Ryujinx/Ryujinx.Graphics.Gpu/Shader/ShaderCacheHashTable.cs

282 lines
11 KiB
C#
Raw Normal View History

New shader cache implementation (#3194) * New shader cache implementation * Remove some debug code * Take transform feedback varying count into account * Create shader cache directory if it does not exist + fragment output map related fixes * Remove debug code * Only check texture descriptors if the constant buffer is bound * Also check CPU VA on GetSpanMapped * Remove more unused code and move cache related code * XML docs + remove more unused methods * Better codegen for TransformFeedbackDescriptor.AsSpan * Support migration from old cache format, remove more unused code Shader cache rebuild now also rewrites the shared toc and data files * Fix migration error with BRX shaders * Add a limit to the async translation queue Avoid async translation threads not being able to keep up and the queue growing very large * Re-create specialization state on recompile This might be required if a new version of the shader translator requires more or less state, or if there is a bug related to the GPU state access * Make shader cache more error resilient * Add some missing XML docs and move GpuAccessor docs to the interface/use inheritdoc * Address early PR feedback * Fix rebase * Remove IRenderer.CompileShader and IShader interface, replace with new ShaderSource struct passed to CreateProgram directly * Handle some missing exceptions * Make shader cache purge delete both old and new shader caches * Register textures on new specialization state * Translate and compile shaders in forward order (eliminates diffs due to different binding numbers) * Limit in-flight shader compilation to the maximum number of compilation threads * Replace ParallelDiskCacheLoader state changed event with a callback function * Better handling for invalid constant buffer 1 data length * Do not create the old cache directory structure if the old cache does not exist * Constant buffer use should be per-stage. This change will invalidate existing new caches (file format version was incremented) * Replace rectangle texture with just coordinate normalization * Skip incompatible shaders that are missing texture information, instead of crashing This is required if we, for example, support new texture instruction to the shader translator, and then they allow access to textures that were not accessed before. In this scenario, the old cache entry is no longer usable * Fix coordinates normalization on cubemap textures * Check if title ID is null before combining shader cache path * More robust constant buffer address validation on spec state * More robust constant buffer address validation on spec state (2) * Regenerate shader cache with one stream, rather than one per shader. * Only create shader cache directory during initialization * Logging improvements * Proper shader program disposal * PR feedback, and add a comment on serialized structs * XML docs for RegisterTexture Co-authored-by: riperiperi <rhy3756547@hotmail.com>
2022-04-10 15:49:44 +02:00
using Ryujinx.Graphics.Gpu.Memory;
using Ryujinx.Graphics.Gpu.Shader.HashTable;
using Ryujinx.Graphics.Shader;
using System;
using System.Collections.Generic;
namespace Ryujinx.Graphics.Gpu.Shader
{
/// <summary>
/// Holds already cached code for a guest shader.
/// </summary>
struct CachedGraphicsGuestCode
{
public byte[] VertexACode;
public byte[] VertexBCode;
public byte[] TessControlCode;
public byte[] TessEvaluationCode;
public byte[] GeometryCode;
public byte[] FragmentCode;
/// <summary>
/// Gets the guest code of a shader stage by its index.
/// </summary>
/// <param name="stageIndex">Index of the shader stage</param>
/// <returns>Guest code, or null if not present</returns>
public byte[] GetByIndex(int stageIndex)
{
return stageIndex switch
{
1 => TessControlCode,
2 => TessEvaluationCode,
3 => GeometryCode,
4 => FragmentCode,
_ => VertexBCode
};
}
}
/// <summary>
/// Graphics shader cache hash table.
/// </summary>
class ShaderCacheHashTable
{
/// <summary>
/// Shader ID cache.
/// </summary>
private struct IdCache
{
private PartitionedHashTable<int> _cache;
private int _id;
/// <summary>
/// Initializes the state.
/// </summary>
public void Initialize()
{
_cache = new PartitionedHashTable<int>();
_id = 0;
}
/// <summary>
/// Adds guest code to the cache.
/// </summary>
/// <remarks>
/// If the code was already cached, it will just return the existing ID.
/// </remarks>
/// <param name="code">Code to add</param>
/// <returns>Unique ID for the guest code</returns>
public int Add(byte[] code)
{
int id = ++_id;
int cachedId = _cache.GetOrAdd(code, id);
if (cachedId != id)
{
--_id;
}
return cachedId;
}
/// <summary>
/// Tries to find cached guest code.
/// </summary>
/// <param name="dataAccessor">Code accessor used to read guest code to find a match on the hash table</param>
/// <param name="id">ID of the guest code, if found</param>
/// <param name="data">Cached guest code, if found</param>
/// <returns>True if found, false otherwise</returns>
public bool TryFind(IDataAccessor dataAccessor, out int id, out byte[] data)
{
return _cache.TryFindItem(dataAccessor, out id, out data);
}
}
/// <summary>
/// Guest code IDs of the guest shaders that when combined forms a single host program.
/// </summary>
private struct IdTable : IEquatable<IdTable>
{
public int VertexAId;
public int VertexBId;
public int TessControlId;
public int TessEvaluationId;
public int GeometryId;
public int FragmentId;
public override bool Equals(object obj)
{
return obj is IdTable other && Equals(other);
}
public bool Equals(IdTable other)
{
return other.VertexAId == VertexAId &&
other.VertexBId == VertexBId &&
other.TessControlId == TessControlId &&
other.TessEvaluationId == TessEvaluationId &&
other.GeometryId == GeometryId &&
other.FragmentId == FragmentId;
}
public override int GetHashCode()
{
return HashCode.Combine(VertexAId, VertexBId, TessControlId, TessEvaluationId, GeometryId, FragmentId);
}
}
private IdCache _vertexACache;
private IdCache _vertexBCache;
private IdCache _tessControlCache;
private IdCache _tessEvaluationCache;
private IdCache _geometryCache;
private IdCache _fragmentCache;
private readonly Dictionary<IdTable, ShaderSpecializationList> _shaderPrograms;
/// <summary>
/// Creates a new graphics shader cache hash table.
/// </summary>
public ShaderCacheHashTable()
{
_vertexACache.Initialize();
_vertexBCache.Initialize();
_tessControlCache.Initialize();
_tessEvaluationCache.Initialize();
_geometryCache.Initialize();
_fragmentCache.Initialize();
_shaderPrograms = new Dictionary<IdTable, ShaderSpecializationList>();
}
/// <summary>
/// Adds a program to the cache.
/// </summary>
/// <param name="program">Program to be added</param>
public void Add(CachedShaderProgram program)
{
IdTable idTable = new IdTable();
foreach (var shader in program.Shaders)
{
if (shader == null)
{
continue;
}
if (shader.Info != null)
{
switch (shader.Info.Stage)
{
case ShaderStage.Vertex:
idTable.VertexBId = _vertexBCache.Add(shader.Code);
break;
case ShaderStage.TessellationControl:
idTable.TessControlId = _tessControlCache.Add(shader.Code);
break;
case ShaderStage.TessellationEvaluation:
idTable.TessEvaluationId = _tessEvaluationCache.Add(shader.Code);
break;
case ShaderStage.Geometry:
idTable.GeometryId = _geometryCache.Add(shader.Code);
break;
case ShaderStage.Fragment:
idTable.FragmentId = _fragmentCache.Add(shader.Code);
break;
}
}
else
{
idTable.VertexAId = _vertexACache.Add(shader.Code);
}
}
if (!_shaderPrograms.TryGetValue(idTable, out ShaderSpecializationList specList))
{
specList = new ShaderSpecializationList();
_shaderPrograms.Add(idTable, specList);
}
specList.Add(program);
}
/// <summary>
/// Tries to find a cached program.
/// </summary>
/// <remarks>
/// Even if false is returned, <paramref name="guestCode"/> might still contain cached guest code.
/// This can be used to avoid additional allocations for guest code that was already cached.
/// </remarks>
/// <param name="channel">GPU channel</param>
/// <param name="poolState">Texture pool state</param>
/// <param name="graphicsState">Graphics state</param>
New shader cache implementation (#3194) * New shader cache implementation * Remove some debug code * Take transform feedback varying count into account * Create shader cache directory if it does not exist + fragment output map related fixes * Remove debug code * Only check texture descriptors if the constant buffer is bound * Also check CPU VA on GetSpanMapped * Remove more unused code and move cache related code * XML docs + remove more unused methods * Better codegen for TransformFeedbackDescriptor.AsSpan * Support migration from old cache format, remove more unused code Shader cache rebuild now also rewrites the shared toc and data files * Fix migration error with BRX shaders * Add a limit to the async translation queue Avoid async translation threads not being able to keep up and the queue growing very large * Re-create specialization state on recompile This might be required if a new version of the shader translator requires more or less state, or if there is a bug related to the GPU state access * Make shader cache more error resilient * Add some missing XML docs and move GpuAccessor docs to the interface/use inheritdoc * Address early PR feedback * Fix rebase * Remove IRenderer.CompileShader and IShader interface, replace with new ShaderSource struct passed to CreateProgram directly * Handle some missing exceptions * Make shader cache purge delete both old and new shader caches * Register textures on new specialization state * Translate and compile shaders in forward order (eliminates diffs due to different binding numbers) * Limit in-flight shader compilation to the maximum number of compilation threads * Replace ParallelDiskCacheLoader state changed event with a callback function * Better handling for invalid constant buffer 1 data length * Do not create the old cache directory structure if the old cache does not exist * Constant buffer use should be per-stage. This change will invalidate existing new caches (file format version was incremented) * Replace rectangle texture with just coordinate normalization * Skip incompatible shaders that are missing texture information, instead of crashing This is required if we, for example, support new texture instruction to the shader translator, and then they allow access to textures that were not accessed before. In this scenario, the old cache entry is no longer usable * Fix coordinates normalization on cubemap textures * Check if title ID is null before combining shader cache path * More robust constant buffer address validation on spec state * More robust constant buffer address validation on spec state (2) * Regenerate shader cache with one stream, rather than one per shader. * Only create shader cache directory during initialization * Logging improvements * Proper shader program disposal * PR feedback, and add a comment on serialized structs * XML docs for RegisterTexture Co-authored-by: riperiperi <rhy3756547@hotmail.com>
2022-04-10 15:49:44 +02:00
/// <param name="addresses">Guest addresses of the shaders to find</param>
/// <param name="program">Cached host program for the given state, if found</param>
/// <param name="guestCode">Cached guest code, if any found</param>
/// <returns>True if a cached host program was found, false otherwise</returns>
public bool TryFind(
GpuChannel channel,
ref GpuChannelPoolState poolState,
ref GpuChannelGraphicsState graphicsState,
New shader cache implementation (#3194) * New shader cache implementation * Remove some debug code * Take transform feedback varying count into account * Create shader cache directory if it does not exist + fragment output map related fixes * Remove debug code * Only check texture descriptors if the constant buffer is bound * Also check CPU VA on GetSpanMapped * Remove more unused code and move cache related code * XML docs + remove more unused methods * Better codegen for TransformFeedbackDescriptor.AsSpan * Support migration from old cache format, remove more unused code Shader cache rebuild now also rewrites the shared toc and data files * Fix migration error with BRX shaders * Add a limit to the async translation queue Avoid async translation threads not being able to keep up and the queue growing very large * Re-create specialization state on recompile This might be required if a new version of the shader translator requires more or less state, or if there is a bug related to the GPU state access * Make shader cache more error resilient * Add some missing XML docs and move GpuAccessor docs to the interface/use inheritdoc * Address early PR feedback * Fix rebase * Remove IRenderer.CompileShader and IShader interface, replace with new ShaderSource struct passed to CreateProgram directly * Handle some missing exceptions * Make shader cache purge delete both old and new shader caches * Register textures on new specialization state * Translate and compile shaders in forward order (eliminates diffs due to different binding numbers) * Limit in-flight shader compilation to the maximum number of compilation threads * Replace ParallelDiskCacheLoader state changed event with a callback function * Better handling for invalid constant buffer 1 data length * Do not create the old cache directory structure if the old cache does not exist * Constant buffer use should be per-stage. This change will invalidate existing new caches (file format version was incremented) * Replace rectangle texture with just coordinate normalization * Skip incompatible shaders that are missing texture information, instead of crashing This is required if we, for example, support new texture instruction to the shader translator, and then they allow access to textures that were not accessed before. In this scenario, the old cache entry is no longer usable * Fix coordinates normalization on cubemap textures * Check if title ID is null before combining shader cache path * More robust constant buffer address validation on spec state * More robust constant buffer address validation on spec state (2) * Regenerate shader cache with one stream, rather than one per shader. * Only create shader cache directory during initialization * Logging improvements * Proper shader program disposal * PR feedback, and add a comment on serialized structs * XML docs for RegisterTexture Co-authored-by: riperiperi <rhy3756547@hotmail.com>
2022-04-10 15:49:44 +02:00
ShaderAddresses addresses,
out CachedShaderProgram program,
out CachedGraphicsGuestCode guestCode)
{
var memoryManager = channel.MemoryManager;
IdTable idTable = new IdTable();
guestCode = new CachedGraphicsGuestCode();
program = null;
bool found = TryGetId(_vertexACache, memoryManager, addresses.VertexA, out idTable.VertexAId, out guestCode.VertexACode);
found &= TryGetId(_vertexBCache, memoryManager, addresses.VertexB, out idTable.VertexBId, out guestCode.VertexBCode);
found &= TryGetId(_tessControlCache, memoryManager, addresses.TessControl, out idTable.TessControlId, out guestCode.TessControlCode);
found &= TryGetId(_tessEvaluationCache, memoryManager, addresses.TessEvaluation, out idTable.TessEvaluationId, out guestCode.TessEvaluationCode);
found &= TryGetId(_geometryCache, memoryManager, addresses.Geometry, out idTable.GeometryId, out guestCode.GeometryCode);
found &= TryGetId(_fragmentCache, memoryManager, addresses.Fragment, out idTable.FragmentId, out guestCode.FragmentCode);
if (found && _shaderPrograms.TryGetValue(idTable, out ShaderSpecializationList specList))
{
return specList.TryFindForGraphics(channel, ref poolState, ref graphicsState, out program);
New shader cache implementation (#3194) * New shader cache implementation * Remove some debug code * Take transform feedback varying count into account * Create shader cache directory if it does not exist + fragment output map related fixes * Remove debug code * Only check texture descriptors if the constant buffer is bound * Also check CPU VA on GetSpanMapped * Remove more unused code and move cache related code * XML docs + remove more unused methods * Better codegen for TransformFeedbackDescriptor.AsSpan * Support migration from old cache format, remove more unused code Shader cache rebuild now also rewrites the shared toc and data files * Fix migration error with BRX shaders * Add a limit to the async translation queue Avoid async translation threads not being able to keep up and the queue growing very large * Re-create specialization state on recompile This might be required if a new version of the shader translator requires more or less state, or if there is a bug related to the GPU state access * Make shader cache more error resilient * Add some missing XML docs and move GpuAccessor docs to the interface/use inheritdoc * Address early PR feedback * Fix rebase * Remove IRenderer.CompileShader and IShader interface, replace with new ShaderSource struct passed to CreateProgram directly * Handle some missing exceptions * Make shader cache purge delete both old and new shader caches * Register textures on new specialization state * Translate and compile shaders in forward order (eliminates diffs due to different binding numbers) * Limit in-flight shader compilation to the maximum number of compilation threads * Replace ParallelDiskCacheLoader state changed event with a callback function * Better handling for invalid constant buffer 1 data length * Do not create the old cache directory structure if the old cache does not exist * Constant buffer use should be per-stage. This change will invalidate existing new caches (file format version was incremented) * Replace rectangle texture with just coordinate normalization * Skip incompatible shaders that are missing texture information, instead of crashing This is required if we, for example, support new texture instruction to the shader translator, and then they allow access to textures that were not accessed before. In this scenario, the old cache entry is no longer usable * Fix coordinates normalization on cubemap textures * Check if title ID is null before combining shader cache path * More robust constant buffer address validation on spec state * More robust constant buffer address validation on spec state (2) * Regenerate shader cache with one stream, rather than one per shader. * Only create shader cache directory during initialization * Logging improvements * Proper shader program disposal * PR feedback, and add a comment on serialized structs * XML docs for RegisterTexture Co-authored-by: riperiperi <rhy3756547@hotmail.com>
2022-04-10 15:49:44 +02:00
}
return false;
}
/// <summary>
/// Tries to get the ID of a single cached shader stage.
/// </summary>
/// <param name="idCache">ID cache of the stage</param>
/// <param name="memoryManager">GPU memory manager</param>
/// <param name="baseAddress">Base address of the shader</param>
/// <param name="id">ID, if found</param>
/// <param name="data">Cached guest code, if found</param>
/// <returns>True if a cached shader is found, false otherwise</returns>
private static bool TryGetId(IdCache idCache, MemoryManager memoryManager, ulong baseAddress, out int id, out byte[] data)
{
if (baseAddress == 0)
{
id = 0;
data = null;
return true;
}
ShaderCodeAccessor codeAccessor = new ShaderCodeAccessor(memoryManager, baseAddress);
return idCache.TryFind(codeAccessor, out id, out data);
}
/// <summary>
/// Gets all programs that have been added to the table.
/// </summary>
/// <returns>Programs added to the table</returns>
public IEnumerable<CachedShaderProgram> GetPrograms()
{
foreach (var specList in _shaderPrograms.Values)
{
foreach (var program in specList)
{
yield return program;
}
}
}
}
}