tsai
/
ImageSharp
mirror of https://github.com/SixLabors/ImageSharp
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
Browse Source

Fix issues with SSE filters

pull/1871/head
Brian Popow 5 years ago
parent
3cc321ca6e
commit
86bf6c31d3
1 changed files with 21 additions and 17 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 38
      src/ImageSharp/Formats/Webp/Lossy/LossyUtils.cs

38
src/ImageSharp/Formats/Webp/Lossy/LossyUtils.cs
View File

@ -22,7 +22,9 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
private static readonly Vector128<sbyte> Three = Vector128.Create((byte)3).AsSByte();
private static readonly Vector128<sbyte> Four = Vector128.Create((byte)4).AsSByte();
private static readonly Vector128<short> FourShort = Vector128.Create((short)4);
private static readonly Vector128<sbyte> FourSByte = Vector128.Create((byte)4).AsSByte();
private static readonly Vector128<sbyte> Nine = Vector128.Create((short)0x0900).AsSByte();
@ -918,7 +920,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
// Horizontal pass and subsequent transpose.
// First pass, c and d calculations are longer because of the "trick" multiplications.
Vector128<short> dc = Sse2.Add(t0.AsInt16(), Four.AsInt16());
Vector128<short> dc = Sse2.Add(t0.AsInt16(), FourShort);
a = Sse2.Add(dc, t2.AsInt16());
b = Sse2.Subtract(dc, t2.AsInt16());
@ -1039,7 +1041,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
// Horizontal pass and subsequent transpose.
// First pass, c and d calculations are longer because of the "trick" multiplications.
Vector128<short> dc = Sse2.Add(t0.AsInt16(), Four.AsInt16());
Vector128<short> dc = Sse2.Add(t0.AsInt16(), FourShort);
a = Sse2.Add(dc, t2.AsInt16());
b = Sse2.Subtract(dc, t2.AsInt16());
@ -1326,12 +1328,12 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
// Store.
ref byte outputRef = ref MemoryMarshal.GetReference(p);
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset - (3 * stride))) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset - (2 * stride))) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset - (3 * stride))) = p2.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset - (2 * stride))) = p1.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset - stride)) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset)) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset + stride)) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset + (2 * stride))) = p0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset)) = q0.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset + stride)) = q1.AsInt32();
Unsafe.As<byte, Vector128<int>>(ref Unsafe.Add(ref outputRef, offset + (2 * stride))) = q2.AsInt32();
}
else
#endif
@ -1363,7 +1365,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
DoFilter6Sse2(ref p2, ref p1, ref p0, ref q0, ref q1, ref q2, mask, hevThresh);
Store16x4(p3, p2, p1, p0, b, b.Slice(8 * stride), stride);
Store16x4(q3, q2, q1, q0, p, p.Slice(8 * stride), stride);
Store16x4(q0, q1, q2, q3, p.Slice(offset), p.Slice(offset + (8 * stride)), stride);
}
else
#endif
@ -1515,7 +1517,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
StoreUv(p0, ref uRef, ref vRef, offset - stride);
StoreUv(q0, ref uRef, ref vRef, offset);
StoreUv(q1, ref uRef, ref vRef, offset + (1 * stride));
StoreUv(p2, ref uRef, ref vRef, offset + (2 * stride));
StoreUv(q2, ref uRef, ref vRef, offset + (2 * stride));
}
else
#endif
@ -1549,7 +1551,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
DoFilter6Sse2(ref p2, ref p1, ref p0, ref q0, ref q1, ref q2, mask, hevThresh);
Store16x4(p3, p2, p1, p0, tu, tv, stride);
Store16x4(q0, p1, q2, q3, u.Slice(offset), v.Slice(offset), stride);
Store16x4(q0, q1, q2, q3, u.Slice(offset), v.Slice(offset), stride);
}
else
#endif
@ -1583,6 +1585,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
Vector128<byte> q1 = LoadUvEdge(ref uRef, ref vRef, offset + stride);
t1 = LoadUvEdge(ref uRef, ref vRef, offset + (stride * 2));
t2 = LoadUvEdge(ref uRef, ref vRef, offset + (stride * 3));
mask = Sse2.Max(mask, Abs(q1, q0));
mask = Sse2.Max(mask, Abs(t2, t1));
mask = Sse2.Max(mask, Abs(t1, q1));
@ -1593,7 +1596,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
// Store.
StoreUv(p1, ref uRef, ref vRef, offset + (-2 * stride));
StoreUv(p0, ref uRef, ref vRef, offset + (-1 * stride));
StoreUv(q1, ref uRef, ref vRef, offset);
StoreUv(q0, ref uRef, ref vRef, offset);
StoreUv(q1, ref uRef, ref vRef, offset + stride);
}
else
@ -1611,7 +1614,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
#if SUPPORTS_RUNTIME_INTRINSICS
if (Sse2.IsSupported)
{
Load16x4(u.Slice(offset), v.Slice(offset), stride, out Vector128<byte> t2, out Vector128<byte> t1, out Vector128<byte> p1, out Vector128<byte> p0);
Load16x4(u.Slice(offset), v.Slice(offset), stride, out Vector128<byte> t2, out Vector128<byte> t1, out Vector128<byte> p1, out Vector128<byte> p0);
Vector128<byte> mask = Abs(p1, p0);
mask = Sse2.Max(mask, Abs(t2, t1));
@ -1621,6 +1624,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
offset += 4;
Load16x4(u.Slice(offset), v.Slice(offset), stride, out Vector128<byte> q0, out Vector128<byte> q1, out t1, out t2);
mask = Sse2.Max(mask, Abs(q1, q0));
mask = Sse2.Max(mask, Abs(t2, t1));
mask = Sse2.Max(mask, Abs(t1, q1));
@ -1856,7 +1860,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
t1 = Sse2.And(t1.AsByte(), mask).AsSByte(); // mask filter values we don't care about.
t2 = Sse2.AddSaturate(t1, Three); // 3 * (q0 - p0) + hev(p1 - q1) + 3
Vector128<sbyte> t3 = Sse2.AddSaturate(t1, Four); // 3 * (q0 - p0) + hev(p1 - q1) + 4
Vector128<sbyte> t3 = Sse2.AddSaturate(t1, FourSByte); // 3 * (q0 - p0) + hev(p1 - q1) + 4
t2 = SignedShift8b(t2.AsByte()); // (3 * (q0 - p0) + hev(p1 - q1) + 3) >> 3
t3 = SignedShift8b(t3.AsByte()); // (3 * (q0 - p0) + hev(p1 - q1) + 4) >> 3
p0 = Sse2.AddSaturate(p0.AsSByte(), t2).AsByte(); // p0 += t2
@ -1903,8 +1907,8 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
Vector128<byte> flow = Sse2.UnpackLow(Vector128<byte>.Zero, f);
Vector128<byte> fhigh = Sse2.UnpackHigh(Vector128<byte>.Zero, f);
Vector128<short> f9High = Sse2.MultiplyHigh(flow.AsInt16(), Nine.AsInt16()); // Filter (lo) * 9
Vector128<short> f9Low = Sse2.MultiplyLow(fhigh.AsInt16(), Nine.AsInt16()); // Filter (hi) * 9
Vector128<short> f9Low = Sse2.MultiplyHigh(flow.AsInt16(), Nine.AsInt16()); // Filter (lo) * 9
Vector128<short> f9High = Sse2.MultiplyHigh(fhigh.AsInt16(), Nine.AsInt16()); // Filter (hi) * 9
Vector128<short> a2Low = Sse2.Add(f9Low, SixtyThree.AsInt16()); // Filter * 9 + 63
Vector128<short> a2High = Sse2.Add(f9High, SixtyThree.AsInt16()); // Filter * 9 + 63
@ -1923,7 +1927,7 @@ namespace SixLabors.ImageSharp.Formats.Webp.Lossy
private static void DoSimpleFilterSse2(ref Vector128<byte> p0, ref Vector128<byte> q0, Vector128<byte> fl)
{
Vector128<sbyte> v3 = Sse2.AddSaturate(fl.AsSByte(), Three);
Vector128<sbyte> v4 = Sse2.AddSaturate(fl.AsSByte(), Four);
Vector128<sbyte> v4 = Sse2.AddSaturate(fl.AsSByte(), FourSByte);
v4 = SignedShift8b(v4.AsByte()).AsSByte(); // v4 >> 3
v3 = SignedShift8b(v3.AsByte()).AsSByte(); // v3 >> 3

Write Preview
Loading…
Cancel
Save