R/Ryujinx.Graphics.Gpu/GpuChannel.cs

149 lines
5.5 KiB
C#
Raw Normal View History

Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu.Engine.GPFifo;
using Ryujinx.Graphics.Gpu.Image;
using Ryujinx.Graphics.Gpu.Memory;
using System;
using System.Threading;
namespace Ryujinx.Graphics.Gpu
{
/// <summary>
/// Represents a GPU channel.
/// </summary>
public class GpuChannel : IDisposable
{
private readonly GpuContext _context;
private readonly GPFifoDevice _device;
private readonly GPFifoProcessor _processor;
private MemoryManager _memoryManager;
/// <summary>
/// Channel buffer bindings manager.
/// </summary>
internal BufferManager BufferManager { get; }
/// <summary>
/// Channel texture bindings manager.
/// </summary>
internal TextureManager TextureManager { get; }
/// <summary>
/// Current channel memory manager.
/// </summary>
internal MemoryManager MemoryManager => _memoryManager;
Vulkan: Add workarounds for MoltenVK (#4202) * Add MVK basics. * Use appropriate output attribute types * 4kb vertex alignment, bunch of fixes * Add reduced shader precision mode for mvk. * Disable ASTC on MVK for now * Only request robustnes2 when it is available. * It's just the one feature actually * Add triangle fan conversion * Allow NullDescriptor on MVK for some reason. * Force safe blit on MoltenVK * Use ASTC only when formats are all available. * Disable multilevel 3d texture views * Filter duplicate render targets (on backend) * Add Automatic MoltenVK Configuration * Do not create color attachment views with formats that are not RT compatible * Make sure that the host format matches the vertex shader input types for invalid/unknown guest formats * FIx rebase for Vertex Attrib State * Fix 4b alignment for vertex * Use asynchronous queue submits for MVK * Ensure color clear shader has correct output type * Update MoltenVK config * Always use MoltenVK workarounds on MacOS * Make MVK supersede all vendors * Fix rebase * Various fixes on rebase * Get portability flags from extension * Fix some minor rebasing issues * Style change * Use LibraryImport for MVKConfiguration * Rename MoltenVK vendor to Apple Intel and AMD GPUs on moltenvk report with the those vendors - only apple silicon reports with vendor 0x106B. * Fix features2 rebase conflict * Rename fragment output type * Add missing check for fragment output types Might have caused the crash in MK8 * Only do fragment output specialization on MoltenVK * Avoid copy when passing capabilities * Self feedback * Address feedback Co-authored-by: gdk <gab.dark.100@gmail.com> Co-authored-by: nastys <nastys@users.noreply.github.com>
2023-01-13 00:31:21 +00:00
/// <summary>
/// Host hardware capabilities from the GPU context.
/// </summary>
internal ref Capabilities Capabilities => ref _context.Capabilities;
/// <summary>
/// Creates a new instance of a GPU channel.
/// </summary>
/// <param name="context">GPU context that the channel belongs to</param>
internal GpuChannel(GpuContext context)
{
_context = context;
_device = context.GPFifo;
_processor = new GPFifoProcessor(context, this);
BufferManager = new BufferManager(context, this);
TextureManager = new TextureManager(context, this);
}
/// <summary>
/// Binds a memory manager to the channel.
/// All submitted and in-flight commands will use the specified memory manager for any memory operations.
/// </summary>
/// <param name="memoryManager">The new memory manager to be bound</param>
public void BindMemory(MemoryManager memoryManager)
{
var oldMemoryManager = Interlocked.Exchange(ref _memoryManager, memoryManager ?? throw new ArgumentNullException(nameof(memoryManager)));
memoryManager.Physical.IncrementReferenceCount();
if (oldMemoryManager != null)
{
oldMemoryManager.Physical.BufferCache.NotifyBuffersModified -= BufferManager.Rebind;
oldMemoryManager.Physical.DecrementReferenceCount();
oldMemoryManager.MemoryUnmapped -= MemoryUnmappedHandler;
}
memoryManager.Physical.BufferCache.NotifyBuffersModified += BufferManager.Rebind;
memoryManager.MemoryUnmapped += MemoryUnmappedHandler;
// Since the memory manager changed, make sure we will get pools from addresses of the new memory manager.
TextureManager.ReloadPools();
memoryManager.Physical.BufferCache.QueuePrune();
}
/// <summary>
/// Memory mappings change event handler.
/// </summary>
/// <param name="sender">Memory manager where the mappings changed</param>
/// <param name="e">Information about the region that is being changed</param>
private void MemoryUnmappedHandler(object sender, UnmapEventArgs e)
{
TextureManager.ReloadPools();
var memoryManager = Volatile.Read(ref _memoryManager);
memoryManager?.Physical.BufferCache.QueuePrune();
}
/// <summary>
/// Writes data directly to the state of the specified class.
/// </summary>
/// <param name="classId">ID of the class to write the data into</param>
/// <param name="offset">State offset in bytes</param>
/// <param name="value">Value to be written</param>
public void Write(ClassId classId, int offset, uint value)
{
_processor.Write(classId, offset, (int)value);
}
/// <summary>
/// Push a GPFIFO entry in the form of a prefetched command buffer.
/// It is intended to be used by nvservices to handle special cases.
/// </summary>
/// <param name="commandBuffer">The command buffer containing the prefetched commands</param>
public void PushHostCommandBuffer(int[] commandBuffer)
{
_device.PushHostCommandBuffer(_processor, commandBuffer);
}
/// <summary>
/// Pushes GPFIFO entries.
/// </summary>
/// <param name="entries">GPFIFO entries</param>
public void PushEntries(ReadOnlySpan<ulong> entries)
{
_device.PushEntries(_processor, entries);
}
/// <summary>
/// Disposes the GPU channel.
/// It's an error to use the GPU channel after disposal.
/// </summary>
public void Dispose()
{
_context.DeferredActions.Enqueue(Destroy);
}
/// <summary>
/// Performs disposal of the host GPU resources used by this channel, that are not shared.
/// This must only be called from the render thread.
/// </summary>
private void Destroy()
{
TextureManager.Dispose();
var oldMemoryManager = Interlocked.Exchange(ref _memoryManager, null);
if (oldMemoryManager != null)
{
oldMemoryManager.Physical.BufferCache.NotifyBuffersModified -= BufferManager.Rebind;
oldMemoryManager.Physical.DecrementReferenceCount();
oldMemoryManager.MemoryUnmapped -= MemoryUnmappedHandler;
}
}
}
}