Ryujinx/Ryujinx.Graphics.Gpu/Shader/GpuAccessor.cs
gdkchan 40b21cc3c4
Separate GPU engines (part 2/2) (#2440)
* 3D engine now uses DeviceState too, plus new state modification tracking

* Remove old methods code

* Remove GpuState and friends

* Optimize DeviceState, force inline some functions

* This change was not supposed to go in

* Proper channel initialization

* Optimize state read/write methods even more

* Fix debug build

* Do not dirty state if the write is redundant

* The YControl register should dirty either the viewport or front face state too, to update the host origin

* Avoid redundant vertex buffer updates

* Move state and get rid of the Ryujinx.Graphics.Gpu.State namespace

* Comments and nits

* Fix rebase

* PR feedback

* Move changed = false to improve codegen

* PR feedback

* Carry RyuJIT a bit more
2021-07-11 17:20:40 -03:00

226 lines
9 KiB
C#

using Ryujinx.Common.Logging;
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Shader;
namespace Ryujinx.Graphics.Gpu.Shader
{
/// <summary>
/// Represents a GPU state and memory accessor.
/// </summary>
class GpuAccessor : TextureDescriptorCapableGpuAccessor, IGpuAccessor
{
private readonly GpuContext _context;
private readonly GpuChannel _channel;
private readonly GpuAccessorState _state;
private readonly int _stageIndex;
private readonly bool _compute;
private readonly int _localSizeX;
private readonly int _localSizeY;
private readonly int _localSizeZ;
private readonly int _localMemorySize;
private readonly int _sharedMemorySize;
/// <summary>
/// Creates a new instance of the GPU state accessor for graphics shader translation.
/// </summary>
/// <param name="context">GPU context</param>
/// <param name="channel">GPU channel</param>
/// <param name="state">Current GPU state</param>
/// <param name="stageIndex">Graphics shader stage index (0 = Vertex, 4 = Fragment)</param>
public GpuAccessor(GpuContext context, GpuChannel channel, GpuAccessorState state, int stageIndex)
{
_context = context;
_channel = channel;
_state = state;
_stageIndex = stageIndex;
}
/// <summary>
/// Creates a new instance of the GPU state accessor for compute shader translation.
/// </summary>
/// <param name="context">GPU context</param>
/// <param name="channel">GPU channel</param>
/// <param name="state">Current GPU state</param>
/// <param name="localSizeX">Local group size X of the compute shader</param>
/// <param name="localSizeY">Local group size Y of the compute shader</param>
/// <param name="localSizeZ">Local group size Z of the compute shader</param>
/// <param name="localMemorySize">Local memory size of the compute shader</param>
/// <param name="sharedMemorySize">Shared memory size of the compute shader</param>
public GpuAccessor(
GpuContext context,
GpuChannel channel,
GpuAccessorState state,
int localSizeX,
int localSizeY,
int localSizeZ,
int localMemorySize,
int sharedMemorySize)
{
_context = context;
_channel = channel;
_state = state;
_compute = true;
_localSizeX = localSizeX;
_localSizeY = localSizeY;
_localSizeZ = localSizeZ;
_localMemorySize = localMemorySize;
_sharedMemorySize = sharedMemorySize;
}
/// <summary>
/// Prints a log message.
/// </summary>
/// <param name="message">Message to print</param>
public void Log(string message)
{
Logger.Warning?.Print(LogClass.Gpu, $"Shader translator: {message}");
}
/// <summary>
/// Reads data from GPU memory.
/// </summary>
/// <typeparam name="T">Type of the data to be read</typeparam>
/// <param name="address">GPU virtual address of the data</param>
/// <returns>Data at the memory location</returns>
public override T MemoryRead<T>(ulong address)
{
return _channel.MemoryManager.Read<T>(address);
}
/// <summary>
/// Checks if a given memory address is mapped.
/// </summary>
/// <param name="address">GPU virtual address to be checked</param>
/// <returns>True if the address is mapped, false otherwise</returns>
public bool MemoryMapped(ulong address)
{
return _channel.MemoryManager.IsMapped(address);
}
/// <summary>
/// Queries Local Size X for compute shaders.
/// </summary>
/// <returns>Local Size X</returns>
public int QueryComputeLocalSizeX() => _localSizeX;
/// <summary>
/// Queries Local Size Y for compute shaders.
/// </summary>
/// <returns>Local Size Y</returns>
public int QueryComputeLocalSizeY() => _localSizeY;
/// <summary>
/// Queries Local Size Z for compute shaders.
/// </summary>
/// <returns>Local Size Z</returns>
public int QueryComputeLocalSizeZ() => _localSizeZ;
/// <summary>
/// Queries Local Memory size in bytes for compute shaders.
/// </summary>
/// <returns>Local Memory size in bytes</returns>
public int QueryComputeLocalMemorySize() => _localMemorySize;
/// <summary>
/// Queries Shared Memory size in bytes for compute shaders.
/// </summary>
/// <returns>Shared Memory size in bytes</returns>
public int QueryComputeSharedMemorySize() => _sharedMemorySize;
/// <summary>
/// Queries Constant Buffer usage information.
/// </summary>
/// <returns>A mask where each bit set indicates a bound constant buffer</returns>
public uint QueryConstantBufferUse()
{
return _compute
? _channel.BufferManager.GetComputeUniformBufferUseMask()
: _channel.BufferManager.GetGraphicsUniformBufferUseMask(_stageIndex);
}
/// <summary>
/// Queries current primitive topology for geometry shaders.
/// </summary>
/// <returns>Current primitive topology</returns>
public InputTopology QueryPrimitiveTopology()
{
return _state.Topology switch
{
PrimitiveTopology.Points => InputTopology.Points,
PrimitiveTopology.Lines or
PrimitiveTopology.LineLoop or
PrimitiveTopology.LineStrip => InputTopology.Lines,
PrimitiveTopology.LinesAdjacency or
PrimitiveTopology.LineStripAdjacency => InputTopology.LinesAdjacency,
PrimitiveTopology.Triangles or
PrimitiveTopology.TriangleStrip or
PrimitiveTopology.TriangleFan => InputTopology.Triangles,
PrimitiveTopology.TrianglesAdjacency or
PrimitiveTopology.TriangleStripAdjacency => InputTopology.TrianglesAdjacency,
_ => InputTopology.Points,
};
}
/// <summary>
/// Queries host storage buffer alignment required.
/// </summary>
/// <returns>Host storage buffer alignment in bytes</returns>
public int QueryStorageBufferOffsetAlignment() => _context.Capabilities.StorageBufferOffsetAlignment;
/// <summary>
/// Queries host support for readable images without a explicit format declaration on the shader.
/// </summary>
/// <returns>True if formatted image load is supported, false otherwise</returns>
public bool QuerySupportsImageLoadFormatted() => _context.Capabilities.SupportsImageLoadFormatted;
/// <summary>
/// Queries host GPU non-constant texture offset support.
/// </summary>
/// <returns>True if the GPU and driver supports non-constant texture offsets, false otherwise</returns>
public bool QuerySupportsNonConstantTextureOffset() => _context.Capabilities.SupportsNonConstantTextureOffset;
/// <summary>
/// Queries host GPU texture shadow LOD support.
/// </summary>
/// <returns>True if the GPU and driver supports texture shadow LOD, false otherwise</returns>
public bool QuerySupportsTextureShadowLod() => _context.Capabilities.SupportsTextureShadowLod;
/// <summary>
/// Gets the texture descriptor for a given texture on the pool.
/// </summary>
/// <param name="handle">Index of the texture (this is the word offset of the handle in the constant buffer)</param>
/// <param name="cbufSlot">Constant buffer slot for the texture handle</param>
/// <returns>Texture descriptor</returns>
public override Image.ITextureDescriptor GetTextureDescriptor(int handle, int cbufSlot)
{
if (_compute)
{
return _channel.TextureManager.GetComputeTextureDescriptor(
_state.TexturePoolGpuVa,
_state.TextureBufferIndex,
_state.TexturePoolMaximumId,
handle,
cbufSlot);
}
else
{
return _channel.TextureManager.GetGraphicsTextureDescriptor(
_state.TexturePoolGpuVa,
_state.TextureBufferIndex,
_state.TexturePoolMaximumId,
_stageIndex,
handle,
cbufSlot);
}
}
/// <summary>
/// Queries if host state forces early depth testing.
/// </summary>
/// <returns>True if early depth testing is forced</returns>
public bool QueryEarlyZForce()
{
return _state.EarlyZForce;
}
}
}