Geometry shader emulation for macOS (#5551)

* Implement vertex and geometry shader conversion to compute

* Call InitializeReservedCounts for compute too

* PR feedback

* Set clip distance mask for geometry and tessellation shaders too

* Transform feedback emulation only for vertex
This commit is contained in:
gdkchan
2023-08-29 21:10:34 -03:00
committed by GitHub
parent 93d78f9ac4
commit f09bba82b9
65 changed files with 3912 additions and 593 deletions

View File

@ -35,7 +35,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
ShaderSpecializationState oldSpecState,
ShaderSpecializationState newSpecState,
ResourceCounts counts,
int stageIndex) : base(context, counts, stageIndex, oldSpecState.TransformFeedbackDescriptors != null)
int stageIndex) : base(context, counts, stageIndex)
{
_data = data;
_cb1Data = cb1Data;

View File

@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
private const ushort FileFormatVersionMajor = 1;
private const ushort FileFormatVersionMinor = 2;
private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor;
private const uint CodeGenVersion = 5609;
private const uint CodeGenVersion = 5551;
private const string SharedTocFileName = "shared.toc";
private const string SharedDataFileName = "shared.data";
@ -140,6 +140,21 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
/// </summary>
public ShaderStage Stage;
/// <summary>
/// Number of vertices that each output primitive has on a geometry shader.
/// </summary>
public byte GeometryVerticesPerPrimitive;
/// <summary>
/// Maximum number of vertices that a geometry shader may generate.
/// </summary>
public ushort GeometryMaxOutputVertices;
/// <summary>
/// Number of invocations per primitive on tessellation or geometry shaders.
/// </summary>
public ushort ThreadsPerInputPrimitive;
/// <summary>
/// Indicates if the fragment shader accesses the fragment coordinate built-in variable.
/// </summary>
@ -783,9 +798,10 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
sBuffers,
textures,
images,
ShaderIdentification.None,
0,
dataInfo.Stage,
dataInfo.GeometryVerticesPerPrimitive,
dataInfo.GeometryMaxOutputVertices,
dataInfo.ThreadsPerInputPrimitive,
dataInfo.UsesFragCoord,
dataInfo.UsesInstanceId,
dataInfo.UsesDrawParameters,
@ -813,6 +829,9 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
TexturesCount = (ushort)info.Textures.Count,
ImagesCount = (ushort)info.Images.Count,
Stage = info.Stage,
GeometryVerticesPerPrimitive = (byte)info.GeometryVerticesPerPrimitive,
GeometryMaxOutputVertices = (ushort)info.GeometryMaxOutputVertices,
ThreadsPerInputPrimitive = (ushort)info.ThreadsPerInputPrimitive,
UsesFragCoord = info.UsesFragCoord,
UsesInstanceId = info.UsesInstanceId,
UsesDrawParameters = info.UsesDrawParameters,

View File

@ -595,6 +595,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
ResourceCounts counts = new();
DiskCacheGpuAccessor[] gpuAccessors = new DiskCacheGpuAccessor[Constants.ShaderStages];
TranslatorContext[] translatorContexts = new TranslatorContext[Constants.ShaderStages + 1];
TranslatorContext nextStage = null;
@ -626,14 +627,22 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
translatorContexts[0] = DecodeGraphicsShader(gpuAccessorA, api, DefaultFlags | TranslationFlags.VertexA, 0);
}
gpuAccessors[stageIndex] = gpuAccessor;
translatorContexts[stageIndex + 1] = currentStage;
nextStage = currentStage;
}
}
if (!_context.Capabilities.SupportsGeometryShader)
bool hasGeometryShader = translatorContexts[4] != null;
bool vertexHasStore = translatorContexts[1] != null && translatorContexts[1].HasStore;
bool geometryHasStore = hasGeometryShader && translatorContexts[4].HasStore;
bool vertexToCompute = ShouldConvertVertexToCompute(_context, vertexHasStore, geometryHasStore, hasGeometryShader);
// We don't support caching shader stages that have been converted to compute currently,
// so just eliminate them if they exist in the cache.
if (vertexToCompute)
{
ShaderCache.TryRemoveGeometryStage(translatorContexts);
return;
}
CachedShaderStage[] shaders = new CachedShaderStage[guestShaders.Length];
@ -647,6 +656,8 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
if (currentStage != null)
{
gpuAccessors[stageIndex].InitializeReservedCounts(specState.TransformFeedbackDescriptors != null, vertexToCompute);
ShaderProgram program;
byte[] guestCode = guestShaders[stageIndex + 1].Value.Code;
@ -701,6 +712,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
ResourceCounts counts = new();
ShaderSpecializationState newSpecState = new(ref specState.ComputeState);
DiskCacheGpuAccessor gpuAccessor = new(_context, shader.Code, shader.Cb1Data, specState, newSpecState, counts, 0);
gpuAccessor.InitializeReservedCounts(tfEnabled: false, vertexAsCompute: false);
TranslatorContext translatorContext = DecodeComputeShader(gpuAccessor, _context.Capabilities.Api, 0);