private void UploadTexture(NvGpuVmm Vmm, long BasePosition, int TexIndex, int HndIndex)
{
long Position = BasePosition + HndIndex * 4;
int TextureHandle = Vmm.ReadInt32(Position);
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 TextureAddress = Vmm.ReadInt64(TicPosition + 4) & 0xffffffffffff;
long Tag = TextureAddress;
TextureAddress = Vmm.GetPhysicalAddress(TextureAddress);
if (IsFrameBufferPosition(TextureAddress))
{
//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.BindFrameBufferTexture(TextureAddress, TexIndex, Sampler);
}
else
{
GalTexture NewTexture = TextureFactory.MakeTexture(Vmm, TicPosition);
long Size = (uint)TextureHelper.GetTextureSize(NewTexture);
if (Gpu.Renderer.TryGetCachedTexture(Tag, Size, out GalTexture Texture))
{
if (NewTexture.Equals(Texture) && !Vmm.IsRegionModified(Tag, Size, NvGpuBufferType.Texture))
{
Gpu.Renderer.BindTexture(Tag, TexIndex);
return;
}
}
byte[] Data = TextureFactory.GetTextureData(Vmm, TicPosition);
Gpu.Renderer.SetTextureAndSampler(Tag, Data, NewTexture, Sampler);
Gpu.Renderer.BindTexture(Tag, TexIndex);
}
}
private static int[] ReadWords(NvGpuVmm Vmm, long Position, int Count)
{
int[] Words = new int[Count];
for (int Index = 0; Index < Count; Index++, Position += 4)
{
Words[Index] = Vmm.ReadInt32(Position);
}
return(Words);
}
public void CallMethod(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
if (Methods.TryGetValue(PBEntry.Method, out NvGpuMethod Method))
{
Method(Vmm, PBEntry);
}
else
{
WriteRegister(PBEntry);
}
}
private void VertexEndGl(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
SetFrameBuffer(Vmm, 0);
long[] Tags = UploadShaders(Vmm);
Gpu.Renderer.BindProgram();
SetAlphaBlending();
UploadTextures(Vmm, Tags);
UploadUniforms(Vmm);
UploadVertexArrays(Vmm);
}
private void ClearBuffers(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
int Arg0 = PBEntry.Arguments[0];
int FbIndex = (Arg0 >> 6) & 0xf;
int Layer = (Arg0 >> 10) & 0x3ff;
GalClearBufferFlags Flags = (GalClearBufferFlags)(Arg0 & 0x3f);
SetFrameBuffer(Vmm, 0);
//TODO: Enable this once the frame buffer problems are fixed.
//Gpu.Renderer.ClearBuffers(Layer, Flags);
}
private void CbData(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
long Position = MakeInt64From2xInt32(NvGpuEngine3dReg.ConstBufferAddress);
int Offset = ReadRegister(NvGpuEngine3dReg.ConstBufferOffset);
foreach (int Arg in PBEntry.Arguments)
{
Vmm.WriteInt32(Position + Offset, Arg);
Offset += 4;
}
WriteRegister(NvGpuEngine3dReg.ConstBufferOffset, Offset);
}
private void SetFrameBuffer(NvGpuVmm Vmm, int FbIndex)
{
long VA = MakeInt64From2xInt32(NvGpuEngine3dReg.FrameBufferNAddress + FbIndex * 0x10);
long PA = Vmm.GetPhysicalAddress(VA);
FrameBuffers.Add(PA);
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.CreateFrameBuffer(PA, 1280, 720);
Gpu.Renderer.BindFrameBuffer(PA);
}
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);
ConstBuffers[Stage][Index].Position = Position;
ConstBuffers[Stage][Index].Enabled = Enabled;
ConstBuffers[Stage][Index].Size = ReadRegister(NvGpuEngine3dReg.ConstBufferSize);
}
private void QueryControl(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
long Position = MakeInt64From2xInt32(NvGpuEngine3dReg.QueryAddress);
int Seq = Registers[(int)NvGpuEngine3dReg.QuerySequence];
int Ctrl = Registers[(int)NvGpuEngine3dReg.QueryControl];
int Mode = Ctrl & 3;
if (Mode == 0)
{
//Write mode.
Vmm.WriteInt32(Position, Seq);
}
WriteRegister(PBEntry);
}
public void Execute(NvGpuVmm Vmm, int[] Mme, int Position, int Param)
{
Reset();
Gprs[1] = Param;
Pc = Position;
FetchOpCode(Mme);
while (Step(Vmm, Mme))
{
;
}
//Due to the delay slot, we still need to execute
//one more instruction before we actually exit.
Step(Vmm, Mme);
}
private long[] UploadShaders(NvGpuVmm Vmm)
{
long[] Tags = new long[5];
long BasePosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
for (int Index = 0; Index < 6; Index++)
{
int Control = ReadRegister(NvGpuEngine3dReg.ShaderNControl + Index * 0x10);
int Offset = ReadRegister(NvGpuEngine3dReg.ShaderNOffset + Index * 0x10);
//Note: Vertex Program (B) is always enabled.
bool Enable = (Control & 1) != 0 || Index == 1;
if (!Enable)
{
continue;
}
long Tag = BasePosition + (uint)Offset;
GalShaderType ShaderType = GetTypeFromProgram(Index);
Tags[(int)ShaderType] = Tag;
Gpu.Renderer.CreateShader(Vmm, Tag, ShaderType);
Gpu.Renderer.BindShader(Tag);
}
int RawSX = ReadRegister(NvGpuEngine3dReg.ViewportScaleX);
int RawSY = ReadRegister(NvGpuEngine3dReg.ViewportScaleY);
float SX = BitConverter.Int32BitsToSingle(RawSX);
float SY = BitConverter.Int32BitsToSingle(RawSY);
float SignX = MathF.Sign(SX);
float SignY = MathF.Sign(SY);
Gpu.Renderer.SetUniform2F(GalConsts.FlipUniformName, SignX, SignY);
return(Tags);
}
public static GalTexture MakeTexture(NvGpuVmm Vmm, long TicPosition)
{
int[] Tic = ReadWords(Vmm, TicPosition, 8);
GalTextureFormat Format = (GalTextureFormat)(Tic[0] & 0x7f);
GalTextureSource XSource = (GalTextureSource)((Tic[0] >> 19) & 7);
GalTextureSource YSource = (GalTextureSource)((Tic[0] >> 22) & 7);
GalTextureSource ZSource = (GalTextureSource)((Tic[0] >> 25) & 7);
GalTextureSource WSource = (GalTextureSource)((Tic[0] >> 28) & 7);
int Width = (Tic[4] & 0xffff) + 1;
int Height = (Tic[5] & 0xffff) + 1;
return(new GalTexture(
Width,
Height,
Format,
XSource,
YSource,
ZSource,
WSource));
}
private void UploadTextures(NvGpuVmm Vmm, long[] Tags)
{
long BaseShPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
int TextureCbIndex = ReadRegister(NvGpuEngine3dReg.TextureCbIndex);
//Note: On the emulator renderer, Texture Unit 0 is
//reserved for drawing the frame buffer.
int TexIndex = 1;
for (int Index = 0; Index < Tags.Length; Index++)
{
foreach (ShaderDeclInfo DeclInfo in Gpu.Renderer.GetTextureUsage(Tags[Index]))
{
long Position = ConstBuffers[Index][TextureCbIndex].Position;
UploadTexture(Vmm, Position, TexIndex, DeclInfo.Index);
Gpu.Renderer.SetUniform1(DeclInfo.Name, TexIndex);
TexIndex++;
}
}
}
private void Execute(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
int Control = PBEntry.Arguments[0];
bool SrcLinear = ((Control >> 7) & 1) != 0;
bool DstLinear = ((Control >> 8) & 1) != 0;
long SrcAddress = MakeInt64From2xInt32(NvGpuEngineDmaReg.SrcAddress);
long DstAddress = MakeInt64From2xInt32(NvGpuEngineDmaReg.DstAddress);
int SrcPitch = ReadRegister(NvGpuEngineDmaReg.SrcPitch);
int DstPitch = ReadRegister(NvGpuEngineDmaReg.DstPitch);
int DstBlkDim = ReadRegister(NvGpuEngineDmaReg.DstBlkDim);
int DstSizeX = ReadRegister(NvGpuEngineDmaReg.DstSizeX);
int DstSizeY = ReadRegister(NvGpuEngineDmaReg.DstSizeY);
int DstSizeZ = ReadRegister(NvGpuEngineDmaReg.DstSizeZ);
int DstPosXY = ReadRegister(NvGpuEngineDmaReg.DstPosXY);
int DstPosZ = ReadRegister(NvGpuEngineDmaReg.DstPosZ);
int SrcBlkDim = ReadRegister(NvGpuEngineDmaReg.SrcBlkDim);
int SrcSizeX = ReadRegister(NvGpuEngineDmaReg.SrcSizeX);
int SrcSizeY = ReadRegister(NvGpuEngineDmaReg.SrcSizeY);
int SrcSizeZ = ReadRegister(NvGpuEngineDmaReg.SrcSizeZ);
int SrcPosXY = ReadRegister(NvGpuEngineDmaReg.SrcPosXY);
int SrcPosZ = ReadRegister(NvGpuEngineDmaReg.SrcPosZ);
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);
ISwizzle SrcSwizzle;
if (SrcLinear)
{
SrcSwizzle = new LinearSwizzle(SrcPitch, 1);
}
else
{
SrcSwizzle = new BlockLinearSwizzle(SrcSizeX, 1, SrcBlockHeight);
}
ISwizzle DstSwizzle;
if (DstLinear)
{
DstSwizzle = new LinearSwizzle(DstPitch, 1);
}
else
{
DstSwizzle = new BlockLinearSwizzle(DstSizeX, 1, DstBlockHeight);
}
for (int Y = 0; Y < DstSizeY; Y++)
{
for (int X = 0; X < DstSizeX; X++)
{
long SrcOffset = SrcAddress + (uint)SrcSwizzle.GetSwizzleOffset(X, Y);
long DstOffset = DstAddress + (uint)DstSwizzle.GetSwizzleOffset(X, Y);
Vmm.WriteByte(DstOffset, Vmm.ReadByte(SrcOffset));
}
}
}
private bool Step(NvGpuVmm Vmm, int[] Mme)
{
int BaseAddr = Pc - 1;
FetchOpCode(Mme);
if ((OpCode & 7) < 7)
{
//Operation produces a value.
AssignmentOperation AsgOp = (AssignmentOperation)((OpCode >> 4) & 7);
int Result = GetAluResult();
switch (AsgOp)
{
//Fetch parameter and ignore result.
case AssignmentOperation.IgnoreAndFetch:
{
SetDstGpr(FetchParam());
break;
}
//Move result.
case AssignmentOperation.Move:
{
SetDstGpr(Result);
break;
}
//Move result and use as Method Address.
case AssignmentOperation.MoveAndSetMaddr:
{
SetDstGpr(Result);
SetMethAddr(Result);
break;
}
//Fetch parameter and send result.
case AssignmentOperation.FetchAndSend:
{
SetDstGpr(FetchParam());
Send(Vmm, Result);
break;
}
//Move and send result.
case AssignmentOperation.MoveAndSend:
{
SetDstGpr(Result);
Send(Vmm, Result);
break;
}
//Fetch parameter and use result as Method Address.
case AssignmentOperation.FetchAndSetMaddr:
{
SetDstGpr(FetchParam());
SetMethAddr(Result);
break;
}
//Move result and use as Method Address, then fetch and send paramter.
case AssignmentOperation.MoveAndSetMaddrThenFetchAndSend:
{
SetDstGpr(Result);
SetMethAddr(Result);
Send(Vmm, FetchParam());
break;
}
//Move result and use as Method Address, then send bits 17:12 of result.
case AssignmentOperation.MoveAndSetMaddrThenSendHigh:
{
SetDstGpr(Result);
SetMethAddr(Result);
Send(Vmm, (Result >> 12) & 0x3f);
break;
}
}
}
else
{
//Branch.
bool OnNotZero = ((OpCode >> 4) & 1) != 0;
bool Taken = OnNotZero
? GetGprA() != 0
: GetGprA() == 0;
if (Taken)
{
Pc = BaseAddr + GetImm();
bool NoDelays = (OpCode & 0x20) != 0;
if (NoDelays)
{
FetchOpCode(Mme);
}
return(true);
}
}
bool Exit = (OpCode & 0x80) != 0;
return(!Exit);
}
private void UploadVertexArrays(NvGpuVmm Vmm)
{
long IndexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.IndexArrayAddress);
int IndexSize = ReadRegister(NvGpuEngine3dReg.IndexArrayFormat);
int IndexFirst = ReadRegister(NvGpuEngine3dReg.IndexBatchFirst);
int IndexCount = ReadRegister(NvGpuEngine3dReg.IndexBatchCount);
GalIndexFormat IndexFormat = (GalIndexFormat)IndexSize;
IndexSize = 1 << IndexSize;
if (IndexSize > 4)
{
throw new InvalidOperationException();
}
if (IndexSize != 0)
{
int IbSize = IndexCount * IndexSize;
bool IboCached = Gpu.Renderer.IsIboCached(IndexPosition, (uint)IbSize);
if (!IboCached || Vmm.IsRegionModified(IndexPosition, (uint)IbSize, NvGpuBufferType.Index))
{
byte[] Data = Vmm.ReadBytes(IndexPosition, (uint)IbSize);
Gpu.Renderer.CreateIbo(IndexPosition, Data);
}
Gpu.Renderer.SetIndexArray(IndexPosition, 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));
}
int VertexFirst = ReadRegister(NvGpuEngine3dReg.VertexArrayFirst);
int VertexCount = ReadRegister(NvGpuEngine3dReg.VertexArrayCount);
int PrimCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
for (int Index = 0; Index < 32; Index++)
{
if (Attribs[Index] == null)
{
continue;
}
int Control = ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + Index * 4);
bool Enable = (Control & 0x1000) != 0;
long VertexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + Index * 4);
long VertexEndPos = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNEndAddr + Index * 2);
if (!Enable)
{
continue;
}
int Stride = Control & 0xfff;
long VbSize = 0;
if (IndexCount != 0)
{
VbSize = (VertexEndPos - VertexPosition) + 1;
}
else
{
VbSize = VertexCount * Stride;
}
bool VboCached = Gpu.Renderer.IsVboCached(VertexPosition, VbSize);
if (!VboCached || Vmm.IsRegionModified(VertexPosition, VbSize, NvGpuBufferType.Vertex))
{
byte[] Data = Vmm.ReadBytes(VertexPosition, VbSize);
Gpu.Renderer.CreateVbo(VertexPosition, Data);
}
Gpu.Renderer.SetVertexArray(Index, Stride, VertexPosition, Attribs[Index].ToArray());
}
GalPrimitiveType PrimType = (GalPrimitiveType)(PrimCtrl & 0xffff);
if (IndexCount != 0)
{
Gpu.Renderer.DrawElements(IndexPosition, IndexFirst, PrimType);
}
else
{
Gpu.Renderer.DrawArrays(VertexFirst, VertexCount, PrimType);
}
}
private void TextureCopy(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
CopyOperation Operation = (CopyOperation)ReadRegister(NvGpuEngine2dReg.CopyOperation);
bool SrcLinear = ReadRegister(NvGpuEngine2dReg.SrcLinear) != 0;
int SrcWidth = ReadRegister(NvGpuEngine2dReg.SrcWidth);
int SrcHeight = ReadRegister(NvGpuEngine2dReg.SrcHeight);
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);
TextureSwizzle DstSwizzle = DstLinear
? TextureSwizzle.Pitch
: TextureSwizzle.BlockLinear;
int DstBlockHeight = 1 << ((DstBlkDim >> 4) & 0xf);
long Tag = Vmm.GetPhysicalAddress(MakeInt64From2xInt32(NvGpuEngine2dReg.SrcAddress));
long SrcAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.SrcAddress);
long DstAddress = MakeInt64From2xInt32(NvGpuEngine2dReg.DstAddress);
bool IsFbTexture = Gpu.Engine3d.IsFrameBufferPosition(Tag);
if (IsFbTexture && DstLinear)
{
DstSwizzle = TextureSwizzle.BlockLinear;
}
Texture DstTexture = new Texture(
DstAddress,
DstWidth,
DstHeight,
DstBlockHeight,
DstBlockHeight,
DstSwizzle,
GalTextureFormat.A8B8G8R8);
if (IsFbTexture)
{
//TODO: Change this when the correct frame buffer resolution is used.
//Currently, the frame buffer size is hardcoded to 1280x720.
SrcWidth = 1280;
SrcHeight = 720;
Gpu.Renderer.GetFrameBufferData(Tag, (byte[] Buffer) =>
{
CopyTexture(
Vmm,
DstTexture,
Buffer,
SrcWidth,
SrcHeight);
});
}
else
{
long Size = SrcWidth * SrcHeight * 4;
byte[] Buffer = Vmm.ReadBytes(SrcAddress, Size);
CopyTexture(
Vmm,
DstTexture,
Buffer,
SrcWidth,
SrcHeight);
}
}