Ryujinx/Ryujinx.Graphics.Gpu/Image/TexturePool.cs

407 lines
15 KiB
C#
Raw Normal View History

2019-12-29 18:41:50 +01:00
using Ryujinx.Common.Logging;
2019-10-13 08:02:07 +02:00
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Texture;
using System;
using System.Collections.Concurrent;
2019-10-13 08:02:07 +02:00
using System.Collections.Generic;
namespace Ryujinx.Graphics.Gpu.Image
{
/// <summary>
/// Texture pool.
/// </summary>
class TexturePool : Pool<Texture, TextureDescriptor>
2019-10-13 08:02:07 +02:00
{
private readonly GpuChannel _channel;
private readonly ConcurrentQueue<Texture> _dereferenceQueue = new ConcurrentQueue<Texture>();
2019-10-13 08:02:07 +02:00
/// <summary>
/// Intrusive linked list node used on the texture pool cache.
/// </summary>
public LinkedListNode<TexturePool> CacheNode { get; set; }
2019-10-13 08:02:07 +02:00
/// <summary>
/// Constructs a new instance of the texture pool.
/// </summary>
/// <param name="context">GPU context that the texture pool belongs to</param>
/// <param name="channel">GPU channel that the texture pool belongs to</param>
/// <param name="address">Address of the texture pool in guest memory</param>
/// <param name="maximumId">Maximum texture ID of the texture pool (equal to maximum textures minus one)</param>
public TexturePool(GpuContext context, GpuChannel channel, ulong address, int maximumId) : base(context, channel.MemoryManager.Physical, address, maximumId)
{
_channel = channel;
}
/// <summary>
/// Gets the texture with the given ID.
/// </summary>
/// <param name="id">ID of the texture. This is effectively a zero-based index</param>
/// <returns>The texture with the given ID</returns>
2019-10-13 08:02:07 +02:00
public override Texture Get(int id)
{
if ((uint)id >= Items.Length)
{
return null;
}
Replace CacheResourceWrite with more general "precise" write (#2684) * Replace CacheResourceWrite with more general "precise" write The goal of CacheResourceWrite was to notify GPU resources when they were modified directly, by looking up the modified address/size in a structure and calling a method on each resource. The downside of this is that each resource cache has to be queried individually, they all have to implement their own way to do this, and it can only signal to resources using the same PhysicalMemory instance. This PR adds the ability to signal a write as "precise" on the tracking, which signals a special handler (if present) which can be used to avoid unnecessary flush actions, or maybe even more. For buffers, precise writes specifically do not flush, and instead punch a hole in the modified range list to indicate that the data on GPU has been replaced. The downside is that precise actions must ignore the page protection bits and always signal - as they need to notify the target resource to ignore the sequence number optimization. I had to reintroduce the sequence number increment after I2M, as removing it was causing issues in rabbids kingdom battle. However - all resources modified by I2M are notified directly to lower their sequence number, so the problem is likely that another unrelated resource is not being properly updated. Thankfully, doing this does not affect performance in the games I tested. This should fix regressions from #2624. Test any games that were broken by that. (RF4, rabbids kingdom battle) I've also added a sequence number increment to ThreedClass.IncrementSyncpoint, as it seems to fix buffer corruption in OpenGL homebrew. (this was a regression from removing sequence number increment from constant buffer update - another unrelated resource thing) * Add tests. * Add XML docs for GpuRegionHandle * Skip UpdateProtection if only precise actions were called This allows precise actions to skip reprotection costs.
2021-09-29 02:27:03 +02:00
if (SequenceNumber != Context.SequenceNumber)
{
Replace CacheResourceWrite with more general "precise" write (#2684) * Replace CacheResourceWrite with more general "precise" write The goal of CacheResourceWrite was to notify GPU resources when they were modified directly, by looking up the modified address/size in a structure and calling a method on each resource. The downside of this is that each resource cache has to be queried individually, they all have to implement their own way to do this, and it can only signal to resources using the same PhysicalMemory instance. This PR adds the ability to signal a write as "precise" on the tracking, which signals a special handler (if present) which can be used to avoid unnecessary flush actions, or maybe even more. For buffers, precise writes specifically do not flush, and instead punch a hole in the modified range list to indicate that the data on GPU has been replaced. The downside is that precise actions must ignore the page protection bits and always signal - as they need to notify the target resource to ignore the sequence number optimization. I had to reintroduce the sequence number increment after I2M, as removing it was causing issues in rabbids kingdom battle. However - all resources modified by I2M are notified directly to lower their sequence number, so the problem is likely that another unrelated resource is not being properly updated. Thankfully, doing this does not affect performance in the games I tested. This should fix regressions from #2624. Test any games that were broken by that. (RF4, rabbids kingdom battle) I've also added a sequence number increment to ThreedClass.IncrementSyncpoint, as it seems to fix buffer corruption in OpenGL homebrew. (this was a regression from removing sequence number increment from constant buffer update - another unrelated resource thing) * Add tests. * Add XML docs for GpuRegionHandle * Skip UpdateProtection if only precise actions were called This allows precise actions to skip reprotection costs.
2021-09-29 02:27:03 +02:00
SequenceNumber = Context.SequenceNumber;
SynchronizeMemory();
}
2019-10-13 08:02:07 +02:00
Texture texture = Items[id];
if (texture == null)
{
TextureDescriptor descriptor = GetDescriptor(id);
2019-10-13 08:02:07 +02:00
TextureInfo info = GetInfo(descriptor, out int layerSize);
2019-10-13 08:02:07 +02:00
ProcessDereferenceQueue();
texture = PhysicalMemory.TextureCache.FindOrCreateTexture(_channel.MemoryManager, TextureSearchFlags.ForSampler, info, layerSize);
// If this happens, then the texture address is invalid, we can't add it to the cache.
if (texture == null)
2019-10-13 08:02:07 +02:00
{
return null;
}
texture.IncrementReferenceCount(this, id);
2019-10-13 08:02:07 +02:00
Items[id] = texture;
DescriptorCache[id] = descriptor;
2019-10-13 08:02:07 +02:00
}
else
{
if (texture.ChangedSize)
{
// Texture changed size at one point - it may be a different size than the sampler expects.
// This can be triggered when the size is changed by a size hint on copy or draw, but the texture has been sampled before.
TextureDescriptor descriptor = GetDescriptor(id);
int baseLevel = descriptor.UnpackBaseLevel();
int width = Math.Max(1, descriptor.UnpackWidth() >> baseLevel);
int height = Math.Max(1, descriptor.UnpackHeight() >> baseLevel);
if (texture.Info.Width != width || texture.Info.Height != height)
{
texture.ChangeSize(width, height, texture.Info.DepthOrLayers);
}
}
2019-10-13 08:02:07 +02:00
// Memory is automatically synchronized on texture creation.
texture.SynchronizeMemory();
}
return texture;
}
/// <summary>
/// Forcibly remove a texture from this pool's items.
/// If deferred, the dereference will be queued to occur on the render thread.
/// </summary>
/// <param name="texture">The texture being removed</param>
/// <param name="id">The ID of the texture in this pool</param>
/// <param name="deferred">If true, queue the dereference to happen on the render thread, otherwise dereference immediately</param>
public void ForceRemove(Texture texture, int id, bool deferred)
{
Items[id] = null;
if (deferred)
{
_dereferenceQueue.Enqueue(texture);
}
else
{
texture.DecrementReferenceCount();
}
}
/// <summary>
/// Process the dereference queue, decrementing the reference count for each texture in it.
/// This is used to ensure that texture disposal happens on the render thread.
/// </summary>
private void ProcessDereferenceQueue()
{
while (_dereferenceQueue.TryDequeue(out Texture toRemove))
{
toRemove.DecrementReferenceCount();
}
}
/// <summary>
/// Implementation of the texture pool range invalidation.
/// </summary>
/// <param name="address">Start address of the range of the texture pool</param>
/// <param name="size">Size of the range being invalidated</param>
2019-10-13 08:02:07 +02:00
protected override void InvalidateRangeImpl(ulong address, ulong size)
{
ProcessDereferenceQueue();
2019-10-13 08:02:07 +02:00
ulong endAddress = address + size;
for (; address < endAddress; address += DescriptorSize)
{
int id = (int)((address - Address) / DescriptorSize);
Texture texture = Items[id];
if (texture != null)
{
TextureDescriptor descriptor = PhysicalMemory.Read<TextureDescriptor>(address);
2019-10-13 08:02:07 +02:00
// If the descriptors are the same, the texture is the same,
// we don't need to remove as it was not modified. Just continue.
if (descriptor.Equals(ref DescriptorCache[id]))
2019-10-13 08:02:07 +02:00
{
continue;
}
texture.DecrementReferenceCount(this, id);
2019-10-13 08:02:07 +02:00
Items[id] = null;
}
}
}
/// <summary>
/// Gets texture information from a texture descriptor.
/// </summary>
/// <param name="descriptor">The texture descriptor</param>
/// <param name="layerSize">Layer size for textures using a sub-range of mipmap levels, otherwise 0</param>
/// <returns>The texture information</returns>
private TextureInfo GetInfo(TextureDescriptor descriptor, out int layerSize)
2019-10-13 08:02:07 +02:00
{
int depthOrLayers = descriptor.UnpackDepth();
int levels = descriptor.UnpackLevels();
TextureMsaaMode msaaMode = descriptor.UnpackTextureMsaaMode();
int samplesInX = msaaMode.SamplesInX();
int samplesInY = msaaMode.SamplesInY();
int stride = descriptor.UnpackStride();
TextureDescriptorType descriptorType = descriptor.UnpackTextureDescriptorType();
bool isLinear = descriptorType == TextureDescriptorType.Linear;
Target target = descriptor.UnpackTextureTarget().Convert((samplesInX | samplesInY) != 1);
int width = target == Target.TextureBuffer ? descriptor.UnpackBufferTextureWidth() : descriptor.UnpackWidth();
int height = descriptor.UnpackHeight();
if (target == Target.Texture2DMultisample || target == Target.Texture2DMultisampleArray)
{
// This is divided back before the backend texture is created.
width *= samplesInX;
height *= samplesInY;
}
// We use 2D targets for 1D textures as that makes texture cache
// management easier. We don't know the target for render target
// and copies, so those would normally use 2D targets, which are
// not compatible with 1D targets. By doing that we also allow those
// to match when looking for compatible textures on the cache.
if (target == Target.Texture1D)
{
target = Target.Texture2D;
height = 1;
}
else if (target == Target.Texture1DArray)
{
target = Target.Texture2DArray;
height = 1;
}
2019-10-13 08:02:07 +02:00
uint format = descriptor.UnpackFormat();
bool srgb = descriptor.UnpackSrgb();
ulong gpuVa = descriptor.UnpackAddress();
2019-10-13 08:02:07 +02:00
if (!FormatTable.TryGetTextureFormat(format, srgb, out FormatInfo formatInfo))
{
if (gpuVa != 0 && (int)format > 0)
Add Profiled Persistent Translation Cache. (#769) * Delete DelegateTypes.cs * Delete DelegateCache.cs * Add files via upload * Update Horizon.cs * Update Program.cs * Update MainWindow.cs * Update Aot.cs * Update RelocEntry.cs * Update Translator.cs * Update MemoryManager.cs * Update InstEmitMemoryHelper.cs * Update Delegates.cs * Nit. * Nit. * Nit. * 10 fewer MSIL bytes for us * Add comment. Nits. * Update Translator.cs * Update Aot.cs * Nits. * Opt.. * Opt.. * Opt.. * Opt.. * Allow to change compression level. * Update MemoryManager.cs * Update Translator.cs * Manage corner cases during the save phase. Nits. * Update Aot.cs * Translator response tweak for Aot disabled. Nit. * Nit. * Nits. * Create DelegateHelpers.cs * Update Delegates.cs * Nit. * Nit. * Nits. * Fix due to #784. * Fixes due to #757 & #841. * Fix due to #846. * Fix due to #847. * Use MethodInfo for managed method calls. Use IR methods instead of managed methods about Max/Min (S/U). Follow-ups & Nits. * Add missing exception messages. Reintroduce slow path for Fmov_Vi. Implement slow path for Fmov_Si. * Switch to the new folder structure. Nits. * Impl. index-based relocation information. Impl. cache file version field. * Nit. * Address gdkchan comments. Mainly: - fixed cache file corruption issue on exit; - exposed a way to disable AOT on the GUI. * Address AcK77 comment. * Address Thealexbarney, jduncanator & emmauss comments. Header magic, CpuId (FI) & Aot -> Ptc. * Adaptation to the new application reloading system. Improvements to the call system of managed methods. Follow-ups. Nits. * Get the same boot times as on master when PTC is disabled. * Profiled Aot. * A32 support (#897). * #975 support (1 of 2). * #975 support (2 of 2). * Rebase fix & nits. * Some fixes and nits (still one bug left). * One fix & nits. * Tests fix (by gdk) & nits. * Support translations not only in high quality and rejit. Nits. * Added possibility to skip translations and continue execution, using `ESC` key. * Update SettingsWindow.cs * Update GLRenderer.cs * Update Ptc.cs * Disabled Profiled PTC by default as requested in the past by gdk. * Fix rejit bug. Increased number of parallel translations. Add stack unwinding stuffs support (1 of 2). Nits. * Add stack unwinding stuffs support (2 of 2). Tuned number of parallel translations. * Restored the ability to assemble jumps with 8-bit offset when Profiled PTC is disabled or during profiling. Modifications due to rebase. Nits. * Limited profiling of the functions to be translated to the addresses belonging to the range of static objects only. * Nits. * Nits. * Update Delegates.cs * Nit. * Update InstEmitSimdArithmetic.cs * Address riperiperi comments. * Fixed the issue of unjustifiably longer boot times at the second boot than at the first boot, measured at the same time or reference point and with the same number of translated functions. * Implemented a simple redundant load/save mechanism. Halved the value of Decoder.MaxInstsPerFunction more appropriate for the current performance of the Translator. Replaced by Logger.PrintError to Logger.PrintDebug in TexturePool.cs about the supposed invalid texture format to avoid the spawn of the log. Nits. * Nit. Improved Logger.PrintError in TexturePool.cs to avoid log spawn. Added missing code for FZ handling (in output) for fp max/min instructions (slow paths). * Add configuration migration for PTC Co-authored-by: Thog <me@thog.eu>
2020-06-16 20:28:02 +02:00
{
Logger.Error?.Print(LogClass.Gpu, $"Invalid texture format 0x{format:X} (sRGB: {srgb}).");
Add Profiled Persistent Translation Cache. (#769) * Delete DelegateTypes.cs * Delete DelegateCache.cs * Add files via upload * Update Horizon.cs * Update Program.cs * Update MainWindow.cs * Update Aot.cs * Update RelocEntry.cs * Update Translator.cs * Update MemoryManager.cs * Update InstEmitMemoryHelper.cs * Update Delegates.cs * Nit. * Nit. * Nit. * 10 fewer MSIL bytes for us * Add comment. Nits. * Update Translator.cs * Update Aot.cs * Nits. * Opt.. * Opt.. * Opt.. * Opt.. * Allow to change compression level. * Update MemoryManager.cs * Update Translator.cs * Manage corner cases during the save phase. Nits. * Update Aot.cs * Translator response tweak for Aot disabled. Nit. * Nit. * Nits. * Create DelegateHelpers.cs * Update Delegates.cs * Nit. * Nit. * Nits. * Fix due to #784. * Fixes due to #757 & #841. * Fix due to #846. * Fix due to #847. * Use MethodInfo for managed method calls. Use IR methods instead of managed methods about Max/Min (S/U). Follow-ups & Nits. * Add missing exception messages. Reintroduce slow path for Fmov_Vi. Implement slow path for Fmov_Si. * Switch to the new folder structure. Nits. * Impl. index-based relocation information. Impl. cache file version field. * Nit. * Address gdkchan comments. Mainly: - fixed cache file corruption issue on exit; - exposed a way to disable AOT on the GUI. * Address AcK77 comment. * Address Thealexbarney, jduncanator & emmauss comments. Header magic, CpuId (FI) & Aot -> Ptc. * Adaptation to the new application reloading system. Improvements to the call system of managed methods. Follow-ups. Nits. * Get the same boot times as on master when PTC is disabled. * Profiled Aot. * A32 support (#897). * #975 support (1 of 2). * #975 support (2 of 2). * Rebase fix & nits. * Some fixes and nits (still one bug left). * One fix & nits. * Tests fix (by gdk) & nits. * Support translations not only in high quality and rejit. Nits. * Added possibility to skip translations and continue execution, using `ESC` key. * Update SettingsWindow.cs * Update GLRenderer.cs * Update Ptc.cs * Disabled Profiled PTC by default as requested in the past by gdk. * Fix rejit bug. Increased number of parallel translations. Add stack unwinding stuffs support (1 of 2). Nits. * Add stack unwinding stuffs support (2 of 2). Tuned number of parallel translations. * Restored the ability to assemble jumps with 8-bit offset when Profiled PTC is disabled or during profiling. Modifications due to rebase. Nits. * Limited profiling of the functions to be translated to the addresses belonging to the range of static objects only. * Nits. * Nits. * Update Delegates.cs * Nit. * Update InstEmitSimdArithmetic.cs * Address riperiperi comments. * Fixed the issue of unjustifiably longer boot times at the second boot than at the first boot, measured at the same time or reference point and with the same number of translated functions. * Implemented a simple redundant load/save mechanism. Halved the value of Decoder.MaxInstsPerFunction more appropriate for the current performance of the Translator. Replaced by Logger.PrintError to Logger.PrintDebug in TexturePool.cs about the supposed invalid texture format to avoid the spawn of the log. Nits. * Nit. Improved Logger.PrintError in TexturePool.cs to avoid log spawn. Added missing code for FZ handling (in output) for fp max/min instructions (slow paths). * Add configuration migration for PTC Co-authored-by: Thog <me@thog.eu>
2020-06-16 20:28:02 +02:00
}
2019-10-13 08:02:07 +02:00
formatInfo = FormatInfo.Default;
}
int gobBlocksInY = descriptor.UnpackGobBlocksInY();
int gobBlocksInZ = descriptor.UnpackGobBlocksInZ();
int gobBlocksInTileX = descriptor.UnpackGobBlocksInTileX();
layerSize = 0;
int minLod = descriptor.UnpackBaseLevel();
int maxLod = descriptor.UnpackMaxLevelInclusive();
// Linear textures don't support mipmaps, so we don't handle this case here.
if ((minLod != 0 || maxLod + 1 != levels) && target != Target.TextureBuffer && !isLinear)
{
int depth = TextureInfo.GetDepth(target, depthOrLayers);
int layers = TextureInfo.GetLayers(target, depthOrLayers);
SizeInfo sizeInfo = SizeCalculator.GetBlockLinearTextureSize(
width,
height,
depth,
levels,
layers,
formatInfo.BlockWidth,
formatInfo.BlockHeight,
formatInfo.BytesPerPixel,
gobBlocksInY,
gobBlocksInZ,
gobBlocksInTileX);
layerSize = sizeInfo.LayerSize;
if (minLod != 0 && minLod < levels)
{
// If the base level is not zero, we additionally add the mip level offset
// to the address, this allows the texture manager to find the base level from the
// address if there is a overlapping texture on the cache that can contain the new texture.
gpuVa += (ulong)sizeInfo.GetMipOffset(minLod);
width = Math.Max(1, width >> minLod);
height = Math.Max(1, height >> minLod);
if (target == Target.Texture3D)
{
depthOrLayers = Math.Max(1, depthOrLayers >> minLod);
}
(gobBlocksInY, gobBlocksInZ) = SizeCalculator.GetMipGobBlockSizes(height, depth, formatInfo.BlockHeight, gobBlocksInY, gobBlocksInZ);
}
levels = (maxLod - minLod) + 1;
}
2019-10-13 08:02:07 +02:00
SwizzleComponent swizzleR = descriptor.UnpackSwizzleR().Convert();
SwizzleComponent swizzleG = descriptor.UnpackSwizzleG().Convert();
SwizzleComponent swizzleB = descriptor.UnpackSwizzleB().Convert();
SwizzleComponent swizzleA = descriptor.UnpackSwizzleA().Convert();
DepthStencilMode depthStencilMode = GetDepthStencilMode(
formatInfo.Format,
swizzleR,
swizzleG,
swizzleB,
swizzleA);
if (formatInfo.Format.IsDepthOrStencil())
{
swizzleR = SwizzleComponent.Red;
swizzleG = SwizzleComponent.Red;
swizzleB = SwizzleComponent.Red;
if (depthStencilMode == DepthStencilMode.Depth)
{
swizzleA = SwizzleComponent.One;
}
else
{
swizzleA = SwizzleComponent.Red;
}
}
2019-10-13 08:02:07 +02:00
return new TextureInfo(
gpuVa,
2019-10-13 08:02:07 +02:00
width,
height,
depthOrLayers,
levels,
samplesInX,
samplesInY,
stride,
isLinear,
gobBlocksInY,
gobBlocksInZ,
gobBlocksInTileX,
target,
formatInfo,
depthStencilMode,
swizzleR,
swizzleG,
swizzleB,
swizzleA);
}
/// <summary>
/// Gets the texture depth-stencil mode, based on the swizzle components of each color channel.
/// The depth-stencil mode is determined based on how the driver sets those parameters.
/// </summary>
/// <param name="format">The format of the texture</param>
/// <param name="components">The texture swizzle components</param>
/// <returns>The depth-stencil mode</returns>
2019-10-13 08:02:07 +02:00
private static DepthStencilMode GetDepthStencilMode(Format format, params SwizzleComponent[] components)
{
// R = Depth, G = Stencil.
// On 24-bits depth formats, this is inverted (Stencil is R etc).
// NVN setup:
// For depth, A is set to 1.0f, the other components are set to Depth.
// For stencil, all components are set to Stencil.
SwizzleComponent component = components[0];
for (int index = 1; index < 4 && !IsRG(component); index++)
{
component = components[index];
}
if (!IsRG(component))
{
return DepthStencilMode.Depth;
}
if (format == Format.D24UnormS8Uint)
2019-10-13 08:02:07 +02:00
{
return component == SwizzleComponent.Red
? DepthStencilMode.Stencil
: DepthStencilMode.Depth;
}
else
{
return component == SwizzleComponent.Red
? DepthStencilMode.Depth
: DepthStencilMode.Stencil;
}
}
/// <summary>
/// Checks if the swizzle component is equal to the red or green channels.
/// </summary>
/// <param name="component">The swizzle component to check</param>
2019-12-30 18:44:22 +01:00
/// <returns>True if the swizzle component is equal to the red or green, false otherwise</returns>
2019-10-13 08:02:07 +02:00
private static bool IsRG(SwizzleComponent component)
{
return component == SwizzleComponent.Red ||
component == SwizzleComponent.Green;
}
/// <summary>
/// Decrements the reference count of the texture.
/// This indicates that the texture pool is not using it anymore.
/// </summary>
/// <param name="item">The texture to be deleted</param>
2019-10-13 08:02:07 +02:00
protected override void Delete(Texture item)
{
item?.DecrementReferenceCount(this);
}
public override void Dispose()
{
ProcessDereferenceQueue();
base.Dispose();
2019-10-13 08:02:07 +02:00
}
}
}