Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
using Ryujinx.Cpu.Tracking;
|
2019-10-13 06:02:07 +00:00
|
|
|
using Ryujinx.Graphics.GAL;
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
using Ryujinx.Memory.Range;
|
2021-01-17 20:08:06 +00:00
|
|
|
using Ryujinx.Memory.Tracking;
|
2019-10-13 06:02:07 +00:00
|
|
|
using System;
|
2021-06-23 23:31:26 +00:00
|
|
|
using System.Collections.Generic;
|
|
|
|
using System.Linq;
|
2019-10-13 06:02:07 +00:00
|
|
|
|
|
|
|
namespace Ryujinx.Graphics.Gpu.Memory
|
|
|
|
{
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Buffer, used to store vertex and index data, uniform and storage buffers, and others.
|
|
|
|
/// </summary>
|
2019-12-31 22:37:00 +00:00
|
|
|
class Buffer : IRange, IDisposable
|
2019-10-13 06:02:07 +00:00
|
|
|
{
|
2021-06-29 17:32:02 +00:00
|
|
|
private const ulong GranularBufferThreshold = 4096;
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
|
2019-12-31 22:37:00 +00:00
|
|
|
private readonly GpuContext _context;
|
2021-06-29 17:32:02 +00:00
|
|
|
private readonly PhysicalMemory _physicalMemory;
|
2019-10-13 06:02:07 +00:00
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
2020-05-23 09:46:09 +00:00
|
|
|
/// Host buffer handle.
|
2019-12-31 03:22:58 +00:00
|
|
|
/// </summary>
|
2020-05-23 09:46:09 +00:00
|
|
|
public BufferHandle Handle { get; }
|
2019-10-27 02:41:01 +00:00
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Start address of the buffer in guest memory.
|
|
|
|
/// </summary>
|
2019-10-13 06:02:07 +00:00
|
|
|
public ulong Address { get; }
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Size of the buffer in bytes.
|
|
|
|
/// </summary>
|
|
|
|
public ulong Size { get; }
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// End address of the buffer in guest memory.
|
|
|
|
/// </summary>
|
2019-10-13 06:02:07 +00:00
|
|
|
public ulong EndAddress => Address + 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 20:52:44 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Increments when the buffer is (partially) unmapped or disposed.
|
|
|
|
/// </summary>
|
|
|
|
public int UnmappedSequence { get; private set; }
|
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Ranges of the buffer that have been modified on the GPU.
|
|
|
|
/// Ranges defined here cannot be updated from CPU until a CPU waiting sync point is reached.
|
|
|
|
/// Then, write tracking will signal, wait for GPU sync (generated at the syncpoint) and flush these regions.
|
|
|
|
/// </summary>
|
|
|
|
/// <remarks>
|
|
|
|
/// This is null until at least one modification occurs.
|
|
|
|
/// </remarks>
|
|
|
|
private BufferModifiedRangeList _modifiedRanges = null;
|
|
|
|
|
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 20:52:44 +00:00
|
|
|
private readonly CpuMultiRegionHandle _memoryTrackingGranular;
|
|
|
|
private readonly CpuRegionHandle _memoryTracking;
|
2021-01-17 20:08:06 +00:00
|
|
|
|
|
|
|
private readonly RegionSignal _externalFlushDelegate;
|
|
|
|
private readonly Action<ulong, ulong> _loadDelegate;
|
2020-12-03 18:34:32 +00:00
|
|
|
private readonly Action<ulong, ulong> _modifiedDelegate;
|
2021-01-17 20:08:06 +00:00
|
|
|
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
private int _sequenceNumber;
|
2020-05-03 22:54:50 +00:00
|
|
|
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
private bool _useGranular;
|
2021-01-17 20:08:06 +00:00
|
|
|
private bool _syncActionRegistered;
|
2019-10-13 06:02:07 +00:00
|
|
|
|
2022-12-01 15:30:13 +00:00
|
|
|
private int _referenceCount = 1;
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Creates a new instance of the buffer.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="context">GPU context that the buffer belongs to</param>
|
2021-06-29 17:32:02 +00:00
|
|
|
/// <param name="physicalMemory">Physical memory where the buffer is mapped</param>
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <param name="address">Start address of the buffer</param>
|
|
|
|
/// <param name="size">Size of the buffer in bytes</param>
|
2021-06-23 23:31:26 +00:00
|
|
|
/// <param name="baseBuffers">Buffers which this buffer contains, and will inherit tracking handles from</param>
|
2021-06-29 17:32:02 +00:00
|
|
|
public Buffer(GpuContext context, PhysicalMemory physicalMemory, ulong address, ulong size, IEnumerable<Buffer> baseBuffers = null)
|
2019-10-13 06:02:07 +00:00
|
|
|
{
|
2021-06-29 17:32:02 +00:00
|
|
|
_context = context;
|
|
|
|
_physicalMemory = physicalMemory;
|
|
|
|
Address = address;
|
|
|
|
Size = size;
|
2019-10-13 06:02:07 +00:00
|
|
|
|
2020-05-23 09:46:09 +00:00
|
|
|
Handle = context.Renderer.CreateBuffer((int)size);
|
2019-10-13 06:02:07 +00:00
|
|
|
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
_useGranular = size > GranularBufferThreshold;
|
2020-05-03 22:54:50 +00:00
|
|
|
|
2021-06-23 23:31:26 +00:00
|
|
|
IEnumerable<IRegionHandle> baseHandles = null;
|
|
|
|
|
|
|
|
if (baseBuffers != null)
|
|
|
|
{
|
|
|
|
baseHandles = baseBuffers.SelectMany(buffer =>
|
|
|
|
{
|
|
|
|
if (buffer._useGranular)
|
|
|
|
{
|
|
|
|
return buffer._memoryTrackingGranular.GetHandles();
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
return Enumerable.Repeat(buffer._memoryTracking.GetHandle(), 1);
|
|
|
|
}
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
if (_useGranular)
|
|
|
|
{
|
2021-06-29 17:32:02 +00:00
|
|
|
_memoryTrackingGranular = physicalMemory.BeginGranularTracking(address, size, baseHandles);
|
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 00:27:03 +00:00
|
|
|
|
|
|
|
_memoryTrackingGranular.RegisterPreciseAction(address, size, PreciseAction);
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2021-06-29 17:32:02 +00:00
|
|
|
_memoryTracking = physicalMemory.BeginTracking(address, size);
|
2021-06-23 23:31:26 +00:00
|
|
|
|
|
|
|
if (baseHandles != null)
|
|
|
|
{
|
|
|
|
_memoryTracking.Reprotect(false);
|
|
|
|
|
|
|
|
foreach (IRegionHandle handle in baseHandles)
|
|
|
|
{
|
|
|
|
if (handle.Dirty)
|
|
|
|
{
|
|
|
|
_memoryTracking.Reprotect(true);
|
|
|
|
}
|
|
|
|
|
|
|
|
handle.Dispose();
|
|
|
|
}
|
|
|
|
}
|
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 00:27:03 +00:00
|
|
|
|
|
|
|
_memoryTracking.RegisterPreciseAction(PreciseAction);
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
}
|
2020-12-03 18:34:32 +00:00
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
_externalFlushDelegate = new RegionSignal(ExternalFlush);
|
|
|
|
_loadDelegate = new Action<ulong, ulong>(LoadRegion);
|
2020-12-03 18:34:32 +00:00
|
|
|
_modifiedDelegate = new Action<ulong, ulong>(RegionModified);
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
|
|
|
|
2021-01-24 22:22:19 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Gets a sub-range from the buffer, from a start address till the end of the buffer.
|
|
|
|
/// </summary>
|
|
|
|
/// <remarks>
|
|
|
|
/// This can be used to bind and use sub-ranges of the buffer on the host API.
|
|
|
|
/// </remarks>
|
|
|
|
/// <param name="address">Start address of the sub-range, must be greater than or equal to the buffer address</param>
|
|
|
|
/// <returns>The buffer sub-range</returns>
|
|
|
|
public BufferRange GetRange(ulong address)
|
|
|
|
{
|
|
|
|
ulong offset = address - Address;
|
|
|
|
|
|
|
|
return new BufferRange(Handle, (int)offset, (int)(Size - offset));
|
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Gets a sub-range from the buffer.
|
|
|
|
/// </summary>
|
2020-01-01 15:39:09 +00:00
|
|
|
/// <remarks>
|
|
|
|
/// This can be used to bind and use sub-ranges of the buffer on the host API.
|
|
|
|
/// </remarks>
|
|
|
|
/// <param name="address">Start address of the sub-range, must be greater than or equal to the buffer address</param>
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <param name="size">Size in bytes of the sub-range, must be less than or equal to the buffer size</param>
|
|
|
|
/// <returns>The buffer sub-range</returns>
|
2019-10-13 06:02:07 +00:00
|
|
|
public BufferRange GetRange(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
int offset = (int)(address - Address);
|
|
|
|
|
2020-05-23 09:46:09 +00:00
|
|
|
return new BufferRange(Handle, offset, (int)size);
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Checks if a given range overlaps with the buffer.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Start address of the range</param>
|
|
|
|
/// <param name="size">Size in bytes of the range</param>
|
|
|
|
/// <returns>True if the range overlaps, false otherwise</returns>
|
2019-10-13 06:02:07 +00:00
|
|
|
public bool OverlapsWith(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
return Address < address + size && address < EndAddress;
|
|
|
|
}
|
|
|
|
|
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 20:52:44 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Checks if a given range is fully contained in the buffer.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Start address of the range</param>
|
|
|
|
/// <param name="size">Size in bytes of the range</param>
|
|
|
|
/// <returns>True if the range is contained, false otherwise</returns>
|
|
|
|
public bool FullyContains(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
return address >= Address && address + size <= EndAddress;
|
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Performs guest to host memory synchronization of the buffer data.
|
2020-01-01 15:39:09 +00:00
|
|
|
/// </summary>
|
|
|
|
/// <remarks>
|
2019-12-31 03:22:58 +00:00
|
|
|
/// This causes the buffer data to be overwritten if a write was detected from the CPU,
|
|
|
|
/// since the last call to this method.
|
2020-01-01 15:39:09 +00:00
|
|
|
/// </remarks>
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <param name="address">Start address of the range to synchronize</param>
|
|
|
|
/// <param name="size">Size in bytes of the range to synchronize</param>
|
2019-10-13 06:02:07 +00:00
|
|
|
public void SynchronizeMemory(ulong address, ulong size)
|
|
|
|
{
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
if (_useGranular)
|
2019-10-13 06:02:07 +00:00
|
|
|
{
|
2020-12-03 18:34:32 +00:00
|
|
|
_memoryTrackingGranular.QueryModified(address, size, _modifiedDelegate, _context.SequenceNumber);
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
}
|
|
|
|
else
|
2019-10-13 06:02:07 +00:00
|
|
|
{
|
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 20:52:44 +00:00
|
|
|
if (_context.SequenceNumber != _sequenceNumber && _memoryTracking.DirtyOrVolatile())
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
{
|
|
|
|
_memoryTracking.Reprotect();
|
2021-01-17 20:08:06 +00:00
|
|
|
|
|
|
|
if (_modifiedRanges != null)
|
|
|
|
{
|
|
|
|
_modifiedRanges.ExcludeModifiedRegions(Address, Size, _loadDelegate);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2021-06-29 17:32:02 +00:00
|
|
|
_context.Renderer.SetBufferData(Handle, 0, _physicalMemory.GetSpan(Address, (int)Size));
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
2021-06-29 17:32:02 +00:00
|
|
|
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
_sequenceNumber = _context.SequenceNumber;
|
|
|
|
}
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Ensure that the modified range list exists.
|
|
|
|
/// </summary>
|
|
|
|
private void EnsureRangeList()
|
|
|
|
{
|
|
|
|
if (_modifiedRanges == null)
|
|
|
|
{
|
2022-12-01 15:30:13 +00:00
|
|
|
_modifiedRanges = new BufferModifiedRangeList(_context, this, Flush);
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Signal that the given region of the buffer has been modified.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">The start address of the modified region</param>
|
|
|
|
/// <param name="size">The size of the modified region</param>
|
|
|
|
public void SignalModified(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
EnsureRangeList();
|
|
|
|
|
|
|
|
_modifiedRanges.SignalModified(address, size);
|
|
|
|
|
|
|
|
if (!_syncActionRegistered)
|
|
|
|
{
|
|
|
|
_context.RegisterSyncAction(SyncAction);
|
|
|
|
_syncActionRegistered = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Indicate that mofifications in a given region of this buffer have been overwritten.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">The start address of the region</param>
|
|
|
|
/// <param name="size">The size of the region</param>
|
|
|
|
public void ClearModified(ulong address, ulong size)
|
|
|
|
{
|
2021-10-04 18:38:59 +00:00
|
|
|
_modifiedRanges?.Clear(address, size);
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Action to be performed when a syncpoint is reached after modification.
|
|
|
|
/// This will register read/write tracking to flush the buffer from GPU when its memory is used.
|
|
|
|
/// </summary>
|
|
|
|
private void SyncAction()
|
|
|
|
{
|
|
|
|
_syncActionRegistered = false;
|
|
|
|
|
|
|
|
if (_useGranular)
|
|
|
|
{
|
2021-10-04 18:38:59 +00:00
|
|
|
_modifiedRanges?.GetRanges(Address, Size, (address, size) =>
|
2021-01-17 20:08:06 +00:00
|
|
|
{
|
|
|
|
_memoryTrackingGranular.RegisterAction(address, size, _externalFlushDelegate);
|
|
|
|
SynchronizeMemory(address, size);
|
|
|
|
});
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
_memoryTracking.RegisterAction(_externalFlushDelegate);
|
|
|
|
SynchronizeMemory(Address, Size);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Inherit modified ranges from another buffer.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="from">The buffer to inherit from</param>
|
|
|
|
public void InheritModifiedRanges(Buffer from)
|
|
|
|
{
|
2022-12-01 15:30:13 +00:00
|
|
|
if (from._modifiedRanges != null && from._modifiedRanges.HasRanges)
|
2021-01-17 20:08:06 +00:00
|
|
|
{
|
|
|
|
if (from._syncActionRegistered && !_syncActionRegistered)
|
|
|
|
{
|
|
|
|
_context.RegisterSyncAction(SyncAction);
|
|
|
|
_syncActionRegistered = true;
|
|
|
|
}
|
|
|
|
|
2021-10-06 23:00:56 +00:00
|
|
|
Action<ulong, ulong> registerRangeAction = (ulong address, ulong size) =>
|
|
|
|
{
|
|
|
|
if (_useGranular)
|
|
|
|
{
|
|
|
|
_memoryTrackingGranular.RegisterAction(address, size, _externalFlushDelegate);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
_memoryTracking.RegisterAction(_externalFlushDelegate);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2022-12-01 15:30:13 +00:00
|
|
|
EnsureRangeList();
|
2021-10-06 23:00:56 +00:00
|
|
|
|
2022-12-01 15:30:13 +00:00
|
|
|
_modifiedRanges.InheritRanges(from._modifiedRanges, registerRangeAction);
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Determine if a given region of the buffer has been modified, and must be flushed.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">The start address of the region</param>
|
|
|
|
/// <param name="size">The size of the region</param>
|
|
|
|
/// <returns></returns>
|
|
|
|
public bool IsModified(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
if (_modifiedRanges != null)
|
|
|
|
{
|
|
|
|
return _modifiedRanges.HasRange(address, size);
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2020-12-03 18:34:32 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Indicate that a region of the buffer was modified, and must be loaded from memory.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="mAddress">Start address of the modified region</param>
|
|
|
|
/// <param name="mSize">Size of the modified region</param>
|
|
|
|
private void RegionModified(ulong mAddress, ulong mSize)
|
|
|
|
{
|
|
|
|
if (mAddress < Address)
|
|
|
|
{
|
|
|
|
mAddress = Address;
|
|
|
|
}
|
|
|
|
|
|
|
|
ulong maxSize = Address + Size - mAddress;
|
|
|
|
|
|
|
|
if (mSize > maxSize)
|
|
|
|
{
|
|
|
|
mSize = maxSize;
|
|
|
|
}
|
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
if (_modifiedRanges != null)
|
|
|
|
{
|
|
|
|
_modifiedRanges.ExcludeModifiedRegions(mAddress, mSize, _loadDelegate);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
LoadRegion(mAddress, mSize);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Load a region of the buffer from memory.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="mAddress">Start address of the modified region</param>
|
|
|
|
/// <param name="mSize">Size of the modified region</param>
|
|
|
|
private void LoadRegion(ulong mAddress, ulong mSize)
|
|
|
|
{
|
2020-12-03 18:34:32 +00:00
|
|
|
int offset = (int)(mAddress - Address);
|
|
|
|
|
2021-06-29 17:32:02 +00:00
|
|
|
_context.Renderer.SetBufferData(Handle, offset, _physicalMemory.GetSpan(mAddress, (int)mSize));
|
2020-12-03 18:34:32 +00:00
|
|
|
}
|
|
|
|
|
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 20:52:44 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Force a region of the buffer to be dirty. Avoids reprotection and nullifies sequence number check.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="mAddress">Start address of the modified region</param>
|
|
|
|
/// <param name="mSize">Size of the region to force dirty</param>
|
|
|
|
public void ForceDirty(ulong mAddress, ulong mSize)
|
|
|
|
{
|
2021-10-04 18:38:59 +00:00
|
|
|
_modifiedRanges?.Clear(mAddress, mSize);
|
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 20:52:44 +00:00
|
|
|
|
|
|
|
if (_useGranular)
|
|
|
|
{
|
|
|
|
_memoryTrackingGranular.ForceDirty(mAddress, mSize);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
_memoryTracking.ForceDirty();
|
|
|
|
_sequenceNumber--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Performs copy of all the buffer data from one buffer to another.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="destination">The destination buffer to copy the data into</param>
|
|
|
|
/// <param name="dstOffset">The offset of the destination buffer to copy into</param>
|
2019-10-13 06:02:07 +00:00
|
|
|
public void CopyTo(Buffer destination, int dstOffset)
|
|
|
|
{
|
2020-05-23 09:46:09 +00:00
|
|
|
_context.Renderer.Pipeline.CopyBuffer(Handle, destination.Handle, 0, dstOffset, (int)Size);
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Flushes a range of the buffer.
|
|
|
|
/// This writes the range data back into guest memory.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Start address of the range</param>
|
|
|
|
/// <param name="size">Size in bytes of the range</param>
|
2019-10-27 02:41:01 +00:00
|
|
|
public void Flush(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
int offset = (int)(address - Address);
|
|
|
|
|
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 22:10:54 +00:00
|
|
|
ReadOnlySpan<byte> data = _context.Renderer.GetBufferData(Handle, offset, (int)size);
|
2019-10-27 02:41:01 +00:00
|
|
|
|
2020-09-10 19:44:04 +00:00
|
|
|
// TODO: When write tracking shaders, they will need to be aware of changes in overlapping buffers.
|
2021-06-29 17:32:02 +00:00
|
|
|
_physicalMemory.WriteUntracked(address, data);
|
2019-10-27 02:41:01 +00:00
|
|
|
}
|
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Align a given address and size region to page boundaries.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">The start address of the region</param>
|
|
|
|
/// <param name="size">The size of the region</param>
|
|
|
|
/// <returns>The page aligned address and size</returns>
|
|
|
|
private static (ulong address, ulong size) PageAlign(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
ulong pageMask = MemoryManager.PageMask;
|
|
|
|
ulong rA = address & ~pageMask;
|
|
|
|
ulong rS = ((address + size + pageMask) & ~pageMask) - rA;
|
|
|
|
return (rA, rS);
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Flush modified ranges of the buffer from another thread.
|
|
|
|
/// This will flush all modifications made before the active SyncNumber was set, and may block to wait for GPU sync.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Address of the memory action</param>
|
|
|
|
/// <param name="size">Size in bytes</param>
|
|
|
|
public void ExternalFlush(ulong address, ulong size)
|
|
|
|
{
|
|
|
|
_context.Renderer.BackgroundContextAction(() =>
|
|
|
|
{
|
|
|
|
var ranges = _modifiedRanges;
|
|
|
|
|
|
|
|
if (ranges != null)
|
|
|
|
{
|
|
|
|
(address, size) = PageAlign(address, size);
|
2022-12-01 15:30:13 +00:00
|
|
|
ranges.WaitForAndFlushRanges(address, size);
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
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. :relieved:
* 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>
2021-08-26 22:31:29 +00:00
|
|
|
}, true);
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
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 00:27:03 +00:00
|
|
|
/// An action to be performed when a precise memory access occurs to this resource.
|
|
|
|
/// For buffers, this skips flush-on-write by punching holes directly into the modified range list.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Address of the memory action</param>
|
|
|
|
/// <param name="size">Size in bytes</param>
|
|
|
|
/// <param name="write">True if the access was a write, false otherwise</param>
|
|
|
|
private bool PreciseAction(ulong address, ulong size, bool write)
|
|
|
|
{
|
|
|
|
if (!write)
|
|
|
|
{
|
|
|
|
// We only want to skip flush-on-write.
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2023-01-17 04:13:24 +00:00
|
|
|
ulong maxAddress = Math.Max(address, Address);
|
|
|
|
ulong minEndAddress = Math.Min(address + size, Address + 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 00:27:03 +00:00
|
|
|
|
2023-01-17 04:13:24 +00:00
|
|
|
if (maxAddress >= minEndAddress)
|
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 00:27:03 +00:00
|
|
|
{
|
2023-01-17 04:13:24 +00:00
|
|
|
// Access doesn't overlap.
|
|
|
|
return false;
|
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 00:27:03 +00:00
|
|
|
}
|
|
|
|
|
2023-01-17 04:13:24 +00:00
|
|
|
ForceDirty(maxAddress, minEndAddress - maxAddress);
|
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 00:27:03 +00:00
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
2021-01-17 20:08:06 +00:00
|
|
|
/// Called when part of the memory for this buffer has been unmapped.
|
|
|
|
/// Calls are from non-GPU threads.
|
|
|
|
/// </summary>
|
|
|
|
/// <param name="address">Start address of the unmapped region</param>
|
|
|
|
/// <param name="size">Size of the unmapped region</param>
|
|
|
|
public void Unmapped(ulong address, ulong size)
|
|
|
|
{
|
2021-10-04 18:38:59 +00:00
|
|
|
BufferModifiedRangeList modifiedRanges = _modifiedRanges;
|
|
|
|
|
|
|
|
modifiedRanges?.Clear(address, 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 20:52:44 +00:00
|
|
|
|
|
|
|
UnmappedSequence++;
|
2021-01-17 20:08:06 +00:00
|
|
|
}
|
|
|
|
|
2022-12-01 15:30:13 +00:00
|
|
|
/// <summary>
|
|
|
|
/// Increments the buffer reference count.
|
|
|
|
/// </summary>
|
|
|
|
public void IncrementReferenceCount()
|
|
|
|
{
|
|
|
|
_referenceCount++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Decrements the buffer reference count.
|
|
|
|
/// </summary>
|
|
|
|
public void DecrementReferenceCount()
|
|
|
|
{
|
|
|
|
if (--_referenceCount == 0)
|
|
|
|
{
|
|
|
|
DisposeData();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-12-31 03:22:58 +00:00
|
|
|
/// <summary>
|
2021-06-23 23:31:26 +00:00
|
|
|
/// Disposes the host buffer's data, not its tracking handles.
|
2019-12-31 03:22:58 +00:00
|
|
|
/// </summary>
|
2021-06-23 23:31:26 +00:00
|
|
|
public void DisposeData()
|
2019-10-13 06:02:07 +00:00
|
|
|
{
|
2021-01-17 20:08:06 +00:00
|
|
|
_modifiedRanges?.Clear();
|
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking
- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.
It works :)
Still a few hacks, messy things, slow things
More work in progress stuff (also move to memory project)
Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)
Move some stuff.
I think we'll eventually just put the dll and so for this in a nuget package.
Fix rebase.
[WIP] MultiRegionHandle variable size ranges
- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity
Fix rebase issue
Commit everything needed for software only tracking.
Remove native components.
Remove more native stuff.
Cleanup
Use a separate window for the background context, update opentk. (fixes linux)
Some experimental changes
Should get things working up to scratch - still need to try some things with flush/modification and res scale.
Include address with the region action.
Initial work to make range tracking work
Still a ton of bugs
Fix some issues with the new stuff.
* Fix texture flush instability
There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)
* Find the destination texture for Buffer->Texture full copy
Greatly improves performance for nvdec videos (with range tracking)
* Further improve texture tracking
* Disable Memory Tracking for view parents
This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)
The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.
* Introduce some tracking tests.
WIP
* Complete base tests.
* Add more tests for multiregion, fix existing test.
* Cleanup Part 1
* Remove unnecessary code from memory tracking
* Fix some inconsistencies with 3D texture rule.
* Add dispose tests.
* Use a background thread for the background context.
Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.
Also nerf the multithreading test a bit.
* Copy to texture with matching alignment
This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.
* Track reads for buffer copies. Synchronize new buffers before copying overlaps.
* Remove old texture flushing mechanisms.
Range tracking all the way, baby.
* Wake the background thread when disposing.
Avoids a deadlock when games are closed.
* Address Feedback 1
* Separate TextureCopy instance for background thread
Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.
* Add missing XML docs.
* Address Feedback
* Maybe I should start drinking coffee.
* Some more feedback.
* Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
|
|
|
|
2021-01-17 20:08:06 +00:00
|
|
|
_context.Renderer.DeleteBuffer(Handle);
|
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 20:52:44 +00:00
|
|
|
|
|
|
|
UnmappedSequence++;
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
2021-06-23 23:31:26 +00:00
|
|
|
|
|
|
|
/// <summary>
|
|
|
|
/// Disposes the host buffer.
|
|
|
|
/// </summary>
|
|
|
|
public void Dispose()
|
|
|
|
{
|
|
|
|
_memoryTrackingGranular?.Dispose();
|
|
|
|
_memoryTracking?.Dispose();
|
|
|
|
|
2022-12-01 15:30:13 +00:00
|
|
|
DecrementReferenceCount();
|
2021-06-23 23:31:26 +00:00
|
|
|
}
|
2019-10-13 06:02:07 +00:00
|
|
|
}
|
|
|
|
}
|