private void CbBind(NvGpuVmm vmm, GpuMethodCall methCall)
{
int stage = (methCall.Method - 0x904) >> 3;
int index = methCall.Argument;
bool enabled = (index & 1) != 0;
index = (index >> 4) & 0x1f;
long position = MakeInt64From2xInt32(NvGpuEngine3dReg.ConstBufferAddress);
long cbKey = vmm.GetPhysicalAddress(position);
int size = ReadRegister(NvGpuEngine3dReg.ConstBufferSize);
if (!_gpu.Renderer.Buffer.IsCached(cbKey, size))
{
_gpu.Renderer.Buffer.Create(cbKey, size);
}
ConstBuffer cb = _constBuffers[stage][index];
if (cb.Position != position || cb.Enabled != enabled || cb.Size != size)
{
_constBuffers[stage][index].Position = position;
_constBuffers[stage][index].Enabled = enabled;
_constBuffers[stage][index].Size = size;
}
}
private void CbBind(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
int Stage = (PBEntry.Method - 0x904) >> 3;
int Index = PBEntry.Arguments[0];
bool Enabled = (Index & 1) != 0;
Index = (Index >> 4) & 0x1f;
long Position = MakeInt64From2xInt32(NvGpuEngine3dReg.ConstBufferAddress);
long CbKey = Vmm.GetPhysicalAddress(Position);
int Size = ReadRegister(NvGpuEngine3dReg.ConstBufferSize);
if (!Gpu.Renderer.Buffer.IsCached(CbKey, Size))
{
Gpu.Renderer.Buffer.Create(CbKey, Size);
}
ConstBuffer Cb = ConstBuffers[Stage][Index];
if (Cb.Position != Position || Cb.Enabled != Enabled || Cb.Size != Size)
{
ConstBuffers[Stage][Index].Position = Position;
ConstBuffers[Stage][Index].Enabled = Enabled;
ConstBuffers[Stage][Index].Size = Size;
}
}
private void SetZeta(NvGpuVmm Vmm)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.ZetaAddress);
int ZetaFormat = ReadRegister(NvGpuEngine3dReg.ZetaFormat);
int BlockDim = ReadRegister(NvGpuEngine3dReg.ZetaBlockDimensions);
int GobBlockHeight = 1 << ((BlockDim >> 4) & 7);
GalMemoryLayout Layout = (GalMemoryLayout)((BlockDim >> 12) & 1); //?
bool ZetaEnable = ReadRegisterBool(NvGpuEngine3dReg.ZetaEnable);
if (VA == 0 || ZetaFormat == 0 || !ZetaEnable)
{
Gpu.Renderer.RenderTarget.UnbindZeta();
return;
}
long Key = Vmm.GetPhysicalAddress(VA);
int Width = ReadRegister(NvGpuEngine3dReg.ZetaHoriz);
int Height = ReadRegister(NvGpuEngine3dReg.ZetaVert);
GalImageFormat Format = ImageUtils.ConvertZeta((GalZetaFormat)ZetaFormat);
GalImage Image = new GalImage(Width, Height, 1, GobBlockHeight, Layout, Format);
Gpu.ResourceManager.SendZetaBuffer(Vmm, Key, Image);
}
private void UploadConstBuffers(NvGpuVmm Vmm, GalPipelineState State, long[] Keys)
{
for (int Stage = 0; Stage < Keys.Length; Stage++)
{
foreach (ShaderDeclInfo DeclInfo in Gpu.Renderer.Shader.GetConstBufferUsage(Keys[Stage]))
{
ConstBuffer Cb = ConstBuffers[Stage][DeclInfo.Cbuf];
if (!Cb.Enabled)
{
continue;
}
long Key = Vmm.GetPhysicalAddress(Cb.Position);
if (Gpu.ResourceManager.MemoryRegionModified(Vmm, Key, Cb.Size, NvGpuBufferType.ConstBuffer))
{
if (Vmm.TryGetHostAddress(Cb.Position, Cb.Size, out IntPtr CbPtr))
{
Gpu.Renderer.Buffer.SetData(Key, Cb.Size, CbPtr);
}
else
{
Gpu.Renderer.Buffer.SetData(Key, Vmm.ReadBytes(Cb.Position, Cb.Size));
}
}
State.ConstBufferKeys[Stage][DeclInfo.Cbuf] = Key;
}
}
}
private void SetZeta(NvGpuVmm vmm)
{
long va = MakeInt64From2xInt32(NvGpuEngine3dReg.ZetaAddress);
int zetaFormat = ReadRegister(NvGpuEngine3dReg.ZetaFormat);
int blockDim = ReadRegister(NvGpuEngine3dReg.ZetaBlockDimensions);
int gobBlockHeight = 1 << ((blockDim >> 4) & 7);
GalMemoryLayout layout = (GalMemoryLayout)((blockDim >> 12) & 1); //?
bool zetaEnable = ReadRegisterBool(NvGpuEngine3dReg.ZetaEnable);
if (va == 0 || zetaFormat == 0 || !zetaEnable)
{
_gpu.Renderer.RenderTarget.UnbindZeta();
return;
}
long key = vmm.GetPhysicalAddress(va);
int width = ReadRegister(NvGpuEngine3dReg.ZetaHoriz);
int height = ReadRegister(NvGpuEngine3dReg.ZetaVert);
GalImageFormat format = ImageUtils.ConvertZeta((GalZetaFormat)zetaFormat);
// TODO: Support non 2D?
GalImage image = new GalImage(width, height, 1, 1, 1, gobBlockHeight, 1, layout, format, GalTextureTarget.TwoD);
_gpu.ResourceManager.SendZetaBuffer(vmm, key, image);
}
private void UploadConstBuffers(NvGpuVmm Vmm, GalPipelineState State, long[] Keys)
{
for (int Stage = 0; Stage < Keys.Length; Stage++)
{
foreach (ShaderDeclInfo DeclInfo in Gpu.Renderer.Shader.GetConstBufferUsage(Keys[Stage]))
{
ConstBuffer Cb = ConstBuffers[Stage][DeclInfo.Cbuf];
if (!Cb.Enabled)
{
continue;
}
long Key = Vmm.GetPhysicalAddress(Cb.Position);
if (QueryKeyUpload(Vmm, Key, Cb.Size, NvGpuBufferType.ConstBuffer))
{
IntPtr Source = Vmm.GetHostAddress(Cb.Position, Cb.Size);
Gpu.Renderer.Buffer.SetData(Key, Cb.Size, Source);
}
State.ConstBufferKeys[Stage][DeclInfo.Cbuf] = Key;
}
}
}
private void DispatchRender(NvGpuVmm Vmm, GalPipelineState State)
{
int IndexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
int PrimCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType PrimType = (GalPrimitiveType)(PrimCtrl & 0xffff);
Gpu.Renderer.Pipeline.Bind(State);
if (IndexCount != 0)
{
int IndexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int IndexFirst = ReadRegister(NvGpuEngine3dReg.IndexBatchFirst);
int VertexBase = ReadRegister(NvGpuEngine3dReg.VertexArrayElemBase);
long IndexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long IboKey = Vmm.GetPhysicalAddress(IndexPosition);
Gpu.Renderer.Rasterizer.DrawElements(IboKey, IndexFirst, VertexBase, PrimType);
}
else
{
int VertexFirst = ReadRegister(NvGpuEngine3dReg.VertexArrayFirst);
int VertexCount = ReadRegister(NvGpuEngine3dReg.VertexArrayCount);
Gpu.Renderer.Rasterizer.DrawArrays(VertexFirst, VertexCount, PrimType);
}
//Is the GPU really clearing those registers after draw?
WriteRegister(NvGpuEngine3dReg.IndexBatchFirst, 0);
WriteRegister(NvGpuEngine3dReg.IndexBatchCount, 0);
}
private void SetZeta(NvGpuVmm Vmm)
{
long ZA = MakeInt64From2xInt32(NvGpuEngine3dReg.ZetaAddress);
int Format = ReadRegister(NvGpuEngine3dReg.ZetaFormat);
bool ZetaEnable = (ReadRegister(NvGpuEngine3dReg.ZetaEnable) & 1) != 0;
if (ZA == 0 || Format == 0 || !ZetaEnable)
{
Gpu.Renderer.FrameBuffer.UnbindZeta();
return;
}
long Key = Vmm.GetPhysicalAddress(ZA);
int Width = ReadRegister(NvGpuEngine3dReg.ZetaHoriz);
int Height = ReadRegister(NvGpuEngine3dReg.ZetaVert);
GalImageFormat ImageFormat = ImageFormatConverter.ConvertZeta((GalZetaFormat)Format);
GalImage Image = new GalImage(Width, Height, ImageFormat);
long Size = TextureHelper.GetTextureSize(Image);
Gpu.Renderer.Texture.CreateFb(Key, Size, Image);
Gpu.Renderer.FrameBuffer.BindZeta(Key);
}
private void SetFrameBuffer(NvGpuVmm Vmm, int FbIndex)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + FbIndex * 0x10);
long Key = Vmm.GetPhysicalAddress(VA);
FrameBuffers.Add(Key);
int Width = ReadRegister(NvGpuEngine3dReg.FrameBufferNWidth + FbIndex * 0x10);
int Height = ReadRegister(NvGpuEngine3dReg.FrameBufferNHeight + FbIndex * 0x10);
float TX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateX + FbIndex * 4);
float TY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateY + FbIndex * 4);
float SX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleX + FbIndex * 4);
float SY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleY + FbIndex * 4);
int VpX = (int)MathF.Max(0, TX - MathF.Abs(SX));
int VpY = (int)MathF.Max(0, TY - MathF.Abs(SY));
int VpW = (int)(TX + MathF.Abs(SX)) - VpX;
int VpH = (int)(TY + MathF.Abs(SY)) - VpY;
Gpu.Renderer.FrameBuffer.Create(Key, Width, Height);
Gpu.Renderer.FrameBuffer.Bind(Key);
Gpu.Renderer.FrameBuffer.SetViewport(VpX, VpY, VpW, VpH);
}
private void UploadConstBuffers(NvGpuVmm vmm, GalPipelineState state, long[] keys)
{
Profile.Begin(Profiles.GPU.Engine3d.UploadConstBuffers);
for (int stage = 0; stage < keys.Length; stage++)
{
foreach (CBufferDescriptor desc in _gpu.Renderer.Shader.GetConstBufferUsage(keys[stage]))
{
ConstBuffer cb = _constBuffers[stage][desc.Slot];
if (!cb.Enabled)
{
continue;
}
long key = vmm.GetPhysicalAddress(cb.Position);
if (_gpu.ResourceManager.MemoryRegionModified(vmm, key, cb.Size, NvGpuBufferType.ConstBuffer))
{
if (vmm.TryGetHostAddress(cb.Position, cb.Size, out IntPtr cbPtr))
{
_gpu.Renderer.Buffer.SetData(key, cb.Size, cbPtr);
}
else
{
_gpu.Renderer.Buffer.SetData(key, vmm.ReadBytes(cb.Position, cb.Size));
}
}
state.ConstBufferKeys[stage][desc.Slot] = key;
}
}
Profile.End(Profiles.GPU.Engine3d.UploadConstBuffers);
}
private byte[] ReadShaderBinary(NvGpuVmm Vmm, long Key)
{
long Size = GetShaderSize(Vmm, Key);
long Address = Vmm.GetPhysicalAddress(Key);
return(Vmm.ReadBytes(Key, Size));
}
private void SetFrameBuffer(NvGpuVmm vmm, int fbIndex)
{
ProfileConfig profile = Profiles.GPU.Engine3d.SetFrameBuffer;
profile.SessionItem = fbIndex.ToString();
Profile.Begin(profile);
long va = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + fbIndex * 0x10);
int surfFormat = ReadRegister(NvGpuEngine3dReg.FrameBufferNFormat + fbIndex * 0x10);
if (va == 0 || surfFormat == 0)
{
_gpu.Renderer.RenderTarget.UnbindColor(fbIndex);
Profile.End(profile);
return;
}
long key = vmm.GetPhysicalAddress(va);
int width = ReadRegister(NvGpuEngine3dReg.FrameBufferNWidth + fbIndex * 0x10);
int height = ReadRegister(NvGpuEngine3dReg.FrameBufferNHeight + fbIndex * 0x10);
int arrayMode = ReadRegister(NvGpuEngine3dReg.FrameBufferNArrayMode + fbIndex * 0x10);
int layerCount = arrayMode & 0xFFFF;
int layerStride = ReadRegister(NvGpuEngine3dReg.FrameBufferNLayerStride + fbIndex * 0x10);
int baseLayer = ReadRegister(NvGpuEngine3dReg.FrameBufferNBaseLayer + fbIndex * 0x10);
int blockDim = ReadRegister(NvGpuEngine3dReg.FrameBufferNBlockDim + fbIndex * 0x10);
int gobBlockHeight = 1 << ((blockDim >> 4) & 7);
GalMemoryLayout layout = (GalMemoryLayout)((blockDim >> 12) & 1);
float tx = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateX + fbIndex * 8);
float ty = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateY + fbIndex * 8);
float sx = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleX + fbIndex * 8);
float sy = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleY + fbIndex * 8);
_viewportX0 = (int)MathF.Max(0, tx - MathF.Abs(sx));
_viewportY0 = (int)MathF.Max(0, ty - MathF.Abs(sy));
_viewportX1 = (int)(tx + MathF.Abs(sx));
_viewportY1 = (int)(ty + MathF.Abs(sy));
GalImageFormat format = ImageUtils.ConvertSurface((GalSurfaceFormat)surfFormat);
GalImage image = new GalImage(width, height, 1, 1, 1, gobBlockHeight, 1, layout, format, GalTextureTarget.TwoD);
_gpu.ResourceManager.SendColorBuffer(vmm, key, fbIndex, image);
_gpu.Renderer.RenderTarget.SetViewport(fbIndex, _viewportX0, _viewportY0, _viewportX1 - _viewportX0, _viewportY1 - _viewportY0);
Profile.End(profile);
}
private (long, GalImage, GalTextureSampler) UploadTexture(NvGpuVmm vmm, int textureHandle)
{
if (textureHandle == 0)
{
// FIXME: Some games like puyo puyo will use handles with the value 0.
// This is a bug, most likely caused by sync issues.
return(0, default(GalImage), default(GalTextureSampler));
}
Profile.Begin(Profiles.GPU.Engine3d.UploadTexture);
bool linkedTsc = ReadRegisterBool(NvGpuEngine3dReg.LinkedTsc);
int ticIndex = (textureHandle >> 0) & 0xfffff;
int tscIndex = linkedTsc ? ticIndex : (textureHandle >> 20) & 0xfff;
long ticPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexHeaderPoolOffset);
long tscPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexSamplerPoolOffset);
ticPosition += ticIndex * 0x20;
tscPosition += tscIndex * 0x20;
GalImage image = TextureFactory.MakeTexture(vmm, ticPosition);
GalTextureSampler sampler = TextureFactory.MakeSampler(_gpu, vmm, tscPosition);
long key = vmm.ReadInt64(ticPosition + 4) & 0xffffffffffff;
if (image.Layout == GalMemoryLayout.BlockLinear)
{
key &= ~0x1ffL;
}
else if (image.Layout == GalMemoryLayout.Pitch)
{
key &= ~0x1fL;
}
key = vmm.GetPhysicalAddress(key);
if (key == -1)
{
Profile.End(Profiles.GPU.Engine3d.UploadTexture);
// FIXME: Shouldn't ignore invalid addresses.
return(0, default(GalImage), default(GalTextureSampler));
}
_gpu.ResourceManager.SendTexture(vmm, key, image);
Profile.End(Profiles.GPU.Engine3d.UploadTexture);
return(key, image, sampler);
}
private void DispatchRender(NvGpuVmm Vmm, GalPipelineState State)
{
int IndexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
int PrimCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType PrimType = (GalPrimitiveType)(PrimCtrl & 0xffff);
bool InstanceNext = ((PrimCtrl >> 26) & 1) != 0;
bool InstanceCont = ((PrimCtrl >> 27) & 1) != 0;
if (InstanceNext && InstanceCont)
{
throw new InvalidOperationException("GPU tried to increase and reset instance count at the same time");
}
if (InstanceNext)
{
CurrentInstance++;
}
else if (!InstanceCont)
{
CurrentInstance = 0;
}
State.Instance = CurrentInstance;
Gpu.Renderer.Pipeline.Bind(State);
Gpu.Renderer.RenderTarget.Bind();
if (IndexCount != 0)
{
int IndexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int IndexFirst = ReadRegister(NvGpuEngine3dReg.IndexBatchFirst);
int VertexBase = ReadRegister(NvGpuEngine3dReg.VertexArrayElemBase);
long IndexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long IboKey = Vmm.GetPhysicalAddress(IndexPosition);
Gpu.Renderer.Rasterizer.DrawElements(IboKey, IndexFirst, VertexBase, PrimType);
}
else
{
int VertexFirst = ReadRegister(NvGpuEngine3dReg.VertexArrayFirst);
int VertexCount = ReadRegister(NvGpuEngine3dReg.VertexArrayCount);
Gpu.Renderer.Rasterizer.DrawArrays(VertexFirst, VertexCount, PrimType);
}
//Is the GPU really clearing those registers after draw?
WriteRegister(NvGpuEngine3dReg.IndexBatchFirst, 0);
WriteRegister(NvGpuEngine3dReg.IndexBatchCount, 0);
}
private bool IsShaderModified(NvGpuVmm Vmm, long Key)
{
long Address = Vmm.GetPhysicalAddress(Key);
if (Gpu.Renderer.Shader.TryGetSize(Address, out long Size))
{
if (!QueryKeyUpload(Vmm, Address, Size, NvGpuBufferType.Shader))
{
return(false);
}
}
return(true);
}
private (long, GalImage, GalTextureSampler) UploadTexture(NvGpuVmm Vmm, int TextureHandle)
{
if (TextureHandle == 0)
{
//FIXME: Some games like puyo puyo will use handles with the value 0.
//This is a bug, most likely caused by sync issues.
return(0, default(GalImage), default(GalTextureSampler));
}
bool LinkedTsc = ReadRegisterBool(NvGpuEngine3dReg.LinkedTsc);
int TicIndex = (TextureHandle >> 0) & 0xfffff;
int TscIndex = LinkedTsc ? TicIndex : (TextureHandle >> 20) & 0xfff;
long TicPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexHeaderPoolOffset);
long TscPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexSamplerPoolOffset);
TicPosition += TicIndex * 0x20;
TscPosition += TscIndex * 0x20;
GalImage Image = TextureFactory.MakeTexture(Vmm, TicPosition);
GalTextureSampler Sampler = TextureFactory.MakeSampler(Gpu, Vmm, TscPosition);
long Key = Vmm.ReadInt64(TicPosition + 4) & 0xffffffffffff;
if (Image.Layout == GalMemoryLayout.BlockLinear)
{
Key &= ~0x1ffL;
}
else if (Image.Layout == GalMemoryLayout.Pitch)
{
Key &= ~0x1fL;
}
Key = Vmm.GetPhysicalAddress(Key);
if (Key == -1)
{
//FIXME: Shouldn't ignore invalid addresses.
return(0, default(GalImage), default(GalTextureSampler));
}
Gpu.ResourceManager.SendTexture(Vmm, Key, Image);
return(Key, Image, Sampler);
}
private void SetFrameBuffer(NvGpuVmm Vmm, int FbIndex)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + FbIndex * 0x10);
long Key = Vmm.GetPhysicalAddress(VA);
FrameBuffers.Add(Key);
int Width = ReadRegister(NvGpuEngine3dReg.FrameBufferNWidth + FbIndex * 0x10);
int Height = ReadRegister(NvGpuEngine3dReg.FrameBufferNHeight + FbIndex * 0x10);
//Note: Using the Width/Height results seems to give incorrect results.
//Maybe the size of all frame buffers is hardcoded to screen size? This seems unlikely.
Gpu.Renderer.FrameBuffer.Create(Key, 1280, 720);
Gpu.Renderer.FrameBuffer.Bind(Key);
}
private void UploadTexture(NvGpuVmm Vmm, int TexIndex, int TextureHandle)
{
if (TextureHandle == 0)
{
//TODO: Is this correct?
//Some games like puyo puyo will have 0 handles.
//It may be just normal behaviour or a bug caused by sync issues.
//The game does initialize the value properly after through.
return;
}
int TicIndex = (TextureHandle >> 0) & 0xfffff;
int TscIndex = (TextureHandle >> 20) & 0xfff;
long TicPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexHeaderPoolOffset);
long TscPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexSamplerPoolOffset);
TicPosition += TicIndex * 0x20;
TscPosition += TscIndex * 0x20;
GalImage Image = TextureFactory.MakeTexture(Vmm, TicPosition);
GalTextureSampler Sampler = TextureFactory.MakeSampler(Gpu, Vmm, TscPosition);
long Key = Vmm.ReadInt64(TicPosition + 4) & 0xffffffffffff;
if (Image.Layout == GalMemoryLayout.BlockLinear)
{
Key &= ~0x1ffL;
}
else if (Image.Layout == GalMemoryLayout.Pitch)
{
Key &= ~0x1fL;
}
Key = Vmm.GetPhysicalAddress(Key);
if (Key == -1)
{
//FIXME: Shouldn't ignore invalid addresses.
return;
}
Gpu.ResourceManager.SendTexture(Vmm, Key, Image, TexIndex);
Gpu.Renderer.Texture.SetSampler(Sampler);
}
private void SetFrameBuffer(NvGpuVmm Vmm, int FbIndex)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + FbIndex * 0x10);
int SurfFormat = ReadRegister(NvGpuEngine3dReg.FrameBufferNFormat + FbIndex * 0x10);
if (VA == 0 || SurfFormat == 0)
{
Gpu.Renderer.RenderTarget.UnbindColor(FbIndex);
return;
}
long Key = Vmm.GetPhysicalAddress(VA);
int Width = ReadRegister(NvGpuEngine3dReg.FrameBufferNWidth + FbIndex * 0x10);
int Height = ReadRegister(NvGpuEngine3dReg.FrameBufferNHeight + FbIndex * 0x10);
int BlockDim = ReadRegister(NvGpuEngine3dReg.FrameBufferNBlockDim + FbIndex * 0x10);
int GobBlockHeight = 1 << ((BlockDim >> 4) & 7);
GalMemoryLayout Layout = (GalMemoryLayout)((BlockDim >> 12) & 1);
float TX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateX + FbIndex * 8);
float TY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateY + FbIndex * 8);
float SX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleX + FbIndex * 8);
float SY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleY + FbIndex * 8);
int VpX = (int)MathF.Max(0, TX - MathF.Abs(SX));
int VpY = (int)MathF.Max(0, TY - MathF.Abs(SY));
int VpW = (int)(TX + MathF.Abs(SX)) - VpX;
int VpH = (int)(TY + MathF.Abs(SY)) - VpY;
GalImageFormat Format = ImageUtils.ConvertSurface((GalSurfaceFormat)SurfFormat);
GalImage Image = new GalImage(Width, Height, 1, GobBlockHeight, Layout, Format);
Gpu.ResourceManager.SendColorBuffer(Vmm, Key, FbIndex, Image);
ViewportHeight = VpH;
Gpu.Renderer.RenderTarget.SetViewport(FbIndex, VpX, VpY, VpW, VpH);
}
private void CopyPlanesRgba8(NvGpuVmm Vmm, SurfaceOutputConfig OutputConfig)
{
FFmpegFrame Frame = FFmpegWrapper.GetFrameRgba();
if ((Frame.Width | Frame.Height) == 0)
{
return;
}
GalImage Image = new GalImage(
OutputConfig.SurfaceWidth,
OutputConfig.SurfaceHeight, 1,
OutputConfig.GobBlockHeight,
GalMemoryLayout.BlockLinear,
GalImageFormat.RGBA8 | GalImageFormat.Unorm);
ImageUtils.WriteTexture(Vmm, Image, Vmm.GetPhysicalAddress(OutputConfig.SurfaceLumaAddress), Frame.Data);
}
private void CopyPlanesRgba8(NvGpuVmm vmm, SurfaceOutputConfig outputConfig)
{
FFmpegFrame frame = FFmpegWrapper.GetFrameRgba();
if ((frame.Width | frame.Height) == 0)
{
return;
}
GalImage image = new GalImage(
outputConfig.SurfaceWidth,
outputConfig.SurfaceHeight, 1, 1, 1,
outputConfig.GobBlockHeight, 1,
GalMemoryLayout.BlockLinear,
GalImageFormat.Rgba8 | GalImageFormat.Unorm,
GalTextureTarget.TwoD);
ImageUtils.WriteTexture(vmm, image, vmm.GetPhysicalAddress(outputConfig.SurfaceLumaAddress), frame.Data);
}
private void SetFrameBuffer(NvGpuVmm Vmm, int FbIndex)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + FbIndex * 0x10);
int Format = ReadRegister(NvGpuEngine3dReg.FrameBufferNFormat + FbIndex * 0x10);
if (VA == 0 || Format == 0)
{
Gpu.Renderer.FrameBuffer.UnbindColor(FbIndex);
return;
}
long Key = Vmm.GetPhysicalAddress(VA);
FrameBuffers.Add(Key);
int Width = ReadRegister(NvGpuEngine3dReg.FrameBufferNWidth + FbIndex * 0x10);
int Height = ReadRegister(NvGpuEngine3dReg.FrameBufferNHeight + FbIndex * 0x10);
float TX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateX + FbIndex * 8);
float TY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNTranslateY + FbIndex * 8);
float SX = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleX + FbIndex * 8);
float SY = ReadRegisterFloat(NvGpuEngine3dReg.ViewportNScaleY + FbIndex * 8);
int VpX = (int)MathF.Max(0, TX - MathF.Abs(SX));
int VpY = (int)MathF.Max(0, TY - MathF.Abs(SY));
int VpW = (int)(TX + MathF.Abs(SX)) - VpX;
int VpH = (int)(TY + MathF.Abs(SY)) - VpY;
GalImageFormat ImageFormat = ImageFormatConverter.ConvertFrameBuffer((GalFrameBufferFormat)Format);
GalImage Image = new GalImage(Width, Height, ImageFormat);
long Size = TextureHelper.GetTextureSize(Image);
Gpu.Renderer.Texture.CreateFb(Key, Size, Image);
Gpu.Renderer.FrameBuffer.BindColor(Key, FbIndex);
Gpu.Renderer.FrameBuffer.SetViewport(VpX, VpY, VpW, VpH);
}
private void TextureCopy(NvGpuVmm vmm)
{
CopyOperation operation = (CopyOperation)ReadRegister(NvGpuEngine2dReg.CopyOperation);
int dstFormat = ReadRegister(NvGpuEngine2dReg.DstFormat);
bool dstLinear = ReadRegister(NvGpuEngine2dReg.DstLinear) != 0;
int dstWidth = ReadRegister(NvGpuEngine2dReg.DstWidth);
int dstHeight = ReadRegister(NvGpuEngine2dReg.DstHeight);
int dstDepth = ReadRegister(NvGpuEngine2dReg.DstDepth);
int dstLayer = ReadRegister(NvGpuEngine2dReg.DstLayer);
int dstPitch = ReadRegister(NvGpuEngine2dReg.DstPitch);
int dstBlkDim = ReadRegister(NvGpuEngine2dReg.DstBlockDimensions);
int srcFormat = ReadRegister(NvGpuEngine2dReg.SrcFormat);
bool srcLinear = ReadRegister(NvGpuEngine2dReg.SrcLinear) != 0;
int srcWidth = ReadRegister(NvGpuEngine2dReg.SrcWidth);
int srcHeight = ReadRegister(NvGpuEngine2dReg.SrcHeight);
int srcDepth = ReadRegister(NvGpuEngine2dReg.SrcDepth);
int srcLayer = ReadRegister(NvGpuEngine2dReg.SrcLayer);
int srcPitch = ReadRegister(NvGpuEngine2dReg.SrcPitch);
int srcBlkDim = ReadRegister(NvGpuEngine2dReg.SrcBlockDimensions);
int dstBlitX = ReadRegister(NvGpuEngine2dReg.BlitDstX);
int dstBlitY = ReadRegister(NvGpuEngine2dReg.BlitDstY);
int dstBlitW = ReadRegister(NvGpuEngine2dReg.BlitDstW);
int dstBlitH = ReadRegister(NvGpuEngine2dReg.BlitDstH);
long blitDuDx = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitDuDxFract);
long blitDvDy = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitDvDyFract);
long srcBlitX = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitSrcXFract);
long srcBlitY = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitSrcYFract);
GalImageFormat srcImgFormat = ImageUtils.ConvertSurface((GalSurfaceFormat)srcFormat);
GalImageFormat dstImgFormat = ImageUtils.ConvertSurface((GalSurfaceFormat)dstFormat);
GalMemoryLayout srcLayout = GetLayout(srcLinear);
GalMemoryLayout dstLayout = GetLayout(dstLinear);
int srcBlockHeight = 1 << ((srcBlkDim >> 4) & 0xf);
int dstBlockHeight = 1 << ((dstBlkDim >> 4) & 0xf);
long srcAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.SrcAddress);
long dstAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.DstAddress);
long srcKey = vmm.GetPhysicalAddress(srcAddress);
long dstKey = vmm.GetPhysicalAddress(dstAddress);
bool isSrcLayered = false;
bool isDstLayered = false;
GalTextureTarget srcTarget = GalTextureTarget.TwoD;
if (srcDepth != 0)
{
srcTarget = GalTextureTarget.TwoDArray;
srcDepth++;
isSrcLayered = true;
}
else
{
srcDepth = 1;
}
GalTextureTarget dstTarget = GalTextureTarget.TwoD;
if (dstDepth != 0)
{
dstTarget = GalTextureTarget.TwoDArray;
dstDepth++;
isDstLayered = true;
}
else
{
dstDepth = 1;
}
GalImage srcTexture = new GalImage(
srcWidth,
srcHeight,
1, srcDepth, 1,
srcBlockHeight, 1,
srcLayout,
srcImgFormat,
srcTarget);
GalImage dstTexture = new GalImage(
dstWidth,
dstHeight,
1, dstDepth, 1,
dstBlockHeight, 1,
dstLayout,
dstImgFormat,
dstTarget);
srcTexture.Pitch = srcPitch;
dstTexture.Pitch = dstPitch;
long GetLayerOffset(GalImage image, int layer)
{
int targetMipLevel = image.MaxMipmapLevel <= 1 ? 1 : image.MaxMipmapLevel - 1;
return(ImageUtils.GetLayerOffset(image, targetMipLevel) * layer);
}
int srcLayerIndex = -1;
if (isSrcLayered && _gpu.ResourceManager.TryGetTextureLayer(srcKey, out srcLayerIndex) && srcLayerIndex != 0)
{
srcKey = srcKey - GetLayerOffset(srcTexture, srcLayerIndex);
}
int dstLayerIndex = -1;
if (isDstLayered && _gpu.ResourceManager.TryGetTextureLayer(dstKey, out dstLayerIndex) && dstLayerIndex != 0)
{
dstKey = dstKey - GetLayerOffset(dstTexture, dstLayerIndex);
}
_gpu.ResourceManager.SendTexture(vmm, srcKey, srcTexture);
_gpu.ResourceManager.SendTexture(vmm, dstKey, dstTexture);
if (isSrcLayered && srcLayerIndex == -1)
{
for (int layer = 0; layer < srcTexture.LayerCount; layer++)
{
_gpu.ResourceManager.SetTextureArrayLayer(srcKey + GetLayerOffset(srcTexture, layer), layer);
}
srcLayerIndex = 0;
}
if (isDstLayered && dstLayerIndex == -1)
{
for (int layer = 0; layer < dstTexture.LayerCount; layer++)
{
_gpu.ResourceManager.SetTextureArrayLayer(dstKey + GetLayerOffset(dstTexture, layer), layer);
}
dstLayerIndex = 0;
}
int srcBlitX1 = (int)(srcBlitX >> 32);
int srcBlitY1 = (int)(srcBlitY >> 32);
int srcBlitX2 = (int)(srcBlitX + dstBlitW * blitDuDx >> 32);
int srcBlitY2 = (int)(srcBlitY + dstBlitH * blitDvDy >> 32);
_gpu.Renderer.RenderTarget.Copy(
srcTexture,
dstTexture,
srcKey,
dstKey,
srcLayerIndex,
dstLayerIndex,
srcBlitX1,
srcBlitY1,
srcBlitX2,
srcBlitY2,
dstBlitX,
dstBlitY,
dstBlitX + dstBlitW,
dstBlitY + dstBlitH);
//Do a guest side copy aswell. This is necessary when
//the texture is modified by the guest, however it doesn't
//work when resources that the gpu can write to are copied,
//like framebuffers.
// FIXME: SUPPORT MULTILAYER CORRECTLY HERE (this will cause weird stuffs on the first layer)
ImageUtils.CopyTexture(
vmm,
srcTexture,
dstTexture,
srcAddress,
dstAddress,
srcBlitX1,
srcBlitY1,
dstBlitX,
dstBlitY,
dstBlitW,
dstBlitH);
vmm.IsRegionModified(dstKey, ImageUtils.GetSize(dstTexture), NvGpuBufferType.Texture);
}
private void UploadVertexArrays(NvGpuVmm Vmm, GalPipelineState State)
{
long IndexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long IboKey = Vmm.GetPhysicalAddress(IndexPosition);
int IndexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int IndexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
GalIndexFormat IndexFormat = (GalIndexFormat)IndexEntryFmt;
int IndexEntrySize = 1 << IndexEntryFmt;
if (IndexEntrySize > 4)
{
throw new InvalidOperationException();
}
if (IndexCount != 0)
{
int IbSize = IndexCount * IndexEntrySize;
bool IboCached = Gpu.Renderer.Rasterizer.IsIboCached(IboKey, (uint)IbSize);
if (!IboCached || QueryKeyUpload(Vmm, IboKey, (uint)IbSize, NvGpuBufferType.Index))
{
IntPtr DataAddress = Vmm.GetHostAddress(IndexPosition, IbSize);
Gpu.Renderer.Rasterizer.CreateIbo(IboKey, IbSize, DataAddress);
}
Gpu.Renderer.Rasterizer.SetIndexArray(IbSize, IndexFormat);
}
List <GalVertexAttrib>[] Attribs = new List <GalVertexAttrib> [32];
for (int Attr = 0; Attr < 16; Attr++)
{
int Packed = ReadRegister(NvGpuEngine3dReg.VertexAttribNFormat + Attr);
int ArrayIndex = Packed & 0x1f;
if (Attribs[ArrayIndex] == null)
{
Attribs[ArrayIndex] = new List <GalVertexAttrib>();
}
Attribs[ArrayIndex].Add(new GalVertexAttrib(
Attr,
((Packed >> 6) & 0x1) != 0,
(Packed >> 7) & 0x3fff,
(GalVertexAttribSize)((Packed >> 21) & 0x3f),
(GalVertexAttribType)((Packed >> 27) & 0x7),
((Packed >> 31) & 0x1) != 0));
}
State.VertexBindings = new GalVertexBinding[32];
for (int Index = 0; Index < 32; Index++)
{
if (Attribs[Index] == null)
{
continue;
}
int Control = ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + Index * 4);
bool Enable = (Control & 0x1000) != 0;
if (!Enable)
{
continue;
}
long VertexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + Index * 4);
long VertexEndPos = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNEndAddr + Index * 2);
int VertexDivisor = ReadRegister(NvGpuEngine3dReg.VertexArrayNDivisor + Index * 4);
bool Instanced = ReadRegisterBool(NvGpuEngine3dReg.VertexArrayNInstance + Index);
int Stride = Control & 0xfff;
if (Instanced && VertexDivisor != 0)
{
VertexPosition += Stride * (CurrentInstance / VertexDivisor);
}
if (VertexPosition > VertexEndPos)
{
//Instance is invalid, ignore the draw call
continue;
}
long VboKey = Vmm.GetPhysicalAddress(VertexPosition);
long VbSize = (VertexEndPos - VertexPosition) + 1;
bool VboCached = Gpu.Renderer.Rasterizer.IsVboCached(VboKey, VbSize);
if (!VboCached || QueryKeyUpload(Vmm, VboKey, VbSize, NvGpuBufferType.Vertex))
{
IntPtr DataAddress = Vmm.GetHostAddress(VertexPosition, VbSize);
Gpu.Renderer.Rasterizer.CreateVbo(VboKey, (int)VbSize, DataAddress);
}
State.VertexBindings[Index].Enabled = true;
State.VertexBindings[Index].Stride = Stride;
State.VertexBindings[Index].VboKey = VboKey;
State.VertexBindings[Index].Instanced = Instanced;
State.VertexBindings[Index].Divisor = VertexDivisor;
State.VertexBindings[Index].Attribs = Attribs[Index].ToArray();
}
}
private void UploadTexture(NvGpuVmm Vmm, int TexIndex, int TextureHandle)
{
if (TextureHandle == 0)
{
//TODO: Is this correct?
//Some games like puyo puyo will have 0 handles.
//It may be just normal behaviour or a bug caused by sync issues.
//The game does initialize the value properly after through.
return;
}
int TicIndex = (TextureHandle >> 0) & 0xfffff;
int TscIndex = (TextureHandle >> 20) & 0xfff;
long TicPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexHeaderPoolOffset);
long TscPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.TexSamplerPoolOffset);
TicPosition += TicIndex * 0x20;
TscPosition += TscIndex * 0x20;
GalTextureSampler Sampler = TextureFactory.MakeSampler(Gpu, Vmm, TscPosition);
long Key = Vmm.ReadInt64(TicPosition + 4) & 0xffffffffffff;
Key = Vmm.GetPhysicalAddress(Key);
if (Key == -1)
{
//FIXME: Should'nt ignore invalid addresses.
return;
}
if (IsFrameBufferPosition(Key))
{
//This texture is a frame buffer texture,
//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.FrameBuffer.BindTexture(Key, TexIndex);
}
else
{
GalImage NewImage = TextureFactory.MakeTexture(Vmm, TicPosition);
long Size = (uint)TextureHelper.GetTextureSize(NewImage);
bool HasCachedTexture = false;
if (Gpu.Renderer.Texture.TryGetCachedTexture(Key, Size, out GalImage Image))
{
if (NewImage.Equals(Image) && !QueryKeyUpload(Vmm, Key, Size, NvGpuBufferType.Texture))
{
Gpu.Renderer.Texture.Bind(Key, TexIndex);
HasCachedTexture = true;
}
}
if (!HasCachedTexture)
{
byte[] Data = TextureFactory.GetTextureData(Vmm, TicPosition);
Gpu.Renderer.Texture.Create(Key, Data, NewImage);
}
Gpu.Renderer.Texture.Bind(Key, TexIndex);
}
Gpu.Renderer.Texture.SetSampler(Sampler);
}
private void UploadVertexArrays(NvGpuVmm Vmm, GalPipelineState State)
{
long IbPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long IboKey = Vmm.GetPhysicalAddress(IbPosition);
int IndexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int IndexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
int PrimCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType PrimType = (GalPrimitiveType)(PrimCtrl & 0xffff);
GalIndexFormat IndexFormat = (GalIndexFormat)IndexEntryFmt;
int IndexEntrySize = 1 << IndexEntryFmt;
if (IndexEntrySize > 4)
{
throw new InvalidOperationException("Invalid index entry size \"" + IndexEntrySize + "\"!");
}
if (IndexCount != 0)
{
int IbSize = IndexCount * IndexEntrySize;
bool IboCached = Gpu.Renderer.Rasterizer.IsIboCached(IboKey, (uint)IbSize);
bool UsesLegacyQuads =
PrimType == GalPrimitiveType.Quads ||
PrimType == GalPrimitiveType.QuadStrip;
if (!IboCached || Gpu.ResourceManager.MemoryRegionModified(Vmm, IboKey, (uint)IbSize, NvGpuBufferType.Index))
{
if (!UsesLegacyQuads)
{
if (Vmm.TryGetHostAddress(IbPosition, IbSize, out IntPtr IbPtr))
{
Gpu.Renderer.Rasterizer.CreateIbo(IboKey, IbSize, IbPtr);
}
else
{
Gpu.Renderer.Rasterizer.CreateIbo(IboKey, IbSize, Vmm.ReadBytes(IbPosition, IbSize));
}
}
else
{
byte[] Buffer = Vmm.ReadBytes(IbPosition, IbSize);
if (PrimType == GalPrimitiveType.Quads)
{
Buffer = QuadHelper.ConvertIbQuadsToTris(Buffer, IndexEntrySize, IndexCount);
}
else /* if (PrimType == GalPrimitiveType.QuadStrip) */
{
Buffer = QuadHelper.ConvertIbQuadStripToTris(Buffer, IndexEntrySize, IndexCount);
}
Gpu.Renderer.Rasterizer.CreateIbo(IboKey, IbSize, Buffer);
}
}
if (!UsesLegacyQuads)
{
Gpu.Renderer.Rasterizer.SetIndexArray(IbSize, IndexFormat);
}
else
{
if (PrimType == GalPrimitiveType.Quads)
{
Gpu.Renderer.Rasterizer.SetIndexArray(QuadHelper.ConvertIbSizeQuadsToTris(IbSize), IndexFormat);
}
else /* if (PrimType == GalPrimitiveType.QuadStrip) */
{
Gpu.Renderer.Rasterizer.SetIndexArray(QuadHelper.ConvertIbSizeQuadStripToTris(IbSize), IndexFormat);
}
}
}
List <GalVertexAttrib>[] Attribs = new List <GalVertexAttrib> [32];
for (int Attr = 0; Attr < 16; Attr++)
{
int Packed = ReadRegister(NvGpuEngine3dReg.VertexAttribNFormat + Attr);
int ArrayIndex = Packed & 0x1f;
if (Attribs[ArrayIndex] == null)
{
Attribs[ArrayIndex] = new List <GalVertexAttrib>();
}
long VbPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + ArrayIndex * 4);
if (VbPosition == 0)
{
continue;
}
bool IsConst = ((Packed >> 6) & 1) != 0;
int Offset = (Packed >> 7) & 0x3fff;
GalVertexAttribSize Size = (GalVertexAttribSize)((Packed >> 21) & 0x3f);
GalVertexAttribType Type = (GalVertexAttribType)((Packed >> 27) & 0x7);
bool IsRgba = ((Packed >> 31) & 1) != 0;
//Note: 16 is the maximum size of an attribute,
//having a component size of 32-bits with 4 elements (a vec4).
byte[] Data = Vmm.ReadBytes(VbPosition + Offset, 16);
Attribs[ArrayIndex].Add(new GalVertexAttrib(Attr, IsConst, Offset, Data, Size, Type, IsRgba));
}
State.VertexBindings = new GalVertexBinding[32];
for (int Index = 0; Index < 32; Index++)
{
if (Attribs[Index] == null)
{
continue;
}
int Control = ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + Index * 4);
bool Enable = (Control & 0x1000) != 0;
if (!Enable)
{
continue;
}
long VbPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + Index * 4);
long VbEndPos = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNEndAddr + Index * 2);
int VertexDivisor = ReadRegister(NvGpuEngine3dReg.VertexArrayNDivisor + Index * 4);
bool Instanced = ReadRegisterBool(NvGpuEngine3dReg.VertexArrayNInstance + Index);
int Stride = Control & 0xfff;
if (Instanced && VertexDivisor != 0)
{
VbPosition += Stride * (CurrentInstance / VertexDivisor);
}
if (VbPosition > VbEndPos)
{
//Instance is invalid, ignore the draw call
continue;
}
long VboKey = Vmm.GetPhysicalAddress(VbPosition);
long VbSize = (VbEndPos - VbPosition) + 1;
bool VboCached = Gpu.Renderer.Rasterizer.IsVboCached(VboKey, VbSize);
if (!VboCached || Gpu.ResourceManager.MemoryRegionModified(Vmm, VboKey, VbSize, NvGpuBufferType.Vertex))
{
if (Vmm.TryGetHostAddress(VbPosition, VbSize, out IntPtr VbPtr))
{
Gpu.Renderer.Rasterizer.CreateVbo(VboKey, (int)VbSize, VbPtr);
}
else
{
Gpu.Renderer.Rasterizer.CreateVbo(VboKey, Vmm.ReadBytes(VbPosition, VbSize));
}
}
State.VertexBindings[Index].Enabled = true;
State.VertexBindings[Index].Stride = Stride;
State.VertexBindings[Index].VboKey = VboKey;
State.VertexBindings[Index].Instanced = Instanced;
State.VertexBindings[Index].Divisor = VertexDivisor;
State.VertexBindings[Index].Attribs = Attribs[Index].ToArray();
}
}
private void TextureCopy(NvGpuVmm Vmm)
{
CopyOperation Operation = (CopyOperation)ReadRegister(NvGpuEngine2dReg.CopyOperation);
int DstFormat = ReadRegister(NvGpuEngine2dReg.DstFormat);
bool DstLinear = ReadRegister(NvGpuEngine2dReg.DstLinear) != 0;
int DstWidth = ReadRegister(NvGpuEngine2dReg.DstWidth);
int DstHeight = ReadRegister(NvGpuEngine2dReg.DstHeight);
int DstPitch = ReadRegister(NvGpuEngine2dReg.DstPitch);
int DstBlkDim = ReadRegister(NvGpuEngine2dReg.DstBlockDimensions);
int SrcFormat = ReadRegister(NvGpuEngine2dReg.SrcFormat);
bool SrcLinear = ReadRegister(NvGpuEngine2dReg.SrcLinear) != 0;
int SrcWidth = ReadRegister(NvGpuEngine2dReg.SrcWidth);
int SrcHeight = ReadRegister(NvGpuEngine2dReg.SrcHeight);
int SrcPitch = ReadRegister(NvGpuEngine2dReg.SrcPitch);
int SrcBlkDim = ReadRegister(NvGpuEngine2dReg.SrcBlockDimensions);
int DstBlitX = ReadRegister(NvGpuEngine2dReg.BlitDstX);
int DstBlitY = ReadRegister(NvGpuEngine2dReg.BlitDstY);
int DstBlitW = ReadRegister(NvGpuEngine2dReg.BlitDstW);
int DstBlitH = ReadRegister(NvGpuEngine2dReg.BlitDstH);
long BlitDuDx = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitDuDxFract);
long BlitDvDy = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitDvDyFract);
long SrcBlitX = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitSrcXFract);
long SrcBlitY = ReadRegisterFixed1_31_32(NvGpuEngine2dReg.BlitSrcYFract);
GalImageFormat SrcImgFormat = ImageUtils.ConvertSurface((GalSurfaceFormat)SrcFormat);
GalImageFormat DstImgFormat = ImageUtils.ConvertSurface((GalSurfaceFormat)DstFormat);
GalMemoryLayout SrcLayout = GetLayout(SrcLinear);
GalMemoryLayout DstLayout = GetLayout(DstLinear);
int SrcBlockHeight = 1 << ((SrcBlkDim >> 4) & 0xf);
int DstBlockHeight = 1 << ((DstBlkDim >> 4) & 0xf);
long SrcAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.SrcAddress);
long DstAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.DstAddress);
long SrcKey = Vmm.GetPhysicalAddress(SrcAddress);
long DstKey = Vmm.GetPhysicalAddress(DstAddress);
GalImage SrcTexture = new GalImage(
SrcWidth,
SrcHeight, 1,
SrcBlockHeight,
SrcLayout,
SrcImgFormat);
GalImage DstTexture = new GalImage(
DstWidth,
DstHeight, 1,
DstBlockHeight,
DstLayout,
DstImgFormat);
SrcTexture.Pitch = SrcPitch;
DstTexture.Pitch = DstPitch;
Gpu.ResourceManager.SendTexture(Vmm, SrcKey, SrcTexture);
Gpu.ResourceManager.SendTexture(Vmm, DstKey, DstTexture);
int SrcBlitX1 = (int)(SrcBlitX >> 32);
int SrcBlitY1 = (int)(SrcBlitY >> 32);
int SrcBlitX2 = (int)(SrcBlitX + DstBlitW * BlitDuDx >> 32);
int SrcBlitY2 = (int)(SrcBlitY + DstBlitH * BlitDvDy >> 32);
Gpu.Renderer.RenderTarget.Copy(
SrcKey,
DstKey,
SrcBlitX1,
SrcBlitY1,
SrcBlitX2,
SrcBlitY2,
DstBlitX,
DstBlitY,
DstBlitX + DstBlitW,
DstBlitY + DstBlitH);
//Do a guest side copy aswell. This is necessary when
//the texture is modified by the guest, however it doesn't
//work when resources that the gpu can write to are copied,
//like framebuffers.
ImageUtils.CopyTexture(
Vmm,
SrcTexture,
DstTexture,
SrcAddress,
DstAddress,
SrcBlitX1,
SrcBlitY1,
DstBlitX,
DstBlitY,
DstBlitW,
DstBlitH);
Vmm.IsRegionModified(DstKey, ImageUtils.GetSize(DstTexture), NvGpuBufferType.Texture);
}
private void UploadVertexArrays(NvGpuVmm vmm, GalPipelineState state)
{
Profile.Begin(Profiles.GPU.Engine3d.UploadVertexArrays);
long ibPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long iboKey = vmm.GetPhysicalAddress(ibPosition);
int indexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int indexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
int primCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType primType = (GalPrimitiveType)(primCtrl & 0xffff);
GalIndexFormat indexFormat = (GalIndexFormat)indexEntryFmt;
int indexEntrySize = 1 << indexEntryFmt;
if (indexEntrySize > 4)
{
throw new InvalidOperationException("Invalid index entry size \"" + indexEntrySize + "\"!");
}
if (indexCount != 0)
{
int ibSize = indexCount * indexEntrySize;
bool iboCached = _gpu.Renderer.Rasterizer.IsIboCached(iboKey, (uint)ibSize);
bool usesLegacyQuads =
primType == GalPrimitiveType.Quads ||
primType == GalPrimitiveType.QuadStrip;
if (!iboCached || _gpu.ResourceManager.MemoryRegionModified(vmm, iboKey, (uint)ibSize, NvGpuBufferType.Index))
{
if (!usesLegacyQuads)
{
if (vmm.TryGetHostAddress(ibPosition, ibSize, out IntPtr ibPtr))
{
_gpu.Renderer.Rasterizer.CreateIbo(iboKey, ibSize, ibPtr);
}
else
{
_gpu.Renderer.Rasterizer.CreateIbo(iboKey, ibSize, vmm.ReadBytes(ibPosition, ibSize));
}
}
else
{
byte[] buffer = vmm.ReadBytes(ibPosition, ibSize);
if (primType == GalPrimitiveType.Quads)
{
buffer = QuadHelper.ConvertQuadsToTris(buffer, indexEntrySize, indexCount);
}
else /* if (PrimType == GalPrimitiveType.QuadStrip) */
{
buffer = QuadHelper.ConvertQuadStripToTris(buffer, indexEntrySize, indexCount);
}
_gpu.Renderer.Rasterizer.CreateIbo(iboKey, ibSize, buffer);
}
}
if (!usesLegacyQuads)
{
_gpu.Renderer.Rasterizer.SetIndexArray(ibSize, indexFormat);
}
else
{
if (primType == GalPrimitiveType.Quads)
{
_gpu.Renderer.Rasterizer.SetIndexArray(QuadHelper.ConvertSizeQuadsToTris(ibSize), indexFormat);
}
else /* if (PrimType == GalPrimitiveType.QuadStrip) */
{
_gpu.Renderer.Rasterizer.SetIndexArray(QuadHelper.ConvertSizeQuadStripToTris(ibSize), indexFormat);
}
}
}
List <GalVertexAttrib>[] attribs = new List <GalVertexAttrib> [32];
for (int attr = 0; attr < 16; attr++)
{
int packed = ReadRegister(NvGpuEngine3dReg.VertexAttribNFormat + attr);
int arrayIndex = packed & 0x1f;
if (attribs[arrayIndex] == null)
{
attribs[arrayIndex] = new List <GalVertexAttrib>();
}
long vbPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + arrayIndex * 4);
if (vbPosition == 0)
{
continue;
}
bool isConst = ((packed >> 6) & 1) != 0;
int offset = (packed >> 7) & 0x3fff;
GalVertexAttribSize size = (GalVertexAttribSize)((packed >> 21) & 0x3f);
GalVertexAttribType type = (GalVertexAttribType)((packed >> 27) & 0x7);
bool isRgba = ((packed >> 31) & 1) != 0;
// Check vertex array is enabled to avoid out of bounds exception when reading bytes
bool enable = (ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + arrayIndex * 4) & 0x1000) != 0;
// Note: 16 is the maximum size of an attribute,
// having a component size of 32-bits with 4 elements (a vec4).
if (enable)
{
byte[] data = vmm.ReadBytes(vbPosition + offset, 16);
attribs[arrayIndex].Add(new GalVertexAttrib(attr, isConst, offset, data, size, type, isRgba));
}
}
state.VertexBindings = new GalVertexBinding[32];
for (int index = 0; index < 32; index++)
{
if (attribs[index] == null)
{
continue;
}
int control = ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + index * 4);
bool enable = (control & 0x1000) != 0;
if (!enable)
{
continue;
}
long vbPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + index * 4);
long vbEndPos = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNEndAddr + index * 2);
int vertexDivisor = ReadRegister(NvGpuEngine3dReg.VertexArrayNDivisor + index * 4);
bool instanced = ReadRegisterBool(NvGpuEngine3dReg.VertexArrayNInstance + index);
int stride = control & 0xfff;
if (instanced && vertexDivisor != 0)
{
vbPosition += stride * (_currentInstance / vertexDivisor);
}
if (vbPosition > vbEndPos)
{
// Instance is invalid, ignore the draw call
continue;
}
long vboKey = vmm.GetPhysicalAddress(vbPosition);
long vbSize = (vbEndPos - vbPosition) + 1;
int modifiedVbSize = (int)vbSize;
// If quads convert size to triangle length
if (stride == 0)
{
if (primType == GalPrimitiveType.Quads)
{
modifiedVbSize = QuadHelper.ConvertSizeQuadsToTris(modifiedVbSize);
}
else if (primType == GalPrimitiveType.QuadStrip)
{
modifiedVbSize = QuadHelper.ConvertSizeQuadStripToTris(modifiedVbSize);
}
}
bool vboCached = _gpu.Renderer.Rasterizer.IsVboCached(vboKey, modifiedVbSize);
if (!vboCached || _gpu.ResourceManager.MemoryRegionModified(vmm, vboKey, vbSize, NvGpuBufferType.Vertex))
{
if ((primType == GalPrimitiveType.Quads | primType == GalPrimitiveType.QuadStrip) && stride != 0)
{
// Convert quad buffer to triangles
byte[] data = vmm.ReadBytes(vbPosition, vbSize);
if (primType == GalPrimitiveType.Quads)
{
data = QuadHelper.ConvertQuadsToTris(data, stride, (int)(vbSize / stride));
}
else
{
data = QuadHelper.ConvertQuadStripToTris(data, stride, (int)(vbSize / stride));
}
_gpu.Renderer.Rasterizer.CreateVbo(vboKey, data);
}
else if (vmm.TryGetHostAddress(vbPosition, vbSize, out IntPtr vbPtr))
{
_gpu.Renderer.Rasterizer.CreateVbo(vboKey, (int)vbSize, vbPtr);
}
else
{
_gpu.Renderer.Rasterizer.CreateVbo(vboKey, vmm.ReadBytes(vbPosition, vbSize));
}
}
state.VertexBindings[index].Enabled = true;
state.VertexBindings[index].Stride = stride;
state.VertexBindings[index].VboKey = vboKey;
state.VertexBindings[index].Instanced = instanced;
state.VertexBindings[index].Divisor = vertexDivisor;
state.VertexBindings[index].Attribs = attribs[index].ToArray();
}
Profile.End(Profiles.GPU.Engine3d.UploadVertexArrays);
}
private void Execute(NvGpuVmm vmm, GpuMethodCall methCall)
{
Profile.Begin(Profiles.GPU.EngineM2mf.Execute);
//TODO: Some registers and copy modes are still not implemented.
int control = methCall.Argument;
bool srcLinear = ((control >> 7) & 1) != 0;
bool dstLinear = ((control >> 8) & 1) != 0;
bool copy2D = ((control >> 9) & 1) != 0;
long srcAddress = MakeInt64From2xInt32(NvGpuEngineM2mfReg.SrcAddress);
long dstAddress = MakeInt64From2xInt32(NvGpuEngineM2mfReg.DstAddress);
int srcPitch = ReadRegister(NvGpuEngineM2mfReg.SrcPitch);
int dstPitch = ReadRegister(NvGpuEngineM2mfReg.DstPitch);
int xCount = ReadRegister(NvGpuEngineM2mfReg.XCount);
int yCount = ReadRegister(NvGpuEngineM2mfReg.YCount);
int swizzle = ReadRegister(NvGpuEngineM2mfReg.Swizzle);
int dstBlkDim = ReadRegister(NvGpuEngineM2mfReg.DstBlkDim);
int dstSizeX = ReadRegister(NvGpuEngineM2mfReg.DstSizeX);
int dstSizeY = ReadRegister(NvGpuEngineM2mfReg.DstSizeY);
int dstSizeZ = ReadRegister(NvGpuEngineM2mfReg.DstSizeZ);
int dstPosXY = ReadRegister(NvGpuEngineM2mfReg.DstPosXY);
int dstPosZ = ReadRegister(NvGpuEngineM2mfReg.DstPosZ);
int srcBlkDim = ReadRegister(NvGpuEngineM2mfReg.SrcBlkDim);
int srcSizeX = ReadRegister(NvGpuEngineM2mfReg.SrcSizeX);
int srcSizeY = ReadRegister(NvGpuEngineM2mfReg.SrcSizeY);
int srcSizeZ = ReadRegister(NvGpuEngineM2mfReg.SrcSizeZ);
int srcPosXY = ReadRegister(NvGpuEngineM2mfReg.SrcPosXY);
int srcPosZ = ReadRegister(NvGpuEngineM2mfReg.SrcPosZ);
int srcCpp = ((swizzle >> 20) & 7) + 1;
int dstCpp = ((swizzle >> 24) & 7) + 1;
int dstPosX = (dstPosXY >> 0) & 0xffff;
int dstPosY = (dstPosXY >> 16) & 0xffff;
int srcPosX = (srcPosXY >> 0) & 0xffff;
int srcPosY = (srcPosXY >> 16) & 0xffff;
int srcBlockHeight = 1 << ((srcBlkDim >> 4) & 0xf);
int dstBlockHeight = 1 << ((dstBlkDim >> 4) & 0xf);
long srcPa = vmm.GetPhysicalAddress(srcAddress);
long dstPa = vmm.GetPhysicalAddress(dstAddress);
if (copy2D)
{
if (srcLinear)
{
srcPosX = srcPosY = srcPosZ = 0;
}
if (dstLinear)
{
dstPosX = dstPosY = dstPosZ = 0;
}
if (srcLinear && dstLinear)
{
for (int y = 0; y < yCount; y++)
{
int srcOffset = (srcPosY + y) * srcPitch + srcPosX * srcCpp;
int dstOffset = (dstPosY + y) * dstPitch + dstPosX * dstCpp;
long src = srcPa + (uint)srcOffset;
long dst = dstPa + (uint)dstOffset;
vmm.Memory.CopyBytes(src, dst, xCount * srcCpp);
}
}
else
{
ISwizzle srcSwizzle;
if (srcLinear)
{
srcSwizzle = new LinearSwizzle(srcPitch, srcCpp, srcSizeX, srcSizeY);
}
else
{
srcSwizzle = new BlockLinearSwizzle(
srcSizeX,
srcSizeY, 1,
srcBlockHeight, 1,
srcCpp);
}
ISwizzle dstSwizzle;
if (dstLinear)
{
dstSwizzle = new LinearSwizzle(dstPitch, dstCpp, srcSizeX, srcSizeY);
}
else
{
dstSwizzle = new BlockLinearSwizzle(
dstSizeX,
dstSizeY, 1,
dstBlockHeight, 1,
dstCpp);
}
// Calculate the bits per pixel
int bpp = srcPitch / xCount;
// Copying all the bits at the same time corrupts the texture, unknown why but probably because the texture isn't linear
// To avoid this we will simply loop more times to cover all the bits,
// this allows up to recalculate the memory locations for each iteration around the loop
xCount *= bpp / srcCpp;
for (int y = 0; y < yCount; y++)
{
for (int x = 0; x < xCount; x++)
{
int srcOffset = srcSwizzle.GetSwizzleOffset(srcPosX + x, srcPosY + y, 0);
int dstOffset = dstSwizzle.GetSwizzleOffset(dstPosX + x, dstPosY + y, 0);
long src = srcPa + (uint)srcOffset;
long dst = dstPa + (uint)dstOffset;
vmm.Memory.CopyBytes(src, dst, srcCpp);
}
}
}
}
else
{
vmm.Memory.CopyBytes(srcPa, dstPa, xCount);
}
Profile.End(Profiles.GPU.EngineM2mf.Execute);
}
private void DispatchRender(NvGpuVmm vmm, GalPipelineState state)
{
int indexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
int primCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType primType = (GalPrimitiveType)(primCtrl & 0xffff);
bool instanceNext = ((primCtrl >> 26) & 1) != 0;
bool instanceCont = ((primCtrl >> 27) & 1) != 0;
if (instanceNext && instanceCont)
{
throw new InvalidOperationException("GPU tried to increase and reset instance count at the same time");
}
if (instanceNext)
{
_currentInstance++;
}
else if (!instanceCont)
{
_currentInstance = 0;
}
state.Instance = _currentInstance;
_gpu.Renderer.Pipeline.Bind(state);
_gpu.Renderer.RenderTarget.Bind();
if (indexCount != 0)
{
int indexEntryFmt = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int indexFirst = ReadRegister(NvGpuEngine3dReg.IndexBatchFirst);
int vertexBase = ReadRegister(NvGpuEngine3dReg.VertexArrayElemBase);
long indexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
long iboKey = vmm.GetPhysicalAddress(indexPosition);
// Quad primitive types were deprecated on OpenGL 3.x,
// they are converted to a triangles index buffer on IB creation,
// so we should use the triangles type here too.
if (primType == GalPrimitiveType.Quads || primType == GalPrimitiveType.QuadStrip)
{
// Note: We assume that index first points to the first
// vertex of a quad, if it points to the middle of a
// quad (First % 4 != 0 for Quads) then it will not work properly.
if (primType == GalPrimitiveType.Quads)
{
indexFirst = QuadHelper.ConvertSizeQuadsToTris(indexFirst);
}
else // QuadStrip
{
indexFirst = QuadHelper.ConvertSizeQuadStripToTris(indexFirst);
}
primType = GalPrimitiveType.Triangles;
}
_gpu.Renderer.Rasterizer.DrawElements(iboKey, indexFirst, vertexBase, primType);
}
else
{
int vertexFirst = ReadRegister(NvGpuEngine3dReg.VertexArrayFirst);
int vertexCount = ReadRegister(NvGpuEngine3dReg.VertexArrayCount);
// Quad primitive types were deprecated on OpenGL 3.x,
// they are converted to a triangles index buffer on IB creation,
// so we should use the triangles type here too.
if (primType == GalPrimitiveType.Quads || primType == GalPrimitiveType.QuadStrip)
{
// Note: We assume that index first points to the first
// vertex of a quad, if it points to the middle of a
// quad (First % 4 != 0 for Quads) then it will not work properly.
if (primType == GalPrimitiveType.Quads)
{
vertexFirst = QuadHelper.ConvertSizeQuadsToTris(vertexFirst);
}
else // QuadStrip
{
vertexFirst = QuadHelper.ConvertSizeQuadStripToTris(vertexFirst);
}
primType = GalPrimitiveType.Triangles;
vertexCount = QuadHelper.ConvertSizeQuadsToTris(vertexCount);
}
_gpu.Renderer.Rasterizer.DrawArrays(vertexFirst, vertexCount, primType);
}
// Reset pipeline for host OpenGL calls
_gpu.Renderer.Pipeline.Unbind(state);
// Is the GPU really clearing those registers after draw?
WriteRegister(NvGpuEngine3dReg.IndexBatchFirst, 0);
WriteRegister(NvGpuEngine3dReg.IndexBatchCount, 0);
}