Ryujinx/Ryujinx.Graphics.Gpu/Shader/HashTable/PartitionedHashTable.cs

245 lines
8.4 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 09:49:44 -04:00
using System;
using System.Collections.Generic;
using System.Diagnostics;
namespace Ryujinx.Graphics.Gpu.Shader.HashTable
{
/// <summary>
/// Partitioned hash table.
/// </summary>
/// <typeparam name="T"></typeparam>
public class PartitionedHashTable<T>
{
/// <summary>
/// Entry for a given data size.
/// </summary>
private struct SizeEntry
{
/// <summary>
/// Size for the data that will be stored on the hash table on this entry.
/// </summary>
public int Size { get; }
/// <summary>
/// Number of entries on the hash table.
/// </summary>
public int TableCount => _table.Count;
private readonly PartitionHashTable<T> _table;
/// <summary>
/// Creates an entry for a given size.
/// </summary>
/// <param name="size">Size of the data to be stored on this entry</param>
public SizeEntry(int size)
{
Size = size;
_table = new PartitionHashTable<T>();
}
/// <summary>
/// Gets an item for existing data, or adds a new one.
/// </summary>
/// <param name="data">Data associated with the item</param>
/// <param name="dataHash">Hash of <paramref name="data"/></param>
/// <param name="item">Item to be added</param>
/// <returns>Existing item, or <paramref name="item"/> if not present</returns>
public T GetOrAdd(byte[] data, uint dataHash, T item)
{
Debug.Assert(data.Length == Size);
return _table.GetOrAdd(data, dataHash, item);
}
/// <summary>
/// Adds a new item.
/// </summary>
/// <param name="data">Data associated with the item</param>
/// <param name="dataHash">Hash of <paramref name="data"/></param>
/// <param name="item">Item to be added</param>
/// <returns>True if added, false otherwise</returns>
public bool Add(byte[] data, uint dataHash, T item)
{
Debug.Assert(data.Length == Size);
return _table.Add(data, dataHash, item);
}
/// <summary>
/// Adds a partial entry.
/// </summary>
/// <param name="ownerData">Full entry data</param>
/// <param name="dataHash">Hash of the sub-region of the data that belongs to this entry</param>
/// <returns>True if added, false otherwise</returns>
public bool AddPartial(byte[] ownerData, uint dataHash)
{
return _table.AddPartial(ownerData, dataHash, Size);
}
/// <summary>
/// Fills a new hash table with "partials" of existing full entries of higher size.
/// </summary>
/// <param name="newEntry">Entry with the new hash table</param>
public void FillPartials(SizeEntry newEntry)
{
Debug.Assert(newEntry.Size < Size);
_table.FillPartials(newEntry._table, newEntry.Size);
}
/// <summary>
/// Tries to find an item on the hash table.
/// </summary>
/// <param name="dataAccessor">Data accessor</param>
/// <param name="item">The item on the table, if found, otherwise unmodified</param>
/// <param name="data">The data on the table, if found, otherwise unmodified</param>
/// <returns>Table lookup result</returns>
public PartitionHashTable<T>.SearchResult TryFindItem(scoped ref SmartDataAccessor dataAccessor, scoped ref T item, scoped ref byte[] data)
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 09:49:44 -04:00
{
return _table.TryFindItem(ref dataAccessor, Size, ref item, ref data);
}
}
private readonly List<SizeEntry> _sizeTable;
/// <summary>
/// Creates a new partitioned hash table.
/// </summary>
public PartitionedHashTable()
{
_sizeTable = new List<SizeEntry>();
}
/// <summary>
/// Adds a new item to the table.
/// </summary>
/// <param name="data">Data</param>
/// <param name="item">Item associated with the data</param>
public void Add(byte[] data, T item)
{
GetOrAdd(data, item);
}
/// <summary>
/// Gets an existing item from the table, or adds a new one if not present.
/// </summary>
/// <param name="data">Data</param>
/// <param name="item">Item associated with the data</param>
/// <returns>Existing item, or <paramref name="item"/> if not present</returns>
public T GetOrAdd(byte[] data, T item)
{
SizeEntry sizeEntry;
int index = BinarySearch(_sizeTable, data.Length);
if (index < _sizeTable.Count && _sizeTable[index].Size == data.Length)
{
sizeEntry = _sizeTable[index];
}
else
{
if (index < _sizeTable.Count && _sizeTable[index].Size < data.Length)
{
index++;
}
sizeEntry = new SizeEntry(data.Length);
_sizeTable.Insert(index, sizeEntry);
for (int i = index + 1; i < _sizeTable.Count; i++)
{
_sizeTable[i].FillPartials(sizeEntry);
}
}
HashState hashState = new HashState();
hashState.Initialize();
for (int i = 0; i < index; i++)
{
ReadOnlySpan<byte> dataSlice = new ReadOnlySpan<byte>(data).Slice(0, _sizeTable[i].Size);
hashState.Continue(dataSlice);
_sizeTable[i].AddPartial(data, hashState.Finalize(dataSlice));
}
hashState.Continue(data);
return sizeEntry.GetOrAdd(data, hashState.Finalize(data), item);
}
/// <summary>
/// Performs binary search on a list of hash tables, each one with a fixed data size.
/// </summary>
/// <param name="entries">List of hash tables</param>
/// <param name="size">Size to search for</param>
/// <returns>Index of the hash table with the given size, or nearest one otherwise</returns>
private static int BinarySearch(List<SizeEntry> entries, int size)
{
int left = 0;
int middle = 0;
int right = entries.Count - 1;
while (left <= right)
{
middle = left + ((right - left) >> 1);
SizeEntry entry = entries[middle];
if (size == entry.Size)
{
break;
}
if (size < entry.Size)
{
right = middle - 1;
}
else
{
left = middle + 1;
}
}
return middle;
}
/// <summary>
/// Tries to find an item on the table.
/// </summary>
/// <param name="dataAccessor">Data accessor</param>
/// <param name="item">Item, if found</param>
/// <param name="data">Data, if found</param>
/// <returns>True if the item was found on the table, false otherwise</returns>
public bool TryFindItem(IDataAccessor dataAccessor, out T item, out byte[] data)
{
SmartDataAccessor sda = new SmartDataAccessor(dataAccessor);
item = default;
data = null;
int left = 0;
int right = _sizeTable.Count;
while (left != right)
{
int index = left + ((right - left) >> 1);
PartitionHashTable<T>.SearchResult result = _sizeTable[index].TryFindItem(ref sda, ref item, ref data);
if (result == PartitionHashTable<T>.SearchResult.FoundFull)
{
return true;
}
if (result == PartitionHashTable<T>.SearchResult.NotFound)
{
right = index;
}
else /* if (result == PartitionHashTable<T>.SearchResult.FoundPartial) */
{
left = index + 1;
}
}
data = null;
return false;
}
}
}