New NVDEC and VIC implementation (#1384)

* 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>

Ryujinx.Graphics.Host1x/ClassId.cs

@@ -0,0 +1,20 @@
+namespace Ryujinx.Graphics.Host1x
+{
+    public enum ClassId
+    {
+        Host1x = 0x1,
+        Mpeg = 0x20,
+        Nvenc = 0x21,
+        Vi = 0x30,
+        Isp = 0x32,
+        Ispb = 0x34,
+        Vii2c = 0x36,
+        Vic = 0x5d,
+        Gr3d = 0x60,
+        Gpu = 0x61,
+        Tsec = 0xe0,
+        Tsecb = 0xe1,
+        Nvjpg = 0xc0,
+        Nvdec = 0xf0
+    }
+}

Ryujinx.Graphics.Host1x/Devices.cs

@@ -0,0 +1,32 @@
+using Ryujinx.Graphics.Device;
+using System;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    class Devices : IDisposable
+    {
+        private readonly Dictionary<ClassId, IDeviceState> _devices = new Dictionary<ClassId, IDeviceState>();
+
+        public void RegisterDevice(ClassId classId, IDeviceState device)
+        {
+            _devices[classId] = device;
+        }
+
+        public IDeviceState GetDevice(ClassId classId)
+        {
+            return _devices.TryGetValue(classId, out IDeviceState device) ? device : null;
+        }
+
+        public void Dispose()
+        {
+            foreach (var device in _devices.Values)
+            {
+                if (device is ThiDevice thi)
+                {
+                    thi.Dispose();
+                }
+            }
+        }
+    }
+}

Ryujinx.Graphics.Host1x/Host1xClass.cs

@@ -0,0 +1,33 @@
+using Ryujinx.Graphics.Device;
+using Ryujinx.Graphics.Gpu.Synchronization;
+using System.Collections.Generic;
+using System.Threading;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    public class Host1xClass : IDeviceState
+    {
+        private readonly SynchronizationManager _syncMgr;
+        private readonly DeviceState<Host1xClassRegisters> _state;
+
+        public Host1xClass(SynchronizationManager syncMgr)
+        {
+            _syncMgr = syncMgr;
+            _state = new DeviceState<Host1xClassRegisters>(new Dictionary<string, RwCallback>
+            {
+                { nameof(Host1xClassRegisters.WaitSyncpt32), new RwCallback(WaitSyncpt32, null) }
+            });
+        }
+
+        public int Read(int offset) => _state.Read(offset);
+        public void Write(int offset, int data) => _state.Write(offset, data);
+
+        private void WaitSyncpt32(int data)
+        {
+            uint syncpointId = (uint)(data & 0xFF);
+            uint threshold = _state.State.LoadSyncptPayload32;
+
+            _syncMgr.WaitOnSyncpoint(syncpointId, threshold, Timeout.InfiniteTimeSpan);
+        }
+    }
+}

Ryujinx.Graphics.Host1x/Host1xClassRegisters.cs

@@ -0,0 +1,41 @@
+using Ryujinx.Common.Memory;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    struct Host1xClassRegisters
+    {
+        public uint IncrSyncpt;
+        public uint IncrSyncptCntrl;
+        public uint IncrSyncptError;
+        public Array5<uint> ReservedC;
+        public uint WaitSyncpt;
+        public uint WaitSyncptBase;
+        public uint WaitSyncptIncr;
+        public uint LoadSyncptBase;
+        public uint IncrSyncptBase;
+        public uint Clear;
+        public uint Wait;
+        public uint WaitWithIntr;
+        public uint DelayUsec;
+        public uint TickcountHi;
+        public uint TickcountLo;
+        public uint Tickctrl;
+        public Array23<uint> Reserved50;
+        public uint Indctrl;
+        public uint Indoff2;
+        public uint Indoff;
+        public Array31<uint> Inddata;
+        public uint Reserved134;
+        public uint LoadSyncptPayload32;
+        public uint Stallctrl;
+        public uint WaitSyncpt32;
+        public uint WaitSyncptBase32;
+        public uint LoadSyncptBase32;
+        public uint IncrSyncptBase32;
+        public uint StallcountHi;
+        public uint StallcountLo;
+        public uint Xrefctrl;
+        public uint ChannelXrefHi;
+        public uint ChannelXrefLo;
+    }
+}

Ryujinx.Graphics.Host1x/Host1xDevice.cs

@@ -0,0 +1,123 @@
+using Ryujinx.Common;
+using Ryujinx.Common.Logging;
+using Ryujinx.Graphics.Device;
+using Ryujinx.Graphics.Gpu.Synchronization;
+using System;
+using System.Numerics;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    public sealed class Host1xDevice : IDisposable
+    {
+        private readonly SyncptIncrManager _syncptIncrMgr;
+        private readonly AsyncWorkQueue<int[]> _commandQueue;
+
+        private readonly Devices _devices = new Devices();
+
+        public Host1xClass Class { get; }
+
+        private IDeviceState _device;
+
+        private int _count;
+        private int _offset;
+        private int _mask;
+        private bool _incrementing;
+
+        public Host1xDevice(SynchronizationManager syncMgr)
+        {
+            _syncptIncrMgr = new SyncptIncrManager(syncMgr);
+            _commandQueue = new AsyncWorkQueue<int[]>(Process, "Ryujinx.Host1xProcessor");
+
+            Class = new Host1xClass(syncMgr);
+
+            _devices.RegisterDevice(ClassId.Host1x, Class);
+        }
+
+        public void RegisterDevice(ClassId classId, IDeviceState device)
+        {
+            var thi = new ThiDevice(classId, device ?? throw new ArgumentNullException(nameof(device)), _syncptIncrMgr);
+            _devices.RegisterDevice(classId, thi);
+        }
+
+        public void Submit(ReadOnlySpan<int> commandBuffer)
+        {
+            _commandQueue.Add(commandBuffer.ToArray());
+        }
+
+        private void Process(int[] commandBuffer)
+        {
+            for (int index = 0; index < commandBuffer.Length; index++)
+            {
+                Step(commandBuffer[index]);
+            }
+        }
+
+        private void Step(int value)
+        {
+            if (_mask != 0)
+            {
+                int lbs = BitOperations.TrailingZeroCount(_mask);
+
+                _mask &= ~(1 << lbs);
+
+                DeviceWrite(_offset + lbs, value);
+
+                return;
+            }
+            else if (_count != 0)
+            {
+                _count--;
+
+                DeviceWrite(_offset, value);
+
+                if (_incrementing)
+                {
+                    _offset++;
+                }
+
+                return;
+            }
+
+            OpCode opCode = (OpCode)((value >> 28) & 0xf);
+
+            switch (opCode)
+            {
+                case OpCode.SetClass:
+                    _mask = value & 0x3f;
+                    ClassId classId = (ClassId)((value >> 6) & 0x3ff);
+                    _offset = (value >> 16) & 0xfff;
+                    _device = _devices.GetDevice(classId);
+                    break;
+                case OpCode.Incr:
+                case OpCode.NonIncr:
+                    _count = value & 0xffff;
+                    _offset = (value >> 16) & 0xfff;
+                    _incrementing = opCode == OpCode.Incr;
+                    break;
+                case OpCode.Mask:
+                    _mask = value & 0xffff;
+                    _offset = (value >> 16) & 0xfff;
+                    break;
+                case OpCode.Imm:
+                    int data = value & 0xfff;
+                    _offset = (value >> 16) & 0xfff;
+                    DeviceWrite(_offset, data);
+                    break;
+                default:
+                    Logger.PrintError(LogClass.Host1x, $"Unsupported opcode \"{opCode}\".");
+                    break;
+            }
+        }
+
+        private void DeviceWrite(int offset, int data)
+        {
+            _device?.Write(offset * 4, data);
+        }
+
+        public void Dispose()
+        {
+            _commandQueue.Dispose();
+            _devices.Dispose();
+        }
+    }
+}

Ryujinx.Graphics.Host1x/OpCode.cs

@@ -0,0 +1,21 @@
+namespace Ryujinx.Graphics.Host1x
+{
+    enum OpCode
+    {
+        SetClass,
+        Incr,
+        NonIncr,
+        Mask,
+        Imm,
+        Restart,
+        Gather,
+        SetStrmId,
+        SetAppId,
+        SetPyld,
+        IncrW,
+        NonIncrW,
+        GatherW,
+        RestartW,
+        Extend
+    }
+}

Ryujinx.Graphics.Host1x/Ryujinx.Graphics.Host1x.csproj

@@ -0,0 +1,20 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <TargetFramework>netcoreapp3.1</TargetFramework>
+  </PropertyGroup>
+
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|AnyCPU'">
+    <AllowUnsafeBlocks>false</AllowUnsafeBlocks>
+  </PropertyGroup>
+
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|AnyCPU'">
+    <AllowUnsafeBlocks>false</AllowUnsafeBlocks>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\Ryujinx.Graphics.Device\Ryujinx.Graphics.Device.csproj" />
+    <ProjectReference Include="..\Ryujinx.Graphics.Gpu\Ryujinx.Graphics.Gpu.csproj" />
+  </ItemGroup>
+
+</Project>

Ryujinx.Graphics.Host1x/SyncptIncrManager.cs

@@ -0,0 +1,99 @@
+using Ryujinx.Graphics.Gpu.Synchronization;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    class SyncptIncrManager
+    {
+        private readonly SynchronizationManager _syncMgr;
+
+        private struct SyncptIncr
+        {
+            public uint Id { get; }
+            public ClassId ClassId { get; }
+            public uint SyncptId { get; }
+            public bool Done { get; }
+
+            public SyncptIncr(uint id, ClassId classId, uint syncptId, bool done = false)
+            {
+                Id = id;
+                ClassId = classId;
+                SyncptId = syncptId;
+                Done = done;
+            }
+        }
+
+        private readonly List<SyncptIncr> _incrs = new List<SyncptIncr>();
+
+        private uint _currentId;
+
+        public SyncptIncrManager(SynchronizationManager syncMgr)
+        {
+            _syncMgr = syncMgr;
+        }
+
+        public void Increment(uint id)
+        {
+            lock (_incrs)
+            {
+                _incrs.Add(new SyncptIncr(0, 0, id, true));
+
+                IncrementAllDone();
+            }
+        }
+
+        public uint IncrementWhenDone(ClassId classId, uint id)
+        {
+            lock (_incrs)
+            {
+                uint handle = _currentId++;
+
+                _incrs.Add(new SyncptIncr(handle, classId, id));
+
+                return handle;
+            }
+        }
+
+        public void SignalDone(uint handle)
+        {
+            lock (_incrs)
+            {
+                // Set pending increment with the given handle to "done".
+                for (int i = 0; i < _incrs.Count; i++)
+                {
+                    SyncptIncr incr = _incrs[i];
+
+                    if (_incrs[i].Id == handle)
+                    {
+                        _incrs[i] = new SyncptIncr(incr.Id, incr.ClassId, incr.SyncptId, true);
+
+                        break;
+                    }
+                }
+
+                IncrementAllDone();
+            }
+        }
+
+        private void IncrementAllDone()
+        {
+            lock (_incrs)
+            {
+                // Increment all sequential pending increments that are already done.
+                int doneCount = 0;
+
+                for (; doneCount < _incrs.Count; doneCount++)
+                {
+                    if (!_incrs[doneCount].Done)
+                    {
+                        break;
+                    }
+
+                    _syncMgr.IncrementSyncpoint(_incrs[doneCount].SyncptId);
+                }
+
+                _incrs.RemoveRange(0, doneCount);
+            }
+        }
+    }
+}

Ryujinx.Graphics.Host1x/ThiDevice.cs

@@ -0,0 +1,96 @@
+using Ryujinx.Common;
+using Ryujinx.Graphics.Device;
+using System;
+using System.Collections.Generic;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    class ThiDevice : IDeviceState, IDisposable
+    {
+        private readonly ClassId _classId;
+        private readonly IDeviceState _device;
+
+        private readonly SyncptIncrManager _syncptIncrMgr;
+
+        private class CommandAction
+        {
+            public int Data { get; }
+
+            public CommandAction(int data)
+            {
+                Data = data;
+            }
+        }
+
+        private class MethodCallAction : CommandAction
+        {
+            public int Method { get; }
+
+            public MethodCallAction(int method, int data) : base(data)
+            {
+                Method = method;
+            }
+        }
+
+        private class SyncptIncrAction : CommandAction
+        {
+            public SyncptIncrAction(uint syncptIncrHandle) : base((int)syncptIncrHandle)
+            {
+            }
+        }
+
+        private readonly AsyncWorkQueue<CommandAction> _commandQueue;
+
+        private readonly DeviceState<ThiRegisters> _state;
+
+        public ThiDevice(ClassId classId, IDeviceState device, SyncptIncrManager syncptIncrMgr)
+        {
+            _classId = classId;
+            _device = device;
+            _syncptIncrMgr = syncptIncrMgr;
+            _commandQueue = new AsyncWorkQueue<CommandAction>(Process, $"Ryujinx.{classId}Processor");
+            _state = new DeviceState<ThiRegisters>(new Dictionary<string, RwCallback>
+            {
+                { nameof(ThiRegisters.IncrSyncpt), new RwCallback(IncrSyncpt, null) },
+                { nameof(ThiRegisters.Method1), new RwCallback(Method1, null) }
+            });
+        }
+
+        public int Read(int offset) => _state.Read(offset);
+        public void Write(int offset, int data) => _state.Write(offset, data);
+
+        private void IncrSyncpt(int data)
+        {
+            uint syncpointId = (uint)(data & 0xFF);
+            uint cond = (uint)((data >> 8) & 0xFF); // 0 = Immediate, 1 = Done
+
+            if (cond == 0)
+            {
+                _syncptIncrMgr.Increment(syncpointId);
+            }
+            else
+            {
+                _commandQueue.Add(new SyncptIncrAction(_syncptIncrMgr.IncrementWhenDone(_classId, syncpointId)));
+            }
+        }
+
+        private void Method1(int data)
+        {
+            _commandQueue.Add(new MethodCallAction((int)_state.State.Method0 * 4, data));
+        }
+
+        private void Process(CommandAction cmdAction)
+        {
+            if (cmdAction is SyncptIncrAction syncptIncrAction)
+            {
+                _syncptIncrMgr.SignalDone((uint)syncptIncrAction.Data);
+            }
+            else if (cmdAction is MethodCallAction methodCallAction)
+            {
+                _device.Write(methodCallAction.Method, methodCallAction.Data);
+            }
+        }
+
+        public void Dispose() => _commandQueue.Dispose();
+    }
+}

Ryujinx.Graphics.Host1x/ThiRegisters.cs

@@ -0,0 +1,22 @@
+using Ryujinx.Common.Memory;
+
+namespace Ryujinx.Graphics.Host1x
+{
+    struct ThiRegisters
+    {
+        public uint IncrSyncpt;
+        public uint Reserved4;
+        public uint IncrSyncptErr;
+        public uint CtxswIncrSyncpt;
+        public Array4<uint> Reserved10;
+        public uint Ctxsw;
+        public uint Reserved24;
+        public uint ContSyncptEof;
+        public Array5<uint> Reserved2C;
+        public uint Method0;
+        public uint Method1;
+        public Array12<uint> Reserved48;
+        public uint IntStatus;
+        public uint IntMask;
+    }
+}