Arm64 SVE: Implement missing scatter methods

[a74nh]

The following scatter methods were approved but are not yet implemented
(It's possible we have implemented these and changed the naming. This needs checking.)

Edit: See #115242 (comment) for the full expanded list

public static unsafe void Scatter16BitNarrowing(Vector<int> mask, short* address, Vector<int> indices, Vector<int> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<uint> mask, ushort* address, Vector<int> indices, Vector<uint> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<long> mask, short* address, Vector<long> indices, Vector<long> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<ulong> mask, ushort* address, Vector<long> indices, Vector<ulong> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<int> mask, short* address, Vector<uint> indices, Vector<int> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<uint> mask, ushort* address, Vector<uint> indices, Vector<uint> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<long> mask, short* address, Vector<ulong> indices, Vector<long> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<ulong> mask, ushort* address, Vector<ulong> indices, Vector<ulong> data); // ST1H

public static unsafe void Scatter32BitNarrowing(Vector<long> mask, int* address, Vector<long> indices, Vector<long> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<ulong> mask, uint* address, Vector<long> indices, Vector<ulong> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<long> mask, int* address, Vector<ulong> indices, Vector<long> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<ulong> mask, uint* address, Vector<ulong> indices, Vector<ulong> data); // ST1W

public static unsafe void ScatterWithByteOffsets(Vector<T> mask, T* address, Vector<T2> offsets, Vector<T> data); // ST1W or ST1D

[dotnet-policy-service]

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

[dotnet-policy-service]

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

[dotnet-policy-service]

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

[teo-tsirpanis]

(moving this and other recently opened stories to https://github.com/dotnet/runtime/labels/area-System.Runtime.Intrinsics, issues must have only one area label and intrinsics is more specific than codegen)

[jonathandavies-arm]

I've had a look at what is currently implemented and this is what I've found.

public static unsafe void Scatter16BitNarrowing(Vector<int> mask, short* address, Vector<int> indices, Vector<int> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<uint> mask, ushort* address, Vector<int> indices, Vector<uint> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<long> mask, short* address, Vector<long> indices, Vector<long> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<ulong> mask, ushort* address, Vector<long> indices, Vector<ulong> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<int> mask, short* address, Vector<uint> indices, Vector<int> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<uint> mask, ushort* address, Vector<uint> indices, Vector<uint> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<long> mask, short* address, Vector<ulong> indices, Vector<long> data); // ST1H
public static unsafe void Scatter16BitNarrowing(Vector<ulong> mask, ushort* address, Vector<ulong> indices, Vector<ulong> data); // ST1H

// Existing functions taken from Sve.cs
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<int> mask, short* address, Vector<int> offsets, Vector<int> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<int> mask, short* address, Vector<uint> offsets, Vector<int> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<long> mask, short* address, Vector<long> offsets, Vector<long> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<long> mask, short* address, Vector<ulong> offsets, Vector<long> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<uint> mask, ushort* address, Vector<int> offsets, Vector<uint> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<uint> mask, ushort* address, Vector<uint> offsets, Vector<uint> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<ulong> mask, ushort* address, Vector<long> offsets, Vector<ulong> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter16BitWithByteOffsetsNarrowing(Vector<ulong> mask, ushort* address, Vector<ulong> offsets, Vector<ulong> data) => Scatter16BitWithByteOffsetsNarrowing(mask, address, offsets, data);

The Scatter16BitNarrowing functions are missing. In the reviewed file they were named Scatter16BitWithByteOffsetsNarrowing. I can't implement them with this name as there are 8 existing Scatter16BitWithByteOffsetsNarrowing functions which have identical function signatures (the 3rd arg is called offsets).

public static unsafe void Scatter32BitNarrowing(Vector<long> mask, int* address, Vector<long> indices, Vector<long> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<ulong> mask, uint* address, Vector<long> indices, Vector<ulong> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<long> mask, int* address, Vector<ulong> indices, Vector<long> data); // ST1W
public static unsafe void Scatter32BitNarrowing(Vector<ulong> mask, uint* address, Vector<ulong> indices, Vector<ulong> data); // ST1W

// Existing functions taken from Sve.cs
public static unsafe void Scatter32BitWithByteOffsetsNarrowing(Vector<long> mask, int* address, Vector<long> offsets, Vector<long> data) => Scatter32BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter32BitWithByteOffsetsNarrowing(Vector<long> mask, int* address, Vector<ulong> offsets, Vector<long> data) => Scatter32BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter32BitWithByteOffsetsNarrowing(Vector<ulong> mask, uint* address, Vector<long> offsets, Vector<ulong> data) => Scatter32BitWithByteOffsetsNarrowing(mask, address, offsets, data);
public static unsafe void Scatter32BitWithByteOffsetsNarrowing(Vector<ulong> mask, uint* address, Vector<ulong> offsets, Vector<ulong> data) => Scatter32BitWithByteOffsetsNarrowing(mask, address, offsets, data);

There is the same issue for the 4 Scatter32BitNarrowing functions above which are named Scatter32BitWithByteOffsetsNarrowing in the review file and there exists 4 functions with the same signature.

The ScatterWithByteOffsets functions have all been been implemented.

[kunalspathak]

I think we are talking about https://docsmirror.github.io/A64/2024-03/st1h_z_p_bz.html, the "scaled offset" and "unscaled offset".