Implement DMA texture copy component shuffle (#4717)

* Implement DMA texture copy component shuffle

* Set UInt24 alignment to 1
This commit is contained in:
gdkchan 2023-04-24 10:28:03 -03:00 committed by GitHub
parent 4dd77316f7
commit db4242c5dc
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23

View file

@ -6,6 +6,7 @@ using Ryujinx.Graphics.Texture;
using System;
using System.Collections.Generic;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
using System.Runtime.Intrinsics;
namespace Ryujinx.Graphics.Gpu.Engine.Dma
@ -32,6 +33,69 @@ namespace Ryujinx.Graphics.Gpu.Engine.Dma
RemapEnable = 1 << 10
}
/// <summary>
/// Texture parameters for copy.
/// </summary>
private struct TextureParams
{
/// <summary>
/// Copy region X coordinate.
/// </summary>
public readonly int RegionX;
/// <summary>
/// Copy region Y coordinate.
/// </summary>
public readonly int RegionY;
/// <summary>
/// Offset from the base pointer of the data in memory.
/// </summary>
public readonly int BaseOffset;
/// <summary>
/// Bytes per pixel.
/// </summary>
public readonly int Bpp;
/// <summary>
/// Whether the texture is linear. If false, the texture is block linear.
/// </summary>
public readonly bool Linear;
/// <summary>
/// Pixel offset from XYZ coordinates calculator.
/// </summary>
public readonly OffsetCalculator Calculator;
/// <summary>
/// Creates texture parameters.
/// </summary>
/// <param name="regionX">Copy region X coordinate</param>
/// <param name="regionY">Copy region Y coordinate</param>
/// <param name="baseOffset">Offset from the base pointer of the data in memory</param>
/// <param name="bpp">Bytes per pixel</param>
/// <param name="linear">Whether the texture is linear. If false, the texture is block linear</param>
/// <param name="calculator">Pixel offset from XYZ coordinates calculator</param>
public TextureParams(int regionX, int regionY, int baseOffset, int bpp, bool linear, OffsetCalculator calculator)
{
RegionX = regionX;
RegionY = regionY;
BaseOffset = baseOffset;
Bpp = bpp;
Linear = linear;
Calculator = calculator;
}
}
[StructLayout(LayoutKind.Sequential, Size = 3, Pack = 1)]
private struct UInt24
{
public byte Byte0;
public byte Byte1;
public byte Byte2;
}
/// <summary>
/// Creates a new instance of the DMA copy engine class.
/// </summary>
@ -154,8 +218,10 @@ namespace Ryujinx.Graphics.Gpu.Engine.Dma
{
// Buffer to texture copy.
int componentSize = (int)_state.State.SetRemapComponentsComponentSize + 1;
int srcBpp = remap ? ((int)_state.State.SetRemapComponentsNumSrcComponents + 1) * componentSize : 1;
int dstBpp = remap ? ((int)_state.State.SetRemapComponentsNumDstComponents + 1) * componentSize : 1;
int srcComponents = (int)_state.State.SetRemapComponentsNumSrcComponents + 1;
int dstComponents = (int)_state.State.SetRemapComponentsNumDstComponents + 1;
int srcBpp = remap ? srcComponents * componentSize : 1;
int dstBpp = remap ? dstComponents * componentSize : 1;
var dst = Unsafe.As<uint, DmaTexture>(ref _state.State.SetDstBlockSize);
var src = Unsafe.As<uint, DmaTexture>(ref _state.State.SetSrcBlockSize);
@ -274,63 +340,51 @@ namespace Ryujinx.Graphics.Gpu.Engine.Dma
}
}
unsafe bool Convert<T>(Span<byte> dstSpan, ReadOnlySpan<byte> srcSpan) where T : unmanaged
{
if (srcLinear && dstLinear && srcBpp == dstBpp)
{
// Optimized path for purely linear copies - we don't need to calculate every single byte offset,
// and we can make use of Span.CopyTo which is very very fast (even compared to pointers)
for (int y = 0; y < yCount; y++)
{
srcCalculator.SetY(srcRegionY + y);
dstCalculator.SetY(dstRegionY + y);
int srcOffset = srcCalculator.GetOffset(srcRegionX);
int dstOffset = dstCalculator.GetOffset(dstRegionX);
srcSpan.Slice(srcOffset - srcBaseOffset, xCount * srcBpp)
.CopyTo(dstSpan.Slice(dstOffset - dstBaseOffset, xCount * dstBpp));
}
}
else
{
fixed (byte* dstPtr = dstSpan, srcPtr = srcSpan)
{
byte* dstBase = dstPtr - dstBaseOffset; // Layout offset is relative to the base, so we need to subtract the span's offset.
byte* srcBase = srcPtr - srcBaseOffset;
for (int y = 0; y < yCount; y++)
{
srcCalculator.SetY(srcRegionY + y);
dstCalculator.SetY(dstRegionY + y);
for (int x = 0; x < xCount; x++)
{
int srcOffset = srcCalculator.GetOffset(srcRegionX + x);
int dstOffset = dstCalculator.GetOffset(dstRegionX + x);
*(T*)(dstBase + dstOffset) = *(T*)(srcBase + srcOffset);
}
}
}
}
return true;
}
// OPT: This allocates a (potentially) huge temporary array and then copies an existing
// region of memory into it, data that might get overwritten entirely anyways. Ideally this should
// all be rewritten to use pooled arrays, but that gets complicated with packed data and strides
Span<byte> dstSpan = memoryManager.GetSpan(dstGpuVa + (ulong)dstBaseOffset, dstSize).ToArray();
bool _ = srcBpp switch
TextureParams srcParams = new TextureParams(srcRegionX, srcRegionY, srcBaseOffset, srcBpp, srcLinear, srcCalculator);
TextureParams dstParams = new TextureParams(dstRegionX, dstRegionY, dstBaseOffset, dstBpp, dstLinear, dstCalculator);
// If remapping is enabled, we always copy the components directly, in order.
// If it's enabled, but the mapping is just XYZW, we also copy them in order.
bool isIdentityRemap = !remap ||
(_state.State.SetRemapComponentsDstX == SetRemapComponentsDst.SrcX &&
(dstComponents < 2 || _state.State.SetRemapComponentsDstY == SetRemapComponentsDst.SrcY) &&
(dstComponents < 3 || _state.State.SetRemapComponentsDstZ == SetRemapComponentsDst.SrcZ) &&
(dstComponents < 4 || _state.State.SetRemapComponentsDstW == SetRemapComponentsDst.SrcW));
if (isIdentityRemap)
{
1 => Convert<byte>(dstSpan, srcSpan),
2 => Convert<ushort>(dstSpan, srcSpan),
4 => Convert<uint>(dstSpan, srcSpan),
8 => Convert<ulong>(dstSpan, srcSpan),
12 => Convert<Bpp12Pixel>(dstSpan, srcSpan),
16 => Convert<Vector128<byte>>(dstSpan, srcSpan),
_ => throw new NotSupportedException($"Unable to copy ${srcBpp} bpp pixel format.")
};
// The order of the components doesn't change, so we can just copy directly
// (with layout conversion if necessary).
switch (srcBpp)
{
case 1: Copy<byte>(dstSpan, srcSpan, dstParams, srcParams); break;
case 2: Copy<ushort>(dstSpan, srcSpan, dstParams, srcParams); break;
case 4: Copy<uint>(dstSpan, srcSpan, dstParams, srcParams); break;
case 8: Copy<ulong>(dstSpan, srcSpan, dstParams, srcParams); break;
case 12: Copy<Bpp12Pixel>(dstSpan, srcSpan, dstParams, srcParams); break;
case 16: Copy<Vector128<byte>>(dstSpan, srcSpan, dstParams, srcParams); break;
default: throw new NotSupportedException($"Unable to copy ${srcBpp} bpp pixel format.");
}
}
else
{
// The order or value of the components might change.
switch (componentSize)
{
case 1: CopyShuffle<byte>(dstSpan, srcSpan, dstParams, srcParams); break;
case 2: CopyShuffle<ushort>(dstSpan, srcSpan, dstParams, srcParams); break;
case 3: CopyShuffle<UInt24>(dstSpan, srcSpan, dstParams, srcParams); break;
case 4: CopyShuffle<uint>(dstSpan, srcSpan, dstParams, srcParams); break;
default: throw new NotSupportedException($"Unable to copy ${componentSize} component size.");
}
}
memoryManager.Write(dstGpuVa + (ulong)dstBaseOffset, dstSpan);
}
@ -372,6 +426,133 @@ namespace Ryujinx.Graphics.Gpu.Engine.Dma
}
}
/// <summary>
/// Copies data from one texture to another, while performing layout conversion if necessary.
/// </summary>
/// <typeparam name="T">Pixel type</typeparam>
/// <param name="dstSpan">Destination texture memory region</param>
/// <param name="srcSpan">Source texture memory region</param>
/// <param name="dst">Destination texture parameters</param>
/// <param name="src">Source texture parameters</param>
private unsafe void Copy<T>(Span<byte> dstSpan, ReadOnlySpan<byte> srcSpan, TextureParams dst, TextureParams src) where T : unmanaged
{
int xCount = (int)_state.State.LineLengthIn;
int yCount = (int)_state.State.LineCount;
if (src.Linear && dst.Linear && src.Bpp == dst.Bpp)
{
// Optimized path for purely linear copies - we don't need to calculate every single byte offset,
// and we can make use of Span.CopyTo which is very very fast (even compared to pointers)
for (int y = 0; y < yCount; y++)
{
src.Calculator.SetY(src.RegionY + y);
dst.Calculator.SetY(dst.RegionY + y);
int srcOffset = src.Calculator.GetOffset(src.RegionX);
int dstOffset = dst.Calculator.GetOffset(dst.RegionX);
srcSpan.Slice(srcOffset - src.BaseOffset, xCount * src.Bpp)
.CopyTo(dstSpan.Slice(dstOffset - dst.BaseOffset, xCount * dst.Bpp));
}
}
else
{
fixed (byte* dstPtr = dstSpan, srcPtr = srcSpan)
{
byte* dstBase = dstPtr - dst.BaseOffset; // Layout offset is relative to the base, so we need to subtract the span's offset.
byte* srcBase = srcPtr - src.BaseOffset;
for (int y = 0; y < yCount; y++)
{
src.Calculator.SetY(src.RegionY + y);
dst.Calculator.SetY(dst.RegionY + y);
for (int x = 0; x < xCount; x++)
{
int srcOffset = src.Calculator.GetOffset(src.RegionX + x);
int dstOffset = dst.Calculator.GetOffset(dst.RegionX + x);
*(T*)(dstBase + dstOffset) = *(T*)(srcBase + srcOffset);
}
}
}
}
}
/// <summary>
/// Sets texture pixel data to a constant value, while performing layout conversion if necessary.
/// </summary>
/// <typeparam name="T">Pixel type</typeparam>
/// <param name="dstSpan">Destination texture memory region</param>
/// <param name="dst">Destination texture parameters</param>
/// <param name="fillValue">Constant pixel value to be set</param>
private unsafe void Fill<T>(Span<byte> dstSpan, TextureParams dst, T fillValue) where T : unmanaged
{
int xCount = (int)_state.State.LineLengthIn;
int yCount = (int)_state.State.LineCount;
fixed (byte* dstPtr = dstSpan)
{
byte* dstBase = dstPtr - dst.BaseOffset; // Layout offset is relative to the base, so we need to subtract the span's offset.
for (int y = 0; y < yCount; y++)
{
dst.Calculator.SetY(dst.RegionY + y);
for (int x = 0; x < xCount; x++)
{
int dstOffset = dst.Calculator.GetOffset(dst.RegionX + x);
*(T*)(dstBase + dstOffset) = fillValue;
}
}
}
}
/// <summary>
/// Copies data from one texture to another, while performing layout conversion and component shuffling if necessary.
/// </summary>
/// <typeparam name="T">Pixel type</typeparam>
/// <param name="dstSpan">Destination texture memory region</param>
/// <param name="srcSpan">Source texture memory region</param>
/// <param name="dst">Destination texture parameters</param>
/// <param name="src">Source texture parameters</param>
private void CopyShuffle<T>(Span<byte> dstSpan, ReadOnlySpan<byte> srcSpan, TextureParams dst, TextureParams src) where T : unmanaged
{
int dstComponents = (int)_state.State.SetRemapComponentsNumDstComponents + 1;
for (int i = 0; i < dstComponents; i++)
{
SetRemapComponentsDst componentsDst = i switch
{
0 => _state.State.SetRemapComponentsDstX,
1 => _state.State.SetRemapComponentsDstY,
2 => _state.State.SetRemapComponentsDstZ,
_ => _state.State.SetRemapComponentsDstW
};
switch (componentsDst)
{
case SetRemapComponentsDst.SrcX:
Copy<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), srcSpan, dst, src);
break;
case SetRemapComponentsDst.SrcY:
Copy<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), srcSpan.Slice(Unsafe.SizeOf<T>()), dst, src);
break;
case SetRemapComponentsDst.SrcZ:
Copy<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), srcSpan.Slice(Unsafe.SizeOf<T>() * 2), dst, src);
break;
case SetRemapComponentsDst.SrcW:
Copy<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), srcSpan.Slice(Unsafe.SizeOf<T>() * 3), dst, src);
break;
case SetRemapComponentsDst.ConstA:
Fill<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), dst, Unsafe.As<uint, T>(ref _state.State.SetRemapConstA));
break;
case SetRemapComponentsDst.ConstB:
Fill<T>(dstSpan.Slice(Unsafe.SizeOf<T>() * i), dst, Unsafe.As<uint, T>(ref _state.State.SetRemapConstB));
break;
}
}
}
/// <summary>
/// Copies block linear data with block linear GOBs to a block linear destination with linear GOBs.
/// </summary>