Optimize System.Text.Unicode using arm64 intrinsics

[BruceForstall]

This item tracks the conversion of the System.Text.Unicode class to use arm64 intrinsics.

Related: #33308

[BruceForstall]

I'm moving this to Future, as there is no current plan to do the work for .NET 5.

[carlossanlop]

@BruceForstall @tannergooding below are the 3 APIs to fix for this area, but I realized that only the first API can be found in Utf16Utility. The other two exist in Utf8Utility instead (there's no Utf16 equivalent):

• System.Text.Unicode.Utf16Utility.GetPointerToFirstInvalidChar() -> Utf16Utility.Validation.cs#L45
• System.Text.Unicode.Utf16Utility.TranscodeToUtf8() -> Utf8Utility.Transcoding.cs#L842
• System.Text.Unicode.Utf16Utility.GetPointerToFirstInvalidByte() -> Utf8Utility.Validation.cs#L26

Was that a typo? Should I work on the Utf8 ones instead?

[danmoseley]

cc @GrabYourPitchforks

[BruceForstall]

It's certainly likely to be a typo. I created the initial list using, essentially, findstr/grep with a little bit of intelligence thrown in. I easily could have made a mistake.

Bottom-line: anything in System.Text.Unicode that has been intrinsics optimized for Intel should be considered tracked by this issue for optimizing for arm64. Feel free to update the list of APIs listed in #33308 to be more correct.

[carlossanlop]

Thanks Bruce. I'll adjust the signatures in the original issue.