API Proposal: Arm64 Simd BitwiseSelect

[sdmaclea]

@eerhardt @CarolEidt @RussKeldorph

Target Framework netcoreapp2.1

namespace System.Runtime.Intrinsics.Arm.Arm64
{
    /// <summary>
    /// This class provides access to the Arm64 AdvSIMD intrinsics
    ///
    /// Arm64 CPUs indicate support for this feature by setting
    /// ID_AA64PFR0_EL1.AdvSIMD == 0 or better.
    /// </summary>
    public static class Simd
    {
        public static bool IsSupported { get { throw null; } }

        /// <summary>
        /// Vector BitwiseSelect
        /// For each bit in the vector result[bit] = sel[bit] ? left[bit] : right[bit]
        /// Corresponds to vector forms of ARM64 BSL (Also BIF & BIT)
        /// </summary>
        public static Vector64<T>  BitwiseSelect<T>(Vector64<T>  sel, Vector64<T>  left, Vector64<T>  right) where T : struct { throw null; }
        public static Vector128<T> BitwiseSelect<T>(Vector128<T> sel, Vector128<T> left, Vector128<T> right) where T : struct { throw null; }
    }
}

[sdmaclea]

My current understanding is the difference between BSL, BIT, BIF is simply which of the 3 source registers is used as the destination register. Therefore I am using the same intrinsic for all of them. This may give more flexibility to the register allocator.

[sdmaclea]

Updated to use generics

[sdmaclea]

@CarolEidt @eerhardt I think this can be marked "API-ready-for-review"

[sdmaclea]

@eerhardt Any reason why this is marked future? The CoreCLR work is complete and ready for 2.1.

[terrajobst]

Looks good as proposed. Feedback:

• The parameter should be spelled out "select", or whatever the spec calls it.
• If we expand the generics, the first argument should be an int-based type