Skip to content

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

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.

Sign up for GitHub

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

Merged
merged 11 commits into from
Jul 20, 2020
Merged

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

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
6e26a21
AdvSimd support for System.Text.Unicode.Utf8Utility.TranscodeToUtf8
carlossanlop Jul 9, 2020
f596eaf
Readd using to prevent build failure.
carlossanlop Jul 10, 2020
f52aaf8
Inverted condition
carlossanlop Jul 14, 2020
c967b87
Address IsSupported order, improve use ExtractNarrowingSaturated usage
carlossanlop Jul 15, 2020
328718b
Rename source to result, second argument utf16Data
carlossanlop Jul 15, 2020
0024ad2
Improve CompareTest
carlossanlop Jul 15, 2020
45d03d9
Add shims causing failures in Linux
carlossanlop Jul 15, 2020
191ec1f
Use unsigned version of ExtractNarrowingSaturate, avoid using MinAcro…
carlossanlop Jul 15, 2020
1e5519c
Missing support check for Sse2.X64
carlossanlop Jul 16, 2020
86ef01f
Add missing case for AdvSimd
carlossanlop Jul 16, 2020
101c4f1
Use MinPairwise for short
carlossanlop Jul 18, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
64 changes: 52 additions & 12 deletions src/libraries/System.Private.CoreLib/src/System/Text/Unicode/Utf8Utility.Transcoding.cs
View file
Open in desktop
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 @@ -882,7 +883,7 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt
// is not enabled.

Unsafe.SkipInit(out Vector128<short> nonAsciiUtf16DataMask);
if (Sse41.X64.IsSupported)
if (Sse41.X64.IsSupported || (AdvSimd.Arm64.IsSupported && BitConverter.IsLittleEndian))
{
nonAsciiUtf16DataMask = Vector128.Create(unchecked((short)0xFF80)); // mask of non-ASCII bits in a UTF-16 char
}
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 (Sse41.X64.IsSupported || (AdvSimd.Arm64.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 +951,30 @@ 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;
}
Vector128<short> isUtf16DataNonAscii = AdvSimd.CompareTest(utf16Data, nonAsciiUtf16DataMask);
bool hasNonAsciiDataInVector = AdvSimd.Arm64.MinPairwise(isUtf16DataNonAscii, isUtf16DataNonAscii).AsUInt64().ToScalar() != 0;

// narrow and write
if (hasNonAsciiDataInVector)
{
goto LoopTerminatedDueToNonAsciiDataInVectorLocal;
}

Sse2.StoreScalar((ulong*)pOutputBuffer /* unaligned */, Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt64());
Vector64<byte> lower = AdvSimd.ExtractNarrowingSaturateUnsignedLower(utf16Data);
AdvSimd.Store(pOutputBuffer, lower);
}
else
{
if (!Sse41.TestZ(utf16Data, nonAsciiUtf16DataMask))
{
goto LoopTerminatedDueToNonAsciiDataInVectorLocal;
}

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

pInputBuffer += 8;
pOutputBuffer += 8;
Expand All @@ -978,7 +993,16 @@ 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);
AdvSimd.StoreSelectedScalar((uint*)pOutputBuffer, lower.AsUInt32(), 0);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should comment on this line that (uint*)pOutputBuffer is unaligned.

}
else
{
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));
}

pInputBuffer += 4;
pOutputBuffer += 4;
Expand All @@ -990,7 +1014,15 @@ public static OperationStatus TranscodeToUtf8(char* pInputBuffer, int inputLengt
LoopTerminatedDueToNonAsciiDataInVectorLocal:

outputBytesRemaining -= 8 * i;
possibleNonAsciiQWord = Sse2.X64.ConvertToUInt64(utf16Data.AsUInt64());

if (Sse2.X64.IsSupported)
{
possibleNonAsciiQWord = Sse2.X64.ConvertToUInt64(utf16Data.AsUInt64());
}
else
{
possibleNonAsciiQWord = utf16Data.AsUInt64().ToScalar();
}

// Temporarily set 'possibleNonAsciiQWord' to be the low 64 bits of the vector,
// then check whether it's all-ASCII. If so, narrow and write to the destination
Expand All @@ -1000,7 +1032,15 @@ 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);
AdvSimd.StoreSelectedScalar((uint*)pOutputBuffer, lower.AsUInt32(), 0);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please add a comment saying (uint*)pOutputBuffer is unaligned.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If that's ok, I can add this to a list of TODOs for the next PR we open for intrinsics changes (there are things that the FC may have to address later). The CI is taking a really long time and I would like to get this merged today. I would like to avoid resetting the CI.

}
else
{
Unsafe.WriteUnaligned<uint>(pOutputBuffer, Sse2.ConvertToUInt32(Sse2.PackUnsignedSaturate(utf16Data, utf16Data).AsUInt32()));
}
pInputBuffer += 4;
pOutputBuffer += 4;
outputBytesRemaining -= 4;
Expand Down
1 change: 1 addition & 0 deletions ...ibraries/System.Utf8String.Experimental/src/System/Runtime/Intrinsics/Intrinsics.Shims.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@ internal static class Vector64
public static Vector64<ulong> Create(ulong value) => throw new PlatformNotSupportedException();
public static Vector64<uint> CreateScalar(uint value) => throw new PlatformNotSupportedException();
public static Vector64<byte> AsByte<T>(this Vector64<T> vector) where T : struct => throw new PlatformNotSupportedException();
public static Vector64<uint> AsUInt32<T>(this Vector64<T> vector) where T : struct => throw new PlatformNotSupportedException();
}
internal readonly struct Vector64<T>
where T : struct
Expand Down