IDE0057: Slice can be simplified does not produce the same code

[msedi]

Currently the IDE suggest with Span not to use slice but instead use the Range indexer (IDE0057: Slice can be simplified to this). So the original code:

new Vector<T>(B.Slice(Idx))

is then translated to

new Vector<T>(B[Idx..])

which would be OK for me but the resulting generated code is different:

    public void WithSlice() {
        ReadOnlySpan<float> A = new float[100];
        
        for (int i=0; i<A.Length; i+= Vector<float>.Count)
            new Vector<float>(A.Slice(i));   
    }
    
       public void WithRange() {
        ReadOnlySpan<float> A = new float[100];
        
        for (int i=0; i<A.Length; i+= Vector<float>.Count)
            new Vector<float>(A[i..]);
    }

and the result taken from sharplab:

    public void WithSlice()
    {
        ReadOnlySpan<float> readOnlySpan = new float[100];
        for (int i = 0; i < readOnlySpan.Length; i += Vector<float>.Count)
        {
            new Vector<float>(readOnlySpan.Slice(i));
        }
    }

    public void WithRange()
    {
        ReadOnlySpan<float> readOnlySpan = new float[100];
        for (int i = 0; i < readOnlySpan.Length; i += Vector<float>.Count)
        {
            ReadOnlySpan<float> readOnlySpan2 = readOnlySpan;
            int length = readOnlySpan2.Length;
            int num = i;
            int length2 = length - num;
            new Vector<float>(readOnlySpan2.Slice(num, length2));
        }
    }

Even the resulting JIT Asm seems to be different. I didn't make any performance comparisons, but for my environment it is crucial that the suggested rewrite does not alter the performance of the code.

[YairHalberstadt]

@msedi

The generated code is expected to be different, only the asm would be the same. This would be an issue for dotnet/runtime.

[msedi]

@YairHalberstadt: Thanks for the info. I wasn't so clear in my text about this. I have put it into dotnet/runtime and will close this one.