POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
using ARMeilleure.Memory;
|
|
|
|
|
|
using Ryujinx.Cpu.Tracking;
|
|
|
|
|
|
using Ryujinx.Memory;
|
|
|
|
|
|
using Ryujinx.Memory.Range;
|
|
|
|
|
|
using Ryujinx.Memory.Tracking;
|
|
|
|
|
|
using System;
|
|
|
|
|
|
using System.Collections.Generic;
|
|
|
|
|
|
using System.Linq;
|
|
|
|
|
|
using System.Runtime.CompilerServices;
|
|
|
|
|
|
using System.Threading;
|
|
|
|
|
|
|
|
|
|
|
|
namespace Ryujinx.Cpu
|
|
|
|
|
|
{
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Represents a CPU memory manager which maps guest virtual memory directly onto a host virtual region.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
public class MemoryManagerHostMapped : MemoryManagerBase, IMemoryManager, IVirtualMemoryManagerTracked
|
|
|
|
|
|
{
|
|
|
|
|
|
public const int PageBits = 12;
|
|
|
|
|
|
public const int PageSize = 1 << PageBits;
|
|
|
|
|
|
public const int PageMask = PageSize - 1;
|
|
|
|
|
|
|
|
|
|
|
|
public const int PageToPteShift = 5; // 32 pages (2 bits each) in one ulong page table entry.
|
|
|
|
|
|
public const ulong BlockMappedMask = 0x5555555555555555; // First bit of each table entry set.
|
|
|
|
|
|
|
|
|
|
|
|
private enum HostMappedPtBits : ulong
|
|
|
|
|
|
{
|
|
|
|
|
|
Unmapped = 0,
|
|
|
|
|
|
Mapped,
|
|
|
|
|
|
WriteTracked,
|
|
|
|
|
|
ReadWriteTracked,
|
|
|
|
|
|
|
|
|
|
|
|
MappedReplicated = 0x5555555555555555,
|
|
|
|
|
|
WriteTrackedReplicated = 0xaaaaaaaaaaaaaaaa,
|
|
|
|
|
|
ReadWriteTrackedReplicated = ulong.MaxValue
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
private readonly InvalidAccessHandler _invalidAccessHandler;
|
|
|
|
|
|
private readonly bool _unsafeMode;
|
|
|
|
|
|
|
|
|
|
|
|
private readonly MemoryBlock _addressSpace;
|
|
|
|
|
|
private readonly MemoryBlock _addressSpaceMirror;
|
|
|
|
|
|
private readonly ulong _addressSpaceSize;
|
|
|
|
|
|
|
|
|
|
|
|
private readonly MemoryEhMeilleure _memoryEh;
|
|
|
|
|
|
|
|
|
|
|
|
private ulong[] _pageTable;
|
|
|
|
|
|
|
|
|
|
|
|
public int AddressSpaceBits { get; }
|
|
|
|
|
|
|
|
|
|
|
|
public IntPtr PageTablePointer => _addressSpace.Pointer;
|
|
|
|
|
|
|
|
|
|
|
|
public MemoryManagerType Type => _unsafeMode ? MemoryManagerType.HostMappedUnsafe : MemoryManagerType.HostMapped;
|
|
|
|
|
|
|
|
|
|
|
|
public MemoryTracking Tracking { get; }
|
|
|
|
|
|
|
|
|
|
|
|
public event Action<ulong, ulong> UnmapEvent;
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Creates a new instance of the host mapped memory manager.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="addressSpaceSize">Size of the address space</param>
|
|
|
|
|
|
/// <param name="unsafeMode">True if unmanaged access should not be masked (unsafe), false otherwise.</param>
|
|
|
|
|
|
/// <param name="invalidAccessHandler">Optional function to handle invalid memory accesses</param>
|
|
|
|
|
|
public MemoryManagerHostMapped(ulong addressSpaceSize, bool unsafeMode, InvalidAccessHandler invalidAccessHandler = null)
|
|
|
|
|
|
{
|
|
|
|
|
|
_invalidAccessHandler = invalidAccessHandler;
|
|
|
|
|
|
_unsafeMode = unsafeMode;
|
|
|
|
|
|
_addressSpaceSize = addressSpaceSize;
|
|
|
|
|
|
|
|
|
|
|
|
ulong asSize = PageSize;
|
|
|
|
|
|
int asBits = PageBits;
|
|
|
|
|
|
|
|
|
|
|
|
while (asSize < addressSpaceSize)
|
|
|
|
|
|
{
|
|
|
|
|
|
asSize <<= 1;
|
|
|
|
|
|
asBits++;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
AddressSpaceBits = asBits;
|
|
|
|
|
|
|
|
|
|
|
|
_pageTable = new ulong[1 << (AddressSpaceBits - (PageBits + PageToPteShift))];
|
|
|
|
|
|
_addressSpace = new MemoryBlock(asSize, MemoryAllocationFlags.Reserve | MemoryAllocationFlags.Mirrorable);
|
|
|
|
|
|
_addressSpaceMirror = _addressSpace.CreateMirror();
|
|
|
|
|
|
Tracking = new MemoryTracking(this, PageSize, invalidAccessHandler);
|
|
|
|
|
|
_memoryEh = new MemoryEhMeilleure(_addressSpace, Tracking);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Checks if the virtual address is part of the addressable space.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual address</param>
|
|
|
|
|
|
/// <returns>True if the virtual address is part of the addressable space</returns>
|
|
|
|
|
|
private bool ValidateAddress(ulong va)
|
|
|
|
|
|
{
|
|
|
|
|
|
return va < _addressSpaceSize;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Checks if the combination of virtual address and size is part of the addressable space.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual address of the range</param>
|
|
|
|
|
|
/// <param name="size">Size of the range in bytes</param>
|
|
|
|
|
|
/// <returns>True if the combination of virtual address and size is part of the addressable space</returns>
|
|
|
|
|
|
private bool ValidateAddressAndSize(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong endVa = va + size;
|
|
|
|
|
|
return endVa >= va && endVa >= size && endVa <= _addressSpaceSize;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Ensures the combination of virtual address and size is part of the addressable space.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual address of the range</param>
|
|
|
|
|
|
/// <param name="size">Size of the range in bytes</param>
|
|
|
|
|
|
/// <exception cref="InvalidMemoryRegionException">Throw when the memory region specified outside the addressable space</exception>
|
|
|
|
|
|
private void AssertValidAddressAndSize(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (!ValidateAddressAndSize(va, size))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw new InvalidMemoryRegionException($"va=0x{va:X16}, size=0x{size:X16}");
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Ensures the combination of virtual address and size is part of the addressable space and fully mapped.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual address of the range</param>
|
|
|
|
|
|
/// <param name="size">Size of the range in bytes</param>
|
|
|
|
|
|
private void AssertMapped(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (!ValidateAddressAndSize(va, size) || !IsRangeMappedImpl(va, size))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw new InvalidMemoryRegionException($"Not mapped: va=0x{va:X16}, size=0x{size:X16}");
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void Map(ulong va, nuint hostAddress, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertValidAddressAndSize(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpace.Commit(va, size);
|
|
|
|
|
|
AddMapping(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
Tracking.Map(va, size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void Unmap(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertValidAddressAndSize(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
UnmapEvent?.Invoke(va, size);
|
|
|
|
|
|
Tracking.Unmap(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
RemoveMapping(va, size);
|
|
|
|
|
|
_addressSpace.Decommit(va, size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public T Read<T>(ulong va) where T : unmanaged
|
|
|
|
|
|
{
|
|
|
|
|
|
try
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertMapped(va, (ulong)Unsafe.SizeOf<T>());
|
|
|
|
|
|
|
|
|
|
|
|
return _addressSpaceMirror.Read<T>(va);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
return default;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public T ReadTracked<T>(ulong va) where T : unmanaged
|
|
|
|
|
|
{
|
|
|
|
|
|
try
|
|
|
|
|
|
{
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)Unsafe.SizeOf<T>(), false);
|
|
|
|
|
|
|
|
|
|
|
|
return Read<T>(va);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
return default;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void Read(ulong va, Span<byte> data)
|
|
|
|
|
|
{
|
|
|
|
|
|
try
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertMapped(va, (ulong)data.Length);
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpaceMirror.Read(va, data);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void Write<T>(ulong va, T value) where T : unmanaged
|
|
|
|
|
|
{
|
|
|
|
|
|
try
|
|
|
|
|
|
{
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)Unsafe.SizeOf<T>(), write: true);
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpaceMirror.Write(va, value);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void Write(ulong va, ReadOnlySpan<byte> data)
|
|
|
|
|
|
{
|
|
|
|
|
|
try {
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)data.Length, write: true);
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpaceMirror.Write(va, data);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void WriteUntracked(ulong va, ReadOnlySpan<byte> data)
|
|
|
|
|
|
{
|
|
|
|
|
|
try
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertMapped(va, (ulong)data.Length);
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpaceMirror.Write(va, data);
|
|
|
|
|
|
}
|
|
|
|
|
|
catch (InvalidMemoryRegionException)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (_invalidAccessHandler == null || !_invalidAccessHandler(va))
|
|
|
|
|
|
{
|
|
|
|
|
|
throw;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public ReadOnlySpan<byte> GetSpan(ulong va, int size, bool tracked = false)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (tracked)
|
|
|
|
|
|
{
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)size, write: false);
|
|
|
|
|
|
}
|
|
|
|
|
|
else
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertMapped(va, (ulong)size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
return _addressSpaceMirror.GetSpan(va, size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
Return mapped buffer pointer directly for flush, WriteableRegion for textures (#2494)
* Return mapped buffer pointer directly for flush, WriteableRegion for textures
A few changes here to generally improve performance, even for platforms not using the persistent buffer flush.
- Texture and buffer flush now return a ReadOnlySpan<byte>. It's guaranteed that this span is pinned in memory, but it will be overwritten on the next flush from that thread, so it is expected that the data is used before calling again.
- As a result, persistent mappings no longer copy to a new array - rather the persistent map is returned directly as a Span<>. A similar host array is used for the glGet flushes instead of allocating new arrays each time.
- Texture flushes now do their layout conversion into a WriteableRegion when the texture is not MultiRange, which allows the flush to happen directly into guest memory rather than into a temporary span, then copied over. This avoids another copy when doing layout conversion.
Overall, this saves 1 data copy for buffer flush, 1 copy for linear textures with matching source/target stride, and 2 copies for block textures or linear textures with mismatching strides.
* Fix tests
* Fix array pointer for Mesa/Intel path
* Address some feedback
* Update method for getting array pointer.
2021-07-19 23:10:54 +01:00
|
|
|
|
public WritableRegion GetWritableRegion(ulong va, int size, bool tracked = false)
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
{
|
Return mapped buffer pointer directly for flush, WriteableRegion for textures (#2494)
* Return mapped buffer pointer directly for flush, WriteableRegion for textures
A few changes here to generally improve performance, even for platforms not using the persistent buffer flush.
- Texture and buffer flush now return a ReadOnlySpan<byte>. It's guaranteed that this span is pinned in memory, but it will be overwritten on the next flush from that thread, so it is expected that the data is used before calling again.
- As a result, persistent mappings no longer copy to a new array - rather the persistent map is returned directly as a Span<>. A similar host array is used for the glGet flushes instead of allocating new arrays each time.
- Texture flushes now do their layout conversion into a WriteableRegion when the texture is not MultiRange, which allows the flush to happen directly into guest memory rather than into a temporary span, then copied over. This avoids another copy when doing layout conversion.
Overall, this saves 1 data copy for buffer flush, 1 copy for linear textures with matching source/target stride, and 2 copies for block textures or linear textures with mismatching strides.
* Fix tests
* Fix array pointer for Mesa/Intel path
* Address some feedback
* Update method for getting array pointer.
2021-07-19 23:10:54 +01:00
|
|
|
|
if (tracked)
|
|
|
|
|
|
{
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)size, true);
|
|
|
|
|
|
}
|
|
|
|
|
|
else
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertMapped(va, (ulong)size);
|
|
|
|
|
|
}
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
|
|
|
|
|
|
return _addressSpaceMirror.GetWritableRegion(va, size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public ref T GetRef<T>(ulong va) where T : unmanaged
|
|
|
|
|
|
{
|
|
|
|
|
|
SignalMemoryTracking(va, (ulong)Unsafe.SizeOf<T>(), true);
|
|
|
|
|
|
|
|
|
|
|
|
return ref _addressSpaceMirror.GetRef<T>(va);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
|
|
|
|
public bool IsMapped(ulong va)
|
|
|
|
|
|
{
|
|
|
|
|
|
return ValidateAddress(va) && IsMappedImpl(va);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
|
|
|
|
private bool IsMappedImpl(ulong va)
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong page = va >> PageBits;
|
|
|
|
|
|
|
|
|
|
|
|
int bit = (int)((page & 31) << 1);
|
|
|
|
|
|
|
|
|
|
|
|
int pageIndex = (int)(page >> PageToPteShift);
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
|
|
|
|
|
|
return ((pte >> bit) & 3) != 0;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public bool IsRangeMapped(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
AssertValidAddressAndSize(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
return IsRangeMappedImpl(va, size);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
|
|
|
|
private void GetPageBlockRange(ulong pageStart, ulong pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
startMask = ulong.MaxValue << ((int)(pageStart & 31) << 1);
|
|
|
|
|
|
endMask = ulong.MaxValue >> (64 - ((int)(pageEnd & 31) << 1));
|
|
|
|
|
|
|
|
|
|
|
|
pageIndex = (int)(pageStart >> PageToPteShift);
|
|
|
|
|
|
pageEndIndex = (int)((pageEnd - 1) >> PageToPteShift);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
private bool IsRangeMappedImpl(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
int pages = GetPagesCount(va, size, out _);
|
|
|
|
|
|
|
|
|
|
|
|
if (pages == 1)
|
|
|
|
|
|
{
|
|
|
|
|
|
return IsMappedImpl(va);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ulong pageStart = va >> PageBits;
|
|
|
|
|
|
ulong pageEnd = pageStart + (ulong)pages;
|
|
|
|
|
|
|
|
|
|
|
|
GetPageBlockRange(pageStart, pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex);
|
|
|
|
|
|
|
|
|
|
|
|
// Check if either bit in each 2 bit page entry is set.
|
|
|
|
|
|
// OR the block with itself shifted down by 1, and check the first bit of each entry.
|
|
|
|
|
|
|
|
|
|
|
|
ulong mask = BlockMappedMask & startMask;
|
|
|
|
|
|
|
|
|
|
|
|
while (pageIndex <= pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (pageIndex == pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
mask &= endMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex++];
|
|
|
|
|
|
ulong pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
|
|
|
|
|
|
pte |= pte >> 1;
|
|
|
|
|
|
if ((pte & mask) != mask)
|
|
|
|
|
|
{
|
|
|
|
|
|
return false;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
mask = BlockMappedMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
return true;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public IEnumerable<HostMemoryRange> GetPhysicalRegions(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (size == 0)
|
|
|
|
|
|
{
|
|
|
|
|
|
return Enumerable.Empty<HostMemoryRange>();
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
AssertMapped(va, size);
|
|
|
|
|
|
|
|
|
|
|
|
return new HostMemoryRange[] { new HostMemoryRange(_addressSpaceMirror.GetPointer(va, size), size) };
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
/// <remarks>
|
|
|
|
|
|
/// This function also validates that the given range is both valid and mapped, and will throw if it is not.
|
|
|
|
|
|
/// </remarks>
|
Replace CacheResourceWrite with more general "precise" write (#2684)
* Replace CacheResourceWrite with more general "precise" write
The goal of CacheResourceWrite was to notify GPU resources when they were modified directly, by looking up the modified address/size in a structure and calling a method on each resource. The downside of this is that each resource cache has to be queried individually, they all have to implement their own way to do this, and it can only signal to resources using the same PhysicalMemory instance.
This PR adds the ability to signal a write as "precise" on the tracking, which signals a special handler (if present) which can be used to avoid unnecessary flush actions, or maybe even more. For buffers, precise writes specifically do not flush, and instead punch a hole in the modified range list to indicate that the data on GPU has been replaced.
The downside is that precise actions must ignore the page protection bits and always signal - as they need to notify the target resource to ignore the sequence number optimization.
I had to reintroduce the sequence number increment after I2M, as removing it was causing issues in rabbids kingdom battle. However - all resources modified by I2M are notified directly to lower their sequence number, so the problem is likely that another unrelated resource is not being properly updated. Thankfully, doing this does not affect performance in the games I tested.
This should fix regressions from #2624. Test any games that were broken by that. (RF4, rabbids kingdom battle)
I've also added a sequence number increment to ThreedClass.IncrementSyncpoint, as it seems to fix buffer corruption in OpenGL homebrew. (this was a regression from removing sequence number increment from constant buffer update - another unrelated resource thing)
* Add tests.
* Add XML docs for GpuRegionHandle
* Skip UpdateProtection if only precise actions were called
This allows precise actions to skip reprotection costs.
2021-09-29 01:27:03 +01:00
|
|
|
|
public void SignalMemoryTracking(ulong va, ulong size, bool write, bool precise = false)
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
{
|
|
|
|
|
|
AssertValidAddressAndSize(va, size);
|
|
|
|
|
|
|
Replace CacheResourceWrite with more general "precise" write (#2684)
* Replace CacheResourceWrite with more general "precise" write
The goal of CacheResourceWrite was to notify GPU resources when they were modified directly, by looking up the modified address/size in a structure and calling a method on each resource. The downside of this is that each resource cache has to be queried individually, they all have to implement their own way to do this, and it can only signal to resources using the same PhysicalMemory instance.
This PR adds the ability to signal a write as "precise" on the tracking, which signals a special handler (if present) which can be used to avoid unnecessary flush actions, or maybe even more. For buffers, precise writes specifically do not flush, and instead punch a hole in the modified range list to indicate that the data on GPU has been replaced.
The downside is that precise actions must ignore the page protection bits and always signal - as they need to notify the target resource to ignore the sequence number optimization.
I had to reintroduce the sequence number increment after I2M, as removing it was causing issues in rabbids kingdom battle. However - all resources modified by I2M are notified directly to lower their sequence number, so the problem is likely that another unrelated resource is not being properly updated. Thankfully, doing this does not affect performance in the games I tested.
This should fix regressions from #2624. Test any games that were broken by that. (RF4, rabbids kingdom battle)
I've also added a sequence number increment to ThreedClass.IncrementSyncpoint, as it seems to fix buffer corruption in OpenGL homebrew. (this was a regression from removing sequence number increment from constant buffer update - another unrelated resource thing)
* Add tests.
* Add XML docs for GpuRegionHandle
* Skip UpdateProtection if only precise actions were called
This allows precise actions to skip reprotection costs.
2021-09-29 01:27:03 +01:00
|
|
|
|
if (precise)
|
|
|
|
|
|
{
|
|
|
|
|
|
Tracking.VirtualMemoryEvent(va, size, write, precise: true);
|
|
|
|
|
|
return;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
// Software table, used for managed memory tracking.
|
|
|
|
|
|
|
|
|
|
|
|
int pages = GetPagesCount(va, size, out _);
|
|
|
|
|
|
ulong pageStart = va >> PageBits;
|
|
|
|
|
|
|
|
|
|
|
|
if (pages == 1)
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong tag = (ulong)(write ? HostMappedPtBits.WriteTracked : HostMappedPtBits.ReadWriteTracked);
|
|
|
|
|
|
|
|
|
|
|
|
int bit = (int)((pageStart & 31) << 1);
|
|
|
|
|
|
|
|
|
|
|
|
int pageIndex = (int)(pageStart >> PageToPteShift);
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
ulong state = ((pte >> bit) & 3);
|
|
|
|
|
|
|
|
|
|
|
|
if (state >= tag)
|
|
|
|
|
|
{
|
|
|
|
|
|
Tracking.VirtualMemoryEvent(va, size, write);
|
|
|
|
|
|
return;
|
|
|
|
|
|
}
|
|
|
|
|
|
else if (state == 0)
|
|
|
|
|
|
{
|
|
|
|
|
|
ThrowInvalidMemoryRegionException($"Not mapped: va=0x{va:X16}, size=0x{size:X16}");
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
else
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong pageEnd = pageStart + (ulong)pages;
|
|
|
|
|
|
|
|
|
|
|
|
GetPageBlockRange(pageStart, pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex);
|
|
|
|
|
|
|
|
|
|
|
|
ulong mask = startMask;
|
|
|
|
|
|
|
|
|
|
|
|
ulong anyTrackingTag = (ulong)HostMappedPtBits.WriteTrackedReplicated;
|
|
|
|
|
|
|
|
|
|
|
|
while (pageIndex <= pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (pageIndex == pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
mask &= endMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex++];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
ulong mappedMask = mask & BlockMappedMask;
|
|
|
|
|
|
|
|
|
|
|
|
ulong mappedPte = pte | (pte >> 1);
|
|
|
|
|
|
if ((mappedPte & mappedMask) != mappedMask)
|
|
|
|
|
|
{
|
|
|
|
|
|
ThrowInvalidMemoryRegionException($"Not mapped: va=0x{va:X16}, size=0x{size:X16}");
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
pte &= mask;
|
|
|
|
|
|
if ((pte & anyTrackingTag) != 0) // Search for any tracking.
|
|
|
|
|
|
{
|
|
|
|
|
|
// Writes trigger any tracking.
|
|
|
|
|
|
// Only trigger tracking from reads if both bits are set on any page.
|
|
|
|
|
|
if (write || (pte & (pte >> 1) & BlockMappedMask) != 0)
|
|
|
|
|
|
{
|
|
|
|
|
|
Tracking.VirtualMemoryEvent(va, size, write);
|
|
|
|
|
|
break;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
mask = ulong.MaxValue;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Computes the number of pages in a virtual address range.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual address of the range</param>
|
|
|
|
|
|
/// <param name="size">Size of the range</param>
|
|
|
|
|
|
/// <param name="startVa">The virtual address of the beginning of the first page</param>
|
|
|
|
|
|
/// <remarks>This function does not differentiate between allocated and unallocated pages.</remarks>
|
|
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
|
|
|
|
private int GetPagesCount(ulong va, ulong size, out ulong startVa)
|
|
|
|
|
|
{
|
|
|
|
|
|
// WARNING: Always check if ulong does not overflow during the operations.
|
|
|
|
|
|
startVa = va & ~(ulong)PageMask;
|
|
|
|
|
|
ulong vaSpan = (va - startVa + size + PageMask) & ~(ulong)PageMask;
|
|
|
|
|
|
|
|
|
|
|
|
return (int)(vaSpan / PageSize);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public void TrackingReprotect(ulong va, ulong size, MemoryPermission protection)
|
|
|
|
|
|
{
|
|
|
|
|
|
// Protection is inverted on software pages, since the default value is 0.
|
|
|
|
|
|
protection = (~protection) & MemoryPermission.ReadAndWrite;
|
|
|
|
|
|
|
|
|
|
|
|
int pages = GetPagesCount(va, size, out va);
|
|
|
|
|
|
ulong pageStart = va >> PageBits;
|
|
|
|
|
|
|
|
|
|
|
|
if (pages == 1)
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong protTag = protection switch
|
|
|
|
|
|
{
|
|
|
|
|
|
MemoryPermission.None => (ulong)HostMappedPtBits.Mapped,
|
|
|
|
|
|
MemoryPermission.Write => (ulong)HostMappedPtBits.WriteTracked,
|
|
|
|
|
|
_ => (ulong)HostMappedPtBits.ReadWriteTracked,
|
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
int bit = (int)((pageStart & 31) << 1);
|
|
|
|
|
|
|
|
|
|
|
|
ulong tagMask = 3UL << bit;
|
|
|
|
|
|
ulong invTagMask = ~tagMask;
|
|
|
|
|
|
|
|
|
|
|
|
ulong tag = protTag << bit;
|
|
|
|
|
|
|
|
|
|
|
|
int pageIndex = (int)(pageStart >> PageToPteShift);
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte;
|
|
|
|
|
|
|
|
|
|
|
|
do
|
|
|
|
|
|
{
|
|
|
|
|
|
pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
}
|
|
|
|
|
|
while ((pte & tagMask) != 0 && Interlocked.CompareExchange(ref pageRef, (pte & invTagMask) | tag, pte) != pte);
|
|
|
|
|
|
}
|
|
|
|
|
|
else
|
|
|
|
|
|
{
|
|
|
|
|
|
ulong pageEnd = pageStart + (ulong)pages;
|
|
|
|
|
|
|
|
|
|
|
|
GetPageBlockRange(pageStart, pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex);
|
|
|
|
|
|
|
|
|
|
|
|
ulong mask = startMask;
|
|
|
|
|
|
|
|
|
|
|
|
ulong protTag = protection switch
|
|
|
|
|
|
{
|
|
|
|
|
|
MemoryPermission.None => (ulong)HostMappedPtBits.MappedReplicated,
|
|
|
|
|
|
MemoryPermission.Write => (ulong)HostMappedPtBits.WriteTrackedReplicated,
|
|
|
|
|
|
_ => (ulong)HostMappedPtBits.ReadWriteTrackedReplicated,
|
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
while (pageIndex <= pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (pageIndex == pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
mask &= endMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex++];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte;
|
|
|
|
|
|
ulong mappedMask;
|
|
|
|
|
|
|
|
|
|
|
|
// Change the protection of all 2 bit entries that are mapped.
|
|
|
|
|
|
do
|
|
|
|
|
|
{
|
|
|
|
|
|
pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
|
|
|
|
|
|
mappedMask = pte | (pte >> 1);
|
|
|
|
|
|
mappedMask |= (mappedMask & BlockMappedMask) << 1;
|
|
|
|
|
|
mappedMask &= mask; // Only update mapped pages within the given range.
|
|
|
|
|
|
}
|
|
|
|
|
|
while (Interlocked.CompareExchange(ref pageRef, (pte & (~mappedMask)) | (protTag & mappedMask), pte) != pte);
|
|
|
|
|
|
|
|
|
|
|
|
mask = ulong.MaxValue;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
protection = protection switch
|
|
|
|
|
|
{
|
|
|
|
|
|
MemoryPermission.None => MemoryPermission.ReadAndWrite,
|
|
|
|
|
|
MemoryPermission.Write => MemoryPermission.Read,
|
|
|
|
|
|
_ => MemoryPermission.None
|
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
_addressSpace.Reprotect(va, size, protection, false);
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public CpuRegionHandle BeginTracking(ulong address, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
return new CpuRegionHandle(Tracking.BeginTracking(address, size));
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
2021-06-24 00:31:26 +01:00
|
|
|
|
public CpuMultiRegionHandle BeginGranularTracking(ulong address, ulong size, IEnumerable<IRegionHandle> handles, ulong granularity)
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
{
|
2021-06-24 00:31:26 +01:00
|
|
|
|
return new CpuMultiRegionHandle(Tracking.BeginGranularTracking(address, size, handles, granularity));
|
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class
* Replace some trivial uses of DRAM address with VA
* Get rid of GetDramAddressFromVa
* Abstracting more operations on derived page table class
* Run auto-format on KPageTableBase
* Managed to make TryConvertVaToPa private, few uses remains now
* Implement guest physical pages ref counting, remove manual freeing
* Make DoMmuOperation private and call new abstract methods only from the base class
* Pass pages count rather than size on Map/UnmapMemory
* Change memory managers to take host pointers
* Fix a guest memory leak and simplify KPageTable
* Expose new methods for host range query and mapping
* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists
* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)
* Add a SharedMemoryStorage class, will be useful for host mapping
* Sayonara AddVaRangeToPageList, you served us well
* Start to implement host memory mapping (WIP)
* Support memory tracking through host exception handling
* Fix some access violations from HLE service guest memory access and CPU
* Fix memory tracking
* Fix mapping list bugs, including a race and a error adding mapping ranges
* Simple page table for memory tracking
* Simple "volatile" region handle mode
* Update UBOs directly (experimental, rough)
* Fix the overlap check
* Only set non-modified buffers as volatile
* Fix some memory tracking issues
* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)
* Write uniform update to memory immediately, only defer the buffer set.
* Fix some memory tracking issues
* Pass correct pages count on shared memory unmap
* Armeilleure Signal Handler v1 + Unix changes
Unix currently behaves like windows, rather than remapping physical
* Actually check if the host platform is unix
* Fix decommit on linux.
* Implement windows 10 placeholder shared memory, fix a buffer issue.
* Make PTC version something that will never match with master
* Remove testing variable for block count
* Add reference count for memory manager, fix dispose
Can still deadlock with OpenAL
* Add address validation, use page table for mapped check, add docs
Might clean up the page table traversing routines.
* Implement batched mapping/tracking.
* Move documentation, fix tests.
* Cleanup uniform buffer update stuff.
* Remove unnecessary assignment.
* Add unsafe host mapped memory switch
On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.
* Remove C# exception handlers
They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.
* Fix MapPhysicalMemory on the software MemoryManager.
* Null check for GetHostAddress, docs
* Add configuration for setting memory manager mode (not in UI yet)
* Add config to UI
* Fix type mismatch on Unix signal handler code emit
* Fix 6GB DRAM mode.
The size can be greater than `uint.MaxValue` when the DRAM is >4GB.
* Address some feedback.
* More detailed error if backing memory cannot be mapped.
* SetLastError on all OS functions for consistency
* Force pages dirty with UBO update instead of setting them directly.
Seems to be much faster across a few games. Need retesting.
* Rebase, configuration rework, fix mem tracking regression
* Fix race in FreePages
* Set memory managers null after decrementing ref count
* Remove readonly keyword, as this is now modified.
* Use a local variable for the signal handler rather than a register.
* Fix bug with buffer resize, and index/uniform buffer binding.
Should fix flickering in games.
* Add InvalidAccessHandler to MemoryTracking
Doesn't do anything yet
* Call invalid access handler on unmapped read/write.
Same rules as the regular memory manager.
* Make unsafe mapped memory its own MemoryManagerType
* Move FlushUboDirty into UpdateState.
* Buffer dirty cache, rather than ubo cache
Much cleaner, may be reusable for Inline2Memory updates.
* This doesn't return anything anymore.
* Add sigaction remove methods, correct a few function signatures.
* Return empty list of physical regions for size 0.
* Also on AddressSpaceManager
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 21:52:44 +01:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <inheritdoc/>
|
|
|
|
|
|
public CpuSmartMultiRegionHandle BeginSmartGranularTracking(ulong address, ulong size, ulong granularity)
|
|
|
|
|
|
{
|
|
|
|
|
|
return new CpuSmartMultiRegionHandle(Tracking.BeginSmartGranularTracking(address, size, granularity));
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Adds the given address mapping to the page table.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual memory address</param>
|
|
|
|
|
|
/// <param name="size">Size to be mapped</param>
|
|
|
|
|
|
private void AddMapping(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
int pages = GetPagesCount(va, size, out _);
|
|
|
|
|
|
ulong pageStart = va >> PageBits;
|
|
|
|
|
|
ulong pageEnd = pageStart + (ulong)pages;
|
|
|
|
|
|
|
|
|
|
|
|
GetPageBlockRange(pageStart, pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex);
|
|
|
|
|
|
|
|
|
|
|
|
ulong mask = startMask;
|
|
|
|
|
|
|
|
|
|
|
|
while (pageIndex <= pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (pageIndex == pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
mask &= endMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex++];
|
|
|
|
|
|
|
|
|
|
|
|
ulong pte;
|
|
|
|
|
|
ulong mappedMask;
|
|
|
|
|
|
|
|
|
|
|
|
// Map all 2-bit entries that are unmapped.
|
|
|
|
|
|
do
|
|
|
|
|
|
{
|
|
|
|
|
|
pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
|
|
|
|
|
|
mappedMask = pte | (pte >> 1);
|
|
|
|
|
|
mappedMask |= (mappedMask & BlockMappedMask) << 1;
|
|
|
|
|
|
mappedMask |= ~mask; // Treat everything outside the range as mapped, thus unchanged.
|
|
|
|
|
|
}
|
|
|
|
|
|
while (Interlocked.CompareExchange(ref pageRef, (pte & mappedMask) | (BlockMappedMask & (~mappedMask)), pte) != pte);
|
|
|
|
|
|
|
|
|
|
|
|
mask = ulong.MaxValue;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Removes the given address mapping from the page table.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
/// <param name="va">Virtual memory address</param>
|
|
|
|
|
|
/// <param name="size">Size to be unmapped</param>
|
|
|
|
|
|
private void RemoveMapping(ulong va, ulong size)
|
|
|
|
|
|
{
|
|
|
|
|
|
int pages = GetPagesCount(va, size, out _);
|
|
|
|
|
|
ulong pageStart = va >> PageBits;
|
|
|
|
|
|
ulong pageEnd = pageStart + (ulong)pages;
|
|
|
|
|
|
|
|
|
|
|
|
GetPageBlockRange(pageStart, pageEnd, out ulong startMask, out ulong endMask, out int pageIndex, out int pageEndIndex);
|
|
|
|
|
|
|
|
|
|
|
|
startMask = ~startMask;
|
|
|
|
|
|
endMask = ~endMask;
|
|
|
|
|
|
|
|
|
|
|
|
ulong mask = startMask;
|
|
|
|
|
|
|
|
|
|
|
|
while (pageIndex <= pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
if (pageIndex == pageEndIndex)
|
|
|
|
|
|
{
|
|
|
|
|
|
mask |= endMask;
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
ref ulong pageRef = ref _pageTable[pageIndex++];
|
|
|
|
|
|
ulong pte;
|
|
|
|
|
|
|
|
|
|
|
|
do
|
|
|
|
|
|
{
|
|
|
|
|
|
pte = Volatile.Read(ref pageRef);
|
|
|
|
|
|
}
|
|
|
|
|
|
while (Interlocked.CompareExchange(ref pageRef, pte & mask, pte) != pte);
|
|
|
|
|
|
|
|
|
|
|
|
mask = 0;
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
|
|
/// Disposes of resources used by the memory manager.
|
|
|
|
|
|
/// </summary>
|
|
|
|
|
|
protected override void Destroy()
|
|
|
|
|
|
{
|
|
|
|
|
|
_addressSpaceMirror.Dispose();
|
|
|
|
|
|
_addressSpace.Dispose();
|
|
|
|
|
|
_memoryEh.Dispose();
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
private void ThrowInvalidMemoryRegionException(string message) => throw new InvalidMemoryRegionException(message);
|
|
|
|
|
|
}
|
|
|
|
|
|
}
|