Mirrors/Ryujinx
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

New NVDEC and VIC implementation (#1384)

Browse source 
* Initial NVDEC and VIC implementation

* Update FFmpeg.AutoGen to 4.3.0

* Add nvdec dependencies for Windows

* Unify some VP9 structures

* Rename VP9 structure fields

* Improvements to Video API

* XML docs for Common.Memory

* Remove now unused or redundant overloads from MemoryAccessor

* NVDEC UV surface read/write scalar paths

* Add FIXME comments about hacky things/stuff that will need to be fixed in the future

* Cleaned up VP9 memory allocation

* Remove some debug logs

* Rename some VP9 structs

* Remove unused struct

* No need to compile Ryujinx.Graphics.Host1x with unsafe anymore

* Name AsyncWorkQueue threads to make debugging easier

* Make Vp9PictureInfo a ref struct

* LayoutConverter no longer needs the depth argument (broken by rebase)

* Pooling of VP9 buffers, plus fix a memory leak on VP9

* Really wish VS could rename projects properly...

* Address feedback

* Remove using

* Catch OperationCanceledException

* Add licensing informations

* Add THIRDPARTY.md to release too

Co-authored-by: Thog <me@thog.eu>
This commit is contained in:
gdkchan 2020-07-12 00:07:01 -03:00 • committed by GitHub
parent 38b26cf424
commit 4d02a2d2c0
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
202 changed files with 20563 additions and 2567 deletions
Download patch file Download diff file Expand all files Collapse all files

418
Ryujinx.Graphics.Nvdec.Vp9/LoopFilter.cs Normal file
View file

@ -0,0 +1,418 @@
using Ryujinx.Common.Memory;
using Ryujinx.Graphics.Nvdec.Vp9.Common;
using Ryujinx.Graphics.Nvdec.Vp9.Types;
using System;
using System.Runtime.InteropServices;
namespace Ryujinx.Graphics.Nvdec.Vp9
{
internal static class LoopFilter
{
public const int MaxLoopFilter = 63;
public const int MaxRefLfDeltas = 4;
public const int MaxModeLfDeltas = 2;
// 64 bit masks for left transform size. Each 1 represents a position where
// we should apply a loop filter across the left border of an 8x8 block
// boundary.
//
// In the case of TX_16X16 -> ( in low order byte first we end up with
// a mask that looks like this
//
// 10101010
// 10101010
// 10101010
// 10101010
// 10101010
// 10101010
// 10101010
// 10101010
//
// A loopfilter should be applied to every other 8x8 horizontally.
private static readonly ulong[] Left64X64TxformMask = new ulong[]
{
0xffffffffffffffffUL, // TX_4X4
0xffffffffffffffffUL, // TX_8x8
0x5555555555555555UL, // TX_16x16
0x1111111111111111UL, // TX_32x32
};
// 64 bit masks for above transform size. Each 1 represents a position where
// we should apply a loop filter across the top border of an 8x8 block
// boundary.
//
// In the case of TX_32x32 -> ( in low order byte first we end up with
// a mask that looks like this
//
// 11111111
// 00000000
// 00000000
// 00000000
// 11111111
// 00000000
// 00000000
// 00000000
//
// A loopfilter should be applied to every other 4 the row vertically.
private static readonly ulong[] Above64X64TxformMask = new ulong[]
{
0xffffffffffffffffUL, // TX_4X4
0xffffffffffffffffUL, // TX_8x8
0x00ff00ff00ff00ffUL, // TX_16x16
0x000000ff000000ffUL, // TX_32x32
};
// 64 bit masks for prediction sizes (left). Each 1 represents a position
// where left border of an 8x8 block. These are aligned to the right most
// appropriate bit, and then shifted into place.
//
// In the case of TX_16x32 -> ( low order byte first ) we end up with
// a mask that looks like this :
//
// 10000000
// 10000000
// 10000000
// 10000000
// 00000000
// 00000000
// 00000000
// 00000000
private static readonly ulong[] LeftPredictionMask = new ulong[]
{
0x0000000000000001UL, // BLOCK_4X4,
0x0000000000000001UL, // BLOCK_4X8,
0x0000000000000001UL, // BLOCK_8X4,
0x0000000000000001UL, // BLOCK_8X8,
0x0000000000000101UL, // BLOCK_8X16,
0x0000000000000001UL, // BLOCK_16X8,
0x0000000000000101UL, // BLOCK_16X16,
0x0000000001010101UL, // BLOCK_16X32,
0x0000000000000101UL, // BLOCK_32X16,
0x0000000001010101UL, // BLOCK_32X32,
0x0101010101010101UL, // BLOCK_32X64,
0x0000000001010101UL, // BLOCK_64X32,
0x0101010101010101UL, // BLOCK_64X64
};
// 64 bit mask to shift and set for each prediction size.
private static readonly ulong[] AbovePredictionMask = new ulong[]
{
0x0000000000000001UL, // BLOCK_4X4
0x0000000000000001UL, // BLOCK_4X8
0x0000000000000001UL, // BLOCK_8X4
0x0000000000000001UL, // BLOCK_8X8
0x0000000000000001UL, // BLOCK_8X16,
0x0000000000000003UL, // BLOCK_16X8
0x0000000000000003UL, // BLOCK_16X16
0x0000000000000003UL, // BLOCK_16X32,
0x000000000000000fUL, // BLOCK_32X16,
0x000000000000000fUL, // BLOCK_32X32,
0x000000000000000fUL, // BLOCK_32X64,
0x00000000000000ffUL, // BLOCK_64X32,
0x00000000000000ffUL, // BLOCK_64X64
};
// 64 bit mask to shift and set for each prediction size. A bit is set for
// each 8x8 block that would be in the left most block of the given block
// size in the 64x64 block.
private static readonly ulong[] SizeMask = new ulong[]
{
0x0000000000000001UL, // BLOCK_4X4
0x0000000000000001UL, // BLOCK_4X8
0x0000000000000001UL, // BLOCK_8X4
0x0000000000000001UL, // BLOCK_8X8
0x0000000000000101UL, // BLOCK_8X16,
0x0000000000000003UL, // BLOCK_16X8
0x0000000000000303UL, // BLOCK_16X16
0x0000000003030303UL, // BLOCK_16X32,
0x0000000000000f0fUL, // BLOCK_32X16,
0x000000000f0f0f0fUL, // BLOCK_32X32,
0x0f0f0f0f0f0f0f0fUL, // BLOCK_32X64,
0x00000000ffffffffUL, // BLOCK_64X32,
0xffffffffffffffffUL, // BLOCK_64X64
};
// These are used for masking the left and above borders.
private const ulong LeftBorder = 0x1111111111111111UL;
private const ulong AboveBorder = 0x000000ff000000ffUL;
// 16 bit masks for uv transform sizes.
private static readonly ushort[] Left64X64TxformMaskUv = new ushort[]
{
0xffff, // TX_4X4
0xffff, // TX_8x8
0x5555, // TX_16x16
0x1111, // TX_32x32
};
private static readonly ushort[] Above64X64TxformMaskUv = new ushort[]
{
0xffff, // TX_4X4
0xffff, // TX_8x8
0x0f0f, // TX_16x16
0x000f, // TX_32x32
};
// 16 bit left mask to shift and set for each uv prediction size.
private static readonly ushort[] LeftPredictionMaskUv = new ushort[]
{
0x0001, // BLOCK_4X4,
0x0001, // BLOCK_4X8,
0x0001, // BLOCK_8X4,
0x0001, // BLOCK_8X8,
0x0001, // BLOCK_8X16,
0x0001, // BLOCK_16X8,
0x0001, // BLOCK_16X16,
0x0011, // BLOCK_16X32,
0x0001, // BLOCK_32X16,
0x0011, // BLOCK_32X32,
0x1111, // BLOCK_32X64
0x0011, // BLOCK_64X32,
0x1111, // BLOCK_64X64
};
// 16 bit above mask to shift and set for uv each prediction size.
private static readonly ushort[] AbovePredictionMaskUv = new ushort[]
{
0x0001, // BLOCK_4X4
0x0001, // BLOCK_4X8
0x0001, // BLOCK_8X4
0x0001, // BLOCK_8X8
0x0001, // BLOCK_8X16,
0x0001, // BLOCK_16X8
0x0001, // BLOCK_16X16
0x0001, // BLOCK_16X32,
0x0003, // BLOCK_32X16,
0x0003, // BLOCK_32X32,
0x0003, // BLOCK_32X64,
0x000f, // BLOCK_64X32,
0x000f, // BLOCK_64X64
};
// 64 bit mask to shift and set for each uv prediction size
private static readonly ushort[] SizeMaskUv = new ushort[]
{
0x0001, // BLOCK_4X4
0x0001, // BLOCK_4X8
0x0001, // BLOCK_8X4
0x0001, // BLOCK_8X8
0x0001, // BLOCK_8X16,
0x0001, // BLOCK_16X8
0x0001, // BLOCK_16X16
0x0011, // BLOCK_16X32,
0x0003, // BLOCK_32X16,
0x0033, // BLOCK_32X32,
0x3333, // BLOCK_32X64,
0x00ff, // BLOCK_64X32,
0xffff, // BLOCK_64X64
};
private const ushort LeftBorderUv = 0x1111;
private const ushort AboveBorderUv = 0x000f;
private static readonly int[] ModeLfLut = new int[]
{
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // INTRA_MODES
1, 1, 0, 1 // INTER_MODES (ZEROMV == 0)
};
private static byte GetFilterLevel(ref LoopFilterInfoN lfiN, ref ModeInfo mi)
{
return lfiN.Lvl[mi.SegmentId][mi.RefFrame[0]][ModeLfLut[(int)mi.Mode]];
}
private static ref LoopFilterMask GetLfm(ref Types.LoopFilter lf, int miRow, int miCol)
{
return ref lf.Lfm[(miCol >> 3) + ((miRow >> 3) * lf.LfmStride)];
}
// 8x8 blocks in a superblock. A "1" represents the first block in a 16x16
// or greater area.
private static readonly byte[][] FirstBlockIn16x16 = new byte[][]
{
new byte[] { 1, 0, 1, 0, 1, 0, 1, 0 }, new byte[] { 0, 0, 0, 0, 0, 0, 0, 0 },
new byte[] { 1, 0, 1, 0, 1, 0, 1, 0 }, new byte[] { 0, 0, 0, 0, 0, 0, 0, 0 },
new byte[] { 1, 0, 1, 0, 1, 0, 1, 0 }, new byte[] { 0, 0, 0, 0, 0, 0, 0, 0 },
new byte[] { 1, 0, 1, 0, 1, 0, 1, 0 }, new byte[] { 0, 0, 0, 0, 0, 0, 0, 0 }
};
// This function sets up the bit masks for a block represented
// by miRow, miCol in a 64x64 region.
public static void BuildMask(ref Vp9Common cm, ref ModeInfo mi, int miRow, int miCol, int bw, int bh)
{
BlockSize blockSize = mi.SbType;
TxSize txSizeY = mi.TxSize;
ref LoopFilterInfoN lfiN = ref cm.LfInfo;
int filterLevel = GetFilterLevel(ref lfiN, ref mi);
TxSize txSizeUv = Luts.UvTxsizeLookup[(int)blockSize][(int)txSizeY][1][1];
ref LoopFilterMask lfm = ref GetLfm(ref cm.Lf, miRow, miCol);
ref ulong leftY = ref lfm.LeftY[(int)txSizeY];
ref ulong aboveY = ref lfm.AboveY[(int)txSizeY];
ref ulong int4X4Y = ref lfm.Int4x4Y;
ref ushort leftUv = ref lfm.LeftUv[(int)txSizeUv];
ref ushort aboveUv = ref lfm.AboveUv[(int)txSizeUv];
ref ushort int4X4Uv = ref lfm.Int4x4Uv;
int rowInSb = (miRow & 7);
int colInSb = (miCol & 7);
int shiftY = colInSb + (rowInSb << 3);
int shiftUv = (colInSb >> 1) + ((rowInSb >> 1) << 2);
int buildUv = FirstBlockIn16x16[rowInSb][colInSb];
if (filterLevel == 0)
{
return;
}
else
{
int index = shiftY;
int i;
for (i = 0; i < bh; i++)
{
MemoryMarshal.CreateSpan(ref lfm.LflY[index], 64 - index).Slice(0, bw).Fill((byte)filterLevel);
index += 8;
}
}
// These set 1 in the current block size for the block size edges.
// For instance if the block size is 32x16, we'll set:
// above = 1111
// 0000
// and
// left = 1000
// = 1000
// NOTE : In this example the low bit is left most ( 1000 ) is stored as
// 1, not 8...
//
// U and V set things on a 16 bit scale.
//
aboveY |= AbovePredictionMask[(int)blockSize] << shiftY;
leftY |= LeftPredictionMask[(int)blockSize] << shiftY;
if (buildUv != 0)
{
aboveUv |= (ushort)(AbovePredictionMaskUv[(int)blockSize] << shiftUv);
leftUv |= (ushort)(LeftPredictionMaskUv[(int)blockSize] << shiftUv);
}
// If the block has no coefficients and is not intra we skip applying
// the loop filter on block edges.
if (mi.Skip != 0 && mi.IsInterBlock())
{
return;
}
// Add a mask for the transform size. The transform size mask is set to
// be correct for a 64x64 prediction block size. Mask to match the size of
// the block we are working on and then shift it into place.
aboveY |= (SizeMask[(int)blockSize] & Above64X64TxformMask[(int)txSizeY]) << shiftY;
leftY |= (SizeMask[(int)blockSize] & Left64X64TxformMask[(int)txSizeY]) << shiftY;
if (buildUv != 0)
{
aboveUv |= (ushort)((SizeMaskUv[(int)blockSize] & Above64X64TxformMaskUv[(int)txSizeUv]) << shiftUv);
leftUv |= (ushort)((SizeMaskUv[(int)blockSize] & Left64X64TxformMaskUv[(int)txSizeUv]) << shiftUv);
}
// Try to determine what to do with the internal 4x4 block boundaries. These
// differ from the 4x4 boundaries on the outside edge of an 8x8 in that the
// internal ones can be skipped and don't depend on the prediction block size.
if (txSizeY == TxSize.Tx4x4)
{
int4X4Y |= SizeMask[(int)blockSize] << shiftY;
}
if (buildUv != 0 && txSizeUv == TxSize.Tx4x4)
{
int4X4Uv |= (ushort)((SizeMaskUv[(int)blockSize] & 0xffff) << shiftUv);
}
}
public static unsafe void ResetLfm(ref Vp9Common cm)
{
if (cm.Lf.FilterLevel != 0)
{
MemoryUtil.Fill(cm.Lf.Lfm.ToPointer(), new LoopFilterMask(), ((cm.MiRows + (Constants.MiBlockSize - 1)) >> 3) * cm.Lf.LfmStride);
}
}
private static void UpdateSharpness(ref LoopFilterInfoN lfi, int sharpnessLvl)
{
int lvl;
// For each possible value for the loop filter fill out limits
for (lvl = 0; lvl <= MaxLoopFilter; lvl++)
{
// Set loop filter parameters that control sharpness.
int blockInsideLimit = lvl >> ((sharpnessLvl > 0 ? 1 : 0) + (sharpnessLvl > 4 ? 1 : 0));
if (sharpnessLvl > 0)
{
if (blockInsideLimit > (9 - sharpnessLvl))
{
blockInsideLimit = (9 - sharpnessLvl);
}
}
if (blockInsideLimit < 1)
{
blockInsideLimit = 1;
}
lfi.Lfthr[lvl].Lim.ToSpan().Fill((byte)blockInsideLimit);
lfi.Lfthr[lvl].Mblim.ToSpan().Fill((byte)(2 * (lvl + 2) + blockInsideLimit));
}
}
public static void LoopFilterFrameInit(ref Vp9Common cm, int defaultFiltLvl)
{
int segId;
// nShift is the multiplier for lfDeltas
// the multiplier is 1 for when filterLvl is between 0 and 31;
// 2 when filterLvl is between 32 and 63
int scale = 1 << (defaultFiltLvl >> 5);
ref LoopFilterInfoN lfi = ref cm.LfInfo;
ref Types.LoopFilter lf = ref cm.Lf;
ref Segmentation seg = ref cm.Seg;
// Update limits if sharpness has changed
if (lf.LastSharpnessLevel != lf.SharpnessLevel)
{
UpdateSharpness(ref lfi, lf.SharpnessLevel);
lf.LastSharpnessLevel = lf.SharpnessLevel;
}
for (segId = 0; segId < Constants.MaxSegments; segId++)
{
int lvlSeg = defaultFiltLvl;
if (seg.IsSegFeatureActive(segId, SegLvlFeatures.SegLvlAltLf) != 0)
{
int data = seg.GetSegData(segId, SegLvlFeatures.SegLvlAltLf);
lvlSeg = Math.Clamp(seg.AbsDelta == Constants.SegmentAbsData ? data : defaultFiltLvl + data, 0, MaxLoopFilter);
}
if (!lf.ModeRefDeltaEnabled)
{
// We could get rid of this if we assume that deltas are set to
// zero when not in use; encoder always uses deltas
MemoryMarshal.Cast<Array2<byte>, byte>(lfi.Lvl[segId].ToSpan()).Fill((byte)lvlSeg);
}
else
{
int refr, mode;
int intraLvl = lvlSeg + lf.RefDeltas[Constants.IntraFrame] * scale;
lfi.Lvl[segId][Constants.IntraFrame][0] = (byte)Math.Clamp(intraLvl, 0, MaxLoopFilter);
for (refr = Constants.LastFrame; refr < Constants.MaxRefFrames; ++refr)
{
for (mode = 0; mode < MaxModeLfDeltas; ++mode)
{
int interLvl = lvlSeg + lf.RefDeltas[refr] * scale + lf.ModeDeltas[mode] * scale;
lfi.Lvl[segId][refr][mode] = (byte)Math.Clamp(interLvl, 0, MaxLoopFilter);
}
}
}
}
}
}
}