Implement Vector512<T> API surface for Vector512 hardware acceleration

[anthonycanino]

The purpose of this task is to implemented hardware accelerated lowerings for the Vector<T> API surface for Vector512. In particular, this would allow Vector512.IsHardwareAccelerated to return true on compatible hardware. It would also allow Vector<T> to select Vector512 for its underlying implementation.

[ghost]

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

Issue Details

---

The purpose of this task is to implemented hardware accelerated lowerings for the Vector<T> API surface for Vector512. In particular, this would allow Vector512.IsHardwareAccelerated to return true on compatible hardware. It would also allow Vector<T> to select Vector512 for its underlying implementation.

Author:	anthonycanino
Assignees:	-
Labels:	area-System.Numerics, untriaged
Milestone:	-

[DeepakRajendrakumaran]

@tannergooding @anthonycanino @DeepakRajendrakumaran

Adding list of API properties/methods/operators that needs to be supported. The plan is to update this as we progress through this task

1. Put Name to task if working or planning to work on it
2. Mark as done once done

ASSIGNED

Intel

• Load(), LoadAligned(), LoadAlignedNonTemporal(T), LoadUnsafe()
• Store(), StoreAligned(), StoreAlignedNonTemporal(), StoreUnsafe()
• AllBitsSet, One, Zero
• Create(), CreateScalar(), CreateScalarUnsafe()
• As(), AsByte(), AsDouble(), AsInt16(), AsInt32(), AsInt64(), AsNInt(), AsNUInt(), AsByte(), AsSingle(), AsUInt16(), AsUInt32(), AsUInt64(), AsVector(), AsVector512()
• GetLower()
• Comparison - GreaterThan(), GreaterThanAll(), GreaterThanAny(), GreaterThanOrEqual(), GreaterThanOrEqualAll(), GreaterThanOrEqualAny(), LessThan(), LessThanAll(), LessThanAny(), LessThanOrEqual(), LessThanOrEqualAll(), LessThanOrEqualAny(), EqualsAny()
• ConditionalSelect()
• Dot(), Sum()
• Shuffle()
• ExtractMostSignificantBits()
• Equality() operator ==
• Inequality() operator !=
• Equals(), EqualsAll()
• ConvertToDouble(), ConvertToInt64(), ConvertToUInt32(), ConvertToUInt64()
• ConvertToInt32(), ConvertToSingle()
• Add(), Subtract()
• Divide(), Multiply()
• Abs(), Negate(), Sqrt(),
• Ceiling(), Floor()
• ShiftLeft(), ShiftRightArithmetic(), ShiftRightLogical()
• Max(), Min()
• operator + (binary), operator - (binary)
• operator /, operator *
• operator <<, operator >>, operator >>>
• operator + (unary), operator - (unary)

Microsoft

• AndNot()
• BitwiseAnd(), BitwiseOr()
• get_Count
• get_IsHardwareAccelerated
• get_IsSupported
• GetElement(), ToScalar()
• GetUpper()
• Narrow()
• OnesComplement()
• operator &, operator |, operator ^, operator ~
• Widen()
• WithElement()
• WidenLower(), WidenUpper()
• WithLower(), WithUpper()
• Xor()

N/A

• CopyTo(), TryCopyTo()
• get_Item[int]
• GetHashCode()
• ToString()

[tannergooding]

Going to close this as we now have functional parity with Vector128/256.

There are a couple of optimizations opportunities that are still available, however, and I'll open explicit issues to track those.