public void CallMethod(NvGpuVmm Vmm, NvGpuPBEntry PBEntry)
{
if (Methods.TryGetValue(PBEntry.Method, out NvGpuMethod Method))
{
Method(Vmm, PBEntry);
}
else
{
WriteRegister(PBEntry);
}
}
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);
}
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 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;
//TODO: Find a better way to calculate the size.
int Size = 0x20000;
byte[] Code = Vmm.ReadBytes(Tag, Size);
GalShaderType ShaderType = GetTypeFromProgram(Index);
Tags[(int)ShaderType] = Tag;
Gpu.Renderer.CreateShader(Tag, ShaderType, Code);
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(NvGpu Gpu, 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);
long TextureAddress = (uint)Tic[1];
TextureAddress |= (long)((ushort)Tic[2]) << 32;
TextureSwizzle Swizzle = (TextureSwizzle)((Tic[2] >> 21) & 7);
int Pitch = (Tic[3] & 0xffff) << 5;
int BlockHeightLog2 = (Tic[3] >> 3) & 7;
int BlockHeight = 1 << BlockHeightLog2;
int Width = (Tic[4] & 0xffff) + 1;
int Height = (Tic[5] & 0xffff) + 1;
Texture Texture = new Texture(
TextureAddress,
Width,
Height,
Pitch,
BlockHeight,
Swizzle,
Format);
byte[] Data = TextureReader.Read(Vmm, Texture);
return(new GalTexture(
Data,
Width,
Height,
Format,
XSource,
YSource,
ZSource,
WSource));
}
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, long Position, int Param)
{
Reset();
Gprs[1] = Param;
Pc = Position;
FetchOpCode(Vmm);
while (Step(Vmm))
{
;
}
//Due to the delay slot, we still need to execute
//one more instruction before we actually exit.
Step(Vmm);
}
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;
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 Texture = TextureFactory.MakeTexture(Gpu, Vmm, TicPosition);
Gpu.Renderer.SetTextureAndSampler(TexIndex, Texture, Sampler);
Gpu.Renderer.BindTexture(TexIndex);
}
}
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 bool Step(NvGpuVmm Vmm)
{
long BaseAddr = Pc - 4;
FetchOpCode(Vmm);
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() << 2);
bool NoDelays = (OpCode & 0x20) != 0;
if (NoDelays)
{
FetchOpCode(Vmm);
}
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 BufferSize = IndexCount * IndexSize;
byte[] Data = Vmm.ReadBytes(IndexPosition, BufferSize);
Gpu.Renderer.SetIndexArray(Data, 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));
}
for (int Index = 0; Index < 32; Index++)
{
int VertexFirst = ReadRegister(NvGpuEngine3dReg.VertexArrayFirst);
int VertexCount = ReadRegister(NvGpuEngine3dReg.VertexArrayCount);
int Control = ReadRegister(NvGpuEngine3dReg.VertexArrayNControl + Index * 4);
bool Enable = (Control & 0x1000) != 0;
long VertexPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.VertexArrayNAddress + Index * 4);
if (!Enable)
{
continue;
}
int Stride = Control & 0xfff;
long Size = 0;
if (IndexCount != 0)
{
Size = GetVertexCountFromIndexBuffer(
Vmm,
IndexPosition,
IndexCount,
IndexSize);
}
else
{
Size = VertexCount;
}
//TODO: Support cases where the Stride is 0.
//In this case, we need to use the size of the attribute.
Size *= Stride;
byte[] Data = Vmm.ReadBytes(VertexPosition, Size);
GalVertexAttrib[] AttribArray = Attribs[Index]?.ToArray() ?? new GalVertexAttrib[0];
Gpu.Renderer.SetVertexArray(Index, Stride, Data, AttribArray);
int PrimCtrl = ReadRegister(NvGpuEngine3dReg.VertexBeginGl);
GalPrimitiveType PrimType = (GalPrimitiveType)(PrimCtrl & 0xffff);
if (IndexCount != 0)
{
Gpu.Renderer.DrawElements(Index, IndexFirst, PrimType);
}
else
{
Gpu.Renderer.DrawArrays(Index, VertexFirst, VertexCount, PrimType);
}
}
}
private void CopyTexture(NvGpuVmm Vmm, Texture Texture, byte[] Buffer)
{
TextureWriter.Write(Vmm, Texture, Buffer);
}
