Add support for rendering raw frame buffer (again), remove shader optimization pass as its not worth the effort to maintain it currently

This commit is contained in:
gdkchan 2018-04-13 03:53:40 -03:00
parent f39ced70c1
commit 120fa26392
17 changed files with 189 additions and 793 deletions

View file

@ -2,6 +2,7 @@ using ChocolArm64.Memory;
using Ryujinx.Core.OsHle.Handles;
using Ryujinx.Core.OsHle.Services.Nv;
using Ryujinx.Graphics.Gal;
using Ryujinx.Graphics.Gpu;
using System;
using System.Collections.Generic;
using System.IO;
@ -63,13 +64,7 @@ namespace Ryujinx.Core.OsHle.Services.Android
private ManualResetEvent WaitBufferFree;
private object RenderQueueLock;
private int RenderQueueCount;
private bool NvFlingerDisposed;
private bool KeepRunning;
private bool Disposed;
public NvFlinger(IGalRenderer Renderer, KEvent ReleaseEvent)
{
@ -92,10 +87,6 @@ namespace Ryujinx.Core.OsHle.Services.Android
BufferQueue = new BufferEntry[0x40];
WaitBufferFree = new ManualResetEvent(false);
RenderQueueLock = new object();
KeepRunning = true;
}
public long ProcessParcelRequest(ServiceCtx Context, byte[] ParcelData, int Code)
@ -296,8 +287,6 @@ namespace Ryujinx.Core.OsHle.Services.Android
NvMapFb MapFb = (NvMapFb)INvDrvServices.NvMapsFb.GetData(Context.Process, 0);
Renderer.SetFrameBuffer(Map.GpuAddress);
long Address = Map.CpuAddress;
if (MapFb.HasBufferOffset(Slot))
@ -369,41 +358,24 @@ namespace Ryujinx.Core.OsHle.Services.Android
Rotate = -MathF.PI * 0.5f;
}
lock (RenderQueueLock)
if (Context.Ns.Gpu.Engine3d.IsFrameBufferPosition(Map.GpuAddress))
{
if (NvFlingerDisposed)
{
return;
}
//Frame buffer is rendered to by the GPU, we can just
//bind the frame buffer texture, it's not necessary to read anything.
Renderer.SetFrameBuffer(Map.GpuAddress);
}
else
{
//Frame buffer is not set on the GPU registers, in this case
//assume that the app is manually writing to it.
Texture Texture = new Texture(Address, FbWidth, FbHeight);
Interlocked.Increment(ref RenderQueueCount);
byte[] Data = TextureReader.Read(Context.Memory, Texture);
Renderer.SetFrameBuffer(Data, FbWidth, FbHeight);
}
byte* Fb = (byte*)Context.Memory.Ram + Address;
Context.Ns.Gpu.Renderer.QueueAction(delegate()
{
Context.Ns.Gpu.Renderer.SetFrameBuffer(
Fb,
FbWidth,
FbHeight,
ScaleX,
ScaleY,
OffsX,
OffsY,
Rotate);
BufferQueue[Slot].State = BufferState.Free;
Interlocked.Decrement(ref RenderQueueCount);
ReleaseEvent.Handle.Set();
lock (WaitBufferFree)
{
WaitBufferFree.Set();
}
});
Context.Ns.Gpu.Renderer.QueueAction(() => ReleaseBuffer(Slot));
}
private NvMap GetNvMap(ServiceCtx Context, int Slot)
@ -422,6 +394,18 @@ namespace Ryujinx.Core.OsHle.Services.Android
return INvDrvServices.NvMaps.GetData<NvMap>(Context.Process, NvMapHandle);
}
private void ReleaseBuffer(int Slot)
{
BufferQueue[Slot].State = BufferState.Free;
ReleaseEvent.Handle.Set();
lock (WaitBufferFree)
{
WaitBufferFree.Set();
}
}
private int GetFreeSlotBlocking(int Width, int Height)
{
int Slot;
@ -437,7 +421,7 @@ namespace Ryujinx.Core.OsHle.Services.Android
Logging.Debug("Waiting for a free BufferQueue slot...");
if (!KeepRunning)
if (Disposed)
{
break;
}
@ -447,7 +431,7 @@ namespace Ryujinx.Core.OsHle.Services.Android
WaitBufferFree.WaitOne();
}
while (KeepRunning);
while (!Disposed);
Logging.Debug($"Found free BufferQueue slot {Slot}!");
@ -487,26 +471,12 @@ namespace Ryujinx.Core.OsHle.Services.Android
protected virtual void Dispose(bool Disposing)
{
if (Disposing && !NvFlingerDisposed)
if (Disposing && !Disposed)
{
lock (RenderQueueLock)
{
NvFlingerDisposed = true;
}
//Ensure that all pending actions was sent before
//we can safely assume that the class was disposed.
while (RenderQueueCount > 0)
{
Thread.Yield();
}
Renderer.ResetFrameBuffer();
Disposed = true;
lock (WaitBufferFree)
{
KeepRunning = false;
WaitBufferFree.Set();
}

View file

@ -6,21 +6,12 @@ namespace Ryujinx.Graphics.Gal
public unsafe interface IGalRenderer
{
void QueueAction(Action ActionMthd);
void RunActions();
void InitializeFrameBuffer();
void ResetFrameBuffer();
void Render();
void SetWindowSize(int Width, int Height);
void SetFrameBuffer(
byte* Fb,
int Width,
int Height,
float ScaleX,
float ScaleY,
float OffsX,
float OffsY,
float Rotate);
//Blend
void SetBlendEnable(bool Enable);
@ -43,10 +34,12 @@ namespace Ryujinx.Graphics.Gal
void BindFrameBuffer(long Tag);
void BindFrameBufferTexture(long Tag, int Index);
void BindFrameBufferTexture(long Tag, int Index, GalTextureSampler Sampler);
void SetFrameBuffer(long Tag);
void SetFrameBuffer(byte[] Data, int Width, int Height);
//Rasterizer
void ClearBuffers(int RtIndex, GalClearBufferFlags Flags);
@ -72,10 +65,8 @@ namespace Ryujinx.Graphics.Gal
void BindProgram();
//Texture
void SetTexture(int Index, GalTexture Tex);
void SetTextureAndSampler(int Index, GalTexture Texture, GalTextureSampler Sampler);
void BindTexture(int Index);
void SetSampler(GalTextureSampler Sampler);
}
}

View file

@ -1,250 +0,0 @@
using OpenTK;
using OpenTK.Graphics.OpenGL;
using System;
namespace Ryujinx.Graphics.Gal.OpenGL
{
unsafe class FrameBuffer
{
public int WindowWidth { get; set; }
public int WindowHeight { get; set; }
private int VtxShaderHandle;
private int FragShaderHandle;
private int PrgShaderHandle;
private int TexHandle;
private int TexWidth;
private int TexHeight;
private int VaoHandle;
private int VboHandle;
private int[] Pixels;
private byte* FbPtr;
private object FbPtrLock;
public FrameBuffer(int Width, int Height)
{
if (Width < 0)
{
throw new ArgumentOutOfRangeException(nameof(Width));
}
if (Height < 0)
{
throw new ArgumentOutOfRangeException(nameof(Height));
}
FbPtrLock = new object();
TexWidth = Width;
TexHeight = Height;
WindowWidth = Width;
WindowHeight = Height;
SetupShaders();
SetupTexture();
SetupVertex();
}
private void SetupShaders()
{
VtxShaderHandle = GL.CreateShader(ShaderType.VertexShader);
FragShaderHandle = GL.CreateShader(ShaderType.FragmentShader);
string VtxShaderSource = EmbeddedResource.GetString("GlFbVtxShader");
string FragShaderSource = EmbeddedResource.GetString("GlFbFragShader");
GL.ShaderSource(VtxShaderHandle, VtxShaderSource);
GL.ShaderSource(FragShaderHandle, FragShaderSource);
GL.CompileShader(VtxShaderHandle);
GL.CompileShader(FragShaderHandle);
PrgShaderHandle = GL.CreateProgram();
GL.AttachShader(PrgShaderHandle, VtxShaderHandle);
GL.AttachShader(PrgShaderHandle, FragShaderHandle);
GL.LinkProgram(PrgShaderHandle);
GL.UseProgram(PrgShaderHandle);
int TexUniformLocation = GL.GetUniformLocation(PrgShaderHandle, "tex");
GL.Uniform1(TexUniformLocation, 0);
int WindowSizeUniformLocation = GL.GetUniformLocation(PrgShaderHandle, "window_size");
GL.Uniform2(WindowSizeUniformLocation, new Vector2(1280.0f, 720.0f));
}
private void SetupTexture()
{
Pixels = new int[TexWidth * TexHeight];
if (TexHandle == 0)
{
TexHandle = GL.GenTexture();
}
GL.BindTexture(TextureTarget.Texture2D, TexHandle);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMinFilter, (int)TextureMinFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMagFilter, (int)TextureMagFilter.Linear);
GL.TexImage2D(TextureTarget.Texture2D,
0,
PixelInternalFormat.Rgba,
TexWidth,
TexHeight,
0,
PixelFormat.Rgba,
PixelType.UnsignedByte,
IntPtr.Zero);
}
private void SetupVertex()
{
VaoHandle = GL.GenVertexArray();
VboHandle = GL.GenBuffer();
float[] Buffer = new float[]
{
-1, 1, 0, 0,
1, 1, 1, 0,
-1, -1, 0, 1,
1, -1, 1, 1
};
IntPtr Length = new IntPtr(Buffer.Length * 4);
GL.BindBuffer(BufferTarget.ArrayBuffer, VboHandle);
GL.BufferData(BufferTarget.ArrayBuffer, Length, Buffer, BufferUsageHint.StreamDraw);
GL.BindBuffer(BufferTarget.ArrayBuffer, 0);
GL.BindVertexArray(VaoHandle);
GL.EnableVertexAttribArray(0);
GL.BindBuffer(BufferTarget.ArrayBuffer, VboHandle);
GL.VertexAttribPointer(0, 2, VertexAttribPointerType.Float, false, 16, 0);
GL.EnableVertexAttribArray(1);
GL.BindBuffer(BufferTarget.ArrayBuffer, VboHandle);
GL.VertexAttribPointer(1, 2, VertexAttribPointerType.Float, false, 16, 8);
GL.BindVertexArray(0);
}
public unsafe void Set(byte* Fb, int Width, int Height, Matrix2 Transform, Vector2 Offs)
{
if (Fb == null)
{
throw new ArgumentNullException(nameof(Fb));
}
if (Width < 0)
{
throw new ArgumentOutOfRangeException(nameof(Width));
}
if (Height < 0)
{
throw new ArgumentOutOfRangeException(nameof(Height));
}
lock (FbPtrLock)
{
FbPtr = Fb;
}
if (Width != TexWidth ||
Height != TexHeight)
{
TexWidth = Width;
TexHeight = Height;
SetupTexture();
}
GL.UseProgram(PrgShaderHandle);
int TransformUniformLocation = GL.GetUniformLocation(PrgShaderHandle, "transform");
GL.UniformMatrix2(TransformUniformLocation, false, ref Transform);
int WindowSizeUniformLocation = GL.GetUniformLocation(PrgShaderHandle, "window_size");
GL.Uniform2(WindowSizeUniformLocation, new Vector2(WindowWidth, WindowHeight));
int OffsetUniformLocation = GL.GetUniformLocation(PrgShaderHandle, "offset");
GL.Uniform2(OffsetUniformLocation, Offs);
}
public void Reset()
{
lock (FbPtrLock)
{
FbPtr = null;
}
}
public void Render()
{
lock (FbPtrLock)
{
if (FbPtr == null)
{
return;
}
for (int Y = 0; Y < TexHeight; Y++)
for (int X = 0; X < TexWidth; X++)
{
Pixels[X + Y * TexWidth] = *((int*)(FbPtr + GetSwizzleOffset(X, Y)));
}
}
GL.BindTexture(TextureTarget.Texture2D, TexHandle);
GL.TexSubImage2D(TextureTarget.Texture2D,
0,
0,
0,
TexWidth,
TexHeight,
PixelFormat.Rgba,
PixelType.UnsignedByte,
Pixels);
GL.ActiveTexture(TextureUnit.Texture0);
GL.BindVertexArray(VaoHandle);
GL.UseProgram(PrgShaderHandle);
GL.DrawArrays(PrimitiveType.TriangleStrip, 0, 4);
}
private int GetSwizzleOffset(int X, int Y)
{
int Pos;
Pos = (Y & 0x7f) >> 4;
Pos += (X >> 4) << 3;
Pos += (Y >> 7) * ((TexWidth >> 4) << 3);
Pos *= 1024;
Pos += ((Y & 0xf) >> 3) << 9;
Pos += ((X & 0xf) >> 3) << 8;
Pos += ((Y & 0x7) >> 1) << 6;
Pos += ((X & 0x7) >> 2) << 5;
Pos += ((Y & 0x1) >> 0) << 4;
Pos += ((X & 0x3) >> 0) << 2;
return Pos;
}
}
}

View file

@ -23,13 +23,13 @@ namespace Ryujinx.Graphics.Gal.OpenGL
private Dictionary<long, FrameBuffer> Fbs;
private FrameBuffer CurrentFb;
private ShaderProgram Shader;
private bool IsInitialized;
private int FbHandle;
private int CurrFbHandle;
private int CurrTexHandle;
private int RawFbTexHandle;
private int VaoHandle;
private int VboHandle;
@ -47,12 +47,17 @@ namespace Ryujinx.Graphics.Gal.OpenGL
return;
}
Width = 1280;
Height = 720;
FrameBuffer Fb = new FrameBuffer();
Fb.Handle = GL.GenFramebuffer();
Fb.Handle = GL.GenFramebuffer();
Fb.RbHandle = GL.GenRenderbuffer();
Fb.TexHandle = GL.GenTexture();
SetupTexture(Fb.TexHandle, Width, Height);
GL.BindFramebuffer(FramebufferTarget.Framebuffer, Fb.Handle);
GL.BindRenderbuffer(RenderbufferTarget.Renderbuffer, Fb.RbHandle);
@ -60,8 +65,8 @@ namespace Ryujinx.Graphics.Gal.OpenGL
GL.RenderbufferStorage(
RenderbufferTarget.Renderbuffer,
RenderbufferStorage.Depth24Stencil8,
1280,
720);
Width,
Height);
GL.FramebufferRenderbuffer(
FramebufferTarget.Framebuffer,
@ -69,14 +74,11 @@ namespace Ryujinx.Graphics.Gal.OpenGL
RenderbufferTarget.Renderbuffer,
Fb.RbHandle);
GL.BindTexture(TextureTarget.Texture2D, Fb.TexHandle);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMinFilter, (int)TextureMinFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMagFilter, (int)TextureMagFilter.Linear);
GL.TexImage2D(TextureTarget.Texture2D, 0, PixelInternalFormat.Rgba, 1280, 720, 0, PixelFormat.Rgba, PixelType.UnsignedByte, IntPtr.Zero);
GL.FramebufferTexture(FramebufferTarget.Framebuffer, FramebufferAttachment.ColorAttachment0, Fb.TexHandle, 0);
GL.FramebufferTexture(
FramebufferTarget.Framebuffer,
FramebufferAttachment.ColorAttachment0,
Fb.TexHandle,
0);
GL.DrawBuffer(DrawBufferMode.ColorAttachment0);
@ -89,7 +91,7 @@ namespace Ryujinx.Graphics.Gal.OpenGL
{
GL.BindFramebuffer(FramebufferTarget.Framebuffer, Fb.Handle);
FbHandle = Fb.Handle;
CurrFbHandle = Fb.Handle;
}
}
@ -107,24 +109,39 @@ namespace Ryujinx.Graphics.Gal.OpenGL
{
if (Fbs.TryGetValue(Tag, out FrameBuffer Fb))
{
CurrentFb = Fb;
CurrTexHandle = Fb.TexHandle;
}
}
public void Set(byte[] Data, int Width, int Height)
{
EnsureInitialized();
GL.ActiveTexture(TextureUnit.Texture0);
GL.BindTexture(TextureTarget.Texture2D, RawFbTexHandle);
(PixelFormat Format, PixelType Type) = OGLEnumConverter.GetTextureFormat(GalTextureFormat.A8B8G8R8);
GL.TexSubImage2D(TextureTarget.Texture2D, 0, 0, 0, Width, Height, Format, Type, Data);
CurrTexHandle = RawFbTexHandle;
}
public void Render()
{
if (CurrentFb.TexHandle != 0)
if (CurrTexHandle != 0)
{
EnsureInitialized();
GL.ActiveTexture(TextureUnit.Texture0);
GL.BindTexture(TextureTarget.Texture2D, CurrTexHandle);
int CurrentProgram = GL.GetInteger(GetPName.CurrentProgram);
GL.BindFramebuffer(FramebufferTarget.Framebuffer, 0);
GL.ActiveTexture(TextureUnit.Texture0);
GL.BindTexture(TextureTarget.Texture2D, CurrentFb.TexHandle);
GL.BindVertexArray(VaoHandle);
GL.UseProgram(Shader.Handle);
@ -132,7 +149,7 @@ namespace Ryujinx.Graphics.Gal.OpenGL
GL.DrawArrays(PrimitiveType.TriangleStrip, 0, 4);
//Restore the original state.
GL.BindFramebuffer(FramebufferTarget.Framebuffer, FbHandle);
GL.BindFramebuffer(FramebufferTarget.Framebuffer, CurrFbHandle);
GL.UseProgram(CurrentProgram);
}
@ -146,6 +163,10 @@ namespace Ryujinx.Graphics.Gal.OpenGL
SetupShader();
SetupVertex();
RawFbTexHandle = GL.GenTexture();
SetupTexture(RawFbTexHandle, 1280, 720);
}
}
@ -217,5 +238,32 @@ namespace Ryujinx.Graphics.Gal.OpenGL
GL.VertexAttribPointer(1, 2, VertexAttribPointerType.Float, false, 16, 8);
}
private void SetupTexture(int Handle, int Width, int Height)
{
GL.BindTexture(TextureTarget.Texture2D, Handle);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMinFilter, (int)TextureMinFilter.Linear);
GL.TexParameter(TextureTarget.Texture2D, TextureParameterName.TextureMagFilter, (int)TextureMagFilter.Linear);
(PixelFormat Format, PixelType Type) = OGLEnumConverter.GetTextureFormat(GalTextureFormat.A8B8G8R8);
const PixelInternalFormat InternalFmt = PixelInternalFormat.Rgba;
const int Level = 0;
const int Border = 0;
GL.TexImage2D(
TextureTarget.Texture2D,
Level,
InternalFmt,
Width,
Height,
Border,
Format,
Type,
IntPtr.Zero);
}
}
}

View file

@ -17,6 +17,7 @@ namespace Ryujinx.Graphics.Gal.OpenGL
Bind(Index);
const int Level = 0; //TODO: Support mipmap textures.
const int Border = 0;
if (IsCompressedTextureFormat(Texture.Format))
@ -25,7 +26,7 @@ namespace Ryujinx.Graphics.Gal.OpenGL
GL.CompressedTexImage2D(
TextureTarget.Texture2D,
0,
Level,
InternalFmt,
Texture.Width,
Texture.Height,
@ -37,17 +38,17 @@ namespace Ryujinx.Graphics.Gal.OpenGL
{
const PixelInternalFormat InternalFmt = PixelInternalFormat.Rgba;
(PixelFormat, PixelType) Format = OGLEnumConverter.GetTextureFormat(Texture.Format);
(PixelFormat Format, PixelType Type) = OGLEnumConverter.GetTextureFormat(Texture.Format);
GL.TexImage2D(
TextureTarget.Texture2D,
0,
Level,
InternalFmt,
Texture.Width,
Texture.Height,
Border,
Format.Item1,
Format.Item2,
Format,
Type,
Texture.Data);
}
}
@ -59,7 +60,7 @@ namespace Ryujinx.Graphics.Gal.OpenGL
GL.BindTexture(TextureTarget.Texture2D, Handle);
}
public void Set(GalTextureSampler Sampler)
public static void Set(GalTextureSampler Sampler)
{
int WrapS = (int)OGLEnumConverter.GetTextureWrapMode(Sampler.AddressU);
int WrapT = (int)OGLEnumConverter.GetTextureWrapMode(Sampler.AddressV);

View file

@ -1,4 +1,3 @@
using OpenTK;
using System;
using System.Collections.Concurrent;
using System.Collections.Generic;
@ -19,8 +18,6 @@ namespace Ryujinx.Graphics.Gal.OpenGL
private ConcurrentQueue<Action> ActionsQueue;
private FrameBuffer FbRenderer;
public OpenGLRenderer()
{
Blend = new OGLBlend();
@ -36,16 +33,6 @@ namespace Ryujinx.Graphics.Gal.OpenGL
ActionsQueue = new ConcurrentQueue<Action>();
}
public void InitializeFrameBuffer()
{
FbRenderer = new FrameBuffer(1280, 720);
}
public void ResetFrameBuffer()
{
FbRenderer.Reset();
}
public void QueueAction(Action ActionMthd)
{
ActionsQueue.Enqueue(ActionMthd);
@ -63,34 +50,12 @@ namespace Ryujinx.Graphics.Gal.OpenGL
public void Render()
{
//FbRenderer.Render();
FrameBuffer.Render();
}
public void SetWindowSize(int Width, int Height)
{
FbRenderer.WindowWidth = Width;
FbRenderer.WindowHeight = Height;
}
public unsafe void SetFrameBuffer(
byte* Fb,
int Width,
int Height,
float ScaleX,
float ScaleY,
float OffsX,
float OffsY,
float Rotate)
{
Matrix2 Transform;
Transform = Matrix2.CreateScale(ScaleX, ScaleY);
Transform *= Matrix2.CreateRotation(Rotate);
Vector2 Offs = new Vector2(OffsX, OffsY);
FbRenderer.Set(Fb, Width, Height, Transform, Offs);
//TODO
}
public void SetBlendEnable(bool Enable)
@ -143,9 +108,14 @@ namespace Ryujinx.Graphics.Gal.OpenGL
ActionsQueue.Enqueue(() => FrameBuffer.Bind(Tag));
}
public void BindFrameBufferTexture(long Tag, int Index)
public void BindFrameBufferTexture(long Tag, int Index, GalTextureSampler Sampler)
{
ActionsQueue.Enqueue(() => FrameBuffer.BindTexture(Tag, Index));
ActionsQueue.Enqueue(() =>
{
FrameBuffer.BindTexture(Tag, Index);
OGLTexture.Set(Sampler);
});
}
public void SetFrameBuffer(long Tag)
@ -153,6 +123,11 @@ namespace Ryujinx.Graphics.Gal.OpenGL
ActionsQueue.Enqueue(() => FrameBuffer.Set(Tag));
}
public void SetFrameBuffer(byte[] Data, int Width, int Height)
{
ActionsQueue.Enqueue(() => FrameBuffer.Set(Data, Width, Height));
}
public void ClearBuffers(int RtIndex, GalClearBufferFlags Flags)
{
ActionsQueue.Enqueue(() => Rasterizer.ClearBuffers(RtIndex, Flags));
@ -245,19 +220,19 @@ namespace Ryujinx.Graphics.Gal.OpenGL
ActionsQueue.Enqueue(() => Shader.BindProgram());
}
public void SetTexture(int Index, GalTexture Texture)
public void SetTextureAndSampler(int Index, GalTexture Texture, GalTextureSampler Sampler)
{
ActionsQueue.Enqueue(() => this.Texture.Set(Index, Texture));
ActionsQueue.Enqueue(() =>
{
this.Texture.Set(Index, Texture);
OGLTexture.Set(Sampler);
});
}
public void BindTexture(int Index)
{
ActionsQueue.Enqueue(() => Texture.Bind(Index));
}
public void SetSampler(GalTextureSampler Sampler)
{
ActionsQueue.Enqueue(() => Texture.Set(Sampler));
}
}
}

View file

@ -81,7 +81,7 @@ namespace Ryujinx.Graphics.Gal.Shader
public GlslProgram Decompile(int[] Code, GalShaderType ShaderType)
{
ShaderIrBlock Block = ShaderDecoder.DecodeBasicBlock(Code, 0, ShaderType);
ShaderIrBlock Block = ShaderDecoder.DecodeBasicBlock(Code, 0);
ShaderIrNode[] Nodes = Block.GetNodes();

View file

@ -74,11 +74,9 @@ namespace Ryujinx.Graphics.Gal.Shader
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Dst = (Ch >> 1) != 0
? GetOperGpr28(OpCode)
: GetOperGpr0 (OpCode);
Dst.Index += Ch & 1;
//Assign it to a temp because the destination registers
//may be used as texture coord input aswell.
ShaderIrOperGpr Dst = new ShaderIrOperGpr(0x100 + Ch);
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
@ -86,6 +84,19 @@ namespace Ryujinx.Graphics.Gal.Shader
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Op), OpCode));
}
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Src = new ShaderIrOperGpr(0x100 + Ch);
ShaderIrOperGpr Dst = (Ch >> 1) != 0
? GetOperGpr28(OpCode)
: GetOperGpr0 (OpCode);
Dst.Index += Ch & 1;
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Src), OpCode));
}
}
}
}

View file

@ -2,7 +2,7 @@ namespace Ryujinx.Graphics.Gal.Shader
{
static class ShaderDecoder
{
public static ShaderIrBlock DecodeBasicBlock(int[] Code, int Offset, GalShaderType ShaderType)
public static ShaderIrBlock DecodeBasicBlock(int[] Code, int Offset)
{
ShaderIrBlock Block = new ShaderIrBlock();
@ -37,8 +37,6 @@ namespace Ryujinx.Graphics.Gal.Shader
}
}
Block.RunOptimizationPasses(ShaderType);
return Block;
}

View file

@ -16,11 +16,6 @@ namespace Ryujinx.Graphics.Gal.Shader
Nodes.Add(Node);
}
public void RunOptimizationPasses(GalShaderType ShaderType)
{
ShaderOptExprProp.Optimize(Nodes, ShaderType);
}
public ShaderIrNode[] GetNodes()
{
return Nodes.ToArray();

View file

@ -1,366 +0,0 @@
using System;
using System.Collections.Generic;
namespace Ryujinx.Graphics.Gal.Shader
{
static class ShaderOptExprProp
{
private struct UseSite
{
public ShaderIrNode Parent { get; private set; }
public ShaderIrCond Cond { get; private set; }
public int UseIndex { get; private set; }
public int OperIndex { get; private set; }
public UseSite(
ShaderIrNode Parent,
ShaderIrCond Cond,
int UseIndex,
int OperIndex)
{
this.Parent = Parent;
this.Cond = Cond;
this.UseIndex = UseIndex;
this.OperIndex = OperIndex;
}
}
private class RegUse
{
public ShaderIrAsg Asg { get; private set; }
public int AsgIndex { get; private set; }
public int LastSiteIndex { get; private set; }
public ShaderIrCond Cond { get; private set; }
private List<UseSite> Sites;
public RegUse()
{
Sites = new List<UseSite>();
}
public void AddUseSite(UseSite Site)
{
if (LastSiteIndex < Site.UseIndex)
{
LastSiteIndex = Site.UseIndex;
}
Sites.Add(Site);
}
public bool TryPropagate()
{
//This happens when a untiliazied register is used,
//this usually indicates a decoding error, but may also
//be caused by bogus programs (?). In any case, we just
//keep the unitialized access and avoid trying to propagate
//the expression (since we can't propagate what doesn't yet exist).
if (Asg == null)
{
return false;
}
if (Cond != null)
{
//If the assignment is conditional, we can only propagate
//to the use sites that shares the same condition of the assignment.
foreach (UseSite Site in Sites)
{
if (!IsSameCondition(Cond, Site.Cond))
{
return false;
}
}
}
if (Sites.Count > 0)
{
foreach (UseSite Site in Sites)
{
if (Site.Parent is ShaderIrCond Cond)
{
switch (Site.OperIndex)
{
case 0: Cond.Pred = Asg.Src; break;
case 1: Cond.Child = Asg.Src; break;
default: throw new InvalidOperationException();
}
}
else if (Site.Parent is ShaderIrOp Op)
{
switch (Site.OperIndex)
{
case 0: Op.OperandA = Asg.Src; break;
case 1: Op.OperandB = Asg.Src; break;
case 2: Op.OperandC = Asg.Src; break;
default: throw new InvalidOperationException();
}
}
else if (Site.Parent is ShaderIrAsg SiteAsg)
{
SiteAsg.Src = Asg.Src;
}
else
{
throw new InvalidOperationException();
}
}
}
return true;
}
public void SetNewAsg(ShaderIrAsg Asg, int AsgIndex, ShaderIrCond Cond)
{
this.Asg = Asg;
this.AsgIndex = AsgIndex;
this.Cond = Cond;
LastSiteIndex = 0;
Sites.Clear();
}
}
public static void Optimize(List<ShaderIrNode> Nodes, GalShaderType ShaderType)
{
Dictionary<int, RegUse> Uses = new Dictionary<int, RegUse>();
RegUse GetUse(int Key)
{
RegUse Use;
if (!Uses.TryGetValue(Key, out Use))
{
Use = new RegUse();
Uses.Add(Key, Use);
}
return Use;
}
int GetGprKey(int GprIndex)
{
return GprIndex;
}
int GetPredKey(int PredIndex)
{
return PredIndex | 0x10000000;
}
RegUse GetGprUse(int GprIndex)
{
return GetUse(GetGprKey(GprIndex));
}
RegUse GetPredUse(int PredIndex)
{
return GetUse(GetPredKey(PredIndex));
}
void RemoveUse(RegUse Use)
{
if (!Nodes.Remove((ShaderIrNode)Use.Cond ?? Use.Asg))
{
throw new InvalidOperationException();
}
}
void FindRegUses(
List<(int, UseSite)> UseList,
ShaderIrNode Parent,
ShaderIrNode Node,
ShaderIrCond CondNode,
int UseIndex,
int OperIndex = 0)
{
if (Node is ShaderIrAsg Asg)
{
FindRegUses(UseList, Asg, Asg.Src, CondNode, UseIndex);
}
else if (Node is ShaderIrCond Cond)
{
FindRegUses(UseList, Cond, Cond.Pred, CondNode, UseIndex, 0);
FindRegUses(UseList, Cond, Cond.Child, CondNode, UseIndex, 1);
}
else if (Node is ShaderIrOp Op)
{
FindRegUses(UseList, Op, Op.OperandA, CondNode, UseIndex, 0);
FindRegUses(UseList, Op, Op.OperandB, CondNode, UseIndex, 1);
FindRegUses(UseList, Op, Op.OperandC, CondNode, UseIndex, 2);
}
else if (Node is ShaderIrOperGpr Gpr && !Gpr.IsConst)
{
UseList.Add((GetGprKey(Gpr.Index), new UseSite(Parent, CondNode, UseIndex, OperIndex)));
}
else if (Node is ShaderIrOperPred Pred)
{
UseList.Add((GetPredKey(Pred.Index), new UseSite(Parent, CondNode, UseIndex, OperIndex)));
}
}
void TryAddRegUseSite(ShaderIrNode Node, ShaderIrCond CondNode, int UseIndex)
{
List<(int, UseSite)> UseList = new List<(int, UseSite)>();
FindRegUses(UseList, null, Node, CondNode, UseIndex);
foreach ((int Key, UseSite Site) in UseList)
{
GetUse(Key).AddUseSite(Site);
}
}
bool TryPropagate(RegUse Use)
{
//We can only propagate if the registers that the expression depends
//on weren't assigned after the original expression assignment
//to a register took place. We traverse the expression tree to find
//all registers being used, if any of those registers was assigned
//after the assignment to be propagated, then we can't propagate.
if (Use?.Asg == null)
{
return false;
}
List<(int, UseSite)> UseList = new List<(int, UseSite)>();
if (Use.Cond != null)
{
FindRegUses(UseList, null, Use.Cond, null, 0);
}
else
{
FindRegUses(UseList, Use.Asg, Use.Asg.Src, null, 0);
}
foreach ((int Key, UseSite Site) in UseList)
{
//TODO: Build an assignment list inside RegUse,
//and check if there is an assignment inside the
//range of Use.AsgIndex and Use.LastSiteIndex,
//and if that's the case, then we should return false.
//The current method is too conservative.
if (GetUse(Key).AsgIndex >= Use.AsgIndex)
{
return false;
}
}
return Use.TryPropagate();
}
for (int Index = 0, IterCount = 0; Index < Nodes.Count; Index++, IterCount++)
{
ShaderIrNode Node = Nodes[Index];
ShaderIrCond CondNode = null;
if (Node is ShaderIrCond)
{
CondNode = (ShaderIrCond)Node;
}
TryAddRegUseSite(Node, CondNode, IterCount);;
while (Node is ShaderIrCond Cond)
{
Node = Cond.Child;
}
if (!(Node is ShaderIrAsg Asg))
{
continue;
}
RegUse Use = null;
if (Asg.Dst is ShaderIrOperGpr Gpr && !Gpr.IsConst)
{
Use = GetGprUse(Gpr.Index);
}
else if (Asg.Dst is ShaderIrOperPred Pred)
{
Use = GetPredUse(Pred.Index);
}
bool CanRemoveAsg = CondNode == null;
CanRemoveAsg |= IsSameCondition(CondNode, Use?.Cond);
if (CanRemoveAsg && TryPropagate(Use))
{
RemoveUse(Use);
//Note: Only decrement if the removal was successful.
//RemoveUse throws when this is not the case so we should be good.
Index--;
}
//All nodes inside conditional nodes can't be propagated,
//as we don't even know if they will be executed to begin with.
Use?.SetNewAsg(Asg, IterCount, CondNode);
}
foreach (RegUse Use in Uses.Values)
{
//Gprs 0-3 are the color output on fragment shaders,
//so we can't remove the last assignments to those registers.
if (ShaderType == GalShaderType.Fragment)
{
if (Use.Asg?.Dst is ShaderIrOperGpr Gpr && Gpr.Index < 4)
{
continue;
}
}
if (TryPropagate(Use))
{
RemoveUse(Use);
}
}
}
private static bool IsSameCondition(ShaderIrCond CondA, ShaderIrCond CondB)
{
if (CondA == null || CondB == null)
{
return CondA == CondB;
}
if (CondA.Not != CondB.Not)
{
return false;
}
if (CondA.Pred is ShaderIrOperPred PredA)
{
if (!(CondB.Pred is ShaderIrOperPred PredB))
{
return false;
}
if (PredA.Index != PredB.Index)
{
return false;
}
}
else if (CondA.Pred != CondB.Pred)
{
return false;
}
return true;
}
}
}

View file

@ -9,9 +9,9 @@ namespace Ryujinx.Graphics.Gpu
internal NsGpuMemoryMgr MemoryMgr { get; private set; }
public NvGpuFifo Fifo;
public NvGpuFifo Fifo { get; private set; }
internal NvGpuEngine3d Engine3d;
public NvGpuEngine3d Engine3d { get; private set; }
private Thread FifoProcessing;
@ -29,7 +29,7 @@ namespace Ryujinx.Graphics.Gpu
KeepRunning = true;
FifoProcessing = new Thread(ProcessFifo);
FifoProcessing = new Thread(ProcessFifo);
FifoProcessing.Start();
}

View file

@ -5,7 +5,7 @@ using System.Collections.Generic;
namespace Ryujinx.Graphics.Gpu
{
class NvGpuEngine3d : INvGpuEngine
public class NvGpuEngine3d : INvGpuEngine
{
public int[] Registers { get; private set; }
@ -249,6 +249,8 @@ namespace Ryujinx.Graphics.Gpu
TicPosition += TicIndex * 0x20;
TscPosition += TscIndex * 0x20;
GalTextureSampler Sampler = TextureFactory.MakeSampler(Gpu, Memory, TscPosition);
long TextureAddress = Memory.ReadInt64(TicPosition + 4) & 0xffffffffffff;
if (FrameBuffers.Contains(TextureAddress))
@ -257,15 +259,15 @@ namespace Ryujinx.Graphics.Gpu
//we shouldn't read anything from memory and bind
//the frame buffer texture instead, since we're not
//really writing anything to memory.
Gpu.Renderer.BindFrameBufferTexture(TextureAddress, TexIndex);
Gpu.Renderer.BindFrameBufferTexture(TextureAddress, TexIndex, Sampler);
}
else
{
Gpu.Renderer.SetTexture(TexIndex, TextureFactory.MakeTexture(Gpu, Memory, TicPosition));
GalTexture Texture = TextureFactory.MakeTexture(Gpu, Memory, TicPosition);
Gpu.Renderer.SetTextureAndSampler(TexIndex, Texture, Sampler);
Gpu.Renderer.BindTexture(TexIndex);
}
Gpu.Renderer.SetSampler(TextureFactory.MakeSampler(Gpu, Memory, TscPosition));
}
private void UploadUniforms(AMemory Memory)
@ -488,5 +490,10 @@ namespace Ryujinx.Graphics.Gpu
{
Registers[(int)Reg] = Value;
}
public bool IsFrameBufferPosition(long Position)
{
return FrameBuffers.Contains(Position);
}
}
}

View file

@ -2,7 +2,7 @@ using Ryujinx.Graphics.Gal;
namespace Ryujinx.Graphics.Gpu
{
struct Texture
public struct Texture
{
public long Position { get; private set; }
@ -16,6 +16,24 @@ namespace Ryujinx.Graphics.Gpu
public GalTextureFormat Format { get; private set; }
public Texture(
long Position,
int Width,
int Height)
{
this.Position = Position;
this.Width = Width;
this.Height = Height;
Pitch = 0;
BlockHeight = 16;
Swizzle = TextureSwizzle.BlockLinear;
Format = GalTextureFormat.A8B8G8R8;
}
public Texture(
long Position,
int Width,

View file

@ -4,7 +4,7 @@ using System;
namespace Ryujinx.Graphics.Gpu
{
static class TextureReader
public static class TextureReader
{
public static byte[] Read(AMemory Memory, Texture Texture)
{

View file

@ -1,6 +1,6 @@
namespace Ryujinx.Graphics.Gpu
{
enum TextureSwizzle
public enum TextureSwizzle
{
_1dBuffer = 0,
PitchColorKey = 1,

View file

@ -38,8 +38,6 @@ namespace Ryujinx
protected override void OnLoad(EventArgs e)
{
VSync = VSyncMode.On;
Renderer.InitializeFrameBuffer();
}
protected override void OnUpdateFrame(FrameEventArgs e)