Why are some x86 intrinsic vector functions missing?

[RamType0]

Where are MOVS(L/H)DUP,and string comparison intrinsic functions?
Why are they missing?

[Dotnet-GitSync-Bot]

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[huoyaoyuan]

I've heard that those instructions are unfriendly for modern CPU architectures (comparing to the time they were introduced), and will have strong negative performance impact.

For their functionality, they should be able to be directly represented by Span methods. Let BCL to choose the most optimal way to do these.

[ghost]

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

[ghost]

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

[benaadams]

Avx.DuplicateEvenIndexed, Avx.DuplicateOddIndexed etc?

runtime/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/X86/Avx.cs

Lines 380 to 395 in ff4f37f

	/// <summary>
	/// __m256 _mm256_moveldup_ps (__m256 a)
	/// VMOVSLDUP ymm, ymm/m256
	/// </summary>
	public static Vector256<float> DuplicateEvenIndexed(Vector256<float> value) => DuplicateEvenIndexed(value);
	/// <summary>
	/// __m256d _mm256_movedup_pd (__m256d a)
	/// VMOVDDUP ymm, ymm/m256
	/// </summary>
	public static Vector256<double> DuplicateEvenIndexed(Vector256<double> value) => DuplicateEvenIndexed(value);
	/// <summary>
	/// __m256 _mm256_movehdup_ps (__m256 a)
	/// VMOVSHDUP ymm, ymm/m256
	/// </summary>
	public static Vector256<float> DuplicateOddIndexed(Vector256<float> value) => DuplicateOddIndexed(value);

Sse3.MoveAndDuplicate, Sse3.MoveHighAndDuplicate, Sse3.MoveLowAndDuplicate

runtime/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/X86/Sse3.cs

Lines 79 to 95 in ff4f37f

	/// <summary>
	/// __m128d _mm_movedup_pd (__m128d a)
	/// MOVDDUP xmm, xmm/m64
	/// </summary>
	public static Vector128<double> MoveAndDuplicate(Vector128<double> source) => MoveAndDuplicate(source);
	/// <summary>
	/// __m128 _mm_movehdup_ps (__m128 a)
	/// MOVSHDUP xmm, xmm/m128
	/// </summary>
	public static Vector128<float> MoveHighAndDuplicate(Vector128<float> source) => MoveHighAndDuplicate(source);
	/// <summary>
	/// __m128 _mm_moveldup_ps (__m128 a)
	/// MOVSLDUP xmm, xmm/m128
	/// </summary>
	public static Vector128<float> MoveLowAndDuplicate(Vector128<float> source) => MoveLowAndDuplicate(source);

[benaadams]

Sse42 only has Crc, no string functions though

[RamType0]

I found SSE3 MOVS(L/H)DUP, but I could not found AVX one,thanks!
But why are their naming is inconsistent?

[tannergooding]

@RamType-0, do you mean why do the .NET names differ from the C/C++ names?

e.g Sse3.MoveAndDuplicate vs _mm_movedup_pd?

[RamType0]

@RamType-0, do you mean why do the .NET names differ from the C/C++ names?

e.g Sse3.MoveAndDuplicate vs _mm_movedup_pd?

No.
I mean Sse3.MoveLowAndDuplicate vs Avx.DuplicateEvenIndexed.

[tannergooding]

Likely just an oversight here.

DuplicateEvenIndexed and DuplicateOddIndexed are the logically better names, while MoveLowAndDuplicate is the description used for the double variant, as it only has two indices for the 128-bit version.

[tannergooding]

Going to close this as resolved. The methods exist, they just have different names.

[RamType0]

Going to close this as resolved. The methods exist, they just have different names.

@tannergooding
Where is string comparison functions?
And also,will we have any changes about inconsistent naming?

[tannergooding]

Where is string comparison functions?

They are tracked by #957 and require someone to propose some better names and API surface to ensure the functionality is exposed.

And also,will we have any changes about inconsistent naming?

At this point, I'd probably say not likely. That would just lead to additional surface area that does the same thing and adds a bit more inconsistency.
If you feel strongly, you should feel free to open an API proposal suggesting we expose them and I can take it to API review: https://github.com/dotnet/runtime/issues/new?assignees=&labels=api-suggestion&template=02_api_proposal.md