private void UploadTextures(NvGpuVmm vmm, GalPipelineState state, long[] keys)
{
Profile.Begin(Profiles.GPU.Engine3d.UploadTextures);
long baseShPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
int textureCbIndex = ReadRegister(NvGpuEngine3dReg.TextureCbIndex);
List <(long, GalImage, GalTextureSampler)> unboundTextures = new List <(long, GalImage, GalTextureSampler)>();
for (int index = 0; index < keys.Length; index++)
{
foreach (TextureDescriptor desc in _gpu.Renderer.Shader.GetTextureUsage(keys[index]))
{
int textureHandle;
if (desc.IsBindless)
{
long position = _constBuffers[index][desc.CbufSlot].Position;
textureHandle = vmm.ReadInt32(position + desc.CbufOffset * 4);
}
else
{
long position = _constBuffers[index][textureCbIndex].Position;
textureHandle = vmm.ReadInt32(position + desc.HandleIndex * 4);
}
unboundTextures.Add(UploadTexture(vmm, textureHandle));
}
}
for (int index = 0; index < unboundTextures.Count; index++)
{
(long key, GalImage image, GalTextureSampler sampler) = unboundTextures[index];
if (key == 0)
{
continue;
}
_gpu.Renderer.Texture.Bind(key, index, image);
_gpu.Renderer.Texture.SetSampler(image, sampler);
}
Profile.End(Profiles.GPU.Engine3d.UploadTextures);
}
private void UploadTextures(NvGpuVmm Vmm, GalPipelineState State, long[] Keys)
{
long BaseShPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
int TextureCbIndex = ReadRegister(NvGpuEngine3dReg.TextureCbIndex);
int TexIndex = 0;
for (int Index = 0; Index < Keys.Length; Index++)
{
foreach (ShaderDeclInfo DeclInfo in Gpu.Renderer.Shader.GetTextureUsage(Keys[Index]))
{
long Position;
if (DeclInfo.IsCb)
{
Position = ConstBuffers[Index][DeclInfo.Cbuf].Position;
}
else
{
Position = ConstBuffers[Index][TextureCbIndex].Position;
}
int TextureHandle = Vmm.ReadInt32(Position + DeclInfo.Index * 4);
UploadTexture(Vmm, TexIndex, TextureHandle);
TexIndex++;
}
}
}
private static long GetShaderSize(NvGpuVmm Vmm, long Position)
{
const int NopInst = 0x50b0;
long Offset = 0x50;
ulong OpCode = 0;
do
{
uint Word0 = (uint)Vmm.ReadInt32(Position + Offset + 0);
uint Word1 = (uint)Vmm.ReadInt32(Position + Offset + 4);
OpCode = Word0 | (ulong)Word1 << 32;
Offset += 8;
}while ((OpCode >> 52 & 0xfff8) != NopInst && OpCode != 0);
return(Offset - 8);
}
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 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 UploadTextures(NvGpuVmm Vmm, GalPipelineState State, long[] Keys)
{
long BaseShPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
int TextureCbIndex = ReadRegister(NvGpuEngine3dReg.TextureCbIndex);
List <(long, GalImage, GalTextureSampler)> UnboundTextures = new List <(long, GalImage, GalTextureSampler)>();
for (int Index = 0; Index < Keys.Length; Index++)
{
foreach (ShaderDeclInfo DeclInfo in Gpu.Renderer.Shader.GetTextureUsage(Keys[Index]))
{
long Position;
if (DeclInfo.IsCb)
{
Position = ConstBuffers[Index][DeclInfo.Cbuf].Position;
}
else
{
Position = ConstBuffers[Index][TextureCbIndex].Position;
}
int TextureHandle = Vmm.ReadInt32(Position + DeclInfo.Index * 4);
UnboundTextures.Add(UploadTexture(Vmm, TextureHandle));
}
}
for (int Index = 0; Index < UnboundTextures.Count; Index++)
{
(long Key, GalImage Image, GalTextureSampler Sampler) = UnboundTextures[Index];
if (Key == 0)
{
continue;
}
Gpu.Renderer.Texture.Bind(Key, Index, Image);
Gpu.Renderer.Texture.SetSampler(Sampler);
}
}
private void UploadTextures(NvGpuVmm vmm, GalPipelineState state, long[] keys)
{
long baseShPosition = MakeInt64From2xInt32(NvGpuEngine3dReg.ShaderAddress);
int textureCbIndex = ReadRegister(NvGpuEngine3dReg.TextureCbIndex);
List <(long, GalImage, GalTextureSampler)> unboundTextures = new List <(long, GalImage, GalTextureSampler)>();
for (int index = 0; index < keys.Length; index++)
{
foreach (ShaderDeclInfo declInfo in _gpu.Renderer.Shader.GetTextureUsage(keys[index]))
{
long position;
if (declInfo.IsCb)
{
position = _constBuffers[index][declInfo.Cbuf].Position;
}
else
{
position = _constBuffers[index][textureCbIndex].Position;
}
int textureHandle = vmm.ReadInt32(position + declInfo.Index * 4);
unboundTextures.Add(UploadTexture(vmm, textureHandle));
}
}
for (int index = 0; index < unboundTextures.Count; index++)
{
(long key, GalImage image, GalTextureSampler sampler) = unboundTextures[index];
if (key == 0)
{
continue;
}
_gpu.Renderer.Texture.Bind(key, index, image);
_gpu.Renderer.Texture.SetSampler(image, sampler);
}
}
private void UploadTexture(NvGpuVmm Vmm, long BasePosition, int TexIndex, int HndIndex)
{
long Position = BasePosition + HndIndex * 4;
int TextureHandle = Vmm.ReadInt32(Position);
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 (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
{
GalTexture NewTexture = TextureFactory.MakeTexture(Vmm, TicPosition);
long Size = (uint)TextureHelper.GetTextureSize(NewTexture);
bool HasCachedTexture = false;
if (Gpu.Renderer.Texture.TryGetCachedTexture(Key, Size, out GalTexture Texture))
{
if (NewTexture.Equals(Texture) && !Vmm.IsRegionModified(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, NewTexture);
}
Gpu.Renderer.Texture.Bind(Key, TexIndex);
}
Gpu.Renderer.Texture.SetSampler(Sampler);
}
private void Execute(NvGpuVmm vmm, int[] arguments)
{
if (_currentVideoCodec == VideoCodec.H264)
{
int frameDataSize = vmm.ReadInt32(_decoderContextAddress + 0x48);
H264ParameterSets Params = MemoryHelper.Read <H264ParameterSets>(vmm.Memory, vmm.GetPhysicalAddress(_decoderContextAddress + 0x58));
H264Matrices matrices = new H264Matrices()
{
ScalingMatrix4 = vmm.ReadBytes(_decoderContextAddress + 0x1c0, 6 * 16),
ScalingMatrix8 = vmm.ReadBytes(_decoderContextAddress + 0x220, 2 * 64)
};
byte[] frameData = vmm.ReadBytes(_frameDataAddress, frameDataSize);
_h264Decoder.Decode(Params, matrices, frameData);
}
else if (_currentVideoCodec == VideoCodec.Vp9)
{
int frameDataSize = vmm.ReadInt32(_decoderContextAddress + 0x30);
Vp9FrameKeys keys = new Vp9FrameKeys()
{
CurrKey = vmm.GetPhysicalAddress(_vpxCurrLumaAddress),
Ref0Key = vmm.GetPhysicalAddress(_vpxRef0LumaAddress),
Ref1Key = vmm.GetPhysicalAddress(_vpxRef1LumaAddress),
Ref2Key = vmm.GetPhysicalAddress(_vpxRef2LumaAddress)
};
Vp9FrameHeader header = MemoryHelper.Read <Vp9FrameHeader>(vmm.Memory, vmm.GetPhysicalAddress(_decoderContextAddress + 0x48));
Vp9ProbabilityTables probs = new Vp9ProbabilityTables()
{
SegmentationTreeProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x387, 0x7),
SegmentationPredProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x38e, 0x3),
Tx8x8Probs = vmm.ReadBytes(_vpxProbTablesAddress + 0x470, 0x2),
Tx16x16Probs = vmm.ReadBytes(_vpxProbTablesAddress + 0x472, 0x4),
Tx32x32Probs = vmm.ReadBytes(_vpxProbTablesAddress + 0x476, 0x6),
CoefProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x5a0, 0x900),
SkipProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x537, 0x3),
InterModeProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x400, 0x1c),
InterpFilterProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x52a, 0x8),
IsInterProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x41c, 0x4),
CompModeProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x532, 0x5),
SingleRefProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x580, 0xa),
CompRefProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x58a, 0x5),
YModeProbs0 = vmm.ReadBytes(_vpxProbTablesAddress + 0x480, 0x20),
YModeProbs1 = vmm.ReadBytes(_vpxProbTablesAddress + 0x47c, 0x4),
PartitionProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x4e0, 0x40),
MvJointProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x53b, 0x3),
MvSignProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x53e, 0x3),
MvClassProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x54c, 0x14),
MvClass0BitProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x540, 0x3),
MvBitsProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x56c, 0x14),
MvClass0FrProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x560, 0xc),
MvFrProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x542, 0x6),
MvClass0HpProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x548, 0x2),
MvHpProbs = vmm.ReadBytes(_vpxProbTablesAddress + 0x54a, 0x2)
};
byte[] frameData = vmm.ReadBytes(_frameDataAddress, frameDataSize);
_vp9Decoder.Decode(keys, header, probs, frameData);
}
else
{
ThrowUnimplementedCodec();
}
}