Ryujinx/Ryujinx.HLE/HOS/Services/Nv/NvMemoryAllocator.cs
sharmander 06aa8a7578
GPU - Improve Memory Allocation (#1722)
* Implement TreeMap from scratch.

Begin implementation of MemoryBlockManager

* Implement GetFreePosition using MemoryBlocks

* Implementation of Memory Management using a Tree.

Still some issues to work around, but promising thus far.

* Resolved invalid mapping issue.

Performance appears promising.

* Add tick metrics

* Use the logger instead

* Use debug loggin instead of info.

* Remove unnecessary code. Add descriptions of added functions.

* Improve memory allocation even further. As well as improve speed of position fetching.

* Add TreeDictionary to Ryujinx Commons

Removed Unnecessary  Usigns

* Add a Performance Profiler + Improve ReserveFixed

* Begin transition to allocation in nvdrv

* Create singleton nvmemallocator

* Moved Allocation into Nv Related Files

As requested by gdkchan, any allocation of memory has been moved into the driver files.

Mapping remains in the GPU MemoryManager.

* Remove unnecessary usings

* Add missing descriptions

* Correct descriptions

* Fix formatting.

* Remove unnecessary whitespace

* Formatting / Convention Updates

* Changes / Fixes

Made syntax and convention changes as requested by gdkchan.

Fixed an issue where IsRegionUsed would return the wrong boolean.

Fixed an issue where GetFreePosition was asked for an address instead of a size.

* Undo commenting of Assert in shader cache

* Update Ryujinx.Common/Collections/TreeDictionary.cs

Co-authored-by: gdkchan <gab.dark.100@gmail.com>

* Resolved many suggestions

* Implement Improved TreeDictionary

Based off of Pseudo code and custom implementations.

* Rename _set to _dictionary

* Remove unused code

* Remove unused code.

* Remove unnecessary MapLow function.

* Resolve data-structure based issues

* Make adjustments to memory management.

Deactive de-allocation for now, it causes more harm than good.

* Minor refactorings + Re-implement deallocation

Also cleaned up unnecessary code.

* Add Tests for TreeDictionary

* Update data structure to properly balance the tree

* Experimental Implementation:

1. Reduce Time to Next Node to O(1) Runtime
2. Reduce While Loop Ct To 2 (In Most Cases)

* Address issues w/ Deallocating Memory

* Final Build

+ Fully Implement Dictionary Interface for new Data Structure
+ Cover All Memory Allocation Edge Cases, particularly w/ Games that De-Allocate a lot.

* Minor Corrections

Give TreeDictionary its own count (do not depend on inner dictionary)

Properly remove adjacent allocations

* Add AsList

* Fix bug where internal dictionary wasn't being updated w/ new node for overwritten key.

* Address comments in review.

* Fix issue where block wouldn't break out (Fixes UE4 issues)

* Update descriptions

* Update descriptions

* Reduce Node visibility to protect TreeDictionary Integrity + Remove usage of struct.

* Update tests to use new TreeDictionary implementation.

* Remove usage of dictionary in TreeDictionary

* Refactoring / Renaming

* Remove unneeded memoryblock class.

* Add space for while

* Add space for if

* Formatting / descriptions

* Clarified some descriptions

* Reduce visibility of memory allocator

* Edit method names to make more sense as memory blocks are no longer in use.

* Make names consistent.

* Protect against npe when sucessorof is called against keys that don't exist. (Not in use by memory manager, this is for other prs that might use this data structure)

* Possible edge-case resolve

* Update Ryujinx.Common/Collections/TreeDictionary.cs

Co-authored-by: gdkchan <gab.dark.100@gmail.com>

* Update Ryujinx.HLE/HOS/Services/Nv/NvMemoryAllocator.cs

Co-authored-by: gdkchan <gab.dark.100@gmail.com>

* Reduce # of unnecessary duplicate variables / Reduce visibility of variables only internally used.

* Rename count to _count

* Update Description of Add method.

* Fix copypasta

* Address comments

* Address comments

* Remove whitespace

* Address comments, condense variables.

* Consolidate vars

* Fix whitespace.

* Nit

* Fix exception msg

* Fix arrayIndex check

* Fix arrayIndex check + indexer

* Remove whitespace from cast

Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2020-12-09 19:26:05 -03:00

282 lines
12 KiB
C#

using Ryujinx.Common.Collections;
using System.Collections.Generic;
using Ryujinx.Common;
using System;
using Ryujinx.Graphics.Gpu.Memory;
namespace Ryujinx.HLE.HOS.Services.Nv.NvDrvServices
{
class NvMemoryAllocator
{
private const ulong AddressSpaceSize = 1UL << 40;
private const ulong DefaultStart = 1UL << 32;
private const ulong InvalidAddress = 0;
private const ulong PageSize = MemoryManager.PageSize;
private const ulong PageMask = MemoryManager.PageMask;
public const ulong PteUnmapped = MemoryManager.PteUnmapped;
// Key --> Start Address of Region
// Value --> End Address of Region
private readonly TreeDictionary<ulong, ulong> _tree = new TreeDictionary<ulong, ulong>();
private readonly Dictionary<ulong, LinkedListNode<ulong>> _dictionary = new Dictionary<ulong, LinkedListNode<ulong>>();
private readonly LinkedList<ulong> _list = new LinkedList<ulong>();
public NvMemoryAllocator()
{
_tree.Add(PageSize, PageSize + AddressSpaceSize);
LinkedListNode<ulong> node = _list.AddFirst(PageSize);
_dictionary[PageSize] = node;
}
/// <summary>
/// Marks a range of memory as consumed by removing it from the tree.
/// This function will split memory regions if there is available space.
/// </summary>
/// <param name="va">Virtual address at which to allocate</param>
/// <param name="size">Size of the allocation in bytes</param>
/// <param name="referenceAddress">Reference to the address of memory where the allocation can take place</param>
#region Memory Allocation
public void AllocateRange(ulong va, ulong size, ulong referenceAddress = InvalidAddress)
{
lock (_tree)
{
if (referenceAddress != InvalidAddress)
{
ulong endAddress = va + size;
ulong referenceEndAddress = _tree.Get(referenceAddress);
if (va >= referenceAddress)
{
// Need Left Node
if (va > referenceAddress)
{
ulong leftEndAddress = va;
// Overwrite existing block with its new smaller range.
_tree.Add(referenceAddress, leftEndAddress);
}
else
{
// We need to get rid of the large chunk.
_tree.Remove(referenceAddress);
}
ulong rightSize = referenceEndAddress - endAddress;
// If leftover space, create a right node.
if (rightSize > 0)
{
_tree.Add(endAddress, referenceEndAddress);
LinkedListNode<ulong> node = _list.AddAfter(_dictionary[referenceAddress], endAddress);
_dictionary[endAddress] = node;
}
if (va == referenceAddress)
{
_list.Remove(_dictionary[referenceAddress]);
_dictionary.Remove(referenceAddress);
}
}
}
}
}
/// <summary>
/// Marks a range of memory as free by adding it to the tree.
/// This function will automatically compact the tree when it determines there are multiple ranges of free memory adjacent to each other.
/// </summary>
/// <param name="va">Virtual address at which to deallocate</param>
/// <param name="size">Size of the allocation in bytes</param>
public void DeallocateRange(ulong va, ulong size)
{
lock (_tree)
{
ulong freeAddressStartPosition = _tree.Floor(va);
if (freeAddressStartPosition != InvalidAddress)
{
LinkedListNode<ulong> node = _dictionary[freeAddressStartPosition];
ulong targetPrevAddress = _dictionary[freeAddressStartPosition].Previous != null ? _dictionary[_dictionary[freeAddressStartPosition].Previous.Value].Value : InvalidAddress;
ulong targetNextAddress = _dictionary[freeAddressStartPosition].Next != null ? _dictionary[_dictionary[freeAddressStartPosition].Next.Value].Value : InvalidAddress;
ulong expandedStart = va;
ulong expandedEnd = va + size;
while (targetPrevAddress != InvalidAddress)
{
ulong prevAddress = targetPrevAddress;
ulong prevEndAddress = _tree.Get(targetPrevAddress);
if (prevEndAddress >= expandedStart)
{
expandedStart = targetPrevAddress;
LinkedListNode<ulong> prevPtr = _dictionary[prevAddress];
if (prevPtr.Previous != null)
{
targetPrevAddress = prevPtr.Previous.Value;
}
else
{
targetPrevAddress = InvalidAddress;
}
node = node.Previous;
_tree.Remove(prevAddress);
_list.Remove(_dictionary[prevAddress]);
_dictionary.Remove(prevAddress);
}
else
{
break;
}
}
while (targetNextAddress != InvalidAddress)
{
ulong nextAddress = targetNextAddress;
ulong nextEndAddress = _tree.Get(targetNextAddress);
if (nextAddress <= expandedEnd)
{
expandedEnd = Math.Max(expandedEnd, nextEndAddress);
LinkedListNode<ulong> nextPtr = _dictionary[nextAddress];
if (nextPtr.Next != null)
{
targetNextAddress = nextPtr.Next.Value;
}
else
{
targetNextAddress = InvalidAddress;
}
_tree.Remove(nextAddress);
_list.Remove(_dictionary[nextAddress]);
_dictionary.Remove(nextAddress);
}
else
{
break;
}
}
_tree.Add(expandedStart, expandedEnd);
LinkedListNode<ulong> nodePtr = _list.AddAfter(node, expandedStart);
_dictionary[expandedStart] = nodePtr;
}
}
}
/// <summary>
/// Gets the address of an unused (free) region of the specified size.
/// </summary>
/// <param name="size">Size of the region in bytes</param>
/// <param name="freeAddressStartPosition">Position at which memory can be allocated</param>
/// <param name="alignment">Required alignment of the region address in bytes</param>
/// <param name="start">Start address of the search on the address space</param>
/// <returns>GPU virtual address of the allocation, or an all ones mask in case of failure</returns>
public ulong GetFreeAddress(ulong size, out ulong freeAddressStartPosition, ulong alignment = 1, ulong start = DefaultStart)
{
// Note: Address 0 is not considered valid by the driver,
// when 0 is returned it's considered a mapping error.
lock (_tree)
{
ulong address = start;
if (alignment == 0)
{
alignment = 1;
}
alignment = (alignment + PageMask) & ~PageMask;
if (address < AddressSpaceSize)
{
bool completedFirstPass = false;
ulong targetAddress;
if(start == DefaultStart)
{
targetAddress = _list.Last.Value;
}
else
{
targetAddress = _tree.Floor(address);
if(targetAddress == InvalidAddress)
{
targetAddress = _tree.Ceiling(address);
}
}
while (address < AddressSpaceSize)
{
if (targetAddress != InvalidAddress)
{
if (address >= targetAddress)
{
if (address + size <= _tree.Get(targetAddress))
{
freeAddressStartPosition = targetAddress;
return address;
}
else
{
LinkedListNode<ulong> nextPtr = _dictionary[targetAddress];
if (nextPtr.Next != null)
{
targetAddress = nextPtr.Next.Value;
}
else
{
if (completedFirstPass)
{
break;
}
else
{
completedFirstPass = true;
address = start;
targetAddress = _tree.Floor(address);
}
}
}
}
else
{
address += PageSize * (targetAddress / PageSize - (address / PageSize));
ulong remainder = address % alignment;
if (remainder != 0)
{
address = (address - remainder) + alignment;
}
}
}
else
{
break;
}
}
}
freeAddressStartPosition = InvalidAddress;
}
return PteUnmapped;
}
/// <summary>
/// Checks if a given memory region is mapped or reserved.
/// </summary>
/// <param name="gpuVa">GPU virtual address of the page</param>
/// <param name="size">Size of the allocation in bytes</param>
/// <param name="freeAddressStartPosition">Nearest lower address that memory can be allocated</param>
/// <returns>True if the page is mapped or reserved, false otherwise</returns>
public bool IsRegionInUse(ulong gpuVa, ulong size, out ulong freeAddressStartPosition)
{
lock (_tree)
{
ulong floorAddress = _tree.Floor(gpuVa);
freeAddressStartPosition = floorAddress;
if (floorAddress != InvalidAddress)
{
return !(gpuVa >= floorAddress && ((gpuVa + size) < _tree.Get(floorAddress)));
}
}
return true;
}
#endregion
}
}