/// <summary>
/// Processes a command buffer.
/// </summary>
/// <param name="baseGpuVa">Base GPU virtual address of the command buffer</param>
/// <param name="commandBuffer">Command buffer</param>
public void Process(ulong baseGpuVa, ReadOnlySpan <int> commandBuffer)
{
for (int index = 0; index < commandBuffer.Length; index++)
{
int command = commandBuffer[index];
ulong gpuVa = baseGpuVa + (ulong)index * 4;
if (_state.MethodCount != 0)
{
if (TryFastI2mBufferUpdate(commandBuffer, ref index))
{
continue;
}
Send(gpuVa, _state.Method, command, _state.SubChannel, _state.MethodCount <= 1);
if (!_state.NonIncrementing)
{
_state.Method++;
}
if (_state.IncrementOnce)
{
_state.NonIncrementing = true;
}
_state.MethodCount--;
}
else
{
CompressedMethod meth = Unsafe.As <int, CompressedMethod>(ref command);
if (TryFastUniformBufferUpdate(meth, commandBuffer, index))
{
index += meth.MethodCount;
continue;
}
switch (meth.SecOp)
{
case SecOp.IncMethod:
case SecOp.NonIncMethod:
case SecOp.OneInc:
_state.Method = meth.MethodAddress;
_state.SubChannel = meth.MethodSubchannel;
_state.MethodCount = meth.MethodCount;
_state.IncrementOnce = meth.SecOp == SecOp.OneInc;
_state.NonIncrementing = meth.SecOp == SecOp.NonIncMethod;
break;
case SecOp.ImmdDataMethod:
Send(gpuVa, meth.MethodAddress, meth.ImmdData, meth.MethodSubchannel, true);
break;
}
}
}
_3dClass.FlushUboDirty();
}
/// <summary>
/// Performs a buffer to buffer, or buffer to texture copy.
/// </summary>
/// <param name="argument">The LaunchDma call argument</param>
private void DmaCopy(int argument)
{
var memoryManager = _channel.MemoryManager;
CopyFlags copyFlags = (CopyFlags)argument;
bool srcLinear = copyFlags.HasFlag(CopyFlags.SrcLinear);
bool dstLinear = copyFlags.HasFlag(CopyFlags.DstLinear);
bool copy2D = copyFlags.HasFlag(CopyFlags.MultiLineEnable);
bool remap = copyFlags.HasFlag(CopyFlags.RemapEnable);
uint size = _state.State.LineLengthIn;
if (size == 0)
{
return;
}
ulong srcGpuVa = ((ulong)_state.State.OffsetInUpperUpper << 32) | _state.State.OffsetInLower;
ulong dstGpuVa = ((ulong)_state.State.OffsetOutUpperUpper << 32) | _state.State.OffsetOutLower;
int xCount = (int)_state.State.LineLengthIn;
int yCount = (int)_state.State.LineCount;
_3dEngine.FlushUboDirty();
if (copy2D)
{
// Buffer to texture copy.
int componentSize = (int)_state.State.SetRemapComponentsComponentSize + 1;
int srcBpp = remap ? ((int)_state.State.SetRemapComponentsNumSrcComponents + 1) * componentSize : 1;
int dstBpp = remap ? ((int)_state.State.SetRemapComponentsNumDstComponents + 1) * componentSize : 1;
var dst = Unsafe.As <uint, DmaTexture>(ref _state.State.SetDstBlockSize);
var src = Unsafe.As <uint, DmaTexture>(ref _state.State.SetSrcBlockSize);
int srcStride = (int)_state.State.PitchIn;
int dstStride = (int)_state.State.PitchOut;
var srcCalculator = new OffsetCalculator(
src.Width,
src.Height,
srcStride,
srcLinear,
src.MemoryLayout.UnpackGobBlocksInY(),
src.MemoryLayout.UnpackGobBlocksInZ(),
srcBpp);
var dstCalculator = new OffsetCalculator(
dst.Width,
dst.Height,
dstStride,
dstLinear,
dst.MemoryLayout.UnpackGobBlocksInY(),
dst.MemoryLayout.UnpackGobBlocksInZ(),
dstBpp);
(int srcBaseOffset, int srcSize) = srcCalculator.GetRectangleRange(src.RegionX, src.RegionY, xCount, yCount);
(int dstBaseOffset, int dstSize) = dstCalculator.GetRectangleRange(dst.RegionX, dst.RegionY, xCount, yCount);
if (srcLinear && srcStride < 0)
{
srcBaseOffset += srcStride * (yCount - 1);
}
if (dstLinear && dstStride < 0)
{
dstBaseOffset += dstStride * (yCount - 1);
}
ReadOnlySpan <byte> srcSpan = memoryManager.GetSpan(srcGpuVa + (ulong)srcBaseOffset, srcSize, true);
Span <byte> dstSpan = memoryManager.GetSpan(dstGpuVa + (ulong)dstBaseOffset, dstSize).ToArray();
bool completeSource = IsTextureCopyComplete(src, srcLinear, srcBpp, srcStride, xCount, yCount);
bool completeDest = IsTextureCopyComplete(dst, dstLinear, dstBpp, dstStride, xCount, yCount);
if (completeSource && completeDest)
{
var target = memoryManager.Physical.TextureCache.FindTexture(
memoryManager,
dst,
dstGpuVa,
dstBpp,
dstStride,
xCount,
yCount,
dstLinear);
if (target != null)
{
ReadOnlySpan <byte> data;
if (srcLinear)
{
data = LayoutConverter.ConvertLinearStridedToLinear(
target.Info.Width,
target.Info.Height,
1,
1,
xCount * srcBpp,
srcStride,
target.Info.FormatInfo.BytesPerPixel,
srcSpan);
}
else
{
data = LayoutConverter.ConvertBlockLinearToLinear(
src.Width,
src.Height,
src.Depth,
1,
1,
1,
1,
1,
srcBpp,
src.MemoryLayout.UnpackGobBlocksInY(),
src.MemoryLayout.UnpackGobBlocksInZ(),
1,
new SizeInfo((int)target.Size),
srcSpan);
}
target.SynchronizeMemory();
target.SetData(data);
target.SignalModified();
return;
}
else if (srcCalculator.LayoutMatches(dstCalculator))
{
srcSpan.CopyTo(dstSpan); // No layout conversion has to be performed, just copy the data entirely.
memoryManager.Write(dstGpuVa + (ulong)dstBaseOffset, dstSpan);
return;
}
}
unsafe bool Convert <T>(Span <byte> dstSpan, ReadOnlySpan <byte> srcSpan) where T : unmanaged
{
fixed(byte *dstPtr = dstSpan, srcPtr = srcSpan)
{
byte *dstBase = dstPtr - dstBaseOffset; // Layout offset is relative to the base, so we need to subtract the span's offset.
byte *srcBase = srcPtr - srcBaseOffset;
for (int y = 0; y < yCount; y++)
{
srcCalculator.SetY(src.RegionY + y);
dstCalculator.SetY(dst.RegionY + y);
for (int x = 0; x < xCount; x++)
{
int srcOffset = srcCalculator.GetOffset(src.RegionX + x);
int dstOffset = dstCalculator.GetOffset(dst.RegionX + x);
*(T *)(dstBase + dstOffset) = *(T *)(srcBase + srcOffset);
}
}
}
return(true);
}
bool _ = srcBpp switch
{
1 => Convert <byte>(dstSpan, srcSpan),
2 => Convert <ushort>(dstSpan, srcSpan),
4 => Convert <uint>(dstSpan, srcSpan),
8 => Convert <ulong>(dstSpan, srcSpan),
12 => Convert <Bpp12Pixel>(dstSpan, srcSpan),
16 => Convert <Vector128 <byte> >(dstSpan, srcSpan),
_ => throw new NotSupportedException($"Unable to copy ${srcBpp} bpp pixel format.")
};
memoryManager.Write(dstGpuVa + (ulong)dstBaseOffset, dstSpan);
}
else
{
if (remap &&
_state.State.SetRemapComponentsDstX == SetRemapComponentsDst.ConstA &&
_state.State.SetRemapComponentsDstY == SetRemapComponentsDst.ConstA &&
_state.State.SetRemapComponentsDstZ == SetRemapComponentsDst.ConstA &&
_state.State.SetRemapComponentsDstW == SetRemapComponentsDst.ConstA &&
_state.State.SetRemapComponentsNumSrcComponents == SetRemapComponentsNumComponents.One &&
_state.State.SetRemapComponentsNumDstComponents == SetRemapComponentsNumComponents.One &&
_state.State.SetRemapComponentsComponentSize == SetRemapComponentsComponentSize.Four)
{
// Fast path for clears when remap is enabled.
memoryManager.Physical.BufferCache.ClearBuffer(memoryManager, dstGpuVa, size * 4, _state.State.SetRemapConstA);
}
else
{
// TODO: Implement remap functionality.
// Buffer to buffer copy.
memoryManager.Physical.BufferCache.CopyBuffer(memoryManager, srcGpuVa, dstGpuVa, size);
}
}
}
/// <summary>
/// Performs the compute dispatch operation.
/// </summary>
/// <param name="argument">Method call argument</param>
private void SendSignalingPcasB(int argument)
{
var memoryManager = _channel.MemoryManager;
_3dEngine.FlushUboDirty();
uint qmdAddress = _state.State.SendPcasA;
var qmd = _channel.MemoryManager.Read <ComputeQmd>((ulong)qmdAddress << 8);
ulong shaderGpuVa = ((ulong)_state.State.SetProgramRegionAAddressUpper << 32) | _state.State.SetProgramRegionB;
shaderGpuVa += (uint)qmd.ProgramOffset;
int localMemorySize = qmd.ShaderLocalMemoryLowSize + qmd.ShaderLocalMemoryHighSize;
int sharedMemorySize = Math.Min(qmd.SharedMemorySize, _context.Capabilities.MaximumComputeSharedMemorySize);
for (int index = 0; index < Constants.TotalCpUniformBuffers; index++)
{
if (!qmd.ConstantBufferValid(index))
{
continue;
}
ulong gpuVa = (uint)qmd.ConstantBufferAddrLower(index) | (ulong)qmd.ConstantBufferAddrUpper(index) << 32;
ulong size = (ulong)qmd.ConstantBufferSize(index);
_channel.BufferManager.SetComputeUniformBuffer(index, gpuVa, size);
}
ulong samplerPoolGpuVa = ((ulong)_state.State.SetTexSamplerPoolAOffsetUpper << 32) | _state.State.SetTexSamplerPoolB;
ulong texturePoolGpuVa = ((ulong)_state.State.SetTexHeaderPoolAOffsetUpper << 32) | _state.State.SetTexHeaderPoolB;
GpuChannelPoolState poolState = new GpuChannelPoolState(
texturePoolGpuVa,
_state.State.SetTexHeaderPoolCMaximumIndex,
_state.State.SetBindlessTextureConstantBufferSlotSelect);
GpuChannelComputeState computeState = new GpuChannelComputeState(
qmd.CtaThreadDimension0,
qmd.CtaThreadDimension1,
qmd.CtaThreadDimension2,
localMemorySize,
sharedMemorySize);
CachedShaderProgram cs = memoryManager.Physical.ShaderCache.GetComputeShader(_channel, poolState, computeState, shaderGpuVa);
_context.Renderer.Pipeline.SetProgram(cs.HostProgram);
_channel.TextureManager.SetComputeSamplerPool(samplerPoolGpuVa, _state.State.SetTexSamplerPoolCMaximumIndex, qmd.SamplerIndex);
_channel.TextureManager.SetComputeTexturePool(texturePoolGpuVa, _state.State.SetTexHeaderPoolCMaximumIndex);
_channel.TextureManager.SetComputeTextureBufferIndex(_state.State.SetBindlessTextureConstantBufferSlotSelect);
ShaderProgramInfo info = cs.Shaders[0].Info;
for (int index = 0; index < info.CBuffers.Count; index++)
{
BufferDescriptor cb = info.CBuffers[index];
// NVN uses the "hardware" constant buffer for anything that is less than 8,
// and those are already bound above.
// Anything greater than or equal to 8 uses the emulated constant buffers.
// They are emulated using global memory loads.
if (cb.Slot < 8)
{
continue;
}
ulong cbDescAddress = _channel.BufferManager.GetComputeUniformBufferAddress(0);
int cbDescOffset = 0x260 + (cb.Slot - 8) * 0x10;
cbDescAddress += (ulong)cbDescOffset;
SbDescriptor cbDescriptor = _channel.MemoryManager.Physical.Read <SbDescriptor>(cbDescAddress);
_channel.BufferManager.SetComputeUniformBuffer(cb.Slot, cbDescriptor.PackAddress(), (uint)cbDescriptor.Size);
}
for (int index = 0; index < info.SBuffers.Count; index++)
{
BufferDescriptor sb = info.SBuffers[index];
ulong sbDescAddress = _channel.BufferManager.GetComputeUniformBufferAddress(0);
int sbDescOffset = 0x310 + sb.Slot * 0x10;
sbDescAddress += (ulong)sbDescOffset;
SbDescriptor sbDescriptor = _channel.MemoryManager.Physical.Read <SbDescriptor>(sbDescAddress);
_channel.BufferManager.SetComputeStorageBuffer(sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size, sb.Flags);
}
_channel.BufferManager.SetComputeStorageBufferBindings(info.SBuffers);
_channel.BufferManager.SetComputeUniformBufferBindings(info.CBuffers);
int maxTextureBinding = -1;
int maxImageBinding = -1;
TextureBindingInfo[] textureBindings = _channel.TextureManager.RentComputeTextureBindings(info.Textures.Count);
for (int index = 0; index < info.Textures.Count; index++)
{
var descriptor = info.Textures[index];
Target target = ShaderTexture.GetTarget(descriptor.Type);
textureBindings[index] = new TextureBindingInfo(
target,
descriptor.Binding,
descriptor.CbufSlot,
descriptor.HandleIndex,
descriptor.Flags);
if (descriptor.Binding > maxTextureBinding)
{
maxTextureBinding = descriptor.Binding;
}
}
TextureBindingInfo[] imageBindings = _channel.TextureManager.RentComputeImageBindings(info.Images.Count);
for (int index = 0; index < info.Images.Count; index++)
{
var descriptor = info.Images[index];
Target target = ShaderTexture.GetTarget(descriptor.Type);
Format format = ShaderTexture.GetFormat(descriptor.Format);
imageBindings[index] = new TextureBindingInfo(
target,
format,
descriptor.Binding,
descriptor.CbufSlot,
descriptor.HandleIndex,
descriptor.Flags);
if (descriptor.Binding > maxImageBinding)
{
maxImageBinding = descriptor.Binding;
}
}
_channel.TextureManager.SetComputeMaxBindings(maxTextureBinding, maxImageBinding);
// Should never return false for mismatching spec state, since the shader was fetched above.
_channel.TextureManager.CommitComputeBindings(cs.SpecializationState);
_channel.BufferManager.CommitComputeBindings();
_context.Renderer.Pipeline.DispatchCompute(qmd.CtaRasterWidth, qmd.CtaRasterHeight, qmd.CtaRasterDepth);
_3dEngine.ForceShaderUpdate();
}
