/// <summary>
/// Gets a compute shader from the cache.
/// </summary>
/// <remarks>
/// This automatically translates, compiles and adds the code to the cache if not present.
/// </remarks>
/// <param name="state">Current GPU state</param>
/// <param name="gpuVa">GPU virtual address of the binary shader code</param>
/// <param name="localSizeX">Local group size X of the computer shader</param>
/// <param name="localSizeY">Local group size Y of the computer shader</param>
/// <param name="localSizeZ">Local group size Z of the computer shader</param>
/// <param name="localMemorySize">Local memory size of the compute shader</param>
/// <param name="sharedMemorySize">Shared memory size of the compute shader</param>
/// <returns>Compiled compute shader code</returns>
public ShaderBundle GetComputeShader(
GpuState state,
ulong gpuVa,
int localSizeX,
int localSizeY,
int localSizeZ,
int localMemorySize,
int sharedMemorySize)
{
bool isCached = _cpPrograms.TryGetValue(gpuVa, out List <ShaderBundle> list);
if (isCached)
{
foreach (ShaderBundle cachedCpShader in list)
{
if (IsShaderEqual(cachedCpShader, gpuVa))
{
return(cachedCpShader);
}
}
}
TranslatorContext[] shaderContexts = new TranslatorContext[1];
shaderContexts[0] = DecodeComputeShader(
state,
gpuVa,
localSizeX,
localSizeY,
localSizeZ,
localMemorySize,
sharedMemorySize);
bool isShaderCacheEnabled = _cacheManager != null;
bool isShaderCacheReadOnly = false;
Hash128 programCodeHash = default;
GuestShaderCacheEntry[] shaderCacheEntries = null;
// Current shader cache doesn't support bindless textures
if (shaderContexts[0].UsedFeatures.HasFlag(FeatureFlags.Bindless))
{
isShaderCacheEnabled = false;
}
if (isShaderCacheEnabled)
{
isShaderCacheReadOnly = _cacheManager.IsReadOnly;
// Compute hash and prepare data for shader disk cache comparison.
shaderCacheEntries = CacheHelper.CreateShaderCacheEntries(_context.MemoryManager, shaderContexts);
programCodeHash = CacheHelper.ComputeGuestHashFromCache(shaderCacheEntries);
}
ShaderBundle cpShader;
// Search for the program hash in loaded shaders.
if (!isShaderCacheEnabled || !_cpProgramsDiskCache.TryGetValue(programCodeHash, out cpShader))
{
if (isShaderCacheEnabled)
{
Logger.Debug?.Print(LogClass.Gpu, $"Shader {programCodeHash} not in cache, compiling!");
}
// The shader isn't currently cached, translate it and compile it.
ShaderCodeHolder shader = TranslateShader(shaderContexts[0]);
shader.HostShader = _context.Renderer.CompileShader(ShaderStage.Compute, shader.Program.Code);
IProgram hostProgram = _context.Renderer.CreateProgram(new IShader[] { shader.HostShader }, null);
hostProgram.CheckProgramLink(true);
byte[] hostProgramBinary = HostShaderCacheEntry.Create(hostProgram.GetBinary(), new ShaderCodeHolder[] { shader });
cpShader = new ShaderBundle(hostProgram, shader);
if (isShaderCacheEnabled)
{
_cpProgramsDiskCache.Add(programCodeHash, cpShader);
if (!isShaderCacheReadOnly)
{
_cacheManager.SaveProgram(ref programCodeHash, CacheHelper.CreateGuestProgramDump(shaderCacheEntries), hostProgramBinary);
}
}
}
if (!isCached)
{
list = new List <ShaderBundle>();
_cpPrograms.Add(gpuVa, list);
}
list.Add(cpShader);
return(cpShader);
}
/// <summary>
/// Invalidates all modified textures on the cache.
/// </summary>
/// <param name="state">Current GPU state (unused)</param>
/// <param name="argument">Method call argument (unused)</param>
private void InvalidateTextures(GpuState state, int argument)
{
TextureManager.Flush();
}
/// <summary>
/// Sets the index buffer count.
/// This also sets internal state that indicates that the next draw is an indexed draw.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void SetIndexBufferCount(GpuState state, int argument)
{
_drawIndexed = true;
}
/// <summary>
/// Dispatches compute work.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
public void Dispatch(GpuState state, int argument)
{
uint qmdAddress = (uint)state.Get <int>(MethodOffset.DispatchParamsAddress);
var qmd = _context.MemoryAccessor.Read <ComputeQmd>((ulong)qmdAddress << 8);
GpuVa shaderBaseAddress = state.Get <GpuVa>(MethodOffset.ShaderBaseAddress);
ulong shaderGpuVa = shaderBaseAddress.Pack() + (uint)qmd.ProgramOffset;
int localMemorySize = qmd.ShaderLocalMemoryLowSize + qmd.ShaderLocalMemoryHighSize;
int sharedMemorySize = Math.Min(qmd.SharedMemorySize, _context.Capabilities.MaximumComputeSharedMemorySize);
ComputeShader cs = ShaderCache.GetComputeShader(
shaderGpuVa,
qmd.CtaThreadDimension0,
qmd.CtaThreadDimension1,
qmd.CtaThreadDimension2,
localMemorySize,
sharedMemorySize);
_context.Renderer.Pipeline.SetProgram(cs.HostProgram);
var samplerPool = state.Get <PoolState>(MethodOffset.SamplerPoolState);
TextureManager.SetComputeSamplerPool(samplerPool.Address.Pack(), samplerPool.MaximumId, qmd.SamplerIndex);
var texturePool = state.Get <PoolState>(MethodOffset.TexturePoolState);
TextureManager.SetComputeTexturePool(texturePool.Address.Pack(), texturePool.MaximumId);
TextureManager.SetComputeTextureBufferIndex(state.Get <int>(MethodOffset.TextureBufferIndex));
ShaderProgramInfo info = cs.Shader.Program.Info;
uint sbEnableMask = 0;
uint ubEnableMask = 0;
for (int index = 0; index < Constants.TotalCpUniformBuffers; index++)
{
if (!qmd.ConstantBufferValid(index))
{
continue;
}
ubEnableMask |= 1u << index;
ulong gpuVa = (uint)qmd.ConstantBufferAddrLower(index) | (ulong)qmd.ConstantBufferAddrUpper(index) << 32;
ulong size = (ulong)qmd.ConstantBufferSize(index);
BufferManager.SetComputeUniformBuffer(index, gpuVa, size);
}
for (int index = 0; index < info.SBuffers.Count; index++)
{
BufferDescriptor sb = info.SBuffers[index];
sbEnableMask |= 1u << sb.Slot;
ulong sbDescAddress = BufferManager.GetComputeUniformBufferAddress(0);
int sbDescOffset = 0x310 + sb.Slot * 0x10;
sbDescAddress += (ulong)sbDescOffset;
ReadOnlySpan <byte> sbDescriptorData = _context.PhysicalMemory.GetSpan(sbDescAddress, 0x10);
SbDescriptor sbDescriptor = MemoryMarshal.Cast <byte, SbDescriptor>(sbDescriptorData)[0];
BufferManager.SetComputeStorageBuffer(sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size);
}
ubEnableMask = 0;
for (int index = 0; index < info.CBuffers.Count; index++)
{
ubEnableMask |= 1u << info.CBuffers[index].Slot;
}
BufferManager.SetComputeStorageBufferEnableMask(sbEnableMask);
BufferManager.SetComputeUniformBufferEnableMask(ubEnableMask);
var textureBindings = new TextureBindingInfo[info.Textures.Count];
for (int index = 0; index < info.Textures.Count; index++)
{
var descriptor = info.Textures[index];
Target target = GetTarget(descriptor.Type);
if (descriptor.IsBindless)
{
textureBindings[index] = new TextureBindingInfo(target, descriptor.CbufOffset, descriptor.CbufSlot);
}
else
{
textureBindings[index] = new TextureBindingInfo(target, descriptor.HandleIndex);
}
}
TextureManager.SetComputeTextures(textureBindings);
var imageBindings = new TextureBindingInfo[info.Images.Count];
for (int index = 0; index < info.Images.Count; index++)
{
var descriptor = info.Images[index];
Target target = GetTarget(descriptor.Type);
imageBindings[index] = new TextureBindingInfo(target, descriptor.HandleIndex);
}
TextureManager.SetComputeImages(imageBindings);
BufferManager.CommitComputeBindings();
TextureManager.CommitComputeBindings();
_context.Renderer.Pipeline.DispatchCompute(
qmd.CtaRasterWidth,
qmd.CtaRasterHeight,
qmd.CtaRasterDepth);
UpdateShaderState(state);
}
/// <summary>
/// Updates host point size based on guest GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdatePointSizeState(GpuState state)
{
float size = state.Get <float>(MethodOffset.PointSize);
_context.Renderer.Pipeline.SetPointSize(size);
}
/// <summary>
/// Updates host viewport transform and clipping state based on current GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdateViewportTransform(GpuState state)
{
DepthMode depthMode = state.Get <DepthMode>(MethodOffset.DepthMode);
_context.Renderer.Pipeline.SetDepthMode(depthMode);
YControl yControl = state.Get <YControl>(MethodOffset.YControl);
bool flipY = yControl.HasFlag(YControl.NegateY);
Origin origin = yControl.HasFlag(YControl.TriangleRastFlip) ? Origin.LowerLeft : Origin.UpperLeft;
_context.Renderer.Pipeline.SetOrigin(origin);
// The triangle rast flip flag only affects rasterization, the viewport is not flipped.
// Setting the origin mode to upper left on the host, however, not only affects rasterization,
// but also flips the viewport.
// We negate the effects of flipping the viewport by flipping it again using the viewport swizzle.
if (origin == Origin.UpperLeft)
{
flipY = !flipY;
}
Span <Viewport> viewports = stackalloc Viewport[Constants.TotalViewports];
for (int index = 0; index < Constants.TotalViewports; index++)
{
var transform = state.Get <ViewportTransform>(MethodOffset.ViewportTransform, index);
var extents = state.Get <ViewportExtents> (MethodOffset.ViewportExtents, index);
float x = transform.TranslateX - MathF.Abs(transform.ScaleX);
float y = transform.TranslateY - MathF.Abs(transform.ScaleY);
float width = MathF.Abs(transform.ScaleX) * 2;
float height = MathF.Abs(transform.ScaleY) * 2;
float scale = TextureManager.RenderTargetScale;
if (scale != 1f)
{
x *= scale;
y *= scale;
width *= scale;
height *= scale;
}
RectangleF region = new RectangleF(x, y, width, height);
ViewportSwizzle swizzleX = transform.UnpackSwizzleX();
ViewportSwizzle swizzleY = transform.UnpackSwizzleY();
ViewportSwizzle swizzleZ = transform.UnpackSwizzleZ();
ViewportSwizzle swizzleW = transform.UnpackSwizzleW();
if (transform.ScaleX < 0)
{
swizzleX ^= ViewportSwizzle.NegativeFlag;
}
if (flipY)
{
swizzleY ^= ViewportSwizzle.NegativeFlag;
}
if (transform.ScaleY < 0)
{
swizzleY ^= ViewportSwizzle.NegativeFlag;
}
if (transform.ScaleZ < 0)
{
swizzleZ ^= ViewportSwizzle.NegativeFlag;
}
viewports[index] = new Viewport(
region,
swizzleX,
swizzleY,
swizzleZ,
swizzleW,
extents.DepthNear,
extents.DepthFar);
}
_context.Renderer.Pipeline.SetViewports(0, viewports);
}
/// <summary>
/// Gets a texture descriptor used on the compute pipeline.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="handle">Shader "fake" handle of the texture</param>
/// <returns>The texture descriptor</returns>
public TextureDescriptor GetComputeTextureDescriptor(GpuState state, int handle)
{
return(_cpBindingsManager.GetTextureDescriptor(state, 0, handle));
}
/// <summary>
/// Finishes the draw call.
/// This draws geometry on the bound buffers based on the current GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void DrawEnd(GpuState state, int argument)
{
var indexBuffer = state.Get <IndexBufferState>(MethodOffset.IndexBufferState);
DrawEnd(state, indexBuffer.First, indexBuffer.Count);
}
/// <summary>
/// Finishes the draw call.
/// This draws geometry on the bound buffers based on the current GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="firstIndex">Index of the first index buffer element used on the draw</param>
/// <param name="indexCount">Number of index buffer elements used on the draw</param>
private void DrawEnd(GpuState state, int firstIndex, int indexCount)
{
ConditionalRenderEnabled renderEnable = GetRenderEnable(state);
if (renderEnable == ConditionalRenderEnabled.False || _instancedDrawPending)
{
if (renderEnable == ConditionalRenderEnabled.False)
{
PerformDeferredDraws();
}
_drawIndexed = false;
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
return;
}
UpdateState(state, firstIndex, indexCount);
bool instanced = _vsUsesInstanceId || _isAnyVbInstanced;
if (instanced)
{
_instancedDrawPending = true;
_instancedIndexed = _drawIndexed;
_instancedFirstIndex = firstIndex;
_instancedFirstVertex = state.Get <int>(MethodOffset.FirstVertex);
_instancedFirstInstance = state.Get <int>(MethodOffset.FirstInstance);
_instancedIndexCount = indexCount;
var drawState = state.Get <VertexBufferDrawState>(MethodOffset.VertexBufferDrawState);
_instancedDrawStateFirst = drawState.First;
_instancedDrawStateCount = drawState.Count;
_drawIndexed = false;
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
return;
}
int firstInstance = state.Get <int>(MethodOffset.FirstInstance);
int inlineIndexCount = _ibStreamer.GetAndResetInlineIndexCount();
if (inlineIndexCount != 0)
{
int firstVertex = state.Get <int>(MethodOffset.FirstVertex);
BufferRange br = new BufferRange(_ibStreamer.GetInlineIndexBuffer(), 0, inlineIndexCount * 4);
state.Channel.BufferManager.SetIndexBuffer(br, IndexType.UInt);
_context.Renderer.Pipeline.DrawIndexed(
inlineIndexCount,
1,
firstIndex,
firstVertex,
firstInstance);
}
else if (_drawIndexed)
{
int firstVertex = state.Get <int>(MethodOffset.FirstVertex);
_context.Renderer.Pipeline.DrawIndexed(
indexCount,
1,
firstIndex,
firstVertex,
firstInstance);
}
else
{
var drawState = state.Get <VertexBufferDrawState>(MethodOffset.VertexBufferDrawState);
_context.Renderer.Pipeline.Draw(
drawState.Count,
1,
drawState.First,
firstInstance);
}
_drawIndexed = false;
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
}
/// <summary>
/// Performs a indexed draw with a low number of index buffer elements,
/// while also pre-incrementing the current instance value.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void DrawIndexedSmallIncInstance2(GpuState state, int argument)
{
DrawIndexedSmallIncInstance(state, argument);
}
/// <summary>
/// Pushes one 32-bit index buffer element.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void VbElementU32(GpuState state, int argument)
{
_ibStreamer.VbElementU32(_context.Renderer, argument);
}
/// <summary>
/// Clears the current color and depth-stencil buffers.
/// Which buffers should be cleared is also specified on the argument.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void Clear(GpuState state, int argument)
{
ConditionalRenderEnabled renderEnable = GetRenderEnable(state);
if (renderEnable == ConditionalRenderEnabled.False)
{
return;
}
// Scissor and rasterizer discard also affect clears.
if (state.QueryModified(MethodOffset.ScissorState))
{
UpdateScissorState(state);
}
if (state.QueryModified(MethodOffset.RasterizeEnable))
{
UpdateRasterizerState(state);
}
int index = (argument >> 6) & 0xf;
UpdateRenderTargetState(state, useControl: false, singleUse: index);
state.Channel.TextureManager.UpdateRenderTargets();
bool clearDepth = (argument & 1) != 0;
bool clearStencil = (argument & 2) != 0;
uint componentMask = (uint)((argument >> 2) & 0xf);
if (componentMask != 0)
{
var clearColor = state.Get <ClearColors>(MethodOffset.ClearColors);
ColorF color = new ColorF(
clearColor.Red,
clearColor.Green,
clearColor.Blue,
clearColor.Alpha);
_context.Renderer.Pipeline.ClearRenderTargetColor(index, componentMask, color);
}
if (clearDepth || clearStencil)
{
float depthValue = state.Get <float>(MethodOffset.ClearDepthValue);
int stencilValue = state.Get <int> (MethodOffset.ClearStencilValue);
int stencilMask = 0;
if (clearStencil)
{
stencilMask = state.Get <StencilTestState>(MethodOffset.StencilTestState).FrontMask;
}
_context.Renderer.Pipeline.ClearRenderTargetDepthStencil(
depthValue,
clearDepth,
stencilValue,
stencilMask);
}
UpdateRenderTargetState(state, useControl: true);
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
}
/// <summary>
/// Finishes draw call.
/// This draws geometry on the bound buffers based on the current GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void DrawEnd(GpuState state, int argument)
{
ConditionalRenderEnabled renderEnable = GetRenderEnable(state);
if (renderEnable == ConditionalRenderEnabled.False || _instancedDrawPending)
{
if (renderEnable == ConditionalRenderEnabled.False)
{
PerformDeferredDraws();
}
_drawIndexed = false;
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
return;
}
UpdateState(state);
bool instanced = _vsUsesInstanceId || _isAnyVbInstanced;
if (instanced)
{
_instancedDrawPending = true;
_instancedIndexed = _drawIndexed;
_instancedFirstIndex = _firstIndex;
_instancedFirstVertex = state.Get <int>(MethodOffset.FirstVertex);
_instancedFirstInstance = state.Get <int>(MethodOffset.FirstInstance);
_instancedIndexCount = _indexCount;
var drawState = state.Get <VertexBufferDrawState>(MethodOffset.VertexBufferDrawState);
_instancedDrawStateFirst = drawState.First;
_instancedDrawStateCount = drawState.Count;
_drawIndexed = false;
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
return;
}
int firstInstance = state.Get <int>(MethodOffset.FirstInstance);
if (_drawIndexed)
{
_drawIndexed = false;
int firstVertex = state.Get <int>(MethodOffset.FirstVertex);
_context.Renderer.Pipeline.DrawIndexed(
_indexCount,
1,
_firstIndex,
firstVertex,
firstInstance);
}
else
{
var drawState = state.Get <VertexBufferDrawState>(MethodOffset.VertexBufferDrawState);
_context.Renderer.Pipeline.Draw(
drawState.Count,
1,
drawState.First,
firstInstance);
}
if (renderEnable == ConditionalRenderEnabled.Host)
{
_context.Renderer.Pipeline.EndHostConditionalRendering();
}
}
/// <summary>
/// Gets a graphics shader program from the shader cache.
/// This includes all the specified shader stages.
/// </summary>
/// <remarks>
/// This automatically translates, compiles and adds the code to the cache if not present.
/// </remarks>
/// <param name="state">Current GPU state</param>
/// <param name="addresses">Addresses of the shaders for each stage</param>
/// <returns>Compiled graphics shader code</returns>
public ShaderBundle GetGraphicsShader(GpuState state, ShaderAddresses addresses)
{
bool isCached = _gpPrograms.TryGetValue(addresses, out List <ShaderBundle> list);
if (isCached)
{
foreach (ShaderBundle cachedGpShaders in list)
{
if (IsShaderEqual(cachedGpShaders, addresses))
{
return(cachedGpShaders);
}
}
}
TranslatorContext[] shaderContexts = new TranslatorContext[Constants.ShaderStages + 1];
TransformFeedbackDescriptor[] tfd = GetTransformFeedbackDescriptors(state);
TranslationFlags flags = DefaultFlags;
if (tfd != null)
{
flags |= TranslationFlags.Feedback;
}
TranslationCounts counts = new TranslationCounts();
if (addresses.VertexA != 0)
{
shaderContexts[0] = DecodeGraphicsShader(state, counts, flags | TranslationFlags.VertexA, ShaderStage.Vertex, addresses.VertexA);
}
shaderContexts[1] = DecodeGraphicsShader(state, counts, flags, ShaderStage.Vertex, addresses.Vertex);
shaderContexts[2] = DecodeGraphicsShader(state, counts, flags, ShaderStage.TessellationControl, addresses.TessControl);
shaderContexts[3] = DecodeGraphicsShader(state, counts, flags, ShaderStage.TessellationEvaluation, addresses.TessEvaluation);
shaderContexts[4] = DecodeGraphicsShader(state, counts, flags, ShaderStage.Geometry, addresses.Geometry);
shaderContexts[5] = DecodeGraphicsShader(state, counts, flags, ShaderStage.Fragment, addresses.Fragment);
bool isShaderCacheEnabled = _cacheManager != null;
bool isShaderCacheReadOnly = false;
Hash128 programCodeHash = default;
GuestShaderCacheEntry[] shaderCacheEntries = null;
// Current shader cache doesn't support bindless textures
for (int i = 0; i < shaderContexts.Length; i++)
{
if (shaderContexts[i] != null && shaderContexts[i].UsedFeatures.HasFlag(FeatureFlags.Bindless))
{
isShaderCacheEnabled = false;
break;
}
}
if (isShaderCacheEnabled)
{
isShaderCacheReadOnly = _cacheManager.IsReadOnly;
// Compute hash and prepare data for shader disk cache comparison.
shaderCacheEntries = CacheHelper.CreateShaderCacheEntries(_context.MemoryManager, shaderContexts);
programCodeHash = CacheHelper.ComputeGuestHashFromCache(shaderCacheEntries, tfd);
}
ShaderBundle gpShaders;
// Search for the program hash in loaded shaders.
if (!isShaderCacheEnabled || !_gpProgramsDiskCache.TryGetValue(programCodeHash, out gpShaders))
{
if (isShaderCacheEnabled)
{
Logger.Debug?.Print(LogClass.Gpu, $"Shader {programCodeHash} not in cache, compiling!");
}
// The shader isn't currently cached, translate it and compile it.
ShaderCodeHolder[] shaders = new ShaderCodeHolder[Constants.ShaderStages];
shaders[0] = TranslateShader(shaderContexts[1], shaderContexts[0]);
shaders[1] = TranslateShader(shaderContexts[2]);
shaders[2] = TranslateShader(shaderContexts[3]);
shaders[3] = TranslateShader(shaderContexts[4]);
shaders[4] = TranslateShader(shaderContexts[5]);
List <IShader> hostShaders = new List <IShader>();
for (int stage = 0; stage < Constants.ShaderStages; stage++)
{
ShaderProgram program = shaders[stage]?.Program;
if (program == null)
{
continue;
}
IShader hostShader = _context.Renderer.CompileShader(program.Stage, program.Code);
shaders[stage].HostShader = hostShader;
hostShaders.Add(hostShader);
}
IProgram hostProgram = _context.Renderer.CreateProgram(hostShaders.ToArray(), tfd);
hostProgram.CheckProgramLink(true);
byte[] hostProgramBinary = HostShaderCacheEntry.Create(hostProgram.GetBinary(), shaders);
gpShaders = new ShaderBundle(hostProgram, shaders);
if (isShaderCacheEnabled)
{
_gpProgramsDiskCache.Add(programCodeHash, gpShaders);
if (!isShaderCacheReadOnly)
{
_cacheManager.SaveProgram(ref programCodeHash, CacheHelper.CreateGuestProgramDump(shaderCacheEntries, tfd), hostProgramBinary);
}
}
}
if (!isCached)
{
list = new List <ShaderBundle>();
_gpPrograms.Add(addresses, list);
}
list.Add(gpShaders);
return(gpShaders);
}
/// <summary>
/// Updates render targets (color and depth-stencil buffers) based on current render target state.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="useControl">Use draw buffers information from render target control register</param>
/// <param name="singleUse">If this is not -1, it indicates that only the given indexed target will be used.</param>
private void UpdateRenderTargetState(GpuState state, bool useControl, int singleUse = -1)
{
var rtControl = state.Get <RtControl>(MethodOffset.RtControl);
int count = useControl ? rtControl.UnpackCount() : Constants.TotalRenderTargets;
var msaaMode = state.Get <TextureMsaaMode>(MethodOffset.RtMsaaMode);
int samplesInX = msaaMode.SamplesInX();
int samplesInY = msaaMode.SamplesInY();
bool changedScale = false;
for (int index = 0; index < Constants.TotalRenderTargets; index++)
{
int rtIndex = useControl ? rtControl.UnpackPermutationIndex(index) : index;
var colorState = state.Get <RtColorState>(MethodOffset.RtColorState, rtIndex);
if (index >= count || !IsRtEnabled(colorState))
{
changedScale |= TextureManager.SetRenderTargetColor(index, null);
continue;
}
Texture color = TextureManager.FindOrCreateTexture(colorState, samplesInX, samplesInY);
changedScale |= TextureManager.SetRenderTargetColor(index, color);
if (color != null)
{
color.SignalModified();
}
}
bool dsEnable = state.Get <Boolean32>(MethodOffset.RtDepthStencilEnable);
Texture depthStencil = null;
if (dsEnable)
{
var dsState = state.Get <RtDepthStencilState>(MethodOffset.RtDepthStencilState);
var dsSize = state.Get <Size3D> (MethodOffset.RtDepthStencilSize);
depthStencil = TextureManager.FindOrCreateTexture(dsState, dsSize, samplesInX, samplesInY);
}
changedScale |= TextureManager.SetRenderTargetDepthStencil(depthStencil);
if (changedScale)
{
TextureManager.UpdateRenderTargetScale(singleUse);
_context.Renderer.Pipeline.SetRenderTargetScale(TextureManager.RenderTargetScale);
UpdateViewportTransform(state);
UpdateScissorState(state);
}
if (depthStencil != null)
{
depthStencil.SignalModified();
}
}
/// <summary>
/// Dispatches compute work.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
public void Dispatch(GpuState state, int argument)
{
uint qmdAddress = (uint)state.Get <int>(MethodOffset.DispatchParamsAddress);
var qmd = _context.MemoryManager.Read <ComputeQmd>((ulong)qmdAddress << 8);
GpuVa shaderBaseAddress = state.Get <GpuVa>(MethodOffset.ShaderBaseAddress);
ulong shaderGpuVa = shaderBaseAddress.Pack() + (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);
BufferManager.SetComputeUniformBuffer(index, gpuVa, size);
}
ShaderBundle cs = ShaderCache.GetComputeShader(
state,
shaderGpuVa,
qmd.CtaThreadDimension0,
qmd.CtaThreadDimension1,
qmd.CtaThreadDimension2,
localMemorySize,
sharedMemorySize);
_context.Renderer.Pipeline.SetProgram(cs.HostProgram);
var samplerPool = state.Get <PoolState>(MethodOffset.SamplerPoolState);
var texturePool = state.Get <PoolState>(MethodOffset.TexturePoolState);
TextureManager.SetComputeSamplerPool(samplerPool.Address.Pack(), samplerPool.MaximumId, qmd.SamplerIndex);
TextureManager.SetComputeTexturePool(texturePool.Address.Pack(), texturePool.MaximumId);
TextureManager.SetComputeTextureBufferIndex(state.Get <int>(MethodOffset.TextureBufferIndex));
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 = BufferManager.GetComputeUniformBufferAddress(0);
int cbDescOffset = 0x260 + (cb.Slot - 8) * 0x10;
cbDescAddress += (ulong)cbDescOffset;
SbDescriptor cbDescriptor = _context.PhysicalMemory.Read <SbDescriptor>(cbDescAddress);
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 = BufferManager.GetComputeUniformBufferAddress(0);
int sbDescOffset = 0x310 + sb.Slot * 0x10;
sbDescAddress += (ulong)sbDescOffset;
SbDescriptor sbDescriptor = _context.PhysicalMemory.Read <SbDescriptor>(sbDescAddress);
BufferManager.SetComputeStorageBuffer(sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size, sb.Flags);
}
BufferManager.SetComputeStorageBufferBindings(info.SBuffers);
BufferManager.SetComputeUniformBufferBindings(info.CBuffers);
var textureBindings = new TextureBindingInfo[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);
}
TextureManager.SetComputeTextures(textureBindings);
var imageBindings = new TextureBindingInfo[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);
}
TextureManager.SetComputeImages(imageBindings);
TextureManager.CommitComputeBindings();
BufferManager.CommitComputeBindings();
_context.Renderer.Pipeline.DispatchCompute(
qmd.CtaRasterWidth,
qmd.CtaRasterHeight,
qmd.CtaRasterDepth);
_forceShaderUpdate = true;
}
/// <summary>
/// Updates host depth clamp state based on current GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdateDepthClampState(GpuState state)
{
ViewVolumeClipControl clip = state.Get <ViewVolumeClipControl>(MethodOffset.ViewVolumeClipControl);
_context.Renderer.Pipeline.SetDepthClamp((clip & ViewVolumeClipControl.DepthClampDisabled) == 0);
}
/// <summary>
/// Executes a macro.
/// </summary>
/// <param name="index">Index of the macro</param>
/// <param name="state">Current GPU state</param>
public void CallMme(int index, GpuState state)
{
_macros[index].Execute(_macroCode, state);
}
/// <summary>
/// Updates host logical operation state, based on guest state.
/// </summary>
/// <param name="state">Current GPU state</param>
public void UpdateLogicOpState(GpuState state)
{
LogicalOpState logicOpState = state.Get <LogicalOpState>(MethodOffset.LogicOpState);
_context.Renderer.Pipeline.SetLogicOpState(logicOpState.Enable, logicOpState.LogicalOp);
}
/// <summary>
/// Performs a buffer to buffer, or buffer to texture copy.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void CopyBuffer(GpuState state, int argument)
{
var cbp = state.Get <CopyBufferParams>(MethodOffset.CopyBufferParams);
var swizzle = state.Get <CopyBufferSwizzle>(MethodOffset.CopyBufferSwizzle);
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);
int size = cbp.XCount;
if (size == 0)
{
return;
}
FlushUboUpdate();
if (copy2D)
{
// Buffer to texture copy.
int srcBpp = remap ? swizzle.UnpackSrcComponentsCount() * swizzle.UnpackComponentSize() : 1;
int dstBpp = remap ? swizzle.UnpackDstComponentsCount() * swizzle.UnpackComponentSize() : 1;
var dst = state.Get <CopyBufferTexture>(MethodOffset.CopyBufferDstTexture);
var src = state.Get <CopyBufferTexture>(MethodOffset.CopyBufferSrcTexture);
var srcCalculator = new OffsetCalculator(
src.Width,
src.Height,
cbp.SrcStride,
srcLinear,
src.MemoryLayout.UnpackGobBlocksInY(),
src.MemoryLayout.UnpackGobBlocksInZ(),
srcBpp);
var dstCalculator = new OffsetCalculator(
dst.Width,
dst.Height,
cbp.DstStride,
dstLinear,
dst.MemoryLayout.UnpackGobBlocksInY(),
dst.MemoryLayout.UnpackGobBlocksInZ(),
dstBpp);
ulong srcBaseAddress = _context.MemoryManager.Translate(cbp.SrcAddress.Pack());
ulong dstBaseAddress = _context.MemoryManager.Translate(cbp.DstAddress.Pack());
(int srcBaseOffset, int srcSize) = srcCalculator.GetRectangleRange(src.RegionX, src.RegionY, cbp.XCount, cbp.YCount);
(int dstBaseOffset, int dstSize) = dstCalculator.GetRectangleRange(dst.RegionX, dst.RegionY, cbp.XCount, cbp.YCount);
ReadOnlySpan <byte> srcSpan = _context.PhysicalMemory.GetSpan(srcBaseAddress + (ulong)srcBaseOffset, srcSize, true);
Span <byte> dstSpan = _context.PhysicalMemory.GetSpan(dstBaseAddress + (ulong)dstBaseOffset, dstSize).ToArray();
bool completeSource = IsTextureCopyComplete(cbp, src, srcLinear, srcBpp, cbp.SrcStride);
bool completeDest = IsTextureCopyComplete(cbp, dst, dstLinear, dstBpp, cbp.DstStride);
if (completeSource && completeDest)
{
Image.Texture target = TextureManager.FindTexture(dst, cbp, swizzle, dstLinear);
if (target != null)
{
ReadOnlySpan <byte> data;
if (srcLinear)
{
data = LayoutConverter.ConvertLinearStridedToLinear(
target.Info.Width,
target.Info.Height,
1,
1,
cbp.SrcStride,
target.Info.FormatInfo.BytesPerPixel,
srcSpan);
}
else
{
data = LayoutConverter.ConvertBlockLinearToLinear(
src.Width,
src.Height,
1,
target.Info.Levels,
1,
1,
1,
srcBpp,
src.MemoryLayout.UnpackGobBlocksInY(),
src.MemoryLayout.UnpackGobBlocksInZ(),
1,
new SizeInfo((int)target.Size),
srcSpan);
}
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.
_context.PhysicalMemory.Write(dstBaseAddress + (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 < cbp.YCount; y++)
{
srcCalculator.SetY(src.RegionY + y);
dstCalculator.SetY(dst.RegionY + y);
for (int x = 0; x < cbp.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.")
};
_context.PhysicalMemory.Write(dstBaseAddress + (ulong)dstBaseOffset, dstSpan);
}
else
{
if (remap &&
swizzle.UnpackDstX() == BufferSwizzleComponent.ConstA &&
swizzle.UnpackDstY() == BufferSwizzleComponent.ConstA &&
swizzle.UnpackDstZ() == BufferSwizzleComponent.ConstA &&
swizzle.UnpackDstW() == BufferSwizzleComponent.ConstA &&
swizzle.UnpackSrcComponentsCount() == 1 &&
swizzle.UnpackDstComponentsCount() == 1 &&
swizzle.UnpackComponentSize() == 4)
{
// Fast path for clears when remap is enabled.
BufferManager.ClearBuffer(cbp.DstAddress, (uint)size * 4, state.Get <uint>(MethodOffset.CopyBufferConstA));
}
else
{
// TODO: Implement remap functionality.
// Buffer to buffer copy.
BufferManager.CopyBuffer(cbp.SrcAddress, cbp.DstAddress, (uint)size);
}
}
}
/// <summary>
/// Gets a texture descriptor used on the graphics pipeline.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="stageIndex">Index of the shader stage where the texture is bound</param>
/// <param name="handle">Shader "fake" handle of the texture</param>
/// <returns>The texture descriptor</returns>
public TextureDescriptor GetGraphicsTextureDescriptor(GpuState state, int stageIndex, int handle)
{
return(_gpBindingsManager.GetTextureDescriptor(state, stageIndex, handle));
}
/// <summary>
/// Creates a new instance of the GPU state accessor for graphics shader translation.
/// </summary>
/// <param name="context">GPU context</param>
/// <param name="state">Current GPU state</param>
/// <param name="stageIndex">Graphics shader stage index (0 = Vertex, 4 = Fragment)</param>
public GpuAccessor(GpuContext context, GpuState state, int stageIndex)
{
_context = context;
_state = state;
_stageIndex = stageIndex;
}
/// <summary>
/// Updates host state based on the current guest GPU state.
/// </summary>
/// <param name="state">Guest GPU state</param>
private void UpdateState(GpuState state)
{
// Shaders must be the first one to be updated if modified, because
// some of the other state depends on information from the currently
// bound shaders.
if (state.QueryModified(MethodOffset.ShaderBaseAddress, MethodOffset.ShaderState))
{
UpdateShaderState(state);
}
if (state.QueryModified(MethodOffset.RtColorState,
MethodOffset.RtDepthStencilState,
MethodOffset.RtControl,
MethodOffset.RtDepthStencilSize,
MethodOffset.RtDepthStencilEnable))
{
UpdateRenderTargetState(state, useControl: true);
}
if (state.QueryModified(MethodOffset.ScissorState))
{
UpdateScissorState(state);
}
if (state.QueryModified(MethodOffset.DepthTestEnable,
MethodOffset.DepthWriteEnable,
MethodOffset.DepthTestFunc))
{
UpdateDepthTestState(state);
}
if (state.QueryModified(MethodOffset.DepthMode, MethodOffset.ViewportTransform, MethodOffset.ViewportExtents))
{
UpdateViewportTransform(state);
}
if (state.QueryModified(MethodOffset.DepthBiasState,
MethodOffset.DepthBiasFactor,
MethodOffset.DepthBiasUnits,
MethodOffset.DepthBiasClamp))
{
UpdateDepthBiasState(state);
}
if (state.QueryModified(MethodOffset.StencilBackMasks,
MethodOffset.StencilTestState,
MethodOffset.StencilBackTestState))
{
UpdateStencilTestState(state);
}
// Pools.
if (state.QueryModified(MethodOffset.SamplerPoolState, MethodOffset.SamplerIndex))
{
UpdateSamplerPoolState(state);
}
if (state.QueryModified(MethodOffset.TexturePoolState))
{
UpdateTexturePoolState(state);
}
// Input assembler state.
if (state.QueryModified(MethodOffset.VertexAttribState))
{
UpdateVertexAttribState(state);
}
if (state.QueryModified(MethodOffset.PointSize))
{
UpdatePointSizeState(state);
}
if (state.QueryModified(MethodOffset.PrimitiveRestartState))
{
UpdatePrimitiveRestartState(state);
}
if (state.QueryModified(MethodOffset.IndexBufferState))
{
UpdateIndexBufferState(state);
}
if (state.QueryModified(MethodOffset.VertexBufferDrawState,
MethodOffset.VertexBufferInstanced,
MethodOffset.VertexBufferState,
MethodOffset.VertexBufferEndAddress))
{
UpdateVertexBufferState(state);
}
if (state.QueryModified(MethodOffset.FaceState))
{
UpdateFaceState(state);
}
if (state.QueryModified(MethodOffset.RtColorMaskShared, MethodOffset.RtColorMask))
{
UpdateRtColorMask(state);
}
if (state.QueryModified(MethodOffset.BlendIndependent,
MethodOffset.BlendStateCommon,
MethodOffset.BlendEnableCommon,
MethodOffset.BlendEnable,
MethodOffset.BlendState))
{
UpdateBlendState(state);
}
CommitBindings();
}
/// <summary>
/// Gets a graphics shader program from the shader cache.
/// This includes all the specified shader stages.
/// </summary>
/// <remarks>
/// This automatically translates, compiles and adds the code to the cache if not present.
/// </remarks>
/// <param name="state">Current GPU state</param>
/// <param name="addresses">Addresses of the shaders for each stage</param>
/// <returns>Compiled graphics shader code</returns>
public ShaderBundle GetGraphicsShader(GpuState state, ShaderAddresses addresses)
{
bool isCached = _gpPrograms.TryGetValue(addresses, out List <ShaderBundle> list);
if (isCached)
{
foreach (ShaderBundle cachedGpShaders in list)
{
if (IsShaderEqual(cachedGpShaders, addresses))
{
return(cachedGpShaders);
}
}
}
ShaderCodeHolder[] shaders = new ShaderCodeHolder[Constants.ShaderStages];
var tfd = GetTransformFeedbackDescriptors(state);
TranslationFlags flags = DefaultFlags;
if (tfd != null)
{
flags |= TranslationFlags.Feedback;
}
if (addresses.VertexA != 0)
{
shaders[0] = TranslateGraphicsShader(state, flags, ShaderStage.Vertex, addresses.Vertex, addresses.VertexA);
}
else
{
shaders[0] = TranslateGraphicsShader(state, flags, ShaderStage.Vertex, addresses.Vertex);
}
shaders[1] = TranslateGraphicsShader(state, flags, ShaderStage.TessellationControl, addresses.TessControl);
shaders[2] = TranslateGraphicsShader(state, flags, ShaderStage.TessellationEvaluation, addresses.TessEvaluation);
shaders[3] = TranslateGraphicsShader(state, flags, ShaderStage.Geometry, addresses.Geometry);
shaders[4] = TranslateGraphicsShader(state, flags, ShaderStage.Fragment, addresses.Fragment);
List <IShader> hostShaders = new List <IShader>();
for (int stage = 0; stage < Constants.ShaderStages; stage++)
{
ShaderProgram program = shaders[stage]?.Program;
if (program == null)
{
continue;
}
IShader hostShader = _context.Renderer.CompileShader(program);
shaders[stage].HostShader = hostShader;
hostShaders.Add(hostShader);
}
IProgram hostProgram = _context.Renderer.CreateProgram(hostShaders.ToArray(), tfd);
ShaderBundle gpShaders = new ShaderBundle(hostProgram, shaders);
if (!isCached)
{
list = new List <ShaderBundle>();
_gpPrograms.Add(addresses, list);
}
list.Add(gpShaders);
return(gpShaders);
}
/// <summary>
/// Updates host shaders based on the guest GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdateShaderState(GpuState state)
{
ShaderAddresses addresses = new ShaderAddresses();
Span <ShaderAddresses> addressesSpan = MemoryMarshal.CreateSpan(ref addresses, 1);
Span <ulong> addressesArray = MemoryMarshal.Cast <ShaderAddresses, ulong>(addressesSpan);
ulong baseAddress = state.Get <GpuVa>(MethodOffset.ShaderBaseAddress).Pack();
for (int index = 0; index < 6; index++)
{
var shader = state.Get <ShaderState>(MethodOffset.ShaderState, index);
if (!shader.UnpackEnable() && index != 1)
{
continue;
}
addressesArray[index] = baseAddress + shader.Offset;
}
GraphicsShader gs = ShaderCache.GetGraphicsShader(state, addresses);
_vsUsesInstanceId = gs.Shaders[0]?.Program.Info.UsesInstanceId ?? false;
for (int stage = 0; stage < Constants.ShaderStages; stage++)
{
ShaderProgramInfo info = gs.Shaders[stage]?.Program.Info;
_currentProgramInfo[stage] = info;
if (info == null)
{
continue;
}
var textureBindings = new TextureBindingInfo[info.Textures.Count];
for (int index = 0; index < info.Textures.Count; index++)
{
var descriptor = info.Textures[index];
Target target = GetTarget(descriptor.Type);
if (descriptor.IsBindless)
{
textureBindings[index] = new TextureBindingInfo(target, descriptor.CbufSlot, descriptor.CbufOffset);
}
else
{
textureBindings[index] = new TextureBindingInfo(target, descriptor.HandleIndex);
}
}
TextureManager.SetGraphicsTextures(stage, textureBindings);
var imageBindings = new TextureBindingInfo[info.Images.Count];
for (int index = 0; index < info.Images.Count; index++)
{
var descriptor = info.Images[index];
Target target = GetTarget(descriptor.Type);
imageBindings[index] = new TextureBindingInfo(target, descriptor.HandleIndex);
}
TextureManager.SetGraphicsImages(stage, imageBindings);
uint sbEnableMask = 0;
uint ubEnableMask = 0;
for (int index = 0; index < info.SBuffers.Count; index++)
{
sbEnableMask |= 1u << info.SBuffers[index].Slot;
}
for (int index = 0; index < info.CBuffers.Count; index++)
{
ubEnableMask |= 1u << info.CBuffers[index].Slot;
}
BufferManager.SetGraphicsStorageBufferEnableMask(stage, sbEnableMask);
BufferManager.SetGraphicsUniformBufferEnableMask(stage, ubEnableMask);
}
_context.Renderer.Pipeline.SetProgram(gs.HostProgram);
}
/// <summary>
/// Updates host state based on the current guest GPU state.
/// </summary>
/// <param name="state">Guest GPU state</param>
private void UpdateState(GpuState state)
{
bool tfEnable = state.Get <Boolean32>(MethodOffset.TfEnable);
if (!tfEnable && _prevTfEnable)
{
_context.Renderer.Pipeline.EndTransformFeedback();
_prevTfEnable = false;
}
// Shaders must be the first one to be updated if modified, because
// some of the other state depends on information from the currently
// bound shaders.
if (state.QueryModified(MethodOffset.ShaderBaseAddress, MethodOffset.ShaderState) || _forceShaderUpdate)
{
_forceShaderUpdate = false;
UpdateShaderState(state);
}
if (state.QueryModified(MethodOffset.TfBufferState))
{
UpdateTfBufferState(state);
}
if (state.QueryModified(MethodOffset.ClipDistanceEnable))
{
UpdateUserClipState(state);
}
if (state.QueryModified(MethodOffset.RasterizeEnable))
{
UpdateRasterizerState(state);
}
if (state.QueryModified(MethodOffset.RtColorState,
MethodOffset.RtDepthStencilState,
MethodOffset.RtControl,
MethodOffset.RtDepthStencilSize,
MethodOffset.RtDepthStencilEnable))
{
UpdateRenderTargetState(state, useControl: true);
}
if (state.QueryModified(MethodOffset.ScissorState))
{
UpdateScissorState(state);
}
if (state.QueryModified(MethodOffset.ViewVolumeClipControl))
{
UpdateDepthClampState(state);
}
if (state.QueryModified(MethodOffset.DepthTestEnable,
MethodOffset.DepthWriteEnable,
MethodOffset.DepthTestFunc))
{
UpdateDepthTestState(state);
}
if (state.QueryModified(MethodOffset.DepthMode,
MethodOffset.ViewportTransform,
MethodOffset.ViewportExtents))
{
UpdateViewportTransform(state);
}
if (state.QueryModified(MethodOffset.DepthBiasState,
MethodOffset.DepthBiasFactor,
MethodOffset.DepthBiasUnits,
MethodOffset.DepthBiasClamp))
{
UpdateDepthBiasState(state);
}
if (state.QueryModified(MethodOffset.StencilBackMasks,
MethodOffset.StencilTestState,
MethodOffset.StencilBackTestState))
{
UpdateStencilTestState(state);
}
// Pools.
if (state.QueryModified(MethodOffset.SamplerPoolState, MethodOffset.SamplerIndex))
{
UpdateSamplerPoolState(state);
}
if (state.QueryModified(MethodOffset.TexturePoolState))
{
UpdateTexturePoolState(state);
}
// Input assembler state.
if (state.QueryModified(MethodOffset.VertexAttribState))
{
UpdateVertexAttribState(state);
}
if (state.QueryModified(MethodOffset.PointSize,
MethodOffset.VertexProgramPointSize,
MethodOffset.PointSpriteEnable,
MethodOffset.PointCoordReplace))
{
UpdatePointState(state);
}
if (state.QueryModified(MethodOffset.PrimitiveRestartState))
{
UpdatePrimitiveRestartState(state);
}
if (state.QueryModified(MethodOffset.IndexBufferState))
{
UpdateIndexBufferState(state);
}
if (state.QueryModified(MethodOffset.VertexBufferDrawState,
MethodOffset.VertexBufferInstanced,
MethodOffset.VertexBufferState,
MethodOffset.VertexBufferEndAddress))
{
UpdateVertexBufferState(state);
}
if (state.QueryModified(MethodOffset.FaceState))
{
UpdateFaceState(state);
}
if (state.QueryModified(MethodOffset.RtColorMaskShared, MethodOffset.RtColorMask))
{
UpdateRtColorMask(state);
}
if (state.QueryModified(MethodOffset.BlendIndependent,
MethodOffset.BlendConstant,
MethodOffset.BlendStateCommon,
MethodOffset.BlendEnableCommon,
MethodOffset.BlendEnable,
MethodOffset.BlendState))
{
UpdateBlendState(state);
}
if (state.QueryModified(MethodOffset.LogicOpState))
{
UpdateLogicOpState(state);
}
CommitBindings();
if (tfEnable && !_prevTfEnable)
{
_context.Renderer.Pipeline.BeginTransformFeedback(PrimitiveType.Convert());
_prevTfEnable = true;
}
}
/// <summary>
/// Issues a texture barrier.
/// This waits until previous texture writes from the GPU to finish, before
/// performing new operations with said textures.
/// This performs a per-tile wait, it is only valid if both the previous write
/// and current access has the same access patterns.
/// This may be faster than the regular barrier on tile-based rasterizers.
/// </summary>
/// <param name="state">Current GPU state (unused)</param>
/// <param name="argument">Method call argument (unused)</param>
private void TextureBarrierTiled(GpuState state, int argument)
{
_context.Renderer.Pipeline.TextureBarrierTiled();
}
/// <summary>
/// Updates Rasterizer primitive discard state based on guest gpu state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdateRasterizerState(GpuState state)
{
Boolean32 enable = state.Get <Boolean32>(MethodOffset.RasterizeEnable);
_context.Renderer.Pipeline.SetRasterizerDiscard(!enable);
}
/// <summary>
/// Performs a indexed draw with a low number of index buffer elements.
/// </summary>
/// <param name="state">Current GPU state</param>
/// <param name="argument">Method call argument</param>
private void DrawIndexedSmall2(GpuState state, int argument)
{
DrawIndexedSmall(state, argument);
}
/// <summary>
/// Updates host shaders based on the guest GPU state.
/// </summary>
/// <param name="state">Current GPU state</param>
private void UpdateShaderState(GpuState state)
{
ShaderAddresses addresses = new ShaderAddresses();
Span <ShaderAddresses> addressesSpan = MemoryMarshal.CreateSpan(ref addresses, 1);
Span <ulong> addressesArray = MemoryMarshal.Cast <ShaderAddresses, ulong>(addressesSpan);
ulong baseAddress = state.Get <GpuVa>(MethodOffset.ShaderBaseAddress).Pack();
for (int index = 0; index < 6; index++)
{
var shader = state.Get <ShaderState>(MethodOffset.ShaderState, index);
if (!shader.UnpackEnable() && index != 1)
{
continue;
}
addressesArray[index] = baseAddress + shader.Offset;
}
ShaderBundle gs = state.Channel.MemoryManager.Physical.ShaderCache.GetGraphicsShader(state, addresses);
byte oldVsClipDistancesWritten = _vsClipDistancesWritten;
_vsUsesInstanceId = gs.Shaders[0]?.Info.UsesInstanceId ?? false;
_vsClipDistancesWritten = gs.Shaders[0]?.Info.ClipDistancesWritten ?? 0;
if (oldVsClipDistancesWritten != _vsClipDistancesWritten)
{
UpdateUserClipState(state);
}
int storageBufferBindingsCount = 0;
int uniformBufferBindingsCount = 0;
for (int stage = 0; stage < Constants.ShaderStages; stage++)
{
ShaderProgramInfo info = gs.Shaders[stage]?.Info;
_currentProgramInfo[stage] = info;
if (info == null)
{
state.Channel.TextureManager.SetGraphicsTextures(stage, Array.Empty <TextureBindingInfo>());
state.Channel.TextureManager.SetGraphicsImages(stage, Array.Empty <TextureBindingInfo>());
state.Channel.BufferManager.SetGraphicsStorageBufferBindings(stage, null);
state.Channel.BufferManager.SetGraphicsUniformBufferBindings(stage, null);
continue;
}
var textureBindings = new TextureBindingInfo[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);
}
state.Channel.TextureManager.SetGraphicsTextures(stage, textureBindings);
var imageBindings = new TextureBindingInfo[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);
}
state.Channel.TextureManager.SetGraphicsImages(stage, imageBindings);
state.Channel.BufferManager.SetGraphicsStorageBufferBindings(stage, info.SBuffers);
state.Channel.BufferManager.SetGraphicsUniformBufferBindings(stage, info.CBuffers);
if (info.SBuffers.Count != 0)
{
storageBufferBindingsCount = Math.Max(storageBufferBindingsCount, info.SBuffers.Max(x => x.Binding) + 1);
}
if (info.CBuffers.Count != 0)
{
uniformBufferBindingsCount = Math.Max(uniformBufferBindingsCount, info.CBuffers.Max(x => x.Binding) + 1);
}
}
state.Channel.BufferManager.SetGraphicsStorageBufferBindingsCount(storageBufferBindingsCount);
state.Channel.BufferManager.SetGraphicsUniformBufferBindingsCount(uniformBufferBindingsCount);
_context.Renderer.Pipeline.SetProgram(gs.HostProgram);
}
