crafterpika/ryujinx
1
0
Fork 0
mirror of https://git.ryujinx.app/ryubing/ryujinx.git synced 2025-06-28 11:06:24 +02:00
Code Issues Projects Releases Packages Wiki Activity

Implement VP9 loop filtering (#550)

Browse source 
Unmerged PR from OG Ryujinx (#4367). From @gdkchan:

> The main goal of this change is porting the loop filtering from
libvpx, which should fix the block artifacts on some VP9 videos on games
using NVDEC to decode them. In addition to that, there are two other
changes:
> 
> - The remaining decoder code required to decode a VP9 video (with
headers included) has been added. That was done because it's much better
to test the decoder standalone with a video file. I decided to keep that
code on the emulator, even if some of it is unused, since it makes
standalone testing easier in the future too, and we can include unit
tests with video files.
> - Large refactoring of both new and existing code to conform with our
conding [sic] styles, done by @TSRBerry (thanks!) Some of it has been
automated.
> 
> Since we had no loop filtering before, this change will make video
decoding slower. That may cause frame drop etc if the decoder is not
fast enough in some games. I plan to optimize the decoder more in the
future to make up for that, but if possible I'd prefer to not do it as
part of this PR, but if the perf loss is too severe I might consider.
> 
> This will need to be tested on games that had the block artifacts, it
would be nice to confirm if they match hardware now, and get some
before/after screenshots etc.

Comment from @Bjorn29512:

> Significantly improves the block artifacts in FE: Engage.
> 
> Before:
>
![](https://user-images.githubusercontent.com/110204265/216882414-ec88dbda-7544-4490-8a47-37f074056ae3.png)
> 
> After:
>
![](https://user-images.githubusercontent.com/110204265/216882478-4e81fead-1033-4877-b282-f9cac6d6aa3b.png)

---------

Co-authored-by: gdkchan <gab.dark.100@gmail.com>
Co-authored-by: TSR Berry <20988865+TSRBerry@users.noreply.github.com>
This commit is contained in:
Keaton 2025-02-18 20:59:36 -06:00 • committed by GitHub
parent 920933bc9f
commit f91cd05260
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
79 changed files with 11343 additions and 3036 deletions
Download patch file Download diff file Expand all files Collapse all files

116
src/Ryujinx.Graphics.Nvdec.Vp9/Decoder.cs
View file

@ -1,4 +1,4 @@
using Ryujinx.Common.Memory;
using Ryujinx.Common.Memory;
using Ryujinx.Graphics.Nvdec.Vp9.Common;
using Ryujinx.Graphics.Nvdec.Vp9.Types;
using Ryujinx.Graphics.Video;
@ -12,14 +12,14 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
private readonly MemoryAllocator _allocator = new();
public ISurface CreateSurface(int width, int height) => new Surface(width, height);
public ISurface CreateSurface(int width, int height)
{
return new Surface(width, height);
}
private static ReadOnlySpan<byte> LiteralToFilter => new byte[]
{
Constants.EightTapSmooth,
Constants.EightTap,
Constants.EightTapSharp,
Constants.Bilinear,
Constants.EightTapSmooth, Constants.EightTap, Constants.EightTapSharp, Constants.Bilinear
};
public unsafe bool Decode(
@ -29,25 +29,24 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
ReadOnlySpan<Vp9MvRef> mvsIn,
Span<Vp9MvRef> mvsOut)
{
Vp9Common cm = new()
{
FrameType = pictureInfo.IsKeyFrame ? FrameType.KeyFrame : FrameType.InterFrame,
IntraOnly = pictureInfo.IntraOnly,
Vp9Common cm = new();
Width = output.Width,
Height = output.Height,
SubsamplingX = 1,
SubsamplingY = 1,
cm.FrameType = pictureInfo.IsKeyFrame ? FrameType.KeyFrame : FrameType.InterFrame;
cm.IntraOnly = pictureInfo.IntraOnly;
UsePrevFrameMvs = pictureInfo.UsePrevInFindMvRefs,
cm.Width = output.Width;
cm.Height = output.Height;
cm.SubsamplingX = 1;
cm.SubsamplingY = 1;
RefFrameSignBias = pictureInfo.RefFrameSignBias,
cm.UsePrevFrameMvs = pictureInfo.UsePrevInFindMvRefs;
BaseQindex = pictureInfo.BaseQIndex,
YDcDeltaQ = pictureInfo.YDcDeltaQ,
UvAcDeltaQ = pictureInfo.UvAcDeltaQ,
UvDcDeltaQ = pictureInfo.UvDcDeltaQ,
};
cm.RefFrameSignBias = pictureInfo.RefFrameSignBias;
cm.BaseQindex = pictureInfo.BaseQIndex;
cm.YDcDeltaQ = pictureInfo.YDcDeltaQ;
cm.UvAcDeltaQ = pictureInfo.UvAcDeltaQ;
cm.UvDcDeltaQ = pictureInfo.UvDcDeltaQ;
cm.Mb.Lossless = pictureInfo.Lossless;
cm.Mb.Bd = 8;
@ -68,6 +67,8 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
cm.CompFixedRef = pictureInfo.CompFixedRef;
cm.CompVarRef = pictureInfo.CompVarRef;
cm.BitDepth = BitDepth.Bits8;
cm.Log2TileCols = pictureInfo.Log2TileCols;
cm.Log2TileRows = pictureInfo.Log2TileRows;
@ -78,6 +79,8 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
cm.Seg.FeatureMask = pictureInfo.SegmentFeatureEnable;
cm.Seg.FeatureData = pictureInfo.SegmentFeatureData;
cm.Lf.FilterLevel = pictureInfo.LoopFilterLevel;
cm.Lf.SharpnessLevel = pictureInfo.LoopFilterSharpnessLevel;
cm.Lf.ModeRefDeltaEnabled = pictureInfo.ModeRefDeltaEnabled;
cm.Lf.RefDeltas = pictureInfo.RefDeltas;
cm.Lf.ModeDeltas = pictureInfo.ModeDeltas;
@ -105,7 +108,12 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
cm.SetupSegmentationDequant();
cm.SetupScaleFactors();
SetMvs(ref cm, mvsIn);
cm.SetMvs(mvsIn);
if (cm.Lf.FilterLevel != 0 && cm.SkipLoopFilter == 0)
{
LoopFilter.LoopFilterFrameInit(ref cm, cm.Lf.FilterLevel);
}
fixed (byte* dataPtr = bitstream)
{
@ -114,10 +122,27 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
if (maxThreads > 1 && tileRows == 1 && tileCols > 1)
{
DecodeFrame.DecodeTilesMt(ref cm, new ArrayPtr<byte>(dataPtr, bitstream.Length), maxThreads);
LoopFilter.LoopFilterFrameMt(
ref cm.Mb.CurBuf,
ref cm,
ref cm.Mb,
cm.Lf.FilterLevel,
false,
false,
maxThreads);
}
else
{
DecodeFrame.DecodeTiles(ref cm, new ArrayPtr<byte>(dataPtr, bitstream.Length));
LoopFilter.LoopFilterFrame(
ref cm.Mb.CurBuf,
ref cm,
ref cm.Mb,
cm.Lf.FilterLevel,
false,
false);
}
}
catch (InternalErrorException)
@ -126,7 +151,7 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
}
}
GetMvs(ref cm, mvsOut);
cm.GetMvs(mvsOut);
cm.FreeTileWorkerData(_allocator);
cm.FreeContextBuffers(_allocator);
@ -134,48 +159,9 @@ namespace Ryujinx.Graphics.Nvdec.Vp9
return true;
}
private static void SetMvs(ref Vp9Common cm, ReadOnlySpan<Vp9MvRef> mvs)
public void Dispose()
{
if (mvs.Length > cm.PrevFrameMvs.Length)
{
throw new ArgumentException($"Size mismatch, expected: {cm.PrevFrameMvs.Length}, but got: {mvs.Length}.");
}
for (int i = 0; i < mvs.Length; i++)
{
ref var mv = ref cm.PrevFrameMvs[i];
mv.Mv[0].Row = mvs[i].Mvs[0].Row;
mv.Mv[0].Col = mvs[i].Mvs[0].Col;
mv.Mv[1].Row = mvs[i].Mvs[1].Row;
mv.Mv[1].Col = mvs[i].Mvs[1].Col;
mv.RefFrame[0] = (sbyte)mvs[i].RefFrames[0];
mv.RefFrame[1] = (sbyte)mvs[i].RefFrames[1];
}
_allocator.Dispose();
}
private static void GetMvs(ref Vp9Common cm, Span<Vp9MvRef> mvs)
{
if (mvs.Length > cm.CurFrameMvs.Length)
{
throw new ArgumentException($"Size mismatch, expected: {cm.CurFrameMvs.Length}, but got: {mvs.Length}.");
}
for (int i = 0; i < mvs.Length; i++)
{
ref var mv = ref cm.CurFrameMvs[i];
mvs[i].Mvs[0].Row = mv.Mv[0].Row;
mvs[i].Mvs[0].Col = mv.Mv[0].Col;
mvs[i].Mvs[1].Row = mv.Mv[1].Row;
mvs[i].Mvs[1].Col = mv.Mv[1].Col;
mvs[i].RefFrames[0] = mv.RefFrame[0];
mvs[i].RefFrames[1] = mv.RefFrame[1];
}
}
public void Dispose() => _allocator.Dispose();
}
}
}