* Initial Implementation
* Further improvements (no support for float/64-bit types)
* Merge atomic and reduce instructions, add missing format switch
* Fix rebase issues.
* Not used.
* Whoops. Fixed.
* Partial implementation of inc/dec, cleanup and TODOs
* Remove testing path
* Address Feedback
* Avoid deleting textures when their data does not overlap.
It's possible that while two textures start and end addresses indicate an overlap, that the actual data contained within them is sparse due to a layer stride. One such possibility is array slices of a cubemap at different mip levels - they overlap on a whole, but the actual texture data fills the gaps between each other's layers rather than actually overlapping.
This fixes issues with UE4 games having incorrect lighting (solid white screen or really dark shadows). There are still remaining issues with games that use the 3D texture prebaked lighting, such as THPS1+2.
This PR also fixes a bug with TexturePool's resized texture handling where the base level in the descriptor was not considered.
* AllRegions granularity for 3d textures is now by level rather than by slice.
* Address feedback
* Only reupload the texture scale array if it changes.
Before, this would be called all the time if any shader needed a scale value. The cost of doing this has increased with threaded-gal, as the scale array is copied to a span pool, and it's was called on pretty much every draw sometimes.
This improves GPU performance in games, scaled or not. Most affected game seems to be Xenoblade Chronicles: Definitive Edition.
* Just use = instead of |=
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. 😌
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
This seems to be the default value when the vertex attribute is disabled, or components aren't defined.
This fixes a regression from #2307 in SMO where a plant in the Wooded Kingdom would draw slightly differently in the depth prepass, leading to depth test failing later on.
GDK has stated that the specific case in Gundam only expects x and y to be 0, and Vulkan's undefined value for z does appear to be 0 when a vertex attribute type does not have that component, hence the value (0, 0, 0, 1).
This worked in Vulkan despite also providing an all 0s attribute due to the vertex attribute binding being R32Float, so the other values were undefined. It should be changed there separately.
This greatly reduces memory usage in games that aggressively reuse memory without removing dead textures from the pool, such as the Xenoblade games, UE3 games, and to a lesser extent, UE4/unity games.
This change stops memory usage from ballooning in xenoblade and some other games. It will also reduce texture view/dependency complexity in some games - for example in MK8D it will reduce the number of surface copies between lighting cubemaps generated for actors.
There shouldn't be any performance impact from doing this, though the deletion and creation of textures could be improved by improving the OpenGL texture storage cache, which is very simple and limited right now. This will be improved in future.
Another potential error has been fixed with the texture cache, which could prevent data loss when data is interchangably written to textures from both the GPU and CPU. It was possible that the dirty flag for a texture would be consumed without the data being synchronized on next use, due to the old overlap check. This check no longer consumes the dirty flag.
Please test a bunch of games to make sure they still work, and there are no performance regressions.
Got this the wrong way round - was causing games to try synchronize mipmap levels of like 52 on a 3d texture with 6 levels. Also, corrected the variable name in the method that _was_ working.
* Use "Undesired" scale mode for certain textures rather than blacklisting
* Nit
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
* 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.
It seems like this method of flushing data is much slower on Mesa drivers, and slightly slower on Intel Windows. Have not tested Intel Mesa, but I'm assuming it is the same as AMD.
This also adds vendor detection for AMD on Unix, which counted as "Unknown" before.
* Honour copy dependencies when switching render target
When switching from one render target to another, when both have a copy dependency to each other, a copy can be deferred on the second target when unbinding the first.
Before, this would not be honoured before binding the new texture, so the copy would stay deferred until the render targets change again, at which point it would copy in old data and essentially clear all the draws done during that time.
This change runs synchronize memory to make sure that copies are honoured. This can cause a redundant copy, but it's better than it breaking for now.
This should fix miiedit on AMD/Intel GPUs on windows. May fix other games, or perhaps rare copy dependency bugs on NVIDIA too.
* Address feedback
This greatly speeds up games that constantly resize buffers, and removes stuttering on games that resize large buffers occasionally:
- Large improvement on Super Mario 3D All-Stars (#1663 needed for best performance)
- Improvement to Hyrule Warriors: AoC, and UE4 games. These games can still stutter due to texture creation/loading.
- Small improvement to other games, potential 1-frame stutters avoided.
`ForceSynchronizeMemory`, which was added with POWER, is no longer needed. Some tests have been added for the MultiRegionHandle.
Improves OpenGL performance on FAST RMX and Xenoblade DE/2. Will probably only work on NVIDIA GPUs, but the emulated quads path will still be valid for other GPUs. Note: SLOW RMX gets a bit faster in handheld mode.
I'd recommend checking on platforms without supported host quads to make sure a GL error is actually thrown when attempting GL.Begin(PrimitiveType.Quads)
* 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>
* Compare aligned size for largest mip level when considering sampler resize
When selecting a texture that's a view for a sampler resize, we should take care that resizing it doesn't change the aligned size of any larger mip levels.
This PR covers two cases:
- When creating a view of the texture, we check that the aligned size of the view shifted up to level 0 still matches the aligned size of the container. If it does not, a copy dependency is created rather than resizing.
- When searching for a texture for sampler, textures that do _not_ match our aligned size when both are shifted up by its base level are not considered an exact match, as resizing the found texture will cause the mip 0 aligned size to change. It will create a copy dependency view instead.
Fixes graphical errors and crashes (on flush) in various Unity games that use render-to-texture.
* Move shared code to its own method.
* Pass CbufSlot when getting info from the texture descriptor
Fixes some issues with bindless textures, when CbufSlot is not equal to the current TextureBufferIndex.
Specifically fixes a random chance of full screen colour flickering in Super Mario Party.
* Apply suggestions from code review
Oops
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
Co-authored-by: gdkchan <gab.dark.100@gmail.com>
* Return focus from controller applet after completion
This fixes controller applet related in Mario Kart 8 Deluxe, in 2 player mode or when opening the applet in character select.
* Return focus for player select
* Improve linear texture compatibility rules
Fixes an issue where small or width-aligned (rather than byte aligned) textures would fail to create a view of existing data. Creates a copy dependency as size change may be risky.
* Minor cleanup
* Remove Size Change for Copy Depenedencies
The copy to the target (potentially different sized) texture can properly deal with cropping by itself.
* Move StrideAlignment and GobAlignment into Constants
This allows bindless handles to be found for image/texture instructions with predicates, when the assignment of the texture handle is within the same predicate.
This seems to cover the remaining bindless handles that compilers seem to be creating due to optimizations.
Will affect newer UE4 games, and games by NdCube (Super Mario Party, Clubhouse Games)
* Improve Buffer Textures and flush Image Stores
Fixes a number of issues with buffer textures:
- Reworked Buffer Textures to create their buffers in the TextureManager, then bind them with the BufferManager later.
- Fixes an issue where a buffer texture's buffer could be invalidated after it is bound, but before use.
- Fixed width unpacking for large buffer textures. The width is now 32-bit rather than 16.
- Force buffer textures to be rebound whenever any buffer is created, as using the handle id wasn't reliable, and the cost of binding isn't too high.
Fixes vertex explosions and flickering animations in UE4 games.
* Set ImageStore flag... for ImageStore.
* Check the offset and size.
* Improve handling for unmapped GPU resources
- Fixed a memory tracking bug that would set protection on empty PTEs
- When a texture's memory is (partially) unmapped, all pool references are forcibly removed and the texture must be rediscovered to draw with it. This will also force the texture discovery to always compare the texture's range for a match.
- RegionHandles now know if they are unmapped, and automatically unset their dirty flag when unmapped.
- Partial texture sync now loads only the region of texture that has been modified. Unmapped memory tracking handles cause dirty flags for a texture group handle to be ignored.
This greatly improves the emulator's stability for newer UE4 games.
* Address feedback, fix MultiRange slice
Fixed an issue where the size of the multi-range slice would be miscalculated.
* Update Ryujinx.Memory/Range/MultiRange.cs (feedback)
Co-authored-by: Mary <thog@protonmail.com>
Co-authored-by: Mary <thog@protonmail.com>
* Initial implementation (3d tex mips broken)
This works rather well for most games, just need to fix 3d texture mips.
* Cleanup
* Address feedback
* Copy Dependencies and various other fixes
* Fix layer/level offset for copy from view<->view.
* Remove dirty flag from dependency
The dirty flag behaviour is not needed - DeferredCopy is all we need.
* Fix tracking mip slices.
* Propagate granularity (fix astral chain)
* Address Feedback pt 1
* Save slice sizes as part of SizeInfo
* Fix nits
* Fix disposing multiple dependencies causing a crash
This list is obviously modified when removing dependencies, so create a copy of it.
* Initial implementation of buffer flush (VERY WIP)
* Host shaders need to be rebuilt for the SSBO write flag.
* New approach with reserved regions and gl sync
* Fix a ton of buffer issues.
* Remove unused buffer unmapped behaviour
* Revert "Remove unused buffer unmapped behaviour"
This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece.
* Delete modified ranges on unmap
Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap.
* Cache some more delegates.
* Dispose Sync on Close
* Also create host sync for GPFifo syncpoint increment.
* Copy buffer optimization, add docs
* Fix race condition with OpenGL Sync
* Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle
* Performance: Only flush individual pages of SSBO at a time
This avoids flushing large amounts of data when only a small amount is actually used.
* Signal Modified rather than flushing after clear
* Fix some docs and code style.
* Introduce a new test for tracking memory protection.
Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master)
* Address Comments
* Add host sync for SetReference
This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise.
* Make PageAlign static
* Re-enable read tracking, for reads.
* Interrupt GPU command processing when a frame's fence is reached.
* Accumulate times rather than %s
* Accurate timer for vsync
Spin wait for the last .667ms of a frame. Avoids issues caused by signalling 16ms vsync. (periodic stutters in smo)
* Use event wait for better timing.
* Fix lazy wait
Windows doesn't seem to want to do 1ms consistently, so force a spin if we're less than 2ms.
* A bit more efficiency on frame waits.
Should now wait the remainder 0.6667 instead of 1.6667 sometimes (odd waits above 1ms are reliable, unlike 1ms waits)
* Better swap interval 0 solution
737 fps without breaking a sweat. Downside: Vsync can no longer be disabled on games that use the event heavily (link's awakening - which is ok since it breaks anyways)
* Fix comment.
* Address Comments.
* Thread create/delete stutter workarounds Pt 1
* As tiered compilation is disabled, disable quick jit too
Should result in tier 1 compilation all the time
* Fix rebase.
* Allow copy destination to have a different scale from source
Will result in more scaled copy destinations, but allows scaling in some games that copy textures to the output framebuffer.
* Support copying multiple levels/layers
Uses glFramebufferTextureLayer to copy multiple layers, copies levels individually (and scales the regions).
Remove CopyArrayScaled, since the backend copy handles it now.
* PBO single layer copy, part 1
Still needs ability to take and set width/height slices. (using pack paramaters)
* PBO Copies pt 2
* Some fixes and cleanup.
* Misc Cleanup
* Move handle into the TextureInfo interface.
This interface is shared between texture storages and views.
* Move unscaled copy to the TextureCopy class.
* Address feedback.
"Screen scissor" is the minimum size of all render targets, and is set when any render target is bound on NVN or OpenGL. Since it works on all active texture's real sizes, it is therefore more reliable than viewport 0's width, and is actually set before clear.
This fixes a regression with Hyrule Warriors: Age Of Calamity's cubemaps, which did not set viewport dimensions before clear. This resulted in attempting to create a cubemap with rectangular sides, which is logically and physically impossible. (also it just fails)
* Size hints for copy regions and viewport dimensions to avoid data loss
* Reword comment.
* Use info for the rule rather than calculating aligned size.
* Reorder min/max, remove spaces
* Only report that commands are available when the queue is not empty.
* Address Feedback
Co-authored-by: FICTURE7 <FICTURE7@gmail.com>
Co-authored-by: FICTURE7 <FICTURE7@gmail.com>
* Add scaling for Texture2DArray when using TexelFetch.
* Should only really trigger for Texture2D. (do not emit texelfetch for
TextureBufferArray(?) and Texture1DArray)
* Address nit.
* 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
* Replace Host FPS with FIFO%
* Change measurement order. Improve calculation.
Now at 100% when FIFO is blocking game exectution, rather than "0".
* Address feedback (1)
* Remove Host FPS
* FIFO rather than Fifo
* Address Ac_k feedback
* Rebase
* Return "NotAvailable" when no UserChannel data is present.
* Return ObjectInvalid for undefined parameter kinds.
* No need to specify which, there's only one.
* Just works as a literal string.
* Add swizzle matching rules.
Improves rules which try to match incompatible formats as perfect, such as D32 float -> R32 float.
Remove Format.HasOneComponent, since this information is now available via the FormatInfo struct.
* Fix this rule.
* Update component counts for depth formats.
* Initial implementation. Still pending better valid-overlap handling,
disposed pool, compressed format flush fix.
* Very messy backend resource cache.
* Oops
* Dispose -> Release
* Improve Release/Dispose.
* More rule refinement.
* View compatibility levels as an enum - you can always know if a view is only copy compatible.
* General cleanup.
Use locking on the resource cache, as it is likely to be used by other threads in future.
* Rename resource cache to resource pool.
* Address some of the smaller nits.
* Fix regression with MK8 lens flare
Texture flushes done the old way should trigger memory tracking.
* Use TextureCreateInfo as a key.
It now implements IEquatable and generates a hashcode based on width/height.
* Fix size change for compressed+non-compressed view combos.
Before, this could set either the compressed or non compressed texture with a size with the wrong size, depending on which texture had its size changed. This caused exceptions when flushing the texture.
Now it correctly takes the block size into account, assuming that these textures are only related because a pixel in the non-compressed texture represents a block in the compressed one.
* Implement JD's suggestion for HashCode Combine
Co-authored-by: jduncanator <1518948+jduncanator@users.noreply.github.com>
* Address feedback
* Address feedback.
Co-authored-by: jduncanator <1518948+jduncanator@users.noreply.github.com>
* Add CRC32 A32 instructions.
* Fix CRC32 instructions.
* Add CRC intrinsic and fast path.
Loop is currently unrolled, will look into adding temp vars after tests are added.
* Begin work on Crc tests
* Fix SSE4.2 path for CRC32C, finialize tests.
* Remove unused IR path.
* Fix spacing between prefix checks.
* This should be Src.
* PTC Version
* OpCodeTable Order
* Integer check improvement. Value and Crc can be either 32 or 64 size.
* This wasn't necessary...
* If size is 3, value type must be I64.
* Fix same src+dest handling for non crc intrinsics.
* Pre-fix (ha) issue with vex encodings
* Initial implementation of Render Target Scaling
Works with most games I have. No GUI option right now, it is hardcoded.
Missing handling for texelFetch operation.
* Realtime Configuration, refactoring.
* texelFetch scaling on fragment shader (WIP)
* Improve Shader-Side changes.
* Fix potential crash when no color/depth bound
* Workaround random uses of textures in compute.
This was blacklisting textures in a few games despite causing no bugs. Will eventually add full support so this doesn't break anything.
* Fix scales oscillating when changing between non-native scales.
* Scaled textures on compute, cleanup, lazier uniform update.
* Cleanup.
* Fix stupidity
* Address Thog Feedback.
* Cover most of GDK's feedback (two comments remain)
* Fix bad rename
* Move IsDepthStencil to FormatExtensions, add docs.
* Fix default config, square texture detection.
* Three final fixes:
- Nearest copy when texture is integer format.
- Texture2D -> Texture3D copy correctly blacklists the texture before trying an unscaled copy (caused driver error)
- Discount small textures.
* Remove scale threshold.
Not needed right now - we'll see if we run into problems.
* All CPU modification blacklists scale.
* Fix comment.
* Add Vmvn (register), tests for both Vmvn variants.
* Add Vpmin, Vpmax, improve Non-FastFp accuracy for Vpadd
* Rebase on top of PTC.
* Add Nopcode
* Increment PTC version.
* Fix nits.
* Improve performance when converting texture formats.
Still more work to do.
* Speed up buffer -> texture copies.
No longer copies byte by byte. Fast path when formats are identical.
* Fix a few things, 64 byte block fast copy.
* Spacing cleanup, unrelated change.
* Fix base offset calculation for region copies.
* Fix Linear -> BlockLinear
* Fix some nits. (part 1 of review feedback)
* Use a generic version of the Convert* functions rather than lambdas.
This is some real monkey's paw shit.
* Remove unnecessary span constructor.
* Revert "Use a generic version of the Convert* functions rather than lambdas."
This reverts commit aa43dcfbe8bba291eea4e10c68569af7a56a5851.
* Fix bug with rectangle destination writing, better rectangle calculation for linear textures.
* Implementation of query queue and host conditional rendering
* Resolve some comments.
* Use overloads instead of passing object.
* Wake the consumer threads when incrementing syncpoints.
Also, do a busy loop when awaiting the counter for a blocking flush, rather than potentially sleeping the thread.
* Ensure there's a command between begin and end query.
* Use the jump table for tail continues.
This path is always reached when a function larger than our current length limit (currently 5000) is compiled.
* Use Call Flag rather than 1L
* Index constant buffer vec4s using ternary expressions.
* Remove indexed path.
We determined that it had negligible impact.
* Revert "Remove indexed path."
This reverts commit 25ec4eddfa441e802bd957dfaabc83b23c6bae38.
* Revert "Revert "Remove indexed path.""
This reverts commit 7cd52fecb529dcb9e1a574533bd38531319f1268.
* Start of JIT garbage collection improvements
- thread static pool for Operand, MemoryOperand, Operation
- Operands and Operations are always to be constructed via their static
helper classes, so they can be pooled.
- removing LinkedList from Node for sources/destinations (replaced with
List<>s for now, but probably could do arrays since size is bounded)
- removing params constructors from Node
- LinkedList<> to List<> with Clear() for Operand assignments/uses
- ThreadStaticPool is very simple and basically just exists for the
purpose of our specific translation allocation problem. Right now it
will stay at the worst case allocation count for that thread (so far) -
the pool can never shrink.
- Still some cases of Operand[] that haven't been removed yet. Will need
to evaluate them (eg. is there a reasonable max number of params for
Calls?)
* ConcurrentStack instead of ConcurrentQueue for Rejit
* Optimize some parts of LSRA
- BitMap now operates on 64-bit int rather than 32-bit
- BitMap is now pooled in a ThreadStatic pool (within lrsa)
- BitMap now is now its own iterator. Marginally speeds up iterating
through the bits.
- A few cases where enumerators were generated have been converted to
forms that generate less garbage.
- New data structure for sorting _usePositions in LiveIntervals. Much
faster split, NextUseAfter, initial insertion. Random insertion is
slightly slower.
- That last one is WIP since you need to insert the values backwards. It
would be ideal if it just flipped it for you, uncomplicating things on
the caller side.
* Use a static pool of thread static pools. (yes.)
Prevents each execution thread creating its own lowCq pool and making me cry.
* Move constant value to top, change naming convention.
* Fix iteration of memory operands.
* Increase max thread count.
* Address Feedback
* Implement Jump Table for Native Calls
NOTE: this slows down rejit considerably! Not recommended to be used
without codegen optimisation or AOT.
- Does not work on Linux
- A32 needs an additional commit.
* A32 Support
(WIP)
* Actually write Direct Call pointers to the table
That would help.
* Direct Calls: Rather than returning to the translator, attempt to keep within the native stack frame.
A return to the translator can still happen, but only by exceptionally
bubbling up to it.
Also:
- Always translate lowCq as a function. Faster interop with the direct
jumps, and this will be useful in future if we want to do speculative
translation.
- Tail Call Detection: after the decoding stage, detect if we do a tail
call, and avoid translating into it. Detected if a jump is made to an
address outwith the contiguous sequence of blocks surrounding the entry
point. The goal is to reduce code touched by jit and rejit.
* A32 Support
* Use smaller max function size for lowCq, fix exceptional returns
When a return has an unexpected value and there is no code block
following this one, we now return the value rather than continuing.
* CompareAndSwap (buggy)
* Ensure CompareAndSwap does not get optimized away.
* Use CompareAndSwap to make the dynamic table thread safe.
* Tail call for linux, throw on too many arguments.
* Combine CompareAndSwap 128 and 32/64.
They emit different IR instructions since their PreAllocator behaviour
is different, but now they just have one function on EmitterContext.
* Fix issues separating from optimisations.
* Use a stub to find and execute missing functions.
This allows us to skip doing many runtime comparisons and branches, and reduces the amount of code we need to emit significantly.
For the indirect call table, this stub also does the work of moving in the highCq address to the table when one is found.
* Make Jump Tables and Jit Cache dynmically resize
Reserve virtual memory, commit as needed.
* Move TailCallRemover to its own class.
* Multithreaded Translation (based on heuristic)
A poor one, at that. Need to get core count for a better one, which
means a lot of OS specific garbage.
* Better priority management for background threads.
* Bound core limit a bit more
Past a certain point the load is not paralellizable and starts stealing from the main thread. Likely due to GC, memory, heap allocation thread contention. Reduce by one core til optimisations come to improve the situation.
* Fix memory management on linux.
* Temporary solution to some sync problems.
This will make sure threads exit correctly, most of the time. There is a potential race where setting the sync counter to 0 does nothing (counter stays at what it was before, thread could take too long to exit), but we need to find a better way to do this anyways. Synchronization frequency has been tightened as we never enter blockwise segments of code. Essentially this means, check every x functions or loop iterations, before lowcq blocks existed and were worth just as much. Ideally it should be done in a better way, since functions can be anywhere from 1 to 5000 instructions. (maybe based on host timer, or an interrupt flag from a scheduler thread)
* Address feedback minus CompareAndSwap change.
* Use default ReservedRegion granularity.
* Merge CompareAndSwap with its V128 variant.
* We already got the source, no need to do it again.
* Make sure all background translation threads exit.
* Fix CompareAndSwap128
Detection criteria was a bit scuffed.
* Address Comments.
* Implement TEQ and MOV (Imm16)
* Initial work on A32 instructions + SVC. No tests yet, hangs in rtld.
* Implement CLZ, fix BFI and BFC
Now stops on SIMD initialization.
* Exclusive access instructions, fix to mul, system instructions.
Now gets to a break after SignalProcessWideKey64.
* Better impl of UBFX, add UDIV and SDIV
Now boots way further - now stuck on VMOV instruction.
* Many more instructions, start on SIMD and testing framework.
* Fix build issues
* svc: Rework 32 bit codepath
Fixing once and for all argument ordering issues.
* Fix 32 bits stacktrace
* hle debug: Add 32 bits dynamic section parsing
* Fix highCq mode, add many tests, fix some instruction bugs
Still suffers from critical malloc failure 😩
* Fix incorrect opcode decoders and a few more instructions.
* Add a few instructions and fix others. re-disable highCq for now.
Disabled the svc memory clear since i'm not sure about it.
* Fix build
* Fix typo in ordered/exclusive stores.
* Implement some more instructions, fix others.
Uxtab16/Sxtab16 are untested.
* Begin impl of pairwise, some other instructions.
* Add a few more instructions, a quick hack to fix svcs for now.
* Add tests and fix issues with VTRN, VZIP, VUZP
* Add a few more instructions, fix Vmul_1 encoding.
* Fix way too many instruction bugs, add tests for some of the more important ones.
* Fix HighCq, enable FastFP paths for some floating point instructions
(not entirely sure why these were disabled, so important to note this
commit exists)
Branching has been removed in A32 shifts until I figure out if it's
worth it
* Cleanup Part 1
There should be no functional change between these next few commits.
Should is the key word. (except for removing break handler)
* Implement 32 bits syscalls
Co-authored-by: riperiperi <rhy3756547@hotmail.com>
Implement all 32 bits counterparts of the 64 bits syscalls we currently
have.
* Refactor part 2: Move index/subindex logic to Operand
May have inadvertently fixed one (1) bug
* Add FlushProcessDataCache32
* Address jd's comments
* Remove 16 bit encodings from OpCodeTable
Still need to catch some edge cases (operands that use the "F" flag) and
make Q encodings with non-even indexes undefined.
* Correct Fpscr handling for FP vector slow paths
WIP
* Add StandardFPSCRValue behaviour for all Arithmetic instructions
* Add StandardFPSCRValue behaviour to compare instructions.
* Force passing of fpcr to FPProcessException and FPUnpack.
Reduces potential for code error significantly
* OpCode cleanup
* Remove urgency from DMB comment in MRRC
DMB is currently a no-op via the instruction, so it should likely still
be a no-op here.
* Test Cleanup
* Fix FPDefaultNaN on Ryzen CPUs
* Improve some tests, fix some shift instructions, add slow path for Vadd
* Fix Typo
* More test cleanup
* Flip order of Fx and index, to indicate that the operand's is the "base"
* Remove Simd32 register type, use Int32 and Int64 for scalars like A64 does.
* Reintroduce alignment to DecoderHelper (removed by accident)
* One more realign as reading diffs is hard
* Use I32 registers in A32 (part 2)
Swap default integer register type based on current execution mode.
* FPSCR flags as Registers (part 1)
Still need to change NativeContext and ExecutionContext to allow
getting/setting with the flag values.
* Use I32 registers in A32 (part 1)
* FPSCR flags as registers (part 2)
Only CMP flags are on the registers right now. It could be useful to use
more of the space in non-fast-float when implementing A32 flags
accurately in the fast path.
* Address Feedback
* Correct FP->Int behaviour (should saturate)
* Make branches made by writing to PC eligible for Rejit
Greatly improves performance in most games.
* Remove unused branching for Vtbl
* RejitRequest as a class rather than a tuple
Makes a lot more sense than storing tuples on a dictionary.
* Add VMOVN, VSHR (imm), VSHRN (imm) and related tests
* Re-order InstEmitSystem32
Alphabetical sorting.
* Address Feedback
Feedback from Ac_K, remove and sort usings.
* Address Feedback 2
* Address Feedback from LDj3SNuD
Opcode table reordered to have alphabetical sorting within groups,
Vmaxnm and Vminnm have split names to be less ambiguous, SoftFloat nits,
Test nits and Test simplification with ValueSource.
* Add Debug Asserts to A32 helpers
Mainly to prevent the shift ones from being used on I64 operands, as
they expect I32 input for most operations (eg. carry flag setting), and
expect I32 input for shift and boolean amounts. Most other helper
functions don't take Operands, throw on out of range values, and take
specific types of OpCode, so didn't need any asserts.
* Use ConstF rather than creating an operand.
(useful for pooling in future)
* Move exclusive load to helper, reference call flag rather than literal 1.
* Address LDj feedback (minus table flatten)
one final look before it's all gone. the world is so beautiful.
* Flatten OpCodeTable
oh no
* Address more table ordering
* Call Flag as int on A32
Co-authored-by: Natalie C. <cyuubiapps@gmail.com>
Co-authored-by: Thog <thog@protonmail.com>
* Only enumarate cached textures that are modified when flushing, rather than all of them.
* Remove locking.
* Add missing clear.
* Remove texture from modified list when data is disposed.
In case the game does not call either flush method at any point.
* Add ReferenceEqualityComparer from jD for the HashSet
riperiperiandGitHub
