 8);

Where the result should be 1, 1, ..., 1.

The same should be true for Vector128<short> using 16, Vector128<int> using 32, and Vector128<long> using 64.

The shiftAmount in all cases should be functionally vector << (shiftAmount & ((sizeof(T) * 8) - 1))

Author:	tannergooding
Assignees:	-
Labels:	area-System.Numerics
Milestone:	-

[vargaz]

Can't reproduce this with master on wasm in interpreter mode:

using System;
using System.Runtime.Intrinsics;

namespace HelloWorld
{
    internal class Program
    {
        private static void Main(string[] args)
        {
            Vector128<byte> vector = Vector128.Create((byte)1);
            Console.WriteLine(vector << 8);
        }
    }
}

Results in:

<1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1>

[tannergooding]

Potentially the testcase wasn't complicated enough. The full repro can be seen by removing the [ActiveIssue] here and running the tests: https://github.com/dotnet/runtime/blob/main/src/tests/JIT/HardwareIntrinsics/General/Shared/VectorImmBinaryOperatorTest.template#L22

Tagging subscribers to this area: @BrzVlad, @kotlarmilos
See info in area-owners.md if you want to be subscribed.

Issue Details

---

As found in #89339, monointerpreter does not currently correctly mask the shiftAmount and produces an incorrect result when the shift amount is larger than the number of bits available in the underlying T.

The simplest repro is:

Vector128<byte> vector = Vector128.Create(1);
Console.WriteLine(vector << 8);

Where the result should be 1, 1, ..., 1.

The same should be true for Vector128<short> using 16, Vector128<int> using 32, and Vector128<long> using 64.

The shiftAmount in all cases should be functionally vector << (shiftAmount & ((sizeof(T) * 8) - 1))

Author:	tannergooding
Assignees:	fanyang-mono
Labels:	untriaged, area-Codegen-Interpreter-mono
Milestone:	-

[SamMonoRT]

Possibly will be fixed via #90666