Skip to content

SSE41 version of Hadamard transform #1810

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 7 commits into from
Nov 9, 2021
Merged

SSE41 version of Hadamard transform #1810

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
5b1720e
Add sse41 version of Hadamard transform
brianpopow Nov 7, 2021
d201793
Add HadamardTransform sse tests
brianpopow Nov 7, 2021
925028c
Merge remote-tracking branch 'origin/master' into bp/hadamardtransfor…
brianpopow Nov 7, 2021
b7e23c1
Merge branch 'master' into bp/hadamardtransformsse
brianpopow Nov 8, 2021
6e135cb
Avoid pinning
brianpopow Nov 9, 2021
d6d1868
Test Hadamard transform only with and without HardwareIntrinsics
brianpopow Nov 9, 2021
2e52890
Merge branch 'master' into bp/hadamardtransformsse
brianpopow Nov 9, 2021
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
147 changes: 142 additions & 5 deletions src/ImageSharp/Formats/Webp/Lossy/LossyUtils.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,11 +4,16 @@
using System;
using System.Buffers.Binary;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
#if SUPPORTS_RUNTIME_INTRINSICS
using System.Runtime.Intrinsics;
using System.Runtime.Intrinsics.X86;
#endif

// ReSharper disable InconsistentNaming
namespace SixLabors.ImageSharp.Formats.Webp.Lossy
{
internal static class LossyUtils
internal static unsafe class LossyUtils
{
[MethodImpl(InliningOptions.ShortMethod)]
public static int Vp8Sse16X16(Span<byte> a, Span<byte> b) => GetSse(a, b, 16, 16);
Expand Down Expand Up @@ -61,11 +66,12 @@ public static void Copy(Span<byte> src, Span<byte> dst, int w, int h)
public static int Vp8Disto16X16(Span<byte> a, Span<byte> b, Span<ushort> w, Span<int> scratch)
{
int d = 0;
int dataSize = (4 * WebpConstants.Bps) - 16;
for (int y = 0; y < 16 * WebpConstants.Bps; y += 4 * WebpConstants.Bps)
{
for (int x = 0; x < 16; x += 4)
{
d += Vp8Disto4X4(a.Slice(x + y), b.Slice(x + y), w, scratch);
d += Vp8Disto4X4(a.Slice(x + y, dataSize), b.Slice(x + y, dataSize), w, scratch);
}
}

Expand All @@ -75,9 +81,19 @@ public static int Vp8Disto16X16(Span<byte> a, Span<byte> b, Span<ushort> w, Span
[MethodImpl(InliningOptions.ShortMethod)]
public static int Vp8Disto4X4(Span<byte> a, Span<byte> b, Span<ushort> w, Span<int> scratch)
{
int sum1 = TTransform(a, w, scratch);
int sum2 = TTransform(b, w, scratch);
return Math.Abs(sum2 - sum1) >> 5;
#if SUPPORTS_RUNTIME_INTRINSICS
if (Sse41.IsSupported)
{
int diffSum = TTransformSse41(a, b, w, scratch);
return Math.Abs(diffSum) >> 5;
}
else
#endif
{
int sum1 = TTransform(a, w, scratch);
int sum2 = TTransform(b, w, scratch);
return Math.Abs(sum2 - sum1) >> 5;
}
}

public static void DC16(Span<byte> dst, Span<byte> yuv, int offset)
Expand Down Expand Up @@ -589,6 +605,127 @@ public static int TTransform(Span<byte> input, Span<ushort> w, Span<int> scratch
return sum;
}

#if SUPPORTS_RUNTIME_INTRINSICS
/// <summary>
/// Hadamard transform
/// Returns the weighted sum of the absolute value of transformed coefficients.
/// w[] contains a row-major 4 by 4 symmetric matrix.
/// </summary>
public static int TTransformSse41(Span<byte> inputA, Span<byte> inputB, Span<ushort> w, Span<int> scratch)
{
Span<int> sum = scratch.Slice(0, 4);
sum.Clear();

// Load and combine inputs.
Vector128<byte> ina0 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputA));
Vector128<byte> ina1 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputA.Slice(WebpConstants.Bps, 16)));
Vector128<byte> ina2 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputA.Slice(WebpConstants.Bps * 2, 16)));
Vector128<long> ina3 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputA.Slice(WebpConstants.Bps * 3, 16))).AsInt64();
Vector128<byte> inb0 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputB));
Vector128<byte> inb1 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputB.Slice(WebpConstants.Bps, 16)));
Vector128<byte> inb2 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputB.Slice(WebpConstants.Bps * 2, 16)));
Vector128<long> inb3 = Unsafe.As<byte, Vector128<byte>>(ref MemoryMarshal.GetReference(inputB.Slice(WebpConstants.Bps * 3, 16))).AsInt64();

// Combine inA and inB (we'll do two transforms in parallel).
Vector128<int> inab0 = Sse2.UnpackLow(ina0.AsInt32(), inb0.AsInt32());
Vector128<int> inab1 = Sse2.UnpackLow(ina1.AsInt32(), inb1.AsInt32());
Vector128<int> inab2 = Sse2.UnpackLow(ina2.AsInt32(), inb2.AsInt32());
Vector128<int> inab3 = Sse2.UnpackLow(ina3.AsInt32(), inb3.AsInt32());
Vector128<short> tmp0 = Sse41.ConvertToVector128Int16(inab0.AsByte());
Vector128<short> tmp1 = Sse41.ConvertToVector128Int16(inab1.AsByte());
Vector128<short> tmp2 = Sse41.ConvertToVector128Int16(inab2.AsByte());
Vector128<short> tmp3 = Sse41.ConvertToVector128Int16(inab3.AsByte());

// a00 a01 a02 a03 b00 b01 b02 b03
// a10 a11 a12 a13 b10 b11 b12 b13
// a20 a21 a22 a23 b20 b21 b22 b23
// a30 a31 a32 a33 b30 b31 b32 b33
// Vertical pass first to avoid a transpose (vertical and horizontal passes
// are commutative because w/kWeightY is symmetric) and subsequent transpose.
// Calculate a and b (two 4x4 at once).
Vector128<short> a0 = Sse2.Add(tmp0, tmp2);
Vector128<short> a1 = Sse2.Add(tmp1, tmp3);
Vector128<short> a2 = Sse2.Subtract(tmp1, tmp3);
Vector128<short> a3 = Sse2.Subtract(tmp0, tmp2);
Vector128<short> b0 = Sse2.Add(a0, a1);
Vector128<short> b1 = Sse2.Add(a3, a2);
Vector128<short> b2 = Sse2.Subtract(a3, a2);
Vector128<short> b3 = Sse2.Subtract(a0, a1);

// a00 a01 a02 a03 b00 b01 b02 b03
// a10 a11 a12 a13 b10 b11 b12 b13
// a20 a21 a22 a23 b20 b21 b22 b23
// a30 a31 a32 a33 b30 b31 b32 b33
// Transpose the two 4x4.
Vector128<short> transpose00 = Sse2.UnpackLow(b0, b1);
Vector128<short> transpose01 = Sse2.UnpackLow(b2, b3);
Vector128<short> transpose02 = Sse2.UnpackHigh(b0, b1);
Vector128<short> transpose03 = Sse2.UnpackHigh(b2, b3);

// a00 a10 a01 a11 a02 a12 a03 a13
// a20 a30 a21 a31 a22 a32 a23 a33
// b00 b10 b01 b11 b02 b12 b03 b13
// b20 b30 b21 b31 b22 b32 b23 b33
Vector128<int> transpose10 = Sse2.UnpackLow(transpose00.AsInt32(), transpose01.AsInt32());
Vector128<int> transpose11 = Sse2.UnpackLow(transpose02.AsInt32(), transpose03.AsInt32());
Vector128<int> transpose12 = Sse2.UnpackHigh(transpose00.AsInt32(), transpose01.AsInt32());
Vector128<int> transpose13 = Sse2.UnpackHigh(transpose02.AsInt32(), transpose03.AsInt32());

// a00 a10 a20 a30 a01 a11 a21 a31
// b00 b10 b20 b30 b01 b11 b21 b31
// a02 a12 a22 a32 a03 a13 a23 a33
// b02 b12 a22 b32 b03 b13 b23 b33
Vector128<long> output0 = Sse2.UnpackLow(transpose10.AsInt64(), transpose11.AsInt64());
Vector128<long> output1 = Sse2.UnpackHigh(transpose10.AsInt64(), transpose11.AsInt64());
Vector128<long> output2 = Sse2.UnpackLow(transpose12.AsInt64(), transpose13.AsInt64());
Vector128<long> output3 = Sse2.UnpackHigh(transpose12.AsInt64(), transpose13.AsInt64());

// a00 a10 a20 a30 b00 b10 b20 b30
// a01 a11 a21 a31 b01 b11 b21 b31
// a02 a12 a22 a32 b02 b12 b22 b32
// a03 a13 a23 a33 b03 b13 b23 b33
// Horizontal pass and difference of weighted sums.
Vector128<ushort> w0 = Unsafe.As<ushort, Vector128<ushort>>(ref MemoryMarshal.GetReference(w));
Vector128<ushort> w8 = Unsafe.As<ushort, Vector128<ushort>>(ref MemoryMarshal.GetReference(w.Slice(8, 8)));

// Calculate a and b (two 4x4 at once).
a0 = Sse2.Add(output0.AsInt16(), output2.AsInt16());
a1 = Sse2.Add(output1.AsInt16(), output3.AsInt16());
a2 = Sse2.Subtract(output1.AsInt16(), output3.AsInt16());
a3 = Sse2.Subtract(output0.AsInt16(), output2.AsInt16());
b0 = Sse2.Add(a0, a1);
b1 = Sse2.Add(a3, a2);
b2 = Sse2.Subtract(a3, a2);
b3 = Sse2.Subtract(a0, a1);

// Separate the transforms of inA and inB.
Vector128<long> ab0 = Sse2.UnpackLow(b0.AsInt64(), b1.AsInt64());
Vector128<long> ab2 = Sse2.UnpackLow(b2.AsInt64(), b3.AsInt64());
Vector128<long> bb0 = Sse2.UnpackHigh(b0.AsInt64(), b1.AsInt64());
Vector128<long> bb2 = Sse2.UnpackHigh(b2.AsInt64(), b3.AsInt64());

Vector128<ushort> ab0Abs = Ssse3.Abs(ab0.AsInt16());
Vector128<ushort> ab2Abs = Ssse3.Abs(ab2.AsInt16());
Vector128<ushort> b0Abs = Ssse3.Abs(bb0.AsInt16());
Vector128<ushort> bb2Abs = Ssse3.Abs(bb2.AsInt16());

// weighted sums.
Vector128<int> ab0mulw0 = Sse2.MultiplyAddAdjacent(ab0Abs.AsInt16(), w0.AsInt16());
Vector128<int> ab2mulw8 = Sse2.MultiplyAddAdjacent(ab2Abs.AsInt16(), w8.AsInt16());
Vector128<int> b0mulw0 = Sse2.MultiplyAddAdjacent(b0Abs.AsInt16(), w0.AsInt16());
Vector128<int> bb2mulw8 = Sse2.MultiplyAddAdjacent(bb2Abs.AsInt16(), w8.AsInt16());
Vector128<int> ab0ab2Sum = Sse2.Add(ab0mulw0, ab2mulw8);
Vector128<int> b0w0bb2w8Sum = Sse2.Add(b0mulw0, bb2mulw8);

// difference of weighted sums.
Vector128<int> result = Sse2.Subtract(ab0ab2Sum.AsInt32(), b0w0bb2w8Sum.AsInt32());

ref int outputRef = ref MemoryMarshal.GetReference(sum);
Unsafe.As<int, Vector128<int>>(ref outputRef) = result.AsInt32();
return sum[3] + sum[2] + sum[1] + sum[0];
}
#endif

public static void TransformTwo(Span<short> src, Span<byte> dst, Span<int> scratch)
{
TransformOne(src, dst, scratch);
Expand Down
52 changes: 52 additions & 0 deletions tests/ImageSharp.Tests/Formats/WebP/LossyUtilsTests.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,52 @@
// Copyright (c) Six Labors.
// Licensed under the Apache License, Version 2.0.

using SixLabors.ImageSharp.Formats.Webp.Lossy;
using SixLabors.ImageSharp.Tests.TestUtilities;
using Xunit;

namespace SixLabors.ImageSharp.Tests.Formats.WebP
{
[Trait("Format", "Webp")]
public class LossyUtilsTests
{
private static void RunHadamardTransformTest()
{
byte[] a =
{
27, 27, 28, 29, 29, 28, 27, 27, 27, 28, 28, 29, 29, 28, 28, 27, 129, 129, 129, 129, 129, 129, 129,
129, 128, 128, 128, 128, 128, 128, 128, 128, 27, 27, 27, 27, 27, 27, 27, 27, 27, 28, 28, 29, 29, 28,
28, 27, 129, 129, 129, 129, 129, 129, 129, 129, 128, 128, 128, 128, 128, 128, 128, 128, 27, 27, 26,
26, 26, 26, 27, 27, 27, 28, 28, 29, 29, 28, 28, 27, 129, 129, 129, 129, 129, 129, 129, 129, 128,
128, 128, 128, 128, 128, 128, 128, 28, 27, 27, 26, 26, 27, 27, 28, 27, 28, 28, 29, 29, 28, 28, 27
};

byte[] b =
{
28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 204, 204, 204, 204, 204, 204, 204,
204, 204, 204, 204, 204, 204, 204, 204, 204, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
28, 28, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 204, 28, 28, 28,
28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 204, 204, 204, 204, 204, 204, 204, 204, 204,
204, 204, 204, 204, 204, 204, 204, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28
};

ushort[] w = { 38, 32, 20, 9, 32, 28, 17, 7, 20, 17, 10, 4, 9, 7, 4, 2 };
int expected = 2;

int actual = LossyUtils.Vp8Disto4X4(a, b, w, new int[16]);
Assert.Equal(expected, actual);
}

[Fact]
public void HadamardTransform_Works() => RunHadamardTransformTest();

#if SUPPORTS_RUNTIME_INTRINSICS
[Fact]
public void HadamardTransform_WithHardwareIntrinsics_Works() => FeatureTestRunner.RunWithHwIntrinsicsFeature(RunHadamardTransformTest, HwIntrinsics.AllowAll);

[Fact]
public void HadamardTransform_WithoutHardwareIntrinsics_Works() => FeatureTestRunner.RunWithHwIntrinsicsFeature(RunHadamardTransformTest, HwIntrinsics.DisableHWIntrinsic);
#endif

}
}