ryujinx-mirror/Ryujinx.HLE/HOS/Services/SurfaceFlinger/NvFlinger.cs
Thog 644de99e86
Implement GPU syncpoints (#980)
* Implement GPU syncpoints

This adds support for GPU syncpoints on the GPU backend & nvservices.

Everything that was implemented here is based on my researches,
hardware testing of the GM20B and reversing of nvservices (8.1.0).

Thanks to @fincs for the informations about some behaviours of the pusher
and for the initial informations about syncpoints.

* syncpoint: address gdkchan's comments

* Add some missing logic to handle SubmitGpfifo correctly

* Handle the NV event API correctly

* evnt => hostEvent

* Finish addressing gdkchan's comments

* nvservices: write the output buffer even when an error is returned

* dma pusher: Implemnet prefetch barrier

lso fix when the commands should be prefetch.

* Partially fix prefetch barrier

* Add a missing syncpoint check in QueryEvent of NvHostSyncPt

* Address Ac_K's comments and fix GetSyncpoint for ChannelResourcePolicy == Channel

* fix SyncptWait & SyncptWaitEx cmds logic

* Address ripinperi's comments

* Address gdkchan's comments

* Move user event management to the control channel

* Fix mm implementation, nvdec works again

* Address ripinperi's comments

* Address gdkchan's comments

* Implement nvhost-ctrl close accurately + make nvservices dispose channels when stopping the emulator

* Fix typo in MultiMediaOperationType
2020-04-19 11:25:57 +10:00

455 lines
No EOL
14 KiB
C#

using Ryujinx.Common;
using Ryujinx.Common.Logging;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu;
using Ryujinx.HLE.HOS.Kernel.Threading;
using Ryujinx.HLE.HOS.Services.Nv.NvDrvServices.NvMap;
using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.IO;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
using System.Text;
using System.Threading;
using static Ryujinx.HLE.HOS.Services.SurfaceFlinger.Parcel;
namespace Ryujinx.HLE.HOS.Services.SurfaceFlinger
{
class NvFlinger : IDisposable
{
private delegate ResultCode ServiceProcessParcel(ServiceCtx context, BinaryReader parcelReader);
private Dictionary<(string, int), ServiceProcessParcel> _commands;
private KEvent _binderEvent;
private IRenderer _renderer;
private const int BufferQueueCount = 0x40;
private const int BufferQueueMask = BufferQueueCount - 1;
private BufferEntry[] _bufferQueue;
private AutoResetEvent _waitBufferFree;
private bool _disposed;
public NvFlinger(IRenderer renderer, KEvent binderEvent)
{
_commands = new Dictionary<(string, int), ServiceProcessParcel>
{
{ ("android.gui.IGraphicBufferProducer", 0x1), GbpRequestBuffer },
{ ("android.gui.IGraphicBufferProducer", 0x3), GbpDequeueBuffer },
{ ("android.gui.IGraphicBufferProducer", 0x4), GbpDetachBuffer },
{ ("android.gui.IGraphicBufferProducer", 0x7), GbpQueueBuffer },
{ ("android.gui.IGraphicBufferProducer", 0x8), GbpCancelBuffer },
{ ("android.gui.IGraphicBufferProducer", 0x9), GbpQuery },
{ ("android.gui.IGraphicBufferProducer", 0xa), GbpConnect },
{ ("android.gui.IGraphicBufferProducer", 0xb), GbpDisconnect },
{ ("android.gui.IGraphicBufferProducer", 0xe), GbpPreallocBuffer }
};
_renderer = renderer;
_binderEvent = binderEvent;
_bufferQueue = new BufferEntry[0x40];
_waitBufferFree = new AutoResetEvent(false);
}
public ResultCode ProcessParcelRequest(ServiceCtx context, byte[] parcelData, int code)
{
using (MemoryStream ms = new MemoryStream(parcelData))
{
BinaryReader reader = new BinaryReader(ms);
ms.Seek(4, SeekOrigin.Current);
int strSize = reader.ReadInt32();
string interfaceName = Encoding.Unicode.GetString(reader.ReadBytes(strSize * 2));
long remainder = ms.Position & 0xf;
if (remainder != 0)
{
ms.Seek(0x10 - remainder, SeekOrigin.Current);
}
ms.Seek(0x50, SeekOrigin.Begin);
if (_commands.TryGetValue((interfaceName, code), out ServiceProcessParcel procReq))
{
Logger.PrintDebug(LogClass.ServiceVi, $"{interfaceName} {procReq.Method.Name}");
return procReq(context, reader);
}
else
{
throw new NotImplementedException($"{interfaceName} {code}");
}
}
}
private ResultCode GbpRequestBuffer(ServiceCtx context, BinaryReader parcelReader)
{
int slot = parcelReader.ReadInt32();
using (MemoryStream ms = new MemoryStream())
{
BinaryWriter writer = new BinaryWriter(ms);
BufferEntry entry = _bufferQueue[slot];
int bufferCount = 1; //?
long bufferSize = entry.Data.Size;
writer.Write(bufferCount);
writer.Write(bufferSize);
entry.Data.Write(writer);
writer.Write(0);
return MakeReplyParcel(context, ms.ToArray());
}
}
private ResultCode GbpDequeueBuffer(ServiceCtx context, BinaryReader parcelReader)
{
// TODO: Errors.
int async = parcelReader.ReadInt32();
int width = parcelReader.ReadInt32();
int height = parcelReader.ReadInt32();
int format = parcelReader.ReadInt32();
int usage = parcelReader.ReadInt32();
int slot = GetFreeSlotBlocking(width, height);
MultiFence multiFence = MultiFence.NoFence;
using (MemoryStream ms = new MemoryStream())
{
BinaryWriter writer = new BinaryWriter(ms);
// Allocated slot
writer.Write(slot);
// Has multi fence
writer.Write(1);
// Write the multi fnece
WriteFlattenedObject(writer, multiFence);
// Padding
writer.Write(0);
// Status
writer.Write(0);
return MakeReplyParcel(context, ms.ToArray());
}
}
private ResultCode GbpQueueBuffer(ServiceCtx context, BinaryReader parcelReader)
{
context.Device.Statistics.RecordGameFrameTime();
// TODO: Errors.
int slot = parcelReader.ReadInt32();
long Position = parcelReader.BaseStream.Position;
QueueBufferObject queueBufferObject = ReadFlattenedObject<QueueBufferObject>(parcelReader);
parcelReader.BaseStream.Position = Position;
_bufferQueue[slot].Transform = queueBufferObject.Transform;
_bufferQueue[slot].Fence = queueBufferObject.Fence;
_bufferQueue[slot].Crop = queueBufferObject.Crop;
_bufferQueue[slot].State = BufferState.Queued;
SendFrameBuffer(context, slot);
if (context.Device.EnableDeviceVsync)
{
context.Device.VsyncEvent.WaitOne();
}
return MakeReplyParcel(context, 1280, 720, 0, 0, 0);
}
private ResultCode GbpDetachBuffer(ServiceCtx context, BinaryReader parcelReader)
{
return MakeReplyParcel(context, 0);
}
private ResultCode GbpCancelBuffer(ServiceCtx context, BinaryReader parcelReader)
{
// TODO: Errors.
int slot = parcelReader.ReadInt32();
MultiFence fence = ReadFlattenedObject<MultiFence>(parcelReader);
_bufferQueue[slot].State = BufferState.Free;
_waitBufferFree.Set();
return MakeReplyParcel(context, 0);
}
private ResultCode GbpQuery(ServiceCtx context, BinaryReader parcelReader)
{
return MakeReplyParcel(context, 0, 0);
}
private ResultCode GbpConnect(ServiceCtx context, BinaryReader parcelReader)
{
return MakeReplyParcel(context, 1280, 720, 0, 0, 0);
}
private ResultCode GbpDisconnect(ServiceCtx context, BinaryReader parcelReader)
{
return MakeReplyParcel(context, 0);
}
private ResultCode GbpPreallocBuffer(ServiceCtx context, BinaryReader parcelReader)
{
int slot = parcelReader.ReadInt32();
bool hasInput = parcelReader.ReadInt32() == 1;
if (hasInput)
{
byte[] graphicBuffer = ReadFlattenedObject(parcelReader);
_bufferQueue[slot].State = BufferState.Free;
using (BinaryReader graphicBufferReader = new BinaryReader(new MemoryStream(graphicBuffer)))
{
_bufferQueue[slot].Data = new GbpBuffer(graphicBufferReader);
}
}
return MakeReplyParcel(context, 0);
}
private byte[] ReadFlattenedObject(BinaryReader reader)
{
long flattenedObjectSize = reader.ReadInt64();
return reader.ReadBytes((int)flattenedObjectSize);
}
private T ReadFlattenedObject<T>(BinaryReader reader) where T: struct
{
long flattenedObjectSize = reader.ReadInt64();
Debug.Assert(flattenedObjectSize == Unsafe.SizeOf<T>());
return reader.ReadStruct<T>();
}
private unsafe void WriteFlattenedObject<T>(BinaryWriter writer, T value) where T : struct
{
writer.Write(Unsafe.SizeOf<T>());
writer.WriteStruct(value);
}
private ResultCode MakeReplyParcel(ServiceCtx context, params int[] ints)
{
using (MemoryStream ms = new MemoryStream())
{
BinaryWriter writer = new BinaryWriter(ms);
foreach (int Int in ints)
{
writer.Write(Int);
}
return MakeReplyParcel(context, ms.ToArray());
}
}
private ResultCode MakeReplyParcel(ServiceCtx context, byte[] data)
{
(long replyPos, long replySize) = context.Request.GetBufferType0x22();
byte[] reply = MakeParcel(data, new byte[0]);
context.Memory.WriteBytes(replyPos, reply);
return ResultCode.Success;
}
private Format ConvertColorFormat(ColorFormat colorFormat)
{
switch (colorFormat)
{
case ColorFormat.A8B8G8R8:
return Format.R8G8B8A8Unorm;
case ColorFormat.X8B8G8R8:
return Format.R8G8B8A8Unorm;
case ColorFormat.R5G6B5:
return Format.B5G6R5Unorm;
case ColorFormat.A8R8G8B8:
return Format.B8G8R8A8Unorm;
case ColorFormat.A4B4G4R4:
return Format.R4G4B4A4Unorm;
default:
throw new NotImplementedException($"Color Format \"{colorFormat}\" not implemented!");
}
}
// TODO: support multi surface
private void SendFrameBuffer(ServiceCtx context, int slot)
{
int fbWidth = _bufferQueue[slot].Data.Header.Width;
int fbHeight = _bufferQueue[slot].Data.Header.Height;
int nvMapHandle = _bufferQueue[slot].Data.Buffer.Surfaces[0].NvMapHandle;
if (nvMapHandle == 0)
{
nvMapHandle = _bufferQueue[slot].Data.Buffer.NvMapId;
}
int bufferOffset = _bufferQueue[slot].Data.Buffer.Surfaces[0].Offset;
NvMapHandle map = NvMapDeviceFile.GetMapFromHandle(context.Process, nvMapHandle);
ulong fbAddr = (ulong)(map.Address + bufferOffset);
_bufferQueue[slot].State = BufferState.Acquired;
Format format = ConvertColorFormat(_bufferQueue[slot].Data.Buffer.Surfaces[0].ColorFormat);
int bytesPerPixel =
format == Format.B5G6R5Unorm ||
format == Format.R4G4B4A4Unorm ? 2 : 4;
int gobBlocksInY = 1 << _bufferQueue[slot].Data.Buffer.Surfaces[0].BlockHeightLog2;
// Note: Rotation is being ignored.
Rect cropRect = _bufferQueue[slot].Crop;
bool flipX = _bufferQueue[slot].Transform.HasFlag(HalTransform.FlipX);
bool flipY = _bufferQueue[slot].Transform.HasFlag(HalTransform.FlipY);
ImageCrop crop = new ImageCrop(
cropRect.Left,
cropRect.Right,
cropRect.Top,
cropRect.Bottom,
flipX,
flipY);
context.Device.Gpu.Window.EnqueueFrameThreadSafe(
fbAddr,
fbWidth,
fbHeight,
0,
false,
gobBlocksInY,
format,
bytesPerPixel,
crop,
AcquireBuffer,
ReleaseBuffer,
slot);
}
private void AcquireBuffer(GpuContext context, object slot)
{
AcquireBuffer(context, (int)slot);
}
private void AcquireBuffer(GpuContext context, int slot)
{
_bufferQueue[slot].Fence.WaitForever(context);
}
private void ReleaseBuffer(object slot)
{
ReleaseBuffer((int)slot);
}
private void ReleaseBuffer(int slot)
{
_bufferQueue[slot].State = BufferState.Free;
_binderEvent.ReadableEvent.Signal();
_waitBufferFree.Set();
}
private int GetFreeSlotBlocking(int width, int height)
{
int slot;
do
{
if ((slot = GetFreeSlot(width, height)) != -1)
{
break;
}
if (_disposed)
{
break;
}
_waitBufferFree.WaitOne();
}
while (!_disposed);
return slot;
}
private int GetFreeSlot(int width, int height)
{
lock (_bufferQueue)
{
for (int slot = 0; slot < _bufferQueue.Length; slot++)
{
if (_bufferQueue[slot].State != BufferState.Free)
{
continue;
}
GbpBuffer data = _bufferQueue[slot].Data;
if (data.Header.Width == width &&
data.Header.Height == height)
{
_bufferQueue[slot].State = BufferState.Dequeued;
return slot;
}
}
}
return -1;
}
public void Dispose()
{
Dispose(true);
}
protected virtual void Dispose(bool disposing)
{
if (disposing && !_disposed)
{
_disposed = true;
_waitBufferFree.Set();
_waitBufferFree.Dispose();
}
}
}
}