Vulkan: Add workarounds for MoltenVK (#4202)

* Add MVK basics.

* Use appropriate output attribute types

* 4kb vertex alignment, bunch of fixes

* Add reduced shader precision mode for mvk.

* Disable ASTC on MVK for now

* Only request robustnes2 when it is available.

* It's just the one feature actually

* Add triangle fan conversion

* Allow NullDescriptor on MVK for some reason.

* Force safe blit on MoltenVK

* Use ASTC only when formats are all available.

* Disable multilevel 3d texture views

* Filter duplicate render targets (on backend)

* Add Automatic MoltenVK Configuration

* Do not create color attachment views with formats that are not RT compatible

* Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats

* FIx rebase for Vertex Attrib State

* Fix 4b alignment for vertex

* Use asynchronous queue submits for MVK

* Ensure color clear shader has correct output type

* Update MoltenVK config

* Always use MoltenVK workarounds on MacOS

* Make MVK supersede all vendors

* Fix rebase

* Various fixes on rebase

* Get portability flags from extension

* Fix some minor rebasing issues

* Style change

* Use LibraryImport for MVKConfiguration

* Rename MoltenVK vendor to Apple

Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B.

* Fix features2 rebase conflict

* Rename fragment output type

* Add missing check for fragment output types

Might have caused the crash in MK8

* Only do fragment output specialization on MoltenVK

* Avoid copy when passing capabilities

* Self feedback

* Address feedback

Co-authored-by: gdk <gab.dark.100@gmail.com>
Co-authored-by: nastys <nastys@users.noreply.github.com>
This commit is contained in:
riperiperi 2023-01-13 00:31:21 +00:00 committed by GitHub
parent 30862b5ffd
commit 8fa248ceb4
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
36 changed files with 739 additions and 61 deletions

View file

@ -9,6 +9,8 @@ namespace Ryujinx.Graphics.GAL
public readonly bool HasFrontFacingBug;
public readonly bool HasVectorIndexingBug;
public readonly bool NeedsFragmentOutputSpecialization;
public readonly bool ReduceShaderPrecision;
public readonly bool SupportsAstcCompression;
public readonly bool SupportsBc123Compression;
@ -49,6 +51,8 @@ namespace Ryujinx.Graphics.GAL
string vendorName,
bool hasFrontFacingBug,
bool hasVectorIndexingBug,
bool needsFragmentOutputSpecialization,
bool reduceShaderPrecision,
bool supportsAstcCompression,
bool supportsBc123Compression,
bool supportsBc45Compression,
@ -85,6 +89,8 @@ namespace Ryujinx.Graphics.GAL
VendorName = vendorName;
HasFrontFacingBug = hasFrontFacingBug;
HasVectorIndexingBug = hasVectorIndexingBug;
NeedsFragmentOutputSpecialization = needsFragmentOutputSpecialization;
ReduceShaderPrecision = reduceShaderPrecision;
SupportsAstcCompression = supportsAstcCompression;
SupportsBc123Compression = supportsBc123Compression;
SupportsBc45Compression = supportsBc45Compression;

View file

@ -1,5 +1,6 @@
using Ryujinx.Common.Memory;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu.Engine.Types;
using Ryujinx.Graphics.Gpu.Shader;
using Ryujinx.Graphics.Shader;
@ -10,6 +11,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
/// </summary>
internal class SpecializationStateUpdater
{
private readonly GpuContext _context;
private GpuChannelGraphicsState _graphics;
private GpuChannelPoolState _pool;
@ -18,6 +20,15 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
private bool _changed;
/// <summary>
/// Creates a new instance of the specialization state updater class.
/// </summary>
/// <param name="context">GPU context</param>
public SpecializationStateUpdater(GpuContext context)
{
_context = context;
}
/// <summary>
/// Signal that the specialization state has changed.
/// </summary>
@ -232,6 +243,42 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
}
}
/// <summary>
/// Updates the type of the outputs produced by the fragment shader based on the current render target state.
/// </summary>
/// <param name="rtControl">The render target control register</param>
/// <param name="state">The color attachment state</param>
public void SetFragmentOutputTypes(RtControl rtControl, ref Array8<RtColorState> state)
{
bool changed = false;
int count = rtControl.UnpackCount();
for (int index = 0; index < Constants.TotalRenderTargets; index++)
{
int rtIndex = rtControl.UnpackPermutationIndex(index);
var colorState = state[rtIndex];
if (index < count && StateUpdater.IsRtEnabled(colorState))
{
Format format = colorState.Format.Convert().Format;
AttributeType type = format.IsInteger() ? (format.IsSint() ? AttributeType.Sint : AttributeType.Uint) : AttributeType.Float;
if (type != _graphics.FragmentOutputTypes[index])
{
_graphics.FragmentOutputTypes[index] = type;
changed = true;
}
}
}
if (changed && _context.Capabilities.NeedsFragmentOutputSpecialization)
{
Signal();
}
}
/// <summary>
/// Indicates that the draw is writing the base vertex, base instance and draw index to Constant Buffer 0.
/// </summary>

View file

@ -138,6 +138,16 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
_dirtyMask = ulong.MaxValue >> ((sizeof(ulong) * 8) - _callbacks.Length);
}
/// <summary>
/// Check if the given register group is dirty without clearing it.
/// </summary>
/// <param name="groupIndex">Index of the group to check</param>
/// <returns>True if dirty, false otherwise</returns>
public bool IsDirty(int groupIndex)
{
return (_dirtyMask & (1UL << groupIndex)) != 0;
}
/// <summary>
/// Check all the groups specified by <paramref name="checkMask"/> for modification, and update if modified.
/// </summary>

View file

@ -20,6 +20,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
public const int ScissorStateIndex = 16;
public const int VertexBufferStateIndex = 0;
public const int PrimitiveRestartStateIndex = 12;
public const int RenderTargetStateIndex = 27;
private readonly GpuContext _context;
private readonly GpuChannel _channel;
@ -264,6 +265,11 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
_prevTfEnable = false;
}
if (_updateTracker.IsDirty(RenderTargetStateIndex))
{
UpdateRenderTargetSpecialization();
}
_updateTracker.Update(ulong.MaxValue);
// If any state that the shader depends on changed,
@ -526,12 +532,20 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
}
}
/// <summary>
/// Updates specialization state based on render target state.
/// </summary>
public void UpdateRenderTargetSpecialization()
{
_currentSpecState.SetFragmentOutputTypes(_state.State.RtControl, ref _state.State.RtColorState);
}
/// <summary>
/// Checks if a render target color buffer is used.
/// </summary>
/// <param name="colorState">Color buffer information</param>
/// <returns>True if the specified buffer is enabled/used, false otherwise</returns>
private static bool IsRtEnabled(RtColorState colorState)
internal static bool IsRtEnabled(RtColorState colorState)
{
// Colors are disabled by writing 0 to the format.
return colorState.Format != 0 && colorState.WidthOrStride != 0;
@ -893,7 +907,12 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
{
Logger.Debug?.Print(LogClass.Gpu, $"Invalid attribute format 0x{vertexAttrib.UnpackFormat():X}.");
format = Format.R32G32B32A32Float;
format = vertexAttrib.UnpackType() switch
{
VertexAttribType.Sint => Format.R32G32B32A32Sint,
VertexAttribType.Uint => Format.R32G32B32A32Uint,
_ => Format.R32G32B32A32Float
};
}
vertexAttribs[index] = new VertexAttribDescriptor(

View file

@ -71,7 +71,7 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
_i2mClass = new InlineToMemoryClass(context, channel, initializeState: false);
var spec = new SpecializationStateUpdater();
var spec = new SpecializationStateUpdater(context);
var drawState = new DrawState();
_drawManager = new DrawManager(context, channel, _state, drawState, spec);

View file

@ -1,4 +1,5 @@
using Ryujinx.Graphics.Gpu.Engine.GPFifo;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu.Engine.GPFifo;
using Ryujinx.Graphics.Gpu.Image;
using Ryujinx.Graphics.Gpu.Memory;
using System;
@ -31,6 +32,11 @@ namespace Ryujinx.Graphics.Gpu
/// </summary>
internal MemoryManager MemoryManager => _memoryManager;
/// <summary>
/// Host hardware capabilities from the GPU context.
/// </summary>
internal ref Capabilities Capabilities => ref _context.Capabilities;
/// <summary>
/// Creates a new instance of a GPU channel.
/// </summary>

View file

@ -107,6 +107,12 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
return _oldSpecState.GraphicsState.AttributeTypes[location];
}
/// <inheritdoc/>
public AttributeType QueryFragmentOutputType(int location)
{
return _oldSpecState.GraphicsState.FragmentOutputTypes[location];
}
/// <inheritdoc/>
public int QueryComputeLocalSizeX() => _oldSpecState.ComputeState.LocalSizeX;

View file

@ -113,6 +113,12 @@ namespace Ryujinx.Graphics.Gpu.Shader
return _state.GraphicsState.AttributeTypes[location];
}
/// <inheritdoc/>
public AttributeType QueryFragmentOutputType(int location)
{
return _state.GraphicsState.FragmentOutputTypes[location];
}
/// <inheritdoc/>
public int QueryComputeLocalSizeX() => _state.ComputeState.LocalSizeX;

View file

@ -112,6 +112,8 @@ namespace Ryujinx.Graphics.Gpu.Shader
};
}
public bool QueryHostReducedPrecision() => _context.Capabilities.ReduceShaderPrecision;
public bool QueryHostHasFrontFacingBug() => _context.Capabilities.HasFrontFacingBug;
public bool QueryHostHasVectorIndexingBug() => _context.Capabilities.HasVectorIndexingBug;

View file

@ -87,6 +87,11 @@ namespace Ryujinx.Graphics.Gpu.Shader
/// </summary>
public bool HasUnalignedStorageBuffer;
/// <summary>
/// Type of the fragment shader outputs.
/// </summary>
public Array8<AttributeType> FragmentOutputTypes;
/// <summary>
/// Creates a new GPU graphics state.
/// </summary>
@ -105,6 +110,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
/// <param name="attributeTypes">Type of the vertex attributes consumed by the shader</param>
/// <param name="hasConstantBufferDrawParameters">Indicates that the draw is writing the base vertex, base instance and draw index to Constant Buffer 0</param>
/// <param name="hasUnalignedStorageBuffer">Indicates that any storage buffer use is unaligned</param>
/// <param name="fragmentOutputTypes">Type of the fragment shader outputs</param>
public GpuChannelGraphicsState(
bool earlyZForce,
PrimitiveTopology topology,
@ -120,7 +126,8 @@ namespace Ryujinx.Graphics.Gpu.Shader
float alphaTestReference,
ref Array32<AttributeType> attributeTypes,
bool hasConstantBufferDrawParameters,
bool hasUnalignedStorageBuffer)
bool hasUnalignedStorageBuffer,
ref Array8<AttributeType> fragmentOutputTypes)
{
EarlyZForce = earlyZForce;
Topology = topology;
@ -137,6 +144,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
AttributeTypes = attributeTypes;
HasConstantBufferDrawParameters = hasConstantBufferDrawParameters;
HasUnalignedStorageBuffer = hasUnalignedStorageBuffer;
FragmentOutputTypes = fragmentOutputTypes;
}
}
}

View file

@ -530,6 +530,11 @@ namespace Ryujinx.Graphics.Gpu.Shader
return false;
}
if (channel.Capabilities.NeedsFragmentOutputSpecialization && !graphicsState.FragmentOutputTypes.AsSpan().SequenceEqual(GraphicsState.FragmentOutputTypes.AsSpan()))
{
return false;
}
return Matches(channel, ref poolState, checkTextures, isCompute: false);
}

View file

@ -106,6 +106,8 @@ namespace Ryujinx.Graphics.OpenGL
vendorName: GpuVendor,
hasFrontFacingBug: HwCapabilities.Vendor == HwCapabilities.GpuVendor.IntelWindows,
hasVectorIndexingBug: HwCapabilities.Vendor == HwCapabilities.GpuVendor.AmdWindows,
needsFragmentOutputSpecialization: false,
reduceShaderPrecision: false,
supportsAstcCompression: HwCapabilities.SupportsAstcCompression,
supportsBc123Compression: HwCapabilities.SupportsTextureCompressionS3tc,
supportsBc45Compression: HwCapabilities.SupportsTextureCompressionRgtc,

View file

@ -346,12 +346,17 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
{
string name = context.OperandManager.DeclareLocal(decl);
context.AppendLine(GetVarTypeName(decl.VarType) + " " + name + ";");
context.AppendLine(GetVarTypeName(context, decl.VarType) + " " + name + ";");
}
}
public static string GetVarTypeName(AggregateType type, bool precise = true)
public static string GetVarTypeName(CodeGenContext context, AggregateType type, bool precise = true)
{
if (context.Config.GpuAccessor.QueryHostReducedPrecision())
{
precise = false;
}
return type switch
{
AggregateType.Void => "void",
@ -666,7 +671,22 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
}
else
{
context.AppendLine($"layout (location = {attr}) out vec4 {name};");
string type = context.Config.Stage != ShaderStage.Fragment ? "vec4" :
context.Config.GpuAccessor.QueryFragmentOutputType(attr) switch
{
AttributeType.Sint => "ivec4",
AttributeType.Uint => "uvec4",
_ => "vec4"
};
if (context.Config.GpuAccessor.QueryHostReducedPrecision() && context.Config.Stage == ShaderStage.Vertex && attr == 0)
{
context.AppendLine($"layout (location = {attr}) invariant out {type} {name};");
}
else
{
context.AppendLine($"layout (location = {attr}) out {type} {name};");
}
}
}

View file

@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
{
for (int i = 1; i < info.Functions.Count; i++)
{
context.AppendLine($"{GetFunctionSignature(info.Functions[i])};");
context.AppendLine($"{GetFunctionSignature(context, info.Functions[i])};");
}
context.AppendLine();
@ -44,7 +44,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
{
context.CurrentFunction = function;
context.AppendLine(GetFunctionSignature(function, funcName));
context.AppendLine(GetFunctionSignature(context, function, funcName));
context.EnterScope();
Declarations.DeclareLocals(context, function);
@ -54,23 +54,23 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl
context.LeaveScope();
}
private static string GetFunctionSignature(StructuredFunction function, string funcName = null)
private static string GetFunctionSignature(CodeGenContext context, StructuredFunction function, string funcName = null)
{
string[] args = new string[function.InArguments.Length + function.OutArguments.Length];
for (int i = 0; i < function.InArguments.Length; i++)
{
args[i] = $"{Declarations.GetVarTypeName(function.InArguments[i])} {OperandManager.GetArgumentName(i)}";
args[i] = $"{Declarations.GetVarTypeName(context, function.InArguments[i])} {OperandManager.GetArgumentName(i)}";
}
for (int i = 0; i < function.OutArguments.Length; i++)
{
int j = i + function.InArguments.Length;
args[j] = $"out {Declarations.GetVarTypeName(function.OutArguments[i])} {OperandManager.GetArgumentName(j)}";
args[j] = $"out {Declarations.GetVarTypeName(context, function.OutArguments[i])} {OperandManager.GetArgumentName(j)}";
}
return $"{Declarations.GetVarTypeName(function.ReturnType)} {funcName ?? function.Name}({string.Join(", ", args)})";
return $"{Declarations.GetVarTypeName(context, function.ReturnType)} {funcName ?? function.Name}({string.Join(", ", args)})";
}
private static void PrintBlock(CodeGenContext context, AstBlock block)

View file

@ -32,7 +32,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
if ((outputType & AggregateType.ElementCountMask) != 0)
{
return $"{Declarations.GetVarTypeName(outputType, precise: false)}({imageConst})";
return $"{Declarations.GetVarTypeName(context, outputType, precise: false)}({imageConst})";
}
return imageConst;
@ -513,7 +513,7 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Glsl.Instructions
if ((outputType & AggregateType.ElementCountMask) != 0)
{
return $"{Declarations.GetVarTypeName(outputType, precise: false)}({scalarValue})";
return $"{Declarations.GetVarTypeName(context, outputType, precise: false)}({scalarValue})";
}
}

View file

@ -577,6 +577,11 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
context.Decorate(spvVar, Decoration.Patch);
}
if (context.Config.GpuAccessor.QueryHostReducedPrecision() && attr == AttributeConsts.PositionX && context.Config.Stage != ShaderStage.Fragment)
{
context.Decorate(spvVar, Decoration.Invariant);
}
context.Decorate(spvVar, Decoration.BuiltIn, (LiteralInteger)GetBuiltIn(context, attrInfo.BaseValue));
if (context.Config.TransformFeedbackEnabled && context.Config.LastInVertexPipeline && isOutAttr)

View file

@ -2194,13 +2194,23 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
if (operation.Inst.HasFlag(Instruction.FP64))
{
var result = emitF(context.TypeFP64(), context.GetFP64(src1), context.GetFP64(src2));
context.Decorate(result, Decoration.NoContraction);
if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
{
context.Decorate(result, Decoration.NoContraction);
}
return new OperationResult(AggregateType.FP64, result);
}
else if (operation.Inst.HasFlag(Instruction.FP32))
{
var result = emitF(context.TypeFP32(), context.GetFP32(src1), context.GetFP32(src2));
context.Decorate(result, Decoration.NoContraction);
if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
{
context.Decorate(result, Decoration.NoContraction);
}
return new OperationResult(AggregateType.FP32, result);
}
else
@ -2255,13 +2265,23 @@ namespace Ryujinx.Graphics.Shader.CodeGen.Spirv
if (operation.Inst.HasFlag(Instruction.FP64))
{
var result = emitF(context.TypeFP64(), context.GetFP64(src1), context.GetFP64(src2), context.GetFP64(src3));
context.Decorate(result, Decoration.NoContraction);
if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
{
context.Decorate(result, Decoration.NoContraction);
}
return new OperationResult(AggregateType.FP64, result);
}
else if (operation.Inst.HasFlag(Instruction.FP32))
{
var result = emitF(context.TypeFP32(), context.GetFP32(src1), context.GetFP32(src2), context.GetFP32(src3));
context.Decorate(result, Decoration.NoContraction);
if (!context.Config.GpuAccessor.QueryHostReducedPrecision())
{
context.Decorate(result, Decoration.NoContraction);
}
return new OperationResult(AggregateType.FP32, result);
}
else

View file

@ -114,6 +114,16 @@ namespace Ryujinx.Graphics.Shader
return index;
}
/// <summary>
/// Queries output type for fragment shaders.
/// </summary>
/// <param name="location">Location of the framgent output</param>
/// <returns>Output location</returns>
AttributeType QueryFragmentOutputType(int location)
{
return AttributeType.Float;
}
/// <summary>
/// Queries Local Size X for compute shaders.
/// </summary>
@ -186,6 +196,15 @@ namespace Ryujinx.Graphics.Shader
return false;
}
/// <summary>
/// Queries host about whether to reduce precision to improve performance.
/// </summary>
/// <returns>True if precision is limited to vertex position, false otherwise</returns>
bool QueryHostReducedPrecision()
{
return false;
}
/// <summary>
/// Queries host about the presence of the FrontFacing built-in variable bug.
/// </summary>

View file

@ -128,7 +128,15 @@ namespace Ryujinx.Graphics.Shader.Translation
}
else if (value >= AttributeConsts.FragmentOutputColorBase && value < AttributeConsts.FragmentOutputColorEnd)
{
return new AttributeInfo(value & ~0xf, (value >> 2) & 3, 4, AggregateType.Vector4 | AggregateType.FP32, false);
int location = (value - AttributeConsts.FragmentOutputColorBase) / 16;
var elemType = config.GpuAccessor.QueryFragmentOutputType(location) switch
{
AttributeType.Sint => AggregateType.S32,
AttributeType.Uint => AggregateType.U32,
_ => AggregateType.FP32
};
return new AttributeInfo(value & ~0xf, (value >> 2) & 3, 4, AggregateType.Vector4 | elemType, false);
}
else if (value == AttributeConsts.SupportBlockViewInverseX || value == AttributeConsts.SupportBlockViewInverseY)
{

View file

@ -140,6 +140,25 @@ namespace Ryujinx.Graphics.Vulkan
return _attachments[index];
}
public ComponentType GetAttachmentComponentType(int index)
{
if (_colors != null && (uint)index < _colors.Length)
{
var format = _colors[index].Info.Format;
if (format.IsSint())
{
return ComponentType.SignedInteger;
}
else if (format.IsUint())
{
return ComponentType.UnsignedInteger;
}
}
return ComponentType.Float;
}
public bool IsValidColorAttachment(int bindIndex)
{
return (uint)bindIndex < Constants.MaxRenderTargets && (_validColorAttachments & (1u << bindIndex)) != 0;

View file

@ -1,7 +1,20 @@
using Silk.NET.Vulkan;
using System;
namespace Ryujinx.Graphics.Vulkan
{
[Flags]
enum PortabilitySubsetFlags
{
None = 0,
VertexBufferAlignment4B = 1,
NoTriangleFans = 1 << 1,
NoPointMode = 1 << 2,
No3DImageView = 1 << 3,
NoLodBias = 1 << 4
}
readonly struct HardwareCapabilities
{
public readonly bool SupportsIndexTypeUint8;
@ -23,6 +36,7 @@ namespace Ryujinx.Graphics.Vulkan
public readonly uint MaxSubgroupSize;
public readonly ShaderStageFlags RequiredSubgroupSizeStages;
public readonly SampleCountFlags SupportedSampleCounts;
public readonly PortabilitySubsetFlags PortabilitySubset;
public HardwareCapabilities(
bool supportsIndexTypeUint8,
@ -43,7 +57,8 @@ namespace Ryujinx.Graphics.Vulkan
uint minSubgroupSize,
uint maxSubgroupSize,
ShaderStageFlags requiredSubgroupSizeStages,
SampleCountFlags supportedSampleCounts)
SampleCountFlags supportedSampleCounts,
PortabilitySubsetFlags portabilitySubset)
{
SupportsIndexTypeUint8 = supportsIndexTypeUint8;
SupportsCustomBorderColor = supportsCustomBorderColor;
@ -64,6 +79,7 @@ namespace Ryujinx.Graphics.Vulkan
MaxSubgroupSize = maxSubgroupSize;
RequiredSubgroupSizeStages = requiredSubgroupSizeStages;
SupportedSampleCounts = supportedSampleCounts;
PortabilitySubset = portabilitySubset;
}
}
}

View file

@ -9,6 +9,13 @@ using VkFormat = Silk.NET.Vulkan.Format;
namespace Ryujinx.Graphics.Vulkan
{
enum ComponentType
{
Float,
SignedInteger,
UnsignedInteger
}
class HelperShader : IDisposable
{
private const int UniformBufferAlignment = 256;
@ -18,7 +25,9 @@ namespace Ryujinx.Graphics.Vulkan
private readonly ISampler _samplerNearest;
private readonly IProgram _programColorBlit;
private readonly IProgram _programColorBlitClearAlpha;
private readonly IProgram _programColorClear;
private readonly IProgram _programColorClearF;
private readonly IProgram _programColorClearSI;
private readonly IProgram _programColorClearUI;
private readonly IProgram _programStrideChange;
private readonly IProgram _programConvertIndexBuffer;
private readonly IProgram _programConvertIndirectData;
@ -63,10 +72,22 @@ namespace Ryujinx.Graphics.Vulkan
Array.Empty<int>(),
Array.Empty<int>());
_programColorClear = gd.CreateProgramWithMinimalLayout(new[]
_programColorClearF = gd.CreateProgramWithMinimalLayout(new[]
{
new ShaderSource(ShaderBinaries.ColorClearVertexShaderSource, colorBlitVertexBindings, ShaderStage.Vertex, TargetLanguage.Spirv),
new ShaderSource(ShaderBinaries.ColorClearFragmentShaderSource, colorClearFragmentBindings, ShaderStage.Fragment, TargetLanguage.Spirv),
new ShaderSource(ShaderBinaries.ColorClearFFragmentShaderSource, colorClearFragmentBindings, ShaderStage.Fragment, TargetLanguage.Spirv),
});
_programColorClearSI = gd.CreateProgramWithMinimalLayout(new[]
{
new ShaderSource(ShaderBinaries.ColorClearVertexShaderSource, colorBlitVertexBindings, ShaderStage.Vertex, TargetLanguage.Spirv),
new ShaderSource(ShaderBinaries.ColorClearSIFragmentShaderSource, colorClearFragmentBindings, ShaderStage.Fragment, TargetLanguage.Spirv),
});
_programColorClearUI = gd.CreateProgramWithMinimalLayout(new[]
{
new ShaderSource(ShaderBinaries.ColorClearVertexShaderSource, colorBlitVertexBindings, ShaderStage.Vertex, TargetLanguage.Spirv),
new ShaderSource(ShaderBinaries.ColorClearUIFragmentShaderSource, colorClearFragmentBindings, ShaderStage.Fragment, TargetLanguage.Spirv),
});
var strideChangeBindings = new ShaderBindings(
@ -242,6 +263,7 @@ namespace Ryujinx.Graphics.Vulkan
int dstWidth,
int dstHeight,
VkFormat dstFormat,
ComponentType type,
Rectangle<int> scissor)
{
const int ClearColorBufferSize = 16;
@ -273,7 +295,22 @@ namespace Ryujinx.Graphics.Vulkan
scissors[0] = scissor;
_pipeline.SetProgram(_programColorClear);
IProgram program;
if (type == ComponentType.SignedInteger)
{
program = _programColorClearSI;
}
else if (type == ComponentType.UnsignedInteger)
{
program = _programColorClearUI;
}
else
{
program = _programColorClearF;
}
_pipeline.SetProgram(program);
_pipeline.SetRenderTarget(dst, (uint)dstWidth, (uint)dstHeight, false, dstFormat);
_pipeline.SetRenderTargetColorMasks(new uint[] { componentMask });
_pipeline.SetViewports(viewports, false);
@ -948,7 +985,9 @@ namespace Ryujinx.Graphics.Vulkan
{
_programColorBlitClearAlpha.Dispose();
_programColorBlit.Dispose();
_programColorClear.Dispose();
_programColorClearF.Dispose();
_programColorClearSI.Dispose();
_programColorClearUI.Dispose();
_programStrideChange.Dispose();
_programConvertIndexBuffer.Dispose();
_programConvertIndirectData.Dispose();

View file

@ -0,0 +1,104 @@
using System;
using System.Runtime.InteropServices;
namespace Ryujinx.Graphics.Vulkan.MoltenVK
{
enum MVKConfigLogLevel : int
{
None = 0,
Error = 1,
Warning = 2,
Info = 3,
Debug = 4
}
enum MVKConfigTraceVulkanCalls : int
{
None = 0,
Enter = 1,
EnterExit = 2,
Duration = 3
}
enum MVKConfigAutoGPUCaptureScope : int
{
None = 0,
Device = 1,
Frame = 2
}
[Flags]
enum MVKConfigAdvertiseExtensions : int
{
All = 0x00000001,
MoltenVK = 0x00000002,
WSI = 0x00000004,
Portability = 0x00000008
}
enum MVKVkSemaphoreSupportStyle : int
{
MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_SINGLE_QUEUE = 0,
MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_METAL_EVENTS_WHERE_SAFE = 1,
MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_METAL_EVENTS = 2,
MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_CALLBACK = 3,
MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_MAX_ENUM = 0x7FFFFFFF
}
readonly struct Bool32
{
uint Value { get; }
public Bool32(uint value)
{
Value = value;
}
public Bool32(bool value)
{
Value = value ? 1u : 0u;
}
public static implicit operator bool(Bool32 val) => val.Value == 1;
public static implicit operator Bool32(bool val) => new Bool32(val);
}
[StructLayout(LayoutKind.Sequential)]
struct MVKConfiguration
{
public Bool32 DebugMode;
public Bool32 ShaderConversionFlipVertexY;
public Bool32 SynchronousQueueSubmits;
public Bool32 PrefillMetalCommandBuffers;
public uint MaxActiveMetalCommandBuffersPerQueue;
public Bool32 SupportLargeQueryPools;
public Bool32 PresentWithCommandBuffer;
public Bool32 SwapchainMagFilterUseNearest;
public ulong MetalCompileTimeout;
public Bool32 PerformanceTracking;
public uint PerformanceLoggingFrameCount;
public Bool32 DisplayWatermark;
public Bool32 SpecializedQueueFamilies;
public Bool32 SwitchSystemGPU;
public Bool32 FullImageViewSwizzle;
public uint DefaultGPUCaptureScopeQueueFamilyIndex;
public uint DefaultGPUCaptureScopeQueueIndex;
public Bool32 FastMathEnabled;
public MVKConfigLogLevel LogLevel;
public MVKConfigTraceVulkanCalls TraceVulkanCalls;
public Bool32 ForceLowPowerGPU;
public Bool32 SemaphoreUseMTLFence;
public MVKVkSemaphoreSupportStyle SemaphoreSupportStyle;
public MVKConfigAutoGPUCaptureScope AutoGPUCaptureScope;
public IntPtr AutoGPUCaptureOutputFilepath;
public Bool32 Texture1DAs2D;
public Bool32 PreallocateDescriptors;
public Bool32 UseCommandPooling;
public Bool32 UseMTLHeap;
public Bool32 LogActivityPerformanceInline;
public uint ApiVersionToAdvertise;
public MVKConfigAdvertiseExtensions AdvertiseExtensions;
public Bool32 ResumeLostDevice;
public Bool32 UseMetalArgumentBuffers;
}
}

View file

@ -0,0 +1,31 @@
using Silk.NET.Vulkan;
using System;
using System.Runtime.Versioning;
using System.Runtime.InteropServices;
namespace Ryujinx.Graphics.Vulkan.MoltenVK
{
[SupportedOSPlatform("macos")]
public static partial class MVKInitialization
{
[LibraryImport("libMoltenVK.dylib")]
private static partial Result vkGetMoltenVKConfigurationMVK(IntPtr unusedInstance, out MVKConfiguration config, in IntPtr configSize);
[LibraryImport("libMoltenVK.dylib")]
private static partial Result vkSetMoltenVKConfigurationMVK(IntPtr unusedInstance, in MVKConfiguration config, in IntPtr configSize);
public static void Initialize()
{
var configSize = (IntPtr)Marshal.SizeOf<MVKConfiguration>();
vkGetMoltenVKConfigurationMVK(IntPtr.Zero, out MVKConfiguration config, configSize);
config.UseMetalArgumentBuffers = true;
config.SemaphoreSupportStyle = MVKVkSemaphoreSupportStyle.MVK_CONFIG_VK_SEMAPHORE_SUPPORT_STYLE_SINGLE_QUEUE;
config.SynchronousQueueSubmits = false;
vkSetMoltenVKConfigurationMVK(IntPtr.Zero, config, configSize);
}
}
}

View file

@ -2,6 +2,7 @@
using Ryujinx.Graphics.Shader;
using Silk.NET.Vulkan;
using System;
using System.Linq;
using System.Numerics;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
@ -50,6 +51,11 @@ namespace Ryujinx.Graphics.Vulkan
private Auto<DisposableRenderPass> _renderPass;
private int _writtenAttachmentCount;
private bool _framebufferUsingColorWriteMask;
private ITexture[] _preMaskColors;
private ITexture _preMaskDepthStencil;
private readonly DescriptorSetUpdater _descriptorSetUpdater;
private IndexBufferState _indexBuffer;
@ -905,22 +911,35 @@ namespace Ryujinx.Graphics.Vulkan
}
}
SignalStateChange();
if (writtenAttachments != _writtenAttachmentCount)
if (_framebufferUsingColorWriteMask)
{
SignalAttachmentChange();
_writtenAttachmentCount = writtenAttachments;
SetRenderTargetsInternal(_preMaskColors, _preMaskDepthStencil, true);
}
else
{
SignalStateChange();
if (writtenAttachments != _writtenAttachmentCount)
{
SignalAttachmentChange();
_writtenAttachmentCount = writtenAttachments;
}
}
}
private void SetRenderTargetsInternal(ITexture[] colors, ITexture depthStencil, bool filterWriteMasked)
{
FramebufferParams?.UpdateModifications();
CreateFramebuffer(colors, depthStencil, filterWriteMasked);
CreateRenderPass();
SignalStateChange();
SignalAttachmentChange();
}
public void SetRenderTargets(ITexture[] colors, ITexture depthStencil)
{
FramebufferParams?.UpdateModifications();
CreateFramebuffer(colors, depthStencil);
CreateRenderPass();
SignalStateChange();
SignalAttachmentChange();
_framebufferUsingColorWriteMask = false;
SetRenderTargetsInternal(colors, depthStencil, Gd.IsTBDR);
}
public void SetRenderTargetScale(float scale)
@ -1102,7 +1121,7 @@ namespace Ryujinx.Graphics.Vulkan
int vbSize = vertexBuffer.Buffer.Size;
if (Gd.Vendor == Vendor.Amd && vertexBuffer.Stride > 0)
if (Gd.Vendor == Vendor.Amd && !Gd.IsMoltenVk && vertexBuffer.Stride > 0)
{
// AMD has a bug where if offset + stride * count is greater than
// the size, then the last attribute will have the wrong value.
@ -1119,7 +1138,8 @@ namespace Ryujinx.Graphics.Vulkan
buffer.Dispose();
if ((vertexBuffer.Stride % FormatExtensions.MaxBufferFormatScalarSize) == 0)
if (!Gd.Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.VertexBufferAlignment4B) &&
(vertexBuffer.Stride % FormatExtensions.MaxBufferFormatScalarSize) == 0)
{
buffer = new VertexBufferState(
vb,
@ -1259,8 +1279,62 @@ namespace Ryujinx.Graphics.Vulkan
_currentPipelineHandle = 0;
}
private void CreateFramebuffer(ITexture[] colors, ITexture depthStencil)
private void CreateFramebuffer(ITexture[] colors, ITexture depthStencil, bool filterWriteMasked)
{
if (filterWriteMasked)
{
// TBDR GPUs don't work properly if the same attachment is bound to multiple targets,
// due to each attachment being a copy of the real attachment, rather than a direct write.
// Just try to remove duplicate attachments.
// Save a copy of the array to rebind when mask changes.
void maskOut()
{
if (!_framebufferUsingColorWriteMask)
{
_preMaskColors = colors.ToArray();
_preMaskDepthStencil = depthStencil;
}
// If true, then the framebuffer must be recreated when the mask changes.
_framebufferUsingColorWriteMask = true;
}
// Look for textures that are masked out.
for (int i = 0; i < colors.Length; i++)
{
if (colors[i] == null)
{
continue;
}
ref var vkBlend = ref _newState.Internal.ColorBlendAttachmentState[i];
for (int j = 0; j < i; j++)
{
// Check each binding for a duplicate binding before it.
if (colors[i] == colors[j])
{
// Prefer the binding with no write mask.
ref var vkBlend2 = ref _newState.Internal.ColorBlendAttachmentState[j];
if (vkBlend.ColorWriteMask == 0)
{
colors[i] = null;
maskOut();
}
else if (vkBlend2.ColorWriteMask == 0)
{
colors[j] = null;
maskOut();
}
}
}
}
}
FramebufferParams = new FramebufferParams(Device, colors, depthStencil);
UpdatePipelineAttachmentFormats();
}

View file

@ -79,6 +79,7 @@ namespace Ryujinx.Graphics.Vulkan
(int)FramebufferParams.Width,
(int)FramebufferParams.Height,
FramebufferParams.AttachmentFormats[index],
FramebufferParams.GetAttachmentComponentType(index),
ClearScissor);
}
else

View file

@ -0,0 +1,9 @@
#version 450 core
layout (location = 0) in vec4 clear_colour;
layout (location = 0) out ivec4 colour;
void main()
{
colour = floatBitsToInt(clear_colour);
}

View file

@ -0,0 +1,9 @@
#version 450 core
layout (location = 0) in vec4 clear_colour;
layout (location = 0) out uvec4 colour;
void main()
{
colour = floatBitsToUint(clear_colour);
}

View file

@ -431,7 +431,7 @@ namespace Ryujinx.Graphics.Vulkan.Shaders
0x3C, 0x00, 0x00, 0x00, 0xFD, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
};
public static readonly byte[] ColorClearFragmentShaderSource = new byte[]
public static readonly byte[] ColorClearFFragmentShaderSource = new byte[]
{
0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x0A, 0x00, 0x08, 0x00, 0x0D, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0B, 0x00, 0x06, 0x00,
@ -459,6 +459,68 @@ namespace Ryujinx.Graphics.Vulkan.Shaders
0x0C, 0x00, 0x00, 0x00, 0xFD, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
};
public static readonly byte[] ColorClearSIFragmentShaderSource = new byte[]
{
0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x0A, 0x00, 0x08, 0x00, 0x10, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0B, 0x00, 0x06, 0x00,
0x01, 0x00, 0x00, 0x00, 0x47, 0x4C, 0x53, 0x4C, 0x2E, 0x73, 0x74, 0x64, 0x2E, 0x34, 0x35, 0x30,
0x00, 0x00, 0x00, 0x00, 0x0E, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x07, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6D, 0x61, 0x69, 0x6E,
0x00, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x10, 0x00, 0x03, 0x00,
0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x03, 0x00, 0x02, 0x00, 0x00, 0x00,
0xC2, 0x01, 0x00, 0x00, 0x05, 0x00, 0x04, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6D, 0x61, 0x69, 0x6E,
0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x63, 0x6F, 0x6C, 0x6F,
0x75, 0x72, 0x00, 0x00, 0x05, 0x00, 0x06, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x63, 0x6C, 0x65, 0x61,
0x72, 0x5F, 0x63, 0x6F, 0x6C, 0x6F, 0x75, 0x72, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00,
0x09, 0x00, 0x00, 0x00, 0x1E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00,
0x0D, 0x00, 0x00, 0x00, 0x1E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00,
0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
0x20, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
0x3B, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
0x16, 0x00, 0x03, 0x00, 0x0A, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00,
0x0B, 0x00, 0x00, 0x00, 0x0A, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
0x0C, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0B, 0x00, 0x00, 0x00, 0x3B, 0x00, 0x04, 0x00,
0x0C, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
0xF8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3D, 0x00, 0x04, 0x00, 0x0B, 0x00, 0x00, 0x00,
0x0E, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x7C, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x00, 0x00, 0x0E, 0x00, 0x00, 0x00, 0x3E, 0x00, 0x03, 0x00, 0x09, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x00, 0x00, 0xFD, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
};
public static readonly byte[] ColorClearUIFragmentShaderSource = new byte[]
{
0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x0A, 0x00, 0x08, 0x00, 0x10, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0B, 0x00, 0x06, 0x00,
0x01, 0x00, 0x00, 0x00, 0x47, 0x4C, 0x53, 0x4C, 0x2E, 0x73, 0x74, 0x64, 0x2E, 0x34, 0x35, 0x30,
0x00, 0x00, 0x00, 0x00, 0x0E, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x07, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6D, 0x61, 0x69, 0x6E,
0x00, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x10, 0x00, 0x03, 0x00,
0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x03, 0x00, 0x03, 0x00, 0x02, 0x00, 0x00, 0x00,
0xC2, 0x01, 0x00, 0x00, 0x05, 0x00, 0x04, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6D, 0x61, 0x69, 0x6E,
0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x63, 0x6F, 0x6C, 0x6F,
0x75, 0x72, 0x00, 0x00, 0x05, 0x00, 0x06, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x63, 0x6C, 0x65, 0x61,
0x72, 0x5F, 0x63, 0x6F, 0x6C, 0x6F, 0x75, 0x72, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00,
0x09, 0x00, 0x00, 0x00, 0x1E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00,
0x0D, 0x00, 0x00, 0x00, 0x1E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00,
0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
0x20, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
0x3B, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
0x16, 0x00, 0x03, 0x00, 0x0A, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00,
0x0B, 0x00, 0x00, 0x00, 0x0A, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
0x0C, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0B, 0x00, 0x00, 0x00, 0x3B, 0x00, 0x04, 0x00,
0x0C, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
0xF8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3D, 0x00, 0x04, 0x00, 0x0B, 0x00, 0x00, 0x00,
0x0E, 0x00, 0x00, 0x00, 0x0D, 0x00, 0x00, 0x00, 0x7C, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x00, 0x00, 0x0E, 0x00, 0x00, 0x00, 0x3E, 0x00, 0x03, 0x00, 0x09, 0x00, 0x00, 0x00,
0x0F, 0x00, 0x00, 0x00, 0xFD, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
};
public static readonly byte[] ColorClearVertexShaderSource = new byte[]
{
0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x0A, 0x00, 0x08, 0x00, 0x36, 0x00, 0x00, 0x00,

View file

@ -106,7 +106,7 @@ namespace Ryujinx.Graphics.Vulkan
flags |= ImageCreateFlags.CreateCubeCompatibleBit;
}
if (type == ImageType.Type3D)
if (type == ImageType.Type3D && !gd.Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.No3DImageView))
{
flags |= ImageCreateFlags.Create2DArrayCompatibleBit;
}

View file

@ -94,8 +94,14 @@ namespace Ryujinx.Graphics.Vulkan
var subresourceRange = new ImageSubresourceRange(aspectFlags, (uint)firstLevel, levels, (uint)firstLayer, layers);
var subresourceRangeDepth = new ImageSubresourceRange(aspectFlagsDepth, (uint)firstLevel, levels, (uint)firstLayer, layers);
unsafe Auto<DisposableImageView> CreateImageView(ComponentMapping cm, ImageSubresourceRange sr, ImageViewType viewType)
unsafe Auto<DisposableImageView> CreateImageView(ComponentMapping cm, ImageSubresourceRange sr, ImageViewType viewType, ImageUsageFlags usageFlags = 0)
{
var usage = new ImageViewUsageCreateInfo()
{
SType = StructureType.ImageViewUsageCreateInfo,
Usage = usageFlags
};
var imageCreateInfo = new ImageViewCreateInfo()
{
SType = StructureType.ImageViewCreateInfo,
@ -103,7 +109,8 @@ namespace Ryujinx.Graphics.Vulkan
ViewType = viewType,
Format = format,
Components = cm,
SubresourceRange = sr
SubresourceRange = sr,
PNext = usageFlags == 0 ? null : &usage
};
gd.Api.CreateImageView(device, imageCreateInfo, null, out var imageView).ThrowOnError();
@ -124,9 +131,21 @@ namespace Ryujinx.Graphics.Vulkan
// Framebuffer attachments also require 3D textures to be bound as 2D array.
if (info.Target == Target.Texture3D)
{
subresourceRange = new ImageSubresourceRange(aspectFlags, (uint)firstLevel, levels, (uint)firstLayer, (uint)info.Depth);
if (gd.Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.No3DImageView))
{
if (levels == 1 && (info.Format.IsRtColorCompatible() || info.Format.IsDepthOrStencil()))
{
subresourceRange = new ImageSubresourceRange(aspectFlags, (uint)firstLevel, levels, (uint)firstLayer, 1);
_imageView2dArray = CreateImageView(identityComponentMapping, subresourceRange, ImageViewType.Type2DArray);
_imageView2dArray = CreateImageView(identityComponentMapping, subresourceRange, ImageViewType.Type2D, ImageUsageFlags.ColorAttachmentBit);
}
}
else
{
subresourceRange = new ImageSubresourceRange(aspectFlags, (uint)firstLevel, levels, (uint)firstLayer, (uint)info.Depth);
_imageView2dArray = CreateImageView(identityComponentMapping, subresourceRange, ImageViewType.Type2DArray);
}
}
Valid = true;
@ -353,7 +372,7 @@ namespace Ryujinx.Graphics.Vulkan
}
if (VulkanConfiguration.UseSlowSafeBlitOnAmd &&
_gd.Vendor == Vendor.Amd &&
(_gd.Vendor == Vendor.Amd || _gd.IsMoltenVk) &&
src.Info.Target == Target.Texture2D &&
dst.Info.Target == Target.Texture2D &&
!dst.Info.Format.IsDepthOrStencil())

View file

@ -5,9 +5,12 @@ namespace Ryujinx.Graphics.Vulkan
enum Vendor
{
Amd,
ImgTec,
Intel,
Nvidia,
ARM,
Qualcomm,
Apple,
Unknown
}
@ -21,7 +24,10 @@ namespace Ryujinx.Graphics.Vulkan
return id switch
{
0x1002 => Vendor.Amd,
0x1010 => Vendor.ImgTec,
0x106B => Vendor.Apple,
0x10DE => Vendor.Nvidia,
0x13B5 => Vendor.ARM,
0x8086 => Vendor.Intel,
0x5143 => Vendor.Qualcomm,
_ => Vendor.Unknown
@ -34,6 +40,7 @@ namespace Ryujinx.Graphics.Vulkan
{
0x1002 => "AMD",
0x1010 => "ImgTec",
0x106B => "Apple",
0x10DE => "NVIDIA",
0x13B5 => "ARM",
0x1AE0 => "Google",

View file

@ -82,9 +82,9 @@ namespace Ryujinx.Graphics.Vulkan
}
_buffer = autoBuffer;
}
state.Internal.VertexBindingDescriptions[DescriptorIndex].Stride = (uint)_stride;
state.Internal.VertexBindingDescriptions[DescriptorIndex].Stride = (uint)stride;
}
return;
}

View file

@ -389,6 +389,18 @@ namespace Ryujinx.Graphics.Vulkan
features2.PNext = &featuresCustomBorderColorSupported;
}
PhysicalDeviceRobustness2FeaturesEXT supportedFeaturesRobustness2 = new PhysicalDeviceRobustness2FeaturesEXT()
{
SType = StructureType.PhysicalDeviceRobustness2FeaturesExt
};
if (supportedExtensions.Contains("VK_EXT_robustness2"))
{
supportedFeaturesRobustness2.PNext = features2.PNext;
features2.PNext = &supportedFeaturesRobustness2;
}
api.GetPhysicalDeviceFeatures2(physicalDevice, &features2);
var supportedFeatures = features2.Features;
@ -428,14 +440,17 @@ namespace Ryujinx.Graphics.Vulkan
pExtendedFeatures = &featuresTransformFeedback;
var featuresRobustness2 = new PhysicalDeviceRobustness2FeaturesEXT()
if (supportedExtensions.Contains("VK_EXT_robustness2"))
{
SType = StructureType.PhysicalDeviceRobustness2FeaturesExt,
PNext = pExtendedFeatures,
NullDescriptor = true
};
var featuresRobustness2 = new PhysicalDeviceRobustness2FeaturesEXT()
{
SType = StructureType.PhysicalDeviceRobustness2FeaturesExt,
PNext = pExtendedFeatures,
NullDescriptor = supportedFeaturesRobustness2.NullDescriptor
};
pExtendedFeatures = &featuresRobustness2;
pExtendedFeatures = &featuresRobustness2;
}
var featuresExtendedDynamicState = new PhysicalDeviceExtendedDynamicStateFeaturesEXT()
{

View file

@ -3,6 +3,7 @@ using Ryujinx.Common.Logging;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Shader;
using Ryujinx.Graphics.Shader.Translation;
using Ryujinx.Graphics.Vulkan.MoltenVK;
using Ryujinx.Graphics.Vulkan.Queries;
using Silk.NET.Vulkan;
using Silk.NET.Vulkan.Extensions.EXT;
@ -77,6 +78,8 @@ namespace Ryujinx.Graphics.Vulkan
internal bool IsAmdWindows { get; private set; }
internal bool IsIntelWindows { get; private set; }
internal bool IsAmdGcn { get; private set; }
internal bool IsMoltenVk { get; private set; }
internal bool IsTBDR { get; private set; }
public string GpuVendor { get; private set; }
public string GpuRenderer { get; private set; }
public string GpuVersion { get; private set; }
@ -93,6 +96,14 @@ namespace Ryujinx.Graphics.Vulkan
Shaders = new HashSet<ShaderCollection>();
Textures = new HashSet<ITexture>();
Samplers = new HashSet<SamplerHolder>();
if (OperatingSystem.IsMacOS())
{
MVKInitialization.Initialize();
// Any device running on MacOS is using MoltenVK, even Intel and AMD vendors.
IsMoltenVk = true;
}
}
private unsafe void LoadFeatures(string[] supportedExtensions, uint maxQueueCount, uint queueFamilyIndex)
@ -161,7 +172,10 @@ namespace Ryujinx.Graphics.Vulkan
properties2.PNext = &propertiesTransformFeedback;
}
Api.GetPhysicalDeviceProperties2(_physicalDevice, &properties2);
PhysicalDevicePortabilitySubsetPropertiesKHR propertiesPortabilitySubset = new PhysicalDevicePortabilitySubsetPropertiesKHR()
{
SType = StructureType.PhysicalDevicePortabilitySubsetPropertiesKhr
};
PhysicalDeviceFeatures2 features2 = new PhysicalDeviceFeatures2()
{
@ -183,6 +197,11 @@ namespace Ryujinx.Graphics.Vulkan
SType = StructureType.PhysicalDeviceCustomBorderColorFeaturesExt
};
PhysicalDevicePortabilitySubsetFeaturesKHR featuresPortabilitySubset = new PhysicalDevicePortabilitySubsetFeaturesKHR()
{
SType = StructureType.PhysicalDevicePortabilitySubsetFeaturesKhr
};
if (supportedExtensions.Contains("VK_EXT_robustness2"))
{
features2.PNext = &featuresRobustness2;
@ -200,8 +219,31 @@ namespace Ryujinx.Graphics.Vulkan
features2.PNext = &featuresCustomBorderColor;
}
bool usePortability = supportedExtensions.Contains("VK_KHR_portability_subset");
if (usePortability)
{
propertiesPortabilitySubset.PNext = properties2.PNext;
properties2.PNext = &propertiesPortabilitySubset;
featuresPortabilitySubset.PNext = features2.PNext;
features2.PNext = &featuresPortabilitySubset;
}
Api.GetPhysicalDeviceProperties2(_physicalDevice, &properties2);
Api.GetPhysicalDeviceFeatures2(_physicalDevice, &features2);
var portabilityFlags = PortabilitySubsetFlags.None;
if (usePortability)
{
portabilityFlags |= propertiesPortabilitySubset.MinVertexInputBindingStrideAlignment > 1 ? PortabilitySubsetFlags.VertexBufferAlignment4B : 0;
portabilityFlags |= featuresPortabilitySubset.TriangleFans ? 0 : PortabilitySubsetFlags.NoTriangleFans;
portabilityFlags |= featuresPortabilitySubset.PointPolygons ? 0 : PortabilitySubsetFlags.NoPointMode;
portabilityFlags |= featuresPortabilitySubset.ImageView2DOn3DImage ? 0 : PortabilitySubsetFlags.No3DImageView;
portabilityFlags |= featuresPortabilitySubset.SamplerMipLodBias ? 0 : PortabilitySubsetFlags.NoLodBias;
}
bool customBorderColorSupported = supportedExtensions.Contains("VK_EXT_custom_border_color") &&
featuresCustomBorderColor.CustomBorderColors &&
featuresCustomBorderColor.CustomBorderColorWithoutFormat;
@ -224,7 +266,7 @@ namespace Ryujinx.Graphics.Vulkan
supportedExtensions.Contains(ExtConditionalRendering.ExtensionName),
supportedExtensions.Contains(ExtExtendedDynamicState.ExtensionName),
features2.Features.MultiViewport,
featuresRobustness2.NullDescriptor,
featuresRobustness2.NullDescriptor || IsMoltenVk,
supportedExtensions.Contains(KhrPushDescriptor.ExtensionName),
supportsTransformFeedback,
propertiesTransformFeedback.TransformFeedbackQueries,
@ -232,7 +274,8 @@ namespace Ryujinx.Graphics.Vulkan
propertiesSubgroupSizeControl.MinSubgroupSize,
propertiesSubgroupSizeControl.MaxSubgroupSize,
propertiesSubgroupSizeControl.RequiredSubgroupSizeStages,
supportedSampleCounts);
supportedSampleCounts,
portabilityFlags);
MemoryAllocator = new MemoryAllocator(Api, _device, properties.Limits.MaxMemoryAllocationCount);
@ -413,6 +456,36 @@ namespace Ryujinx.Graphics.Vulkan
bool supportsR4G4B4A4Format = FormatCapabilities.OptimalFormatsSupport(compressedFormatFeatureFlags,
GAL.Format.R4G4B4A4Unorm);
bool supportsAstcFormats = FormatCapabilities.OptimalFormatsSupport(compressedFormatFeatureFlags,
GAL.Format.Astc4x4Unorm,
GAL.Format.Astc5x4Unorm,
GAL.Format.Astc5x5Unorm,
GAL.Format.Astc6x5Unorm,
GAL.Format.Astc6x6Unorm,
GAL.Format.Astc8x5Unorm,
GAL.Format.Astc8x6Unorm,
GAL.Format.Astc8x8Unorm,
GAL.Format.Astc10x5Unorm,
GAL.Format.Astc10x6Unorm,
GAL.Format.Astc10x8Unorm,
GAL.Format.Astc10x10Unorm,
GAL.Format.Astc12x10Unorm,
GAL.Format.Astc12x12Unorm,
GAL.Format.Astc4x4Srgb,
GAL.Format.Astc5x4Srgb,
GAL.Format.Astc5x5Srgb,
GAL.Format.Astc6x5Srgb,
GAL.Format.Astc6x6Srgb,
GAL.Format.Astc8x5Srgb,
GAL.Format.Astc8x6Srgb,
GAL.Format.Astc8x8Srgb,
GAL.Format.Astc10x5Srgb,
GAL.Format.Astc10x6Srgb,
GAL.Format.Astc10x8Srgb,
GAL.Format.Astc10x10Srgb,
GAL.Format.Astc12x10Srgb,
GAL.Format.Astc12x12Srgb);
PhysicalDeviceVulkan12Features featuresVk12 = new PhysicalDeviceVulkan12Features()
{
SType = StructureType.PhysicalDeviceVulkan12Features
@ -434,7 +507,9 @@ namespace Ryujinx.Graphics.Vulkan
GpuVendor,
hasFrontFacingBug: IsIntelWindows,
hasVectorIndexingBug: Vendor == Vendor.Qualcomm,
supportsAstcCompression: features2.Features.TextureCompressionAstcLdr,
needsFragmentOutputSpecialization: IsMoltenVk,
reduceShaderPrecision: IsMoltenVk,
supportsAstcCompression: features2.Features.TextureCompressionAstcLdr && supportsAstcFormats,
supportsBc123Compression: supportsBc123CompressionFormat,
supportsBc45Compression: supportsBc45CompressionFormat,
supportsBc67Compression: supportsBc67CompressionFormat,
@ -515,12 +590,13 @@ namespace Ryujinx.Graphics.Vulkan
IsAmdWindows = Vendor == Vendor.Amd && RuntimeInformation.IsOSPlatform(OSPlatform.Windows);
IsIntelWindows = Vendor == Vendor.Intel && RuntimeInformation.IsOSPlatform(OSPlatform.Windows);
IsTBDR = IsMoltenVk || Vendor == Vendor.Qualcomm || Vendor == Vendor.ARM || Vendor == Vendor.ImgTec;
GpuVendor = vendorName;
GpuRenderer = Marshal.PtrToStringAnsi((IntPtr)properties.DeviceName);
GpuVersion = $"Vulkan v{ParseStandardVulkanVersion(properties.ApiVersion)}, Driver v{ParseDriverVersion(ref properties)}";
IsAmdGcn = Vendor == Vendor.Amd && VendorUtils.AmdGcnRegex().IsMatch(GpuRenderer);
IsAmdGcn = !IsMoltenVk && Vendor == Vendor.Amd && VendorUtils.AmdGcnRegex().IsMatch(GpuRenderer);
Logger.Notice.Print(LogClass.Gpu, $"{GpuVendor} {GpuRenderer} ({GpuVersion})");
}
@ -531,6 +607,7 @@ namespace Ryujinx.Graphics.Vulkan
{
GAL.PrimitiveTopology.Quads => GAL.PrimitiveTopology.Triangles,
GAL.PrimitiveTopology.QuadStrip => GAL.PrimitiveTopology.TriangleStrip,
GAL.PrimitiveTopology.TriangleFan => Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.NoTriangleFans) ? GAL.PrimitiveTopology.Triangles : topology,
_ => topology
};
}
@ -540,6 +617,7 @@ namespace Ryujinx.Graphics.Vulkan
return topology switch
{
GAL.PrimitiveTopology.Quads => true,
GAL.PrimitiveTopology.TriangleFan => Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.NoTriangleFans),
_ => false
};
}
@ -553,7 +631,13 @@ namespace Ryujinx.Graphics.Vulkan
public bool NeedsVertexBufferAlignment(int attrScalarAlignment, out int alignment)
{
if (Vendor != Vendor.Nvidia)
if (Capabilities.PortabilitySubset.HasFlag(PortabilitySubsetFlags.VertexBufferAlignment4B))
{
alignment = 4;
return true;
}
else if (Vendor != Vendor.Nvidia)
{
// Vulkan requires that vertex attributes are globally aligned by their component size,
// so buffer strides that don't divide by the largest scalar element are invalid.