Ryujinx/src/Ryujinx.Graphics.Gpu/Image/TextureCache.cs
riperiperi 95c06de4c1
GPU: Remove swizzle undefined matching and rework depth aliasing (#4896)
* GPU: Remove swizzle undefined matching and rework depth aliasing

@gdkchan pointed out that UI textures in TOTK seemed to be setting their texture swizzle incorrectly (texture was RGB but was sampling A, swizzle for A was wrong), so I determined that SwizzleComponentMatches was the problem and set on eliminating it. This PR combines existing work to select the most recently modified texture (now used when selecting which aliased texture to use) with some additional changes to remove the swizzle check and support aliased view creation.

The original observation (#1538) was that we wanted to match depth textures for the purposes of aliasing with color textures, but they often had different swizzle from what was sampled (as it's generally the identity swizzle once rendered). At the time, I decided to allow swizzles to match if only the defined components matched, which fixed the issue in all known cases but could easily be broken by a game _expecting_ a given swizzle, such as a 1/0 value on a component.

This error case could also occur in textures that don't even depth alias, such as R11G11B10, as the rule was created to generally apply to all cases.

The solution is now to fail this exact match test, and allow the search for an R32 texture to create a swizzled view of a D32 texture (and other such cases). This allows the creation of a view that mismatches the requested format, which wasn't present before and was the reason for the swizzle matching approach.

The exact match and view creation rules now follow the same rules over what textures to select when there are multiple options (such as a "perfect" match and an "aliased" match at the same time). It now selects the most recently modified texture, which is done with a new sequence number in the GpuContext (because we don't have enough of these).

Reportedly fixes UI having weird coloured backgrounds in TOTK. This also fixes an issue in MK8D where returning from a race resulted in the character selection cubemaps being broken. May work around issues introduced by the "short texture cache" PR due to modification ordering, though they won't be truly fixed.

Should allow (#4365) to avoid copies in more cases. Need to test that.

I tested a bunch of games #1538 originally affected and they seem to be fine. This change affects all games so it would be good to get some wide testing on it.

* Address feedback 1, fix an issue

* Workaround: Do not allow copies for format alias.

These should be removed when D32<->R32 copy dependencies become legal
2023-05-11 21:30:47 -03:00

1279 lines
52 KiB
C#

using Ryujinx.Common;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu.Engine.Threed;
using Ryujinx.Graphics.Gpu.Engine.Twod;
using Ryujinx.Graphics.Gpu.Engine.Types;
using Ryujinx.Graphics.Gpu.Memory;
using Ryujinx.Graphics.Texture;
using Ryujinx.Memory.Range;
using System;
using System.Collections.Generic;
namespace Ryujinx.Graphics.Gpu.Image
{
/// <summary>
/// Texture cache.
/// </summary>
class TextureCache : IDisposable
{
private readonly struct OverlapInfo
{
public TextureViewCompatibility Compatibility { get; }
public int FirstLayer { get; }
public int FirstLevel { get; }
public OverlapInfo(TextureViewCompatibility compatibility, int firstLayer, int firstLevel)
{
Compatibility = compatibility;
FirstLayer = firstLayer;
FirstLevel = firstLevel;
}
}
private const int OverlapsBufferInitialCapacity = 10;
private const int OverlapsBufferMaxCapacity = 10000;
private readonly GpuContext _context;
private readonly PhysicalMemory _physicalMemory;
private readonly MultiRangeList<Texture> _textures;
private readonly HashSet<Texture> _partiallyMappedTextures;
private Texture[] _textureOverlaps;
private OverlapInfo[] _overlapInfo;
private readonly AutoDeleteCache _cache;
/// <summary>
/// Constructs a new instance of the texture manager.
/// </summary>
/// <param name="context">The GPU context that the texture manager belongs to</param>
/// <param name="physicalMemory">Physical memory where the textures managed by this cache are mapped</param>
public TextureCache(GpuContext context, PhysicalMemory physicalMemory)
{
_context = context;
_physicalMemory = physicalMemory;
_textures = new MultiRangeList<Texture>();
_partiallyMappedTextures = new HashSet<Texture>();
_textureOverlaps = new Texture[OverlapsBufferInitialCapacity];
_overlapInfo = new OverlapInfo[OverlapsBufferInitialCapacity];
_cache = new AutoDeleteCache();
}
/// <summary>
/// Handles marking of textures written to a memory region being (partially) remapped.
/// </summary>
/// <param name="sender">Sender object</param>
/// <param name="e">Event arguments</param>
public void MemoryUnmappedHandler(object sender, UnmapEventArgs e)
{
Texture[] overlaps = new Texture[10];
int overlapCount;
MultiRange unmapped = ((MemoryManager)sender).GetPhysicalRegions(e.Address, e.Size);
lock (_textures)
{
overlapCount = _textures.FindOverlaps(unmapped, ref overlaps);
}
if (overlapCount > 0)
{
for (int i = 0; i < overlapCount; i++)
{
overlaps[i].Unmapped(unmapped);
}
}
lock (_partiallyMappedTextures)
{
if (overlapCount > 0 || _partiallyMappedTextures.Count > 0)
{
e.AddRemapAction(() =>
{
lock (_partiallyMappedTextures)
{
if (overlapCount > 0)
{
for (int i = 0; i < overlapCount; i++)
{
_partiallyMappedTextures.Add(overlaps[i]);
}
}
// Any texture that has been unmapped at any point or is partially unmapped
// should update their pool references after the remap completes.
MultiRange unmapped = ((MemoryManager)sender).GetPhysicalRegions(e.Address, e.Size);
foreach (var texture in _partiallyMappedTextures)
{
texture.UpdatePoolMappings();
}
}
});
}
}
}
/// <summary>
/// Determines if a given texture is eligible for upscaling from its info.
/// </summary>
/// <param name="info">The texture info to check</param>
/// <param name="withUpscale">True if the user of the texture would prefer it to be upscaled immediately</param>
/// <returns>True if eligible</returns>
private static TextureScaleMode IsUpscaleCompatible(TextureInfo info, bool withUpscale)
{
if ((info.Target == Target.Texture2D || info.Target == Target.Texture2DArray || info.Target == Target.Texture2DMultisample) && !info.FormatInfo.IsCompressed)
{
return UpscaleSafeMode(info) ? (withUpscale ? TextureScaleMode.Scaled : TextureScaleMode.Eligible) : TextureScaleMode.Undesired;
}
return TextureScaleMode.Blacklisted;
}
/// <summary>
/// Determines if a given texture is "safe" for upscaling from its info.
/// Note that this is different from being compatible - this elilinates targets that would have detrimental effects when scaled.
/// </summary>
/// <param name="info">The texture info to check</param>
/// <returns>True if safe</returns>
private static bool UpscaleSafeMode(TextureInfo info)
{
// While upscaling works for all targets defined by IsUpscaleCompatible, we additionally blacklist targets here that
// may have undesirable results (upscaling blur textures) or simply waste GPU resources (upscaling texture atlas).
if (info.Levels > 3)
{
// Textures with more than 3 levels are likely to be game textures, rather than render textures.
// Small textures with full mips are likely to be removed by the next check.
return false;
}
if (info.Width < 8 || info.Height < 8)
{
// Discount textures with small dimensions.
return false;
}
int widthAlignment = (info.IsLinear ? Constants.StrideAlignment : Constants.GobAlignment) / info.FormatInfo.BytesPerPixel;
if (!(info.FormatInfo.Format.IsDepthOrStencil() || info.FormatInfo.Components == 1))
{
// Discount square textures that aren't depth-stencil like. (excludes game textures, cubemap faces, most 3D texture LUT, texture atlas)
// Detect if the texture is possibly square. Widths may be aligned, so to remove the uncertainty we align both the width and height.
bool possiblySquare = BitUtils.AlignUp(info.Width, widthAlignment) == BitUtils.AlignUp(info.Height, widthAlignment);
if (possiblySquare)
{
return false;
}
}
if (info.Height < 360)
{
int aspectWidth = (int)MathF.Ceiling((info.Height / 9f) * 16f);
int aspectMaxWidth = BitUtils.AlignUp(aspectWidth, widthAlignment);
int aspectMinWidth = BitUtils.AlignDown(aspectWidth, widthAlignment);
if (info.Width >= aspectMinWidth && info.Width <= aspectMaxWidth && info.Height < 360)
{
// Targets that are roughly 16:9 can only be rescaled if they're equal to or above 360p. (excludes blur and bloom textures)
return false;
}
}
if (info.Width == info.Height * info.Height)
{
// Possibly used for a "3D texture" drawn onto a 2D surface.
// Some games do this to generate a tone mapping LUT without rendering into 3D texture slices.
return false;
}
return true;
}
/// <summary>
/// Lifts the texture to the top of the AutoDeleteCache. This is primarily used to enforce that
/// data written to a target will be flushed to memory should the texture be deleted, but also
/// keeps rendered textures alive without a pool reference.
/// </summary>
/// <param name="texture">Texture to lift</param>
public void Lift(Texture texture)
{
_cache.Lift(texture);
}
/// <summary>
/// Attempts to update a texture's physical memory range.
/// Returns false if there is an existing texture that matches with the updated range.
/// </summary>
/// <param name="texture">Texture to update</param>
/// <param name="range">New physical memory range</param>
/// <returns>True if the mapping was updated, false otherwise</returns>
public bool UpdateMapping(Texture texture, MultiRange range)
{
// There cannot be an existing texture compatible with this mapping in the texture cache already.
int overlapCount = _textures.FindOverlaps(range, ref _textureOverlaps);
for (int i = 0; i < overlapCount; i++)
{
var other = _textureOverlaps[i];
if (texture != other &&
(texture.IsViewCompatible(other.Info, other.Range, true, other.LayerSize, _context.Capabilities, out _, out _) != TextureViewCompatibility.Incompatible ||
other.IsViewCompatible(texture.Info, texture.Range, true, texture.LayerSize, _context.Capabilities, out _, out _) != TextureViewCompatibility.Incompatible))
{
return false;
}
}
_textures.Remove(texture);
texture.ReplaceRange(range);
_textures.Add(texture);
return true;
}
/// <summary>
/// Tries to find an existing texture, or create a new one if not found.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="copyTexture">Copy texture to find or create</param>
/// <param name="offset">Offset to be added to the physical texture address</param>
/// <param name="formatInfo">Format information of the copy texture</param>
/// <param name="depthAlias">Indicates if aliasing between color and depth format should be allowed</param>
/// <param name="shouldCreate">Indicates if a new texture should be created if none is found on the cache</param>
/// <param name="preferScaling">Indicates if the texture should be scaled from the start</param>
/// <param name="sizeHint">A hint indicating the minimum used size for the texture</param>
/// <returns>The texture</returns>
public Texture FindOrCreateTexture(
MemoryManager memoryManager,
TwodTexture copyTexture,
ulong offset,
FormatInfo formatInfo,
bool depthAlias,
bool shouldCreate,
bool preferScaling,
Size sizeHint)
{
int gobBlocksInY = copyTexture.MemoryLayout.UnpackGobBlocksInY();
int gobBlocksInZ = copyTexture.MemoryLayout.UnpackGobBlocksInZ();
int width;
if (copyTexture.LinearLayout)
{
width = copyTexture.Stride / formatInfo.BytesPerPixel;
}
else
{
width = copyTexture.Width;
}
TextureInfo info = new TextureInfo(
copyTexture.Address.Pack() + offset,
GetMinimumWidthInGob(width, sizeHint.Width, formatInfo.BytesPerPixel, copyTexture.LinearLayout),
copyTexture.Height,
copyTexture.Depth,
1,
1,
1,
copyTexture.Stride,
copyTexture.LinearLayout,
gobBlocksInY,
gobBlocksInZ,
1,
Target.Texture2D,
formatInfo);
TextureSearchFlags flags = TextureSearchFlags.ForCopy;
if (depthAlias)
{
flags |= TextureSearchFlags.DepthAlias;
}
if (preferScaling)
{
flags |= TextureSearchFlags.WithUpscale;
}
if (!shouldCreate)
{
flags |= TextureSearchFlags.NoCreate;
}
Texture texture = FindOrCreateTexture(memoryManager, flags, info, 0);
texture?.SynchronizeMemory();
return texture;
}
/// <summary>
/// Tries to find an existing texture, or create a new one if not found.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="colorState">Color buffer texture to find or create</param>
/// <param name="layered">Indicates if the texture might be accessed with a non-zero layer index</param>
/// <param name="samplesInX">Number of samples in the X direction, for MSAA</param>
/// <param name="samplesInY">Number of samples in the Y direction, for MSAA</param>
/// <param name="sizeHint">A hint indicating the minimum used size for the texture</param>
/// <returns>The texture</returns>
public Texture FindOrCreateTexture(
MemoryManager memoryManager,
RtColorState colorState,
bool layered,
int samplesInX,
int samplesInY,
Size sizeHint)
{
bool isLinear = colorState.MemoryLayout.UnpackIsLinear();
int gobBlocksInY = colorState.MemoryLayout.UnpackGobBlocksInY();
int gobBlocksInZ = colorState.MemoryLayout.UnpackGobBlocksInZ();
Target target;
if (colorState.MemoryLayout.UnpackIsTarget3D())
{
target = Target.Texture3D;
}
else if ((samplesInX | samplesInY) != 1)
{
target = colorState.Depth > 1 && layered
? Target.Texture2DMultisampleArray
: Target.Texture2DMultisample;
}
else
{
target = colorState.Depth > 1 && layered
? Target.Texture2DArray
: Target.Texture2D;
}
FormatInfo formatInfo = colorState.Format.Convert();
int width, stride;
// For linear textures, the width value is actually the stride.
// We can easily get the width by dividing the stride by the bpp,
// since the stride is the total number of bytes occupied by a
// line. The stride should also meet alignment constraints however,
// so the width we get here is the aligned width.
if (isLinear)
{
width = colorState.WidthOrStride / formatInfo.BytesPerPixel;
stride = colorState.WidthOrStride;
}
else
{
width = colorState.WidthOrStride;
stride = 0;
}
TextureInfo info = new TextureInfo(
colorState.Address.Pack(),
GetMinimumWidthInGob(width, sizeHint.Width, formatInfo.BytesPerPixel, isLinear),
colorState.Height,
colorState.Depth,
1,
samplesInX,
samplesInY,
stride,
isLinear,
gobBlocksInY,
gobBlocksInZ,
1,
target,
formatInfo);
int layerSize = !isLinear ? colorState.LayerSize * 4 : 0;
Texture texture = FindOrCreateTexture(memoryManager, TextureSearchFlags.WithUpscale, info, layerSize);
texture?.SynchronizeMemory();
return texture;
}
/// <summary>
/// Tries to find an existing texture, or create a new one if not found.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="dsState">Depth-stencil buffer texture to find or create</param>
/// <param name="size">Size of the depth-stencil texture</param>
/// <param name="layered">Indicates if the texture might be accessed with a non-zero layer index</param>
/// <param name="samplesInX">Number of samples in the X direction, for MSAA</param>
/// <param name="samplesInY">Number of samples in the Y direction, for MSAA</param>
/// <param name="sizeHint">A hint indicating the minimum used size for the texture</param>
/// <returns>The texture</returns>
public Texture FindOrCreateTexture(
MemoryManager memoryManager,
RtDepthStencilState dsState,
Size3D size,
bool layered,
int samplesInX,
int samplesInY,
Size sizeHint)
{
int gobBlocksInY = dsState.MemoryLayout.UnpackGobBlocksInY();
int gobBlocksInZ = dsState.MemoryLayout.UnpackGobBlocksInZ();
Target target;
if (dsState.MemoryLayout.UnpackIsTarget3D())
{
target = Target.Texture3D;
}
else if ((samplesInX | samplesInY) != 1)
{
target = size.Depth > 1 && layered
? Target.Texture2DMultisampleArray
: Target.Texture2DMultisample;
}
else
{
target = size.Depth > 1 && layered
? Target.Texture2DArray
: Target.Texture2D;
}
FormatInfo formatInfo = dsState.Format.Convert();
TextureInfo info = new TextureInfo(
dsState.Address.Pack(),
GetMinimumWidthInGob(size.Width, sizeHint.Width, formatInfo.BytesPerPixel, false),
size.Height,
size.Depth,
1,
samplesInX,
samplesInY,
0,
false,
gobBlocksInY,
gobBlocksInZ,
1,
target,
formatInfo);
Texture texture = FindOrCreateTexture(memoryManager, TextureSearchFlags.WithUpscale, info, dsState.LayerSize * 4);
texture?.SynchronizeMemory();
return texture;
}
/// <summary>
/// For block linear textures, gets the minimum width of the texture
/// that would still have the same number of GOBs per row as the original width.
/// </summary>
/// <param name="width">The possibly aligned texture width</param>
/// <param name="minimumWidth">The minimum width that the texture may have without losing data</param>
/// <param name="bytesPerPixel">Bytes per pixel of the texture format</param>
/// <param name="isLinear">True if the texture is linear, false for block linear</param>
/// <returns>The minimum width of the texture with the same amount of GOBs per row</returns>
private static int GetMinimumWidthInGob(int width, int minimumWidth, int bytesPerPixel, bool isLinear)
{
if (isLinear || (uint)minimumWidth >= (uint)width)
{
return width;
}
// Calculate the minimum possible that would not cause data loss
// and would be still within the same GOB (aligned size would be the same).
// This is useful for render and copy operations, where we don't know the
// exact width of the texture, but it doesn't matter, as long the texture is
// at least as large as the region being rendered or copied.
int alignment = 64 / bytesPerPixel;
int widthAligned = BitUtils.AlignUp(width, alignment);
return Math.Clamp(widthAligned - alignment + 1, minimumWidth, widthAligned);
}
/// <summary>
/// Tries to find an existing texture, or create a new one if not found.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="flags">The texture search flags, defines texture comparison rules</param>
/// <param name="info">Texture information of the texture to be found or created</param>
/// <param name="layerSize">Size in bytes of a single texture layer</param>
/// <param name="range">Optional ranges of physical memory where the texture data is located</param>
/// <returns>The texture</returns>
public Texture FindOrCreateTexture(
MemoryManager memoryManager,
TextureSearchFlags flags,
TextureInfo info,
int layerSize = 0,
MultiRange? range = null)
{
bool isSamplerTexture = (flags & TextureSearchFlags.ForSampler) != 0;
TextureScaleMode scaleMode = IsUpscaleCompatible(info, (flags & TextureSearchFlags.WithUpscale) != 0);
ulong address;
if (range != null)
{
address = range.Value.GetSubRange(0).Address;
}
else
{
address = memoryManager.Translate(info.GpuAddress);
// If the start address is unmapped, let's try to find a page of memory that is mapped.
if (address == MemoryManager.PteUnmapped)
{
// Make sure that the dimensions are valid before calculating the texture size.
if (info.Width < 1 || info.Height < 1 || info.Levels < 1)
{
return null;
}
if ((info.Target == Target.Texture3D ||
info.Target == Target.Texture2DArray ||
info.Target == Target.Texture2DMultisampleArray ||
info.Target == Target.CubemapArray) && info.DepthOrLayers < 1)
{
return null;
}
ulong dataSize = (ulong)info.CalculateSizeInfo(layerSize).TotalSize;
address = memoryManager.TranslateFirstMapped(info.GpuAddress, dataSize);
}
// If address is still invalid, the texture is fully unmapped, so it has no data, just return null.
if (address == MemoryManager.PteUnmapped)
{
return null;
}
}
int sameAddressOverlapsCount;
lock (_textures)
{
// Try to find a perfect texture match, with the same address and parameters.
sameAddressOverlapsCount = _textures.FindOverlaps(address, ref _textureOverlaps);
}
Texture texture = null;
long bestSequence = 0;
for (int index = 0; index < sameAddressOverlapsCount; index++)
{
Texture overlap = _textureOverlaps[index];
TextureMatchQuality matchQuality = overlap.IsExactMatch(info, flags);
if (matchQuality != TextureMatchQuality.NoMatch)
{
// If the parameters match, we need to make sure the texture is mapped to the same memory regions.
if (range != null)
{
// If a range of memory was supplied, just check if the ranges match.
if (!overlap.Range.Equals(range.Value))
{
continue;
}
}
else
{
// If no range was supplied, we can check if the GPU virtual address match. If they do,
// we know the textures are located at the same memory region.
// If they don't, it may still be mapped to the same physical region, so we
// do a more expensive check to tell if they are mapped into the same physical regions.
// If the GPU VA for the texture has ever been unmapped, then the range must be checked regardless.
if ((overlap.Info.GpuAddress != info.GpuAddress || overlap.ChangedMapping) &&
!memoryManager.CompareRange(overlap.Range, info.GpuAddress))
{
continue;
}
}
if (texture == null || overlap.Group.ModifiedSequence - bestSequence > 0)
{
texture = overlap;
bestSequence = overlap.Group.ModifiedSequence;
}
}
}
if (texture != null)
{
texture.SynchronizeMemory();
return texture;
}
else if (flags.HasFlag(TextureSearchFlags.NoCreate))
{
return null;
}
// Calculate texture sizes, used to find all overlapping textures.
SizeInfo sizeInfo = info.CalculateSizeInfo(layerSize);
ulong size = (ulong)sizeInfo.TotalSize;
bool partiallyMapped = false;
if (range == null)
{
range = memoryManager.GetPhysicalRegions(info.GpuAddress, size);
for (int i = 0; i < range.Value.Count; i++)
{
if (range.Value.GetSubRange(i).Address == MemoryManager.PteUnmapped)
{
partiallyMapped = true;
break;
}
}
}
// Find view compatible matches.
int overlapsCount;
lock (_textures)
{
overlapsCount = _textures.FindOverlaps(range.Value, ref _textureOverlaps);
}
if (_overlapInfo.Length != _textureOverlaps.Length)
{
Array.Resize(ref _overlapInfo, _textureOverlaps.Length);
}
// =============== Find Texture View of Existing Texture ===============
int fullyCompatible = 0;
// Evaluate compatibility of overlaps, add temporary references
int preferredOverlap = -1;
for (int index = 0; index < overlapsCount; index++)
{
Texture overlap = _textureOverlaps[index];
TextureViewCompatibility overlapCompatibility = overlap.IsViewCompatible(
info,
range.Value,
isSamplerTexture,
sizeInfo.LayerSize,
_context.Capabilities,
out int firstLayer,
out int firstLevel,
flags);
if (overlapCompatibility >= TextureViewCompatibility.FormatAlias)
{
if (overlap.IsView)
{
overlapCompatibility = overlapCompatibility == TextureViewCompatibility.FormatAlias ?
TextureViewCompatibility.Incompatible :
TextureViewCompatibility.CopyOnly;
}
else
{
fullyCompatible++;
if (preferredOverlap == -1 || overlap.Group.ModifiedSequence - bestSequence > 0)
{
preferredOverlap = index;
bestSequence = overlap.Group.ModifiedSequence;
}
}
}
_overlapInfo[index] = new OverlapInfo(overlapCompatibility, firstLayer, firstLevel);
overlap.IncrementReferenceCount();
}
// Search through the overlaps to find a compatible view and establish any copy dependencies.
if (preferredOverlap != -1)
{
Texture overlap = _textureOverlaps[preferredOverlap];
OverlapInfo oInfo = _overlapInfo[preferredOverlap];
bool aliased = oInfo.Compatibility == TextureViewCompatibility.FormatAlias;
if (!isSamplerTexture)
{
// If this is not a sampler texture, the size might be different from the requested size,
// so we need to make sure the texture information has the correct size for this base texture,
// before creating the view.
info = info.CreateInfoForLevelView(overlap, oInfo.FirstLevel, aliased);
}
else if (aliased)
{
// The format must be changed to match the parent.
info = info.CreateInfoWithFormat(overlap.Info.FormatInfo);
}
texture = overlap.CreateView(info, sizeInfo, range.Value, oInfo.FirstLayer, oInfo.FirstLevel);
texture.SynchronizeMemory();
}
else
{
for (int index = 0; index < overlapsCount; index++)
{
Texture overlap = _textureOverlaps[index];
OverlapInfo oInfo = _overlapInfo[index];
if (oInfo.Compatibility == TextureViewCompatibility.CopyOnly && fullyCompatible == 0)
{
// Only copy compatible. If there's another choice for a FULLY compatible texture, choose that instead.
texture = new Texture(_context, _physicalMemory, info, sizeInfo, range.Value, scaleMode);
texture.InitializeGroup(true, true, new List<TextureIncompatibleOverlap>());
texture.InitializeData(false, false);
overlap.SynchronizeMemory();
overlap.CreateCopyDependency(texture, oInfo.FirstLayer, oInfo.FirstLevel, true);
break;
}
}
}
if (texture != null)
{
// This texture could be a view of multiple parent textures with different storages, even if it is a view.
// When a texture is created, make sure all possible dependencies to other textures are created as copies.
// (even if it could be fulfilled without a copy)
for (int index = 0; index < overlapsCount; index++)
{
Texture overlap = _textureOverlaps[index];
OverlapInfo oInfo = _overlapInfo[index];
if (oInfo.Compatibility <= TextureViewCompatibility.LayoutIncompatible || oInfo.Compatibility == TextureViewCompatibility.FormatAlias)
{
if (!overlap.IsView && texture.DataOverlaps(overlap, oInfo.Compatibility))
{
texture.Group.RegisterIncompatibleOverlap(new TextureIncompatibleOverlap(overlap.Group, oInfo.Compatibility), true);
}
}
else if (overlap.Group != texture.Group)
{
overlap.SynchronizeMemory();
overlap.CreateCopyDependency(texture, oInfo.FirstLayer, oInfo.FirstLevel, true);
}
}
texture.SynchronizeMemory();
}
// =============== Create a New Texture ===============
// No match, create a new texture.
if (texture == null)
{
texture = new Texture(_context, _physicalMemory, info, sizeInfo, range.Value, scaleMode);
// Step 1: Find textures that are view compatible with the new texture.
// Any textures that are incompatible will contain garbage data, so they should be removed where possible.
int viewCompatible = 0;
fullyCompatible = 0;
bool setData = isSamplerTexture || overlapsCount == 0 || flags.HasFlag(TextureSearchFlags.ForCopy);
bool hasLayerViews = false;
bool hasMipViews = false;
var incompatibleOverlaps = new List<TextureIncompatibleOverlap>();
for (int index = 0; index < overlapsCount; index++)
{
Texture overlap = _textureOverlaps[index];
bool overlapInCache = overlap.CacheNode != null;
TextureViewCompatibility compatibility = texture.IsViewCompatible(
overlap.Info,
overlap.Range,
exactSize: true,
overlap.LayerSize,
_context.Capabilities,
out int firstLayer,
out int firstLevel);
if (overlap.IsView && compatibility == TextureViewCompatibility.Full)
{
compatibility = TextureViewCompatibility.CopyOnly;
}
if (compatibility > TextureViewCompatibility.LayoutIncompatible)
{
_overlapInfo[viewCompatible] = new OverlapInfo(compatibility, firstLayer, firstLevel);
_textureOverlaps[index] = _textureOverlaps[viewCompatible];
_textureOverlaps[viewCompatible] = overlap;
if (compatibility == TextureViewCompatibility.Full)
{
if (viewCompatible != fullyCompatible)
{
// Swap overlaps so that the fully compatible views have priority.
_overlapInfo[viewCompatible] = _overlapInfo[fullyCompatible];
_textureOverlaps[viewCompatible] = _textureOverlaps[fullyCompatible];
_overlapInfo[fullyCompatible] = new OverlapInfo(compatibility, firstLayer, firstLevel);
_textureOverlaps[fullyCompatible] = overlap;
}
fullyCompatible++;
}
viewCompatible++;
hasLayerViews |= overlap.Info.GetSlices() < texture.Info.GetSlices();
hasMipViews |= overlap.Info.Levels < texture.Info.Levels;
}
else
{
bool dataOverlaps = texture.DataOverlaps(overlap, compatibility);
if (!overlap.IsView && dataOverlaps && !incompatibleOverlaps.Exists(incompatible => incompatible.Group == overlap.Group))
{
incompatibleOverlaps.Add(new TextureIncompatibleOverlap(overlap.Group, compatibility));
}
bool removeOverlap;
bool modified = overlap.CheckModified(false);
if (overlapInCache || !setData)
{
if (!dataOverlaps)
{
// Allow textures to overlap if their data does not actually overlap.
// This typically happens when mip level subranges of a layered texture are used. (each texture fills the gaps of the others)
continue;
}
// The overlap texture is going to contain garbage data after we draw, or is generally incompatible.
// The texture group will obtain copy dependencies for any subresources that are compatible between the two textures,
// but sometimes its data must be flushed regardless.
// If the texture was modified since its last use, then that data is probably meant to go into this texture.
// If the data has been modified by the CPU, then it also shouldn't be flushed.
bool flush = overlapInCache && !modified && overlap.AlwaysFlushOnOverlap;
setData |= modified || flush;
if (overlapInCache)
{
if (flush || overlap.HadPoolOwner || overlap.IsView)
{
_cache.Remove(overlap, flush);
}
else
{
// This texture has only ever been referenced in the AutoDeleteCache.
// Keep this texture alive with the short duration cache, as it may be used often but not sampled.
_cache.AddShortCache(overlap);
}
}
removeOverlap = modified;
}
else
{
// If an incompatible overlapping texture has been modified, then it's data is likely destined for this texture,
// and the overlapped texture will contain garbage. In this case, it should be removed to save memory.
removeOverlap = modified;
}
if (removeOverlap && overlap.Info.Target != Target.TextureBuffer)
{
overlap.RemoveFromPools(false);
}
}
}
texture.InitializeGroup(hasLayerViews, hasMipViews, incompatibleOverlaps);
// We need to synchronize before copying the old view data to the texture,
// otherwise the copied data would be overwritten by a future synchronization.
texture.InitializeData(false, setData);
texture.Group.InitializeOverlaps();
for (int index = 0; index < viewCompatible; index++)
{
Texture overlap = _textureOverlaps[index];
OverlapInfo oInfo = _overlapInfo[index];
if (overlap.Group == texture.Group)
{
// If the texture group is equal, then this texture (or its parent) is already a view.
continue;
}
// Note: If we allow different sizes for those overlaps,
// we need to make sure that the "info" has the correct size for the parent texture here.
// Since this is not allowed right now, we don't need to do it.
TextureInfo overlapInfo = overlap.Info;
if (texture.ScaleFactor != overlap.ScaleFactor)
{
// A bit tricky, our new texture may need to contain an existing texture that is upscaled, but isn't itself.
// In that case, we prefer the higher scale only if our format is render-target-like, otherwise we scale the view down before copy.
texture.PropagateScale(overlap);
}
if (oInfo.Compatibility != TextureViewCompatibility.Full)
{
// Copy only compatibility, or target texture is already a view.
overlap.SynchronizeMemory();
texture.CreateCopyDependency(overlap, oInfo.FirstLayer, oInfo.FirstLevel, false);
}
else
{
TextureCreateInfo createInfo = GetCreateInfo(overlapInfo, _context.Capabilities, overlap.ScaleFactor);
ITexture newView = texture.HostTexture.CreateView(createInfo, oInfo.FirstLayer, oInfo.FirstLevel);
overlap.SynchronizeMemory();
overlap.HostTexture.CopyTo(newView, 0, 0);
overlap.ReplaceView(texture, overlapInfo, newView, oInfo.FirstLayer, oInfo.FirstLevel);
}
}
texture.SynchronizeMemory();
}
// Sampler textures are managed by the texture pool, all other textures
// are managed by the auto delete cache.
if (!isSamplerTexture)
{
_cache.Add(texture);
}
lock (_textures)
{
_textures.Add(texture);
}
if (partiallyMapped)
{
lock (_partiallyMappedTextures)
{
_partiallyMappedTextures.Add(texture);
}
}
ShrinkOverlapsBufferIfNeeded();
for (int i = 0; i < overlapsCount; i++)
{
_textureOverlaps[i].DecrementReferenceCount();
}
return texture;
}
/// <summary>
/// Attempt to find a texture on the short duration cache.
/// </summary>
/// <param name="descriptor">The texture descriptor</param>
/// <returns>The texture if found, null otherwise</returns>
public Texture FindShortCache(in TextureDescriptor descriptor)
{
return _cache.FindShortCache(descriptor);
}
/// <summary>
/// Tries to find an existing texture matching the given buffer copy destination. If none is found, returns null.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="gpuVa">GPU virtual address of the texture</param>
/// <param name="bpp">Bytes per pixel</param>
/// <param name="stride">If <paramref name="linear"/> is true, should have the texture stride, otherwise ignored</param>
/// <param name="height">If <paramref name="linear"/> is false, should have the texture height, otherwise ignored</param>
/// <param name="xCount">Number of pixels to be copied per line</param>
/// <param name="yCount">Number of lines to be copied</param>
/// <param name="linear">True if the texture has a linear layout, false otherwise</param>
/// <param name="gobBlocksInY">If <paramref name="linear"/> is false, the amount of GOB blocks in the Y axis</param>
/// <param name="gobBlocksInZ">If <paramref name="linear"/> is false, the amount of GOB blocks in the Z axis</param>
/// <returns>A matching texture, or null if there is no match</returns>
public Texture FindTexture(
MemoryManager memoryManager,
ulong gpuVa,
int bpp,
int stride,
int height,
int xCount,
int yCount,
bool linear,
int gobBlocksInY,
int gobBlocksInZ)
{
ulong address = memoryManager.Translate(gpuVa);
if (address == MemoryManager.PteUnmapped)
{
return null;
}
int addressMatches = _textures.FindOverlaps(address, ref _textureOverlaps);
Texture textureMatch = null;
for (int i = 0; i < addressMatches; i++)
{
Texture texture = _textureOverlaps[i];
FormatInfo format = texture.Info.FormatInfo;
if (texture.Info.DepthOrLayers > 1 || texture.Info.Levels > 1 || texture.Info.FormatInfo.IsCompressed)
{
// Don't support direct buffer copies to anything that isn't a single 2D image, uncompressed.
continue;
}
bool match;
if (linear)
{
// Size is not available for linear textures. Use the stride and end of the copy region instead.
match = texture.Info.IsLinear && texture.Info.Stride == stride && yCount == texture.Info.Height;
}
else
{
// Bpp may be a mismatch between the target texture and the param.
// Due to the way linear strided and block layouts work, widths can be multiplied by Bpp for comparison.
// Note: tex.Width is the aligned texture size. Prefer param.XCount, as the destination should be a texture with that exact size.
bool sizeMatch = xCount * bpp == texture.Info.Width * format.BytesPerPixel && height == texture.Info.Height;
bool formatMatch = !texture.Info.IsLinear &&
texture.Info.GobBlocksInY == gobBlocksInY &&
texture.Info.GobBlocksInZ == gobBlocksInZ;
match = sizeMatch && formatMatch;
}
if (match)
{
if (textureMatch == null)
{
textureMatch = texture;
}
else if (texture.Group != textureMatch.Group)
{
return null; // It's ambiguous which texture should match between multiple choices, so leave it up to the slow path.
}
}
}
return textureMatch;
}
/// <summary>
/// Resizes the temporary buffer used for range list intersection results, if it has grown too much.
/// </summary>
private void ShrinkOverlapsBufferIfNeeded()
{
if (_textureOverlaps.Length > OverlapsBufferMaxCapacity)
{
Array.Resize(ref _textureOverlaps, OverlapsBufferMaxCapacity);
}
}
/// <summary>
/// Gets a texture creation information from texture information.
/// This can be used to create new host textures.
/// </summary>
/// <param name="info">Texture information</param>
/// <param name="caps">GPU capabilities</param>
/// <param name="scale">Texture scale factor, to be applied to the texture size</param>
/// <returns>The texture creation information</returns>
public static TextureCreateInfo GetCreateInfo(TextureInfo info, Capabilities caps, float scale)
{
FormatInfo formatInfo = TextureCompatibility.ToHostCompatibleFormat(info, caps);
if (info.Target == Target.TextureBuffer && !caps.SupportsSnormBufferTextureFormat)
{
// If the host does not support signed normalized formats, we use a signed integer format instead.
// The shader will need the appropriate conversion code to compensate.
switch (formatInfo.Format)
{
case Format.R8Snorm:
formatInfo = new FormatInfo(Format.R8Sint, 1, 1, 1, 1);
break;
case Format.R16Snorm:
formatInfo = new FormatInfo(Format.R16Sint, 1, 1, 2, 1);
break;
case Format.R8G8Snorm:
formatInfo = new FormatInfo(Format.R8G8Sint, 1, 1, 2, 2);
break;
case Format.R16G16Snorm:
formatInfo = new FormatInfo(Format.R16G16Sint, 1, 1, 4, 2);
break;
case Format.R8G8B8A8Snorm:
formatInfo = new FormatInfo(Format.R8G8B8A8Sint, 1, 1, 4, 4);
break;
case Format.R16G16B16A16Snorm:
formatInfo = new FormatInfo(Format.R16G16B16A16Sint, 1, 1, 8, 4);
break;
}
}
int width = info.Width / info.SamplesInX;
int height = info.Height / info.SamplesInY;
int depth = info.GetDepth() * info.GetLayers();
if (scale != 1f)
{
width = (int)MathF.Ceiling(width * scale);
height = (int)MathF.Ceiling(height * scale);
}
return new TextureCreateInfo(
width,
height,
depth,
info.Levels,
info.Samples,
formatInfo.BlockWidth,
formatInfo.BlockHeight,
formatInfo.BytesPerPixel,
formatInfo.Format,
info.DepthStencilMode,
info.Target,
info.SwizzleR,
info.SwizzleG,
info.SwizzleB,
info.SwizzleA);
}
/// <summary>
/// Removes a texture from the cache.
/// </summary>
/// <remarks>
/// This only removes the texture from the internal list, not from the auto-deletion cache.
/// It may still have live references after the removal.
/// </remarks>
/// <param name="texture">The texture to be removed</param>
public void RemoveTextureFromCache(Texture texture)
{
lock (_textures)
{
_textures.Remove(texture);
}
lock (_partiallyMappedTextures)
{
_partiallyMappedTextures.Remove(texture);
}
}
/// <summary>
/// Queries a texture's memory range and marks it as partially mapped or not.
/// Partially mapped textures re-evaluate their memory range after each time GPU memory is mapped.
/// </summary>
/// <param name="memoryManager">GPU memory manager where the texture is mapped</param>
/// <param name="address">The virtual address of the texture</param>
/// <param name="texture">The texture to be marked</param>
/// <returns>The physical regions for the texture, found when evaluating whether the texture was partially mapped</returns>
public MultiRange UpdatePartiallyMapped(MemoryManager memoryManager, ulong address, Texture texture)
{
MultiRange range;
lock (_partiallyMappedTextures)
{
range = memoryManager.GetPhysicalRegions(address, texture.Size);
bool partiallyMapped = false;
for (int i = 0; i < range.Count; i++)
{
if (range.GetSubRange(i).Address == MemoryManager.PteUnmapped)
{
partiallyMapped = true;
break;
}
}
if (partiallyMapped)
{
_partiallyMappedTextures.Add(texture);
}
else
{
_partiallyMappedTextures.Remove(texture);
}
}
return range;
}
/// <summary>
/// Adds a texture to the short duration cache. This typically keeps it alive for two ticks.
/// </summary>
/// <param name="texture">Texture to add to the short cache</param>
/// <param name="descriptor">Last used texture descriptor</param>
public void AddShortCache(Texture texture, ref TextureDescriptor descriptor)
{
_cache.AddShortCache(texture, ref descriptor);
}
/// <summary>
/// Adds a texture to the short duration cache without a descriptor. This typically keeps it alive for two ticks.
/// On expiry, it will be removed from the AutoDeleteCache.
/// </summary>
/// <param name="texture">Texture to add to the short cache</param>
public void AddShortCache(Texture texture)
{
_cache.AddShortCache(texture);
}
/// <summary>
/// Removes a texture from the short duration cache.
/// </summary>
/// <param name="texture">Texture to remove from the short cache</param>
public void RemoveShortCache(Texture texture)
{
_cache.RemoveShortCache(texture);
}
/// <summary>
/// Ticks periodic elements of the texture cache.
/// </summary>
public void Tick()
{
_cache.ProcessShortCache();
}
/// <summary>
/// Disposes all textures and samplers in the cache.
/// It's an error to use the texture cache after disposal.
/// </summary>
public void Dispose()
{
lock (_textures)
{
foreach (Texture texture in _textures)
{
texture.Dispose();
}
}
}
}
}