Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

AdvSimd support for System.Text.Unicode.Utf8Utility.TranscodeToUtf8 #39041

Merged
merged 11 commits into from
Jul 20, 2020
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@
using System.Numerics;
using System.Runtime.CompilerServices;
using System.Runtime.Intrinsics;
using System.Runtime.Intrinsics.Arm;
using System.Runtime.Intrinsics.X86;

#if SYSTEM_PRIVATE_CORELIB
Expand Down Expand Up @@ -940,10 +941,8 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt
uint inputCharsRemaining = (uint)(pFinalPosWhereCanReadDWordFromInputBuffer - pInputBuffer) + 2;
uint minElementsRemaining = (uint)Math.Min(inputCharsRemaining, outputBytesRemaining);

if (Sse41.X64.IsSupported)
if ((AdvSimd.IsSupported || Sse41.X64.IsSupported) && BitConverter.IsLittleEndian)
{
Debug.Assert(BitConverter.IsLittleEndian, "SSE41 requires little-endian.");

// Try reading and writing 8 elements per iteration.
uint maxIters = minElementsRemaining / 8;
ulong possibleNonAsciiQWord;
Expand All @@ -952,14 +951,28 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt
for (i = 0; (uint)i < maxIters; i++)
{
utf16Data = Unsafe.ReadUnaligned<Vector128<short>>(pInputBuffer);
if (!Sse41.TestZ(utf16Data, nonAsciiUtf16DataMask))

if (AdvSimd.IsSupported)
{
goto LoopTerminatedDueToNonAsciiDataInVectorLocal;
}
if (AdvSimd.CompareTest(utf16Data, nonAsciiUtf16DataMask).ToScalar() == 0)
{
goto LoopTerminatedDueToNonAsciiDataInVectorLocal;
}

// narrow and write
Vector64<sbyte> lower = AdvSimd.ExtractNarrowingSaturateLower(utf16Data);
Vector128<sbyte> source = AdvSimd.ExtractNarrowingSaturateUpper(lower, AdvSimd.LoadVector128((short*)pInputBuffer));
AdvSimd.Store((ulong*)pOutputBuffer, source.AsUInt64());
}
else
{
if (!Sse41.TestZ(utf16Data, nonAsciiUtf16DataMask))
{
goto LoopTerminatedDueToNonAsciiDataInVectorLocal;
}

Sse2.StoreScalar((ulong*)pOutputBuffer /* unaligned */, Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt64());
// narrow and write
Sse2.StoreScalar((ulong*)pOutputBuffer /* unaligned */, Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt64());
}

pInputBuffer += 8;
pOutputBuffer += 8;
Expand All @@ -978,7 +991,17 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt
}

utf16Data = Vector128.CreateScalarUnsafe(possibleNonAsciiQWord).AsInt16();
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));

if (AdvSimd.IsSupported)
{
Vector64<byte> lower = AdvSimd.ExtractNarrowingSaturateUnsignedLower(utf16Data);
Vector128<byte> source = AdvSimd.ExtractNarrowingSaturateUnsignedUpper(lower, AdvSimd.LoadVector128((short*)pInputBuffer + Vector128<short>.Count));
AdvSimd.StoreSelectedScalar((uint*)pOutputBuffer, source.AsUInt32(), 0);
}
else
{
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));
}

pInputBuffer += 4;
pOutputBuffer += 4;
Expand All @@ -1000,7 +1023,16 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt

if (Utf16Utility.AllCharsInUInt64AreAscii(possibleNonAsciiQWord)) // all chars in first QWORD are ASCII
{
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));
if (AdvSimd.IsSupported)
{
Vector64<byte> lower = AdvSimd.ExtractNarrowingSaturateUnsignedLower(utf16Data);
Vector128<byte> source = AdvSimd.ExtractNarrowingSaturateUnsignedUpper(lower, AdvSimd.LoadVector128((short*)pOutputBuffer + Vector128<short>.Count));
Unsafe.WriteUnaligned<uint>(pOutputBuffer, source.AsUInt32().ToScalar());
}
else
{
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));
}
pInputBuffer += 4;
pOutputBuffer += 4;
outputBytesRemaining -= 4;
Expand Down