Ryujinx/Ryujinx.Graphics.Vic/Image/SurfaceReader.cs
gdkchan cb250162cb
Accelerate NVDEC VIC surface read/write and colorspace conversion with Arm64 HW intrinsics (#4351)
* Accelerate NVDEC VIC surface read/write and colorspace conversion with Arm64 HW intrinsics

* Improve ReadNv12 x86 SSE path
2023-02-07 02:38:54 +00:00

496 lines
21 KiB
C#

using Ryujinx.Common.Logging;
using Ryujinx.Common.Memory;
using Ryujinx.Graphics.Texture;
using Ryujinx.Graphics.Vic.Types;
using System;
using System.Runtime.CompilerServices;
using System.Runtime.Intrinsics;
using System.Runtime.Intrinsics.Arm;
using System.Runtime.Intrinsics.X86;
using static Ryujinx.Graphics.Vic.Image.SurfaceCommon;
namespace Ryujinx.Graphics.Vic.Image
{
static class SurfaceReader
{
public static Surface Read(
ResourceManager rm,
ref SlotConfig config,
ref SlotSurfaceConfig surfaceConfig,
ref Array8<PlaneOffsets> offsets)
{
switch (surfaceConfig.SlotPixelFormat)
{
case PixelFormat.Y8___V8U8_N420: return ReadNv12(rm, ref config, ref surfaceConfig, ref offsets);
}
Logger.Error?.Print(LogClass.Vic, $"Unsupported pixel format \"{surfaceConfig.SlotPixelFormat}\".");
int lw = surfaceConfig.SlotLumaWidth + 1;
int lh = surfaceConfig.SlotLumaHeight + 1;
return new Surface(rm.SurfacePool, lw, lh);
}
private unsafe static Surface ReadNv12(
ResourceManager rm,
ref SlotConfig config,
ref SlotSurfaceConfig surfaceConfig,
ref Array8<PlaneOffsets> offsets)
{
InputSurface input = ReadSurface(rm, ref config, ref surfaceConfig, ref offsets, 1, 2);
int width = input.Width;
int height = input.Height;
int yStride = GetPitch(width, 1);
int uvStride = GetPitch(input.UvWidth, 2);
Surface output = new Surface(rm.SurfacePool, width, height);
if (Sse41.IsSupported)
{
Vector128<byte> shufMask = Vector128.Create(
(byte)0, (byte)2, (byte)3, (byte)1,
(byte)4, (byte)6, (byte)7, (byte)5,
(byte)8, (byte)10, (byte)11, (byte)9,
(byte)12, (byte)14, (byte)15, (byte)13);
Vector128<short> alphaMask = Vector128.Create(0xff << 24).AsInt16();
int yStrideGap = yStride - width;
int uvStrideGap = uvStride - input.UvWidth;
int widthTrunc = width & ~0xf;
fixed (Pixel* dstPtr = output.Data)
{
Pixel* op = dstPtr;
fixed (byte* src0Ptr = input.Buffer0, src1Ptr = input.Buffer1)
{
byte* i0p = src0Ptr;
for (int y = 0; y < height; y++)
{
byte* i1p = src1Ptr + (y >> 1) * uvStride;
int x = 0;
for (; x < widthTrunc; x += 16, i0p += 16, i1p += 16)
{
Vector128<short> ya0 = Sse41.ConvertToVector128Int16(i0p);
Vector128<short> ya1 = Sse41.ConvertToVector128Int16(i0p + 8);
Vector128<byte> uv = Sse2.LoadVector128(i1p);
Vector128<short> uv0 = Sse2.UnpackLow(uv.AsInt16(), uv.AsInt16());
Vector128<short> uv1 = Sse2.UnpackHigh(uv.AsInt16(), uv.AsInt16());
Vector128<short> rgba0 = Sse2.UnpackLow(ya0, uv0);
Vector128<short> rgba1 = Sse2.UnpackHigh(ya0, uv0);
Vector128<short> rgba2 = Sse2.UnpackLow(ya1, uv1);
Vector128<short> rgba3 = Sse2.UnpackHigh(ya1, uv1);
rgba0 = Ssse3.Shuffle(rgba0.AsByte(), shufMask).AsInt16();
rgba1 = Ssse3.Shuffle(rgba1.AsByte(), shufMask).AsInt16();
rgba2 = Ssse3.Shuffle(rgba2.AsByte(), shufMask).AsInt16();
rgba3 = Ssse3.Shuffle(rgba3.AsByte(), shufMask).AsInt16();
rgba0 = Sse2.Or(rgba0, alphaMask);
rgba1 = Sse2.Or(rgba1, alphaMask);
rgba2 = Sse2.Or(rgba2, alphaMask);
rgba3 = Sse2.Or(rgba3, alphaMask);
Vector128<short> rgba16_0 = Sse41.ConvertToVector128Int16(rgba0.AsByte());
Vector128<short> rgba16_1 = Sse41.ConvertToVector128Int16(HighToLow(rgba0.AsByte()));
Vector128<short> rgba16_2 = Sse41.ConvertToVector128Int16(rgba1.AsByte());
Vector128<short> rgba16_3 = Sse41.ConvertToVector128Int16(HighToLow(rgba1.AsByte()));
Vector128<short> rgba16_4 = Sse41.ConvertToVector128Int16(rgba2.AsByte());
Vector128<short> rgba16_5 = Sse41.ConvertToVector128Int16(HighToLow(rgba2.AsByte()));
Vector128<short> rgba16_6 = Sse41.ConvertToVector128Int16(rgba3.AsByte());
Vector128<short> rgba16_7 = Sse41.ConvertToVector128Int16(HighToLow(rgba3.AsByte()));
rgba16_0 = Sse2.ShiftLeftLogical(rgba16_0, 2);
rgba16_1 = Sse2.ShiftLeftLogical(rgba16_1, 2);
rgba16_2 = Sse2.ShiftLeftLogical(rgba16_2, 2);
rgba16_3 = Sse2.ShiftLeftLogical(rgba16_3, 2);
rgba16_4 = Sse2.ShiftLeftLogical(rgba16_4, 2);
rgba16_5 = Sse2.ShiftLeftLogical(rgba16_5, 2);
rgba16_6 = Sse2.ShiftLeftLogical(rgba16_6, 2);
rgba16_7 = Sse2.ShiftLeftLogical(rgba16_7, 2);
Sse2.Store((short*)(op + (uint)x + 0), rgba16_0);
Sse2.Store((short*)(op + (uint)x + 2), rgba16_1);
Sse2.Store((short*)(op + (uint)x + 4), rgba16_2);
Sse2.Store((short*)(op + (uint)x + 6), rgba16_3);
Sse2.Store((short*)(op + (uint)x + 8), rgba16_4);
Sse2.Store((short*)(op + (uint)x + 10), rgba16_5);
Sse2.Store((short*)(op + (uint)x + 12), rgba16_6);
Sse2.Store((short*)(op + (uint)x + 14), rgba16_7);
}
for (; x < width; x++, i1p += (x & 1) * 2)
{
Pixel* px = op + (uint)x;
px->R = Upsample(*i0p++);
px->G = Upsample(*i1p);
px->B = Upsample(*(i1p + 1));
px->A = 0x3ff;
}
op += width;
i0p += yStrideGap;
i1p += uvStrideGap;
}
}
}
}
else if (AdvSimd.Arm64.IsSupported)
{
Vector128<int> alphaMask = Vector128.Create(0xffu << 24).AsInt32();
int yStrideGap = yStride - width;
int uvStrideGap = uvStride - input.UvWidth;
int widthTrunc = width & ~0xf;
fixed (Pixel* dstPtr = output.Data)
{
Pixel* op = dstPtr;
fixed (byte* src0Ptr = input.Buffer0, src1Ptr = input.Buffer1)
{
byte* i0p = src0Ptr;
for (int y = 0; y < height; y++)
{
byte* i1p = src1Ptr + (y >> 1) * uvStride;
int x = 0;
for (; x < widthTrunc; x += 16, i0p += 16, i1p += 16)
{
Vector128<byte> ya = AdvSimd.LoadVector128(i0p);
Vector128<byte> uv = AdvSimd.LoadVector128(i1p);
Vector128<short> ya0 = AdvSimd.ZeroExtendWideningLower(ya.GetLower()).AsInt16();
Vector128<short> ya1 = AdvSimd.ZeroExtendWideningUpper(ya).AsInt16();
Vector128<short> uv0 = AdvSimd.Arm64.ZipLow(uv.AsInt16(), uv.AsInt16());
Vector128<short> uv1 = AdvSimd.Arm64.ZipHigh(uv.AsInt16(), uv.AsInt16());
ya0 = AdvSimd.ShiftLeftLogical(ya0, 8);
ya1 = AdvSimd.ShiftLeftLogical(ya1, 8);
Vector128<short> rgba0 = AdvSimd.Arm64.ZipLow(ya0, uv0);
Vector128<short> rgba1 = AdvSimd.Arm64.ZipHigh(ya0, uv0);
Vector128<short> rgba2 = AdvSimd.Arm64.ZipLow(ya1, uv1);
Vector128<short> rgba3 = AdvSimd.Arm64.ZipHigh(ya1, uv1);
rgba0 = AdvSimd.ShiftRightLogicalAdd(alphaMask, rgba0.AsInt32(), 8).AsInt16();
rgba1 = AdvSimd.ShiftRightLogicalAdd(alphaMask, rgba1.AsInt32(), 8).AsInt16();
rgba2 = AdvSimd.ShiftRightLogicalAdd(alphaMask, rgba2.AsInt32(), 8).AsInt16();
rgba3 = AdvSimd.ShiftRightLogicalAdd(alphaMask, rgba3.AsInt32(), 8).AsInt16();
Vector128<short> rgba16_0 = AdvSimd.ZeroExtendWideningLower(rgba0.AsByte().GetLower()).AsInt16();
Vector128<short> rgba16_1 = AdvSimd.ZeroExtendWideningUpper(rgba0.AsByte()).AsInt16();
Vector128<short> rgba16_2 = AdvSimd.ZeroExtendWideningLower(rgba1.AsByte().GetLower()).AsInt16();
Vector128<short> rgba16_3 = AdvSimd.ZeroExtendWideningUpper(rgba1.AsByte()).AsInt16();
Vector128<short> rgba16_4 = AdvSimd.ZeroExtendWideningLower(rgba2.AsByte().GetLower()).AsInt16();
Vector128<short> rgba16_5 = AdvSimd.ZeroExtendWideningUpper(rgba2.AsByte()).AsInt16();
Vector128<short> rgba16_6 = AdvSimd.ZeroExtendWideningLower(rgba3.AsByte().GetLower()).AsInt16();
Vector128<short> rgba16_7 = AdvSimd.ZeroExtendWideningUpper(rgba3.AsByte()).AsInt16();
rgba16_0 = AdvSimd.ShiftLeftLogical(rgba16_0, 2);
rgba16_1 = AdvSimd.ShiftLeftLogical(rgba16_1, 2);
rgba16_2 = AdvSimd.ShiftLeftLogical(rgba16_2, 2);
rgba16_3 = AdvSimd.ShiftLeftLogical(rgba16_3, 2);
rgba16_4 = AdvSimd.ShiftLeftLogical(rgba16_4, 2);
rgba16_5 = AdvSimd.ShiftLeftLogical(rgba16_5, 2);
rgba16_6 = AdvSimd.ShiftLeftLogical(rgba16_6, 2);
rgba16_7 = AdvSimd.ShiftLeftLogical(rgba16_7, 2);
AdvSimd.Store((short*)(op + (uint)x + 0), rgba16_0);
AdvSimd.Store((short*)(op + (uint)x + 2), rgba16_1);
AdvSimd.Store((short*)(op + (uint)x + 4), rgba16_2);
AdvSimd.Store((short*)(op + (uint)x + 6), rgba16_3);
AdvSimd.Store((short*)(op + (uint)x + 8), rgba16_4);
AdvSimd.Store((short*)(op + (uint)x + 10), rgba16_5);
AdvSimd.Store((short*)(op + (uint)x + 12), rgba16_6);
AdvSimd.Store((short*)(op + (uint)x + 14), rgba16_7);
}
for (; x < width; x++, i1p += (x & 1) * 2)
{
Pixel* px = op + (uint)x;
px->R = Upsample(*i0p++);
px->G = Upsample(*i1p);
px->B = Upsample(*(i1p + 1));
px->A = 0x3ff;
}
op += width;
i0p += yStrideGap;
i1p += uvStrideGap;
}
}
}
}
else
{
for (int y = 0; y < height; y++)
{
int uvBase = (y >> 1) * uvStride;
for (int x = 0; x < width; x++)
{
output.SetR(x, y, Upsample(input.Buffer0[y * yStride + x]));
int uvOffs = uvBase + (x & ~1);
output.SetG(x, y, Upsample(input.Buffer1[uvOffs]));
output.SetB(x, y, Upsample(input.Buffer1[uvOffs + 1]));
output.SetA(x, y, 0x3ff);
}
}
}
input.Return(rm.BufferPool);
return output;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static Vector128<byte> HighToLow(Vector128<byte> value)
{
return Sse.MoveHighToLow(value.AsSingle(), value.AsSingle()).AsByte();
}
private static InputSurface ReadSurface(
ResourceManager rm,
ref SlotConfig config,
ref SlotSurfaceConfig surfaceConfig,
ref Array8<PlaneOffsets> offsets,
int bytesPerPixel,
int planes)
{
InputSurface surface = new InputSurface();
surface.Initialize();
int gobBlocksInY = 1 << surfaceConfig.SlotBlkHeight;
bool linear = surfaceConfig.SlotBlkKind == 0;
int lw = surfaceConfig.SlotLumaWidth + 1;
int lh = surfaceConfig.SlotLumaHeight + 1;
int cw = surfaceConfig.SlotChromaWidth + 1;
int ch = surfaceConfig.SlotChromaHeight + 1;
// Interlaced inputs have double the height when deinterlaced.
int heightShift = config.FrameFormat.IsField() ? 1 : 0;
surface.Width = lw;
surface.Height = lh << heightShift;
surface.UvWidth = cw;
surface.UvHeight = ch << heightShift;
if (planes > 0)
{
surface.SetBuffer0(ReadBuffer(rm, ref config, ref offsets, linear, 0, lw, lh, bytesPerPixel, gobBlocksInY));
}
if (planes > 1)
{
surface.SetBuffer1(ReadBuffer(rm, ref config, ref offsets, linear, 1, cw, ch, planes == 2 ? 2 : 1, gobBlocksInY));
}
if (planes > 2)
{
surface.SetBuffer2(ReadBuffer(rm, ref config, ref offsets, linear, 2, cw, ch, 1, gobBlocksInY));
}
return surface;
}
private static RentedBuffer ReadBuffer(
ResourceManager rm,
scoped ref SlotConfig config,
scoped ref Array8<PlaneOffsets> offsets,
bool linear,
int plane,
int width,
int height,
int bytesPerPixel,
int gobBlocksInY)
{
FrameFormat frameFormat = config.FrameFormat;
bool isLuma = plane == 0;
bool isField = frameFormat.IsField();
bool isTopField = frameFormat.IsTopField(isLuma);
int stride = GetPitch(width, bytesPerPixel);
uint offset = GetOffset(ref offsets[0], plane);
int dstStart = 0;
int dstStride = stride;
if (isField)
{
dstStart = isTopField ? 0 : stride;
dstStride = stride * 2;
}
RentedBuffer buffer;
if (linear)
{
buffer = ReadBufferLinear(rm, offset, width, height, dstStart, dstStride, bytesPerPixel);
}
else
{
buffer = ReadBufferBlockLinear(rm, offset, width, height, dstStart, dstStride, bytesPerPixel, gobBlocksInY);
}
if (isField || frameFormat.IsInterlaced())
{
RentedBuffer prevBuffer = RentedBuffer.Empty;
RentedBuffer nextBuffer = RentedBuffer.Empty;
if (config.PrevFieldEnable)
{
prevBuffer = ReadBufferNoDeinterlace(rm, ref offsets[1], linear, plane, width, height, bytesPerPixel, gobBlocksInY);
}
if (config.NextFieldEnable)
{
nextBuffer = ReadBufferNoDeinterlace(rm, ref offsets[2], linear, plane, width, height, bytesPerPixel, gobBlocksInY);
}
int w = width * bytesPerPixel;
switch (config.DeinterlaceMode)
{
case DeinterlaceMode.Weave:
Scaler.DeinterlaceWeave(buffer.Data, prevBuffer.Data, w, stride, isTopField);
break;
case DeinterlaceMode.BobField:
Scaler.DeinterlaceBob(buffer.Data, w, stride, isTopField);
break;
case DeinterlaceMode.Bob:
bool isCurrentTop = isLuma ? config.IsEven : config.ChromaEven;
Scaler.DeinterlaceBob(buffer.Data, w, stride, isCurrentTop ^ frameFormat.IsInterlacedBottomFirst());
break;
case DeinterlaceMode.NewBob:
case DeinterlaceMode.Disi1:
Scaler.DeinterlaceMotionAdaptive(buffer.Data, prevBuffer.Data, nextBuffer.Data, w, stride, isTopField);
break;
case DeinterlaceMode.WeaveLumaBobFieldChroma:
if (isLuma)
{
Scaler.DeinterlaceWeave(buffer.Data, prevBuffer.Data, w, stride, isTopField);
}
else
{
Scaler.DeinterlaceBob(buffer.Data, w, stride, isTopField);
}
break;
default:
Logger.Error?.Print(LogClass.Vic, $"Unsupported deinterlace mode \"{config.DeinterlaceMode}\".");
break;
}
prevBuffer.Return(rm.BufferPool);
nextBuffer.Return(rm.BufferPool);
}
return buffer;
}
private static uint GetOffset(ref PlaneOffsets offsets, int plane)
{
return plane switch
{
0 => offsets.LumaOffset,
1 => offsets.ChromaUOffset,
2 => offsets.ChromaVOffset,
_ => throw new ArgumentOutOfRangeException(nameof(plane))
};
}
private static RentedBuffer ReadBufferNoDeinterlace(
ResourceManager rm,
ref PlaneOffsets offsets,
bool linear,
int plane,
int width,
int height,
int bytesPerPixel,
int gobBlocksInY)
{
int stride = GetPitch(width, bytesPerPixel);
uint offset = GetOffset(ref offsets, plane);
if (linear)
{
return ReadBufferLinear(rm, offset, width, height, 0, stride, bytesPerPixel);
}
return ReadBufferBlockLinear(rm, offset, width, height, 0, stride, bytesPerPixel, gobBlocksInY);
}
private static RentedBuffer ReadBufferLinear(
ResourceManager rm,
uint offset,
int width,
int height,
int dstStart,
int dstStride,
int bytesPerPixel)
{
int srcStride = GetPitch(width, bytesPerPixel);
int inSize = srcStride * height;
ReadOnlySpan<byte> src = rm.Gmm.GetSpan(ExtendOffset(offset), inSize);
int outSize = dstStride * height;
int bufferIndex = rm.BufferPool.RentMinimum(outSize, out byte[] buffer);
Span<byte> dst = buffer;
dst = dst.Slice(0, outSize);
for (int y = 0; y < height; y++)
{
src.Slice(y * srcStride, srcStride).CopyTo(dst.Slice(dstStart + y * dstStride, srcStride));
}
return new RentedBuffer(dst, bufferIndex);
}
private static RentedBuffer ReadBufferBlockLinear(
ResourceManager rm,
uint offset,
int width,
int height,
int dstStart,
int dstStride,
int bytesPerPixel,
int gobBlocksInY)
{
int inSize = GetBlockLinearSize(width, height, bytesPerPixel, gobBlocksInY);
ReadOnlySpan<byte> src = rm.Gmm.GetSpan(ExtendOffset(offset), inSize);
int outSize = dstStride * height;
int bufferIndex = rm.BufferPool.RentMinimum(outSize, out byte[] buffer);
Span<byte> dst = buffer;
dst = dst.Slice(0, outSize);
LayoutConverter.ConvertBlockLinearToLinear(dst.Slice(dstStart), width, height, dstStride, bytesPerPixel, gobBlocksInY, src);
return new RentedBuffer(dst, bufferIndex);
}
}
}