ryujinx/Ryujinx.Memory/Tracking/MemoryTracking.cs
riperiperi ec3e848d79
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation

About as fast as nvidia GL multithreading, can be improved with faster command queuing.

* Struct based command list

Speeds up a bit. Still a lot of time lost to resource copy.

* Do shader init while the render thread is active.

* Introduce circular span pool V1

Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.

* Refactor SpanRef some more

Use a struct to represent SpanRef, rather than a reference.

* Flush buffers on background thread

* Use a span for UpdateRenderScale.

Much faster than copying the array.

* Calculate command size using reflection

* WIP parallel shaders

* Some minor optimisation

* Only 2 max refs per command now.

The command with 3 refs is gone. 😌

* Don't cast on the GPU side

* Remove redundant casts, force sync on window present

* Fix Shader Cache

* Fix host shader save.

* Fixup to work with new renderer stuff

* Make command Run static, use array of delegates as lookup

Profile says this takes less time than the previous way.

* Bring up to date

* Add settings toggle. Fix Muiltithreading Off mode.

* Fix warning.

* Release tracking lock for flushes

* Fix Conditional Render fast path with threaded gal

* Make handle iteration safe when releasing the lock

This is mostly temporary.

* Attempt to set backend threading on driver

Only really works on nvidia before launching a game.

* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions

* Update buffer set commands

* Some cleanup

* Only use stutter workaround when using opengl renderer non-threaded

* Add host-conditional reservation of counter events

There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.

* Address Feedback

* Make counter flush tracked again.

Hopefully does not cause any issues this time.

* Wait for FlushTo on the main queue thread.

Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)

* Add SDL2 headless integration

* Add HLE macro commands.

Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00

258 lines
11 KiB
C#

using Ryujinx.Common.Pools;
using Ryujinx.Memory.Range;
using System;
using System.Collections.Generic;
namespace Ryujinx.Memory.Tracking
{
/// <summary>
/// Manages memory tracking for a given virutal/physical memory block.
/// </summary>
public class MemoryTracking
{
private readonly IVirtualMemoryManager _memoryManager;
private readonly InvalidAccessHandler _invalidAccessHandler;
// Only use these from within the lock.
private readonly NonOverlappingRangeList<VirtualRegion> _virtualRegions;
private readonly int _pageSize;
/// <summary>
/// This lock must be obtained when traversing or updating the region-handle hierarchy.
/// It is not required when reading dirty flags.
/// </summary>
internal object TrackingLock = new object();
/// <summary>
/// Create a new tracking structure for the given "physical" memory block,
/// with a given "virtual" memory manager that will provide mappings and virtual memory protection.
/// </summary>
/// <param name="memoryManager">Virtual memory manager</param>
/// <param name="block">Physical memory block</param>
/// <param name="pageSize">Page size of the virtual memory space</param>
public MemoryTracking(IVirtualMemoryManager memoryManager, int pageSize, InvalidAccessHandler invalidAccessHandler = null)
{
_memoryManager = memoryManager;
_pageSize = pageSize;
_invalidAccessHandler = invalidAccessHandler;
_virtualRegions = new NonOverlappingRangeList<VirtualRegion>();
}
private (ulong address, ulong size) PageAlign(ulong address, ulong size)
{
ulong pageMask = (ulong)_pageSize - 1;
ulong rA = address & ~pageMask;
ulong rS = ((address + size + pageMask) & ~pageMask) - rA;
return (rA, rS);
}
/// <summary>
/// Indicate that a virtual region has been mapped, and which physical region it has been mapped to.
/// Should be called after the mapping is complete.
/// </summary>
/// <param name="va">Virtual memory address</param>
/// <param name="size">Size to be mapped</param>
public void Map(ulong va, ulong size)
{
// A mapping may mean we need to re-evaluate each VirtualRegion's affected area.
// Find all handles that overlap with the range, we need to recalculate their physical regions
lock (TrackingLock)
{
ref var overlaps = ref ThreadStaticArray<VirtualRegion>.Get();
int count = _virtualRegions.FindOverlapsNonOverlapping(va, size, ref overlaps);
for (int i = 0; i < count; i++)
{
VirtualRegion region = overlaps[i];
// If the region has been fully remapped, signal that it has been mapped again.
bool remapped = _memoryManager.IsRangeMapped(region.Address, region.Size);
if (remapped)
{
region.SignalMappingChanged(true);
}
region.UpdateProtection();
}
}
}
/// <summary>
/// Indicate that a virtual region has been unmapped.
/// Should be called before the unmapping is complete.
/// </summary>
/// <param name="va">Virtual memory address</param>
/// <param name="size">Size to be unmapped</param>
public void Unmap(ulong va, ulong size)
{
// An unmapping may mean we need to re-evaluate each VirtualRegion's affected area.
// Find all handles that overlap with the range, we need to notify them that the region was unmapped.
lock (TrackingLock)
{
ref var overlaps = ref ThreadStaticArray<VirtualRegion>.Get();
int count = _virtualRegions.FindOverlapsNonOverlapping(va, size, ref overlaps);
for (int i = 0; i < count; i++)
{
VirtualRegion region = overlaps[i];
region.SignalMappingChanged(false);
}
}
}
/// <summary>
/// Get a list of virtual regions that a handle covers.
/// </summary>
/// <param name="va">Starting virtual memory address of the handle</param>
/// <param name="size">Size of the handle's memory region</param>
/// <returns>A list of virtual regions within the given range</returns>
internal List<VirtualRegion> GetVirtualRegionsForHandle(ulong va, ulong size)
{
List<VirtualRegion> result = new List<VirtualRegion>();
_virtualRegions.GetOrAddRegions(result, va, size, (va, size) => new VirtualRegion(this, va, size));
return result;
}
/// <summary>
/// Remove a virtual region from the range list. This assumes that the lock has been acquired.
/// </summary>
/// <param name="region">Region to remove</param>
internal void RemoveVirtual(VirtualRegion region)
{
_virtualRegions.Remove(region);
}
/// <summary>
/// Obtains a memory tracking handle for the given virtual region, with a specified granularity. This should be disposed when finished with.
/// </summary>
/// <param name="address">CPU virtual address of the region</param>
/// <param name="size">Size of the region</param>
/// <param name="handles">Handles to inherit state from or reuse. When none are present, provide null</param>
/// <param name="granularity">Desired granularity of write tracking</param>
/// <returns>The memory tracking handle</returns>
public MultiRegionHandle BeginGranularTracking(ulong address, ulong size, IEnumerable<IRegionHandle> handles, ulong granularity)
{
(address, size) = PageAlign(address, size);
return new MultiRegionHandle(this, address, size, handles, granularity);
}
/// <summary>
/// Obtains a smart memory tracking handle for the given virtual region, with a specified granularity. This should be disposed when finished with.
/// </summary>
/// <param name="address">CPU virtual address of the region</param>
/// <param name="size">Size of the region</param>
/// <param name="granularity">Desired granularity of write tracking</param>
/// <returns>The memory tracking handle</returns>
public SmartMultiRegionHandle BeginSmartGranularTracking(ulong address, ulong size, ulong granularity)
{
(address, size) = PageAlign(address, size);
return new SmartMultiRegionHandle(this, address, size, granularity);
}
/// <summary>
/// Obtains a memory tracking handle for the given virtual region. This should be disposed when finished with.
/// </summary>
/// <param name="address">CPU virtual address of the region</param>
/// <param name="size">Size of the region</param>
/// <returns>The memory tracking handle</returns>
public RegionHandle BeginTracking(ulong address, ulong size)
{
(address, size) = PageAlign(address, size);
lock (TrackingLock)
{
RegionHandle handle = new RegionHandle(this, address, size, _memoryManager.IsRangeMapped(address, size));
return handle;
}
}
/// <summary>
/// Signal that a virtual memory event happened at the given location (one byte).
/// </summary>
/// <param name="address">Virtual address accessed</param>
/// <param name="write">Whether the address was written to or read</param>
/// <returns>True if the event triggered any tracking regions, false otherwise</returns>
public bool VirtualMemoryEventTracking(ulong address, bool write)
{
return VirtualMemoryEvent(address, 1, write);
}
/// <summary>
/// Signal that a virtual memory event happened at the given location.
/// </summary>
/// <param name="address">Virtual address accessed</param>
/// <param name="size">Size of the region affected in bytes</param>
/// <param name="write">Whether the region was written to or read</param>
/// <returns>True if the event triggered any tracking regions, false otherwise</returns>
public bool VirtualMemoryEvent(ulong address, ulong size, bool write)
{
// Look up the virtual region using the region list.
// Signal up the chain to relevant handles.
lock (TrackingLock)
{
ref var overlaps = ref ThreadStaticArray<VirtualRegion>.Get();
int count = _virtualRegions.FindOverlapsNonOverlapping(address, size, ref overlaps);
if (count == 0)
{
if (!_memoryManager.IsMapped(address))
{
_invalidAccessHandler?.Invoke(address);
// We can't continue - it's impossible to remove protection from the page.
// Even if the access handler wants us to continue, we wouldn't be able to.
throw new InvalidMemoryRegionException();
}
_memoryManager.TrackingReprotect(address & ~(ulong)(_pageSize - 1), (ulong)_pageSize, MemoryPermission.ReadAndWrite);
return false; // We can't handle this - it's probably a real invalid access.
}
for (int i = 0; i < count; i++)
{
VirtualRegion region = overlaps[i];
region.Signal(address, size, write);
}
}
return true;
}
/// <summary>
/// Reprotect a given virtual region. The virtual memory manager will handle this.
/// </summary>
/// <param name="region">Region to reprotect</param>
/// <param name="permission">Memory permission to protect with</param>
internal void ProtectVirtualRegion(VirtualRegion region, MemoryPermission permission)
{
_memoryManager.TrackingReprotect(region.Address, region.Size, permission);
}
/// <summary>
/// Returns the number of virtual regions currently being tracked.
/// Useful for tests and metrics.
/// </summary>
/// <returns>The number of virtual regions</returns>
public int GetRegionCount()
{
lock (TrackingLock)
{
return _virtualRegions.Count;
}
}
}
}