Ryujinx/Ryujinx.Graphics.Gpu/Engine/GPFifo/GPFifoClass.cs
riperiperi e20abbf9cc
Vulkan: Don't flush commands when creating most sync (#4087)
* Vulkan: Don't flush commands when creating most sync

When the WaitForIdle method is called, we create sync as some internal GPU method may read back written buffer data. Some games randomly intersperse compute dispatch into their render passes, which result in this happening an unbounded number of times depending on how many times they run compute.

Creating sync in Vulkan is expensive, as we need to flush the current command buffer so that it can be waited on. We have a limited number of active command buffers due to how we track resource usage, so submitting too many command buffers will force us to wait for them to return to the pool.

This PR allows less "important" sync (things which are less likely to be waited on) to wait on a command buffer's result without submitting it, instead relying on AutoFlush or another, more important sync to flush it later on.

Because of the possibility of us waiting for a command buffer that hasn't submitted yet, any thread needs to be able to force the active command buffer to submit. The ability to do this has been added to the backend multithreading via an "Interrupt", though it is not supported without multithreading.

OpenGL drivers should already be doing something similar so they don't blow up when creating lots of sync, which is why this hasn't been a problem for these games over there.

Improves Vulkan performance on Xenoblade DE, Pokemon Scarlet/Violet, and Zelda BOTW (still another large issue here)

* Add strict argument

This is technically a separate concern from whether the sync is a host syncpoint.

* Remove _interrupted variable

* Actually wait for the invoke

This is required by AMD GPUs, and also may have caused some issues on other GPUs.

* Remove unused using.

* I don't know why it added these ones.

* Address Feedback

* Fix typo
2022-12-29 15:39:04 +01:00

249 lines
9.3 KiB
C#

using Ryujinx.Graphics.Device;
using Ryujinx.Graphics.Gpu.Engine.MME;
using System;
using System.Collections.Generic;
using System.Threading;
namespace Ryujinx.Graphics.Gpu.Engine.GPFifo
{
/// <summary>
/// Represents a GPU General Purpose FIFO class.
/// </summary>
class GPFifoClass : IDeviceState
{
private readonly GpuContext _context;
private readonly GPFifoProcessor _parent;
private readonly DeviceState<GPFifoClassState> _state;
private int _previousSubChannel;
private bool _createSyncPending;
private const int MacrosCount = 0x80;
// Note: The size of the macro memory is unknown, we just make
// a guess here and use 256kb as the size. Increase if needed.
private const int MacroCodeSize = 256 * 256;
private readonly Macro[] _macros;
private readonly int[] _macroCode;
/// <summary>
/// Creates a new instance of the GPU General Purpose FIFO class.
/// </summary>
/// <param name="context">GPU context</param>
/// <param name="parent">Parent GPU General Purpose FIFO processor</param>
public GPFifoClass(GpuContext context, GPFifoProcessor parent)
{
_context = context;
_parent = parent;
_state = new DeviceState<GPFifoClassState>(new Dictionary<string, RwCallback>
{
{ nameof(GPFifoClassState.Semaphored), new RwCallback(Semaphored, null) },
{ nameof(GPFifoClassState.Syncpointb), new RwCallback(Syncpointb, null) },
{ nameof(GPFifoClassState.WaitForIdle), new RwCallback(WaitForIdle, null) },
{ nameof(GPFifoClassState.SetReference), new RwCallback(SetReference, null) },
{ nameof(GPFifoClassState.LoadMmeInstructionRam), new RwCallback(LoadMmeInstructionRam, null) },
{ nameof(GPFifoClassState.LoadMmeStartAddressRam), new RwCallback(LoadMmeStartAddressRam, null) },
{ nameof(GPFifoClassState.SetMmeShadowRamControl), new RwCallback(SetMmeShadowRamControl, null) }
});
_macros = new Macro[MacrosCount];
_macroCode = new int[MacroCodeSize];
}
/// <summary>
/// Create any syncs from WaitForIdle command that are currently pending.
/// </summary>
public void CreatePendingSyncs()
{
if (_createSyncPending)
{
_createSyncPending = false;
_context.CreateHostSyncIfNeeded(false, false);
}
}
/// <summary>
/// Reads data from the class registers.
/// </summary>
/// <param name="offset">Register byte offset</param>
/// <returns>Data at the specified offset</returns>
public int Read(int offset) => _state.Read(offset);
/// <summary>
/// Writes data to the class registers.
/// </summary>
/// <param name="offset">Register byte offset</param>
/// <param name="data">Data to be written</param>
public void Write(int offset, int data) => _state.Write(offset, data);
/// <summary>
/// Writes a GPU counter to guest memory.
/// </summary>
/// <param name="argument">Method call argument</param>
public void Semaphored(int argument)
{
ulong address = ((ulong)_state.State.SemaphorebOffsetLower << 2) |
((ulong)_state.State.SemaphoreaOffsetUpper << 32);
int value = _state.State.SemaphorecPayload;
SemaphoredOperation operation = _state.State.SemaphoredOperation;
if (_state.State.SemaphoredReleaseSize == SemaphoredReleaseSize.SixteenBytes)
{
_parent.MemoryManager.Write(address + 4, 0);
_parent.MemoryManager.Write(address + 8, _context.GetTimestamp());
}
// TODO: Acquire operations (Wait), interrupts for invalid combinations.
if (operation == SemaphoredOperation.Release)
{
_parent.MemoryManager.Write(address, value);
}
else if (operation == SemaphoredOperation.Reduction)
{
bool signed = _state.State.SemaphoredFormat == SemaphoredFormat.Signed;
int mem = _parent.MemoryManager.Read<int>(address);
switch (_state.State.SemaphoredReduction)
{
case SemaphoredReduction.Min:
value = signed ? Math.Min(mem, value) : (int)Math.Min((uint)mem, (uint)value);
break;
case SemaphoredReduction.Max:
value = signed ? Math.Max(mem, value) : (int)Math.Max((uint)mem, (uint)value);
break;
case SemaphoredReduction.Xor:
value ^= mem;
break;
case SemaphoredReduction.And:
value &= mem;
break;
case SemaphoredReduction.Or:
value |= mem;
break;
case SemaphoredReduction.Add:
value += mem;
break;
case SemaphoredReduction.Inc:
value = (uint)mem < (uint)value ? mem + 1 : 0;
break;
case SemaphoredReduction.Dec:
value = (uint)mem > 0 && (uint)mem <= (uint)value ? mem - 1 : value;
break;
}
_parent.MemoryManager.Write(address, value);
}
}
/// <summary>
/// Apply a fence operation on a syncpoint.
/// </summary>
/// <param name="argument">Method call argument</param>
public void Syncpointb(int argument)
{
SyncpointbOperation operation = _state.State.SyncpointbOperation;
uint syncpointId = (uint)_state.State.SyncpointbSyncptIndex;
if (operation == SyncpointbOperation.Wait)
{
uint threshold = (uint)_state.State.SyncpointaPayload;
_context.Synchronization.WaitOnSyncpoint(syncpointId, threshold, Timeout.InfiniteTimeSpan);
}
else if (operation == SyncpointbOperation.Incr)
{
_context.CreateHostSyncIfNeeded(true, true);
_context.Synchronization.IncrementSyncpoint(syncpointId);
}
_context.AdvanceSequence();
}
/// <summary>
/// Waits for the GPU to be idle.
/// </summary>
/// <param name="argument">Method call argument</param>
public void WaitForIdle(int argument)
{
_parent.PerformDeferredDraws();
_context.Renderer.Pipeline.Barrier();
_createSyncPending = true;
}
/// <summary>
/// Used as an indirect data barrier on NVN. When used, access to previously written data must be coherent.
/// </summary>
/// <param name="argument">Method call argument</param>
public void SetReference(int argument)
{
_context.Renderer.Pipeline.CommandBufferBarrier();
_context.CreateHostSyncIfNeeded(false, true);
}
/// <summary>
/// Sends macro code/data to the MME.
/// </summary>
/// <param name="argument">Method call argument</param>
public void LoadMmeInstructionRam(int argument)
{
_macroCode[_state.State.LoadMmeInstructionRamPointer++] = argument;
}
/// <summary>
/// Binds a macro index to a position for the MME
/// </summary>
/// <param name="argument">Method call argument</param>
public void LoadMmeStartAddressRam(int argument)
{
_macros[_state.State.LoadMmeStartAddressRamPointer++] = new Macro(argument);
}
/// <summary>
/// Changes the shadow RAM control.
/// </summary>
/// <param name="argument">Method call argument</param>
public void SetMmeShadowRamControl(int argument)
{
_parent.SetShadowRamControl(argument);
}
/// <summary>
/// Pushes an argument to a macro.
/// </summary>
/// <param name="index">Index of the macro</param>
/// <param name="gpuVa">GPU virtual address where the command word is located</param>
/// <param name="argument">Argument to be pushed to the macro</param>
public void MmePushArgument(int index, ulong gpuVa, int argument)
{
_macros[index].PushArgument(gpuVa, argument);
}
/// <summary>
/// Prepares a macro for execution.
/// </summary>
/// <param name="index">Index of the macro</param>
/// <param name="argument">Initial argument passed to the macro</param>
public void MmeStart(int index, int argument)
{
_macros[index].StartExecution(_context, _parent, _macroCode, argument);
}
/// <summary>
/// Executes a macro.
/// </summary>
/// <param name="index">Index of the macro</param>
/// <param name="state">Current GPU state</param>
public void CallMme(int index, IDeviceState state)
{
_macros[index].Execute(_macroCode, state);
}
}
}