- Switch to Y X Z ordering
- In theory this will be more coherent since the first lut step on the
GPU will have a more constrained range of values in the worst case
- LightLut now does incremental updates java-side
- Still requires a full upload when changed, though it does not take up
much space
- ShaderLightVisualStorage now actually triggers removal of light
section from the lut
- Don't initialize instancers if they have 0 instances
- Never shrink the index pool
- Actually process recently allocated meshes to avoid growing the list
forever
- Fix hiz test bounds clamping, good ol off by 1 error
- Remove check for useMin since it's never used and the depth reduce
shader is hard-coded already
- Replace monolithic lut building function with a class representing a
layer of the lut
- Actually need 2 classes because int[] and Object[] aren't trivial to
make a type parameter, and we don't really want to be boxing ints here
- No longer need sorted inputs
- Should fix index out of bounds crash caused by reserving space for the
wrong index layer
- This will make it much easier to change the coordinate ordering scheme
- Clean up VertexWriter to closer match BufferBuilder
- Clean up MeshEmitter on Fabric/NeoForge and remove BufferBuilderAccessor
- Remove model data and model data lookup parameters in NeoForge model builders; use IBlockGetterExtension#getModelData instead
- Fix artifact Minecraft version
* Automated testing
* Testing testing
- Use 2 spaces for indents yaml
- Move setupTestMod to PlatformExtension
- Allow specifying the sourceSet for the testMod artifact
- Rename things to camelCase
- Use rootCompile from transitiveSourceSets for the testMod source sets
- Use a blanket remapTestModJar task in the gh actions build
* Fail slowly
- We want to know the results of both tests regardless
* Add workflow dispatch
* Shoes should be steel toed, dangerous stuff
* Update build.yml
* fix modid
* Update FlywheelTestModClient.java
* add debug logging
* fix syntax issues
* fix issues
* Update build.yml
* Add debug logging
* more logging
* get testmod from correct dir
* switch to env var
* Why wait?
- Immediately audit on client tick
* DidObfuscate
- Fix RenderSystemMixin on fabric
- setShaderFogShape's arguments need to be remapped, but the name of the
function should not be. Fortunately mixin allows matching by function
name alone
* Clever commit title
- Change the Fabric mod ID to match Forge
- Move "Flywheel Test Mod" to static
- Cleanup start/stop messages
- Use the client start event on Fabric
---------
Co-authored-by: Jozufozu <jozsefaug@gmail.com>
- Generic pain relief
- Use new Instance[] rather than capturing the class object of the
instance type
- Make InstancePage static, but manually track the instancer parent so
we can check when stealing
- Simplify array creation helpers and make them static
- Mark InstanceHandleImpl#state as UnknownNullability
- Only upload changed page frame descriptors
- In the instancer, track changed contents separately from changed
validity, so we can separately upload objects and descriptors
- Combine pages only when they're at most half full, and not empty
- This guarantees that we'll fully empty a page, allowing us to free the memory for use by other instancers
- Track mergeable pages via a separate bitset
- Try to shuffle over instances into pages with space
- Clear out now-unused logic from ObjectStorage
- Some cleanup and more comments in IndirectInstancer
- Make AbstractInstancer much more slim and move logic to BaseInstancer
- Extend paging concept to the indirect instancer
- Extend ObjectStorage to support more interesting layouts
- Instance creation on indirect is now entirely lock free and deletions
no longer require re-uploading the entire instancer
- MaterialEncoder would trigger an indexing of CutoutShaders.OFF, though
PipelineCompiler would explicitly not index OFF
- This caused a crash on instancing when MaterialEncoder would delete
all pipeline shaders while instancing was trying to upload the packed
ubershader uniform
* Backport changes from 1.21.1
* fix
* Fix building
* fix compile error
* fix
* fix build for real
* address reviews
* Fix sodium compat
* address requested changes
* mark rubidium as incompatible
* add missed call
* Should have worn steel toe boots
- Add "stub" sourceset to each subproject
- Directly pass vararg sourcesets to methods in PlatformExtension to
avoid automatically shipping jars with the api stubs
- We may have to include stubs in setupLoomMod, but I don't think so
- A lot of this can be stripped back out if we don't need stub sources
for the forge/fabric subprojects
* Guarded stubs
- Add Sodium 0.6 and Iris API stubs to stubs source set and remove Gradle dependencies on local Sodium jar, Iris, and Oculus
- Ensure usage of APIs that may not exist at runtime is in private classes and access is always guarded
- Change ShadersModHandler
- Rename to ShadersModHelper
- Convert methods to check for Iris' and Optifine's presence into static final fields
- Move implementation to impl source set in form of IrisCompat and OptifineCompat classes
- Rename CompatMods to CompatMod and add public field to access mod ID
- Set BlockEntityType's Sodium predicate to null after it is removed
- Update repository links
- Remove local libs repository
---------
Co-authored-by: Jozufozu <jozsefaug@gmail.com>
Co-authored-by: PepperCode1 <44146161+PepperCode1@users.noreply.github.com>
- Flatten TextVisual API to be less of a parameter container
- This is mostly a surface change
- Do not accept x and y positions, instead force the user to translate
in their matrix
- Track light from light updates to apply when setting up text
- Add (unoptimized) updateLight, backgroundColor, and updateObfuscated
methods to TextVisual
- Save 8 bytes per glyph instance by switching uvs to
FloatRepr.NORMALIZED_UNSIGNED_SHORT
- With vanilla's 256x glyph atlases we could technically save another 4
bytes, but that would make compat with caxton, which uses 4096x
atlases, much more difficult later
- TextVisual renders an arbitrary number of layers
- Layers have relatively fine control over how the glyph instances are
set up
- Encapsulate all fields in TextVisual
- Move Sinks to a thread local
- Fix glowing sign rendering
- Use 2 instances per character
- Move most glyph mesh state into the GlyphMode enum
- GlyphSettings now has a GlyphMode and a bold boolean
- Disable diffuse for text
- Render normal text with polygon offset and bias the instancers for
normal text so they appear in front of shadows or outlines
- Make sink static
- Fix glyph instance color being unnormalized
- Get text instancing minimally working
- Use special glyph instance type to instance all characters under the
same mesh
- Unfortunately MC's font stuff is not threading friendly so hacks were
required to allow visuals to query glyphs from their beginFrame
- Mixin to CodePointMap to overwrite most methods and synchronize
- Mixin to FontSet to use a custom AsyncFontTexture that defers texture
init and upload
- Trying to mixin directly to FontTexture or subclass it is very
complicated because of the calls to getId, but I think that can be
improved
- Use vanilla light directions for diffuse lighting
- Copy mc's glsl code for it, but assume directions are normalized
- Add command/config to toggle use of light directions vs chunk accurate
diffuse
- Always use shade in getItemMaterial
- Do not reload resource packs when updating light smoothness config,
we don't need to anymore with lazy compilation
- Decide not to render entities directly in the renderEntity method
- Prevents allocating large lists every frame to filter entities from
the client level
- Add pick glint material and system time uniform
- Move _FlwCullData to beginning of uniform block to ensure alignment
- Add helper to convert item rendertype into flywheel material
- Make UberShaderComponent#build NotNull
- Move index update and key creation logic to PipelineCompiler
- Always update index when a resource location is requested to fix
MaterialEncoder misses
- Indices trigger pipeline compiler deletion when updated
- Make everything in the compiler chain's results not null
- Throw errors immediately when encountered
- Log error messages when falling back
- Do not eagerly grab utility programs in IndirectDrawManager so we can
actually catch errors and fall back
- Remove CompilerStats
- Remove fog shader registry
- Remove Registry and RegistryImpl
- Make shader indices mutable
- Track fog uber component in a static field in PipelineCompiler
- When a new fog source is added, delete the pipeline compilation
harness and recreate the fog uber component
- Inline SourceLoader
- Strip out almost all source registries
- Fog will be dealt with in a follow-up commit
- Remove most static #init methods
- Remove old ubershader indices from shaders
- Hidden state now tracks the Instance object to keep the handle small
- Make the recreate supplier an explicit record to allow comparisons
- Add setVisible method to Instance
- Use state machine interface in InstanceHandleImpl
- 3 states: deleted, visible, hidden
- Visible is directly implemented by AbstractInstancer
- Hidden stores the instancer supplier to recreate an instancer
- Eagerly load ALL shaders in ShaderSources, resolving imports there
- Compile and cache programs on-demand
- Move gl state try blocks to EngineImpl
- EngineImpl catches shader exceptions and triggers a fallback
- Fix crumbling on indirect
- Directly use the baseInstance as instance index without indirection
- #define base instance and draw id variables to simplify usage
- Fix null pointer looking up culling group
- Add method to map an instancer's local instance index to a global
index in the page file
- Remove ModelHolder and ModelCache
- Remove lib/util.FlwUtil
- Remove lib/util.Pair and replace usages with com.mojang.datafixers.util.Pair
- Remove lib/util.Unit and replace usages with net.minecraft.util.Unit
- Make ResourceReloadHolder and ResourceReloadCache final and move to util
- Clean up code in backend/glsl
- Move LightSmoothnessArgument to impl
- Remove LoweringVisitor
- Move functionality of four main static methods in LoweringVisitor to new ModelTrees class
- Return ModelTree directly
- Accept Material instead of TextureAtlasSprite for efficiency, so visuals don't need to look up the sprite to get the ModelTree
- Use ResourceReloadCache for MeshTree.CACHE
- Implement instance hiding by deleting/stealing
- Work around instancer persistence by storing a recreation supplier in
the instance handle
- Rework instancer ctors to just take an InstancerKey
- Parameterize InstanceHandle by I extends Instance so the steal method
and the supplier can be safely assigned
- IndirectInstancer#uploadInstances: 46% of render thread to 26%
- Inline #enqueueCopy to avoid allocating LongConsumers
- Do not even bother to track individual changed indices, instead rely
on just the changedPage set
- Convert ShulkerBoxVisual to use InstanceTree
- Add "pruning" helper visitors
- Remove ModelPartConverter
- Remove TextureMapper and related code from VertexWriter
- Add ModelTree
- Add LoweringVisitor to traverse a MeshTree and emit ModelTree nodes
and Models
- Provide some default visitor creation methods
- Abstract ModelCache -> ResourceReloadCache
- Abstract ModelHolder -> ResourceReloadHolder
- Add ModelTreeCache to hide lookup cost if it gets extreme
- Use InstanceTrees in BellVisual and MinecartVisual
- Use JOML Matrix4fStack instead of PoseStack
- Directly transform Matrix4f instead of using PoseStack to compute initial pose
- Track if the transforms for an InstanceTree have changed
- Pass a boolean down the tree and || it with our changed flag to force
updates
- Expose the force flag to visuals so they can hint to us if their root
transforms never change
- Add 2 wrapper methods to make the distinction more clear
- Store a pose matrix in each InstanceTree, equivalent to its instance's pose matrix if the instance exists
- Directly transform the current InstanceTree's pose matrix instead of transforming a PoseStack and copying its matrix to the instance, eliminating the need to push and pop stack entries
- Remove InstanceTree.rotation
- Add more InstanceTree methods to allow full inspection of children
- Rename TransformedInstance -> PosedInstance
- Add TransformedInstance, which does not have a normal matrix
- Use mat3(transpose(inverse(i.pose))) in the vertex shader, but for
most cases that will be overkill
- Use arrrays in the trees
- MeshTree tracks parallel arrays of children and keys
- InstanceTree tracks which mesh tree it came from for lookup purposes
- Remove walker object
- Make RecyclingPoseStack use add/removeLast instead of push/pop
- Add MeshTree and InstanceTree
- Deprecate ModelPartConverter for removal
- Refactor ChestVisual to use InstanceTree
- Combine double chest light in ChestVisual
- Cherry-pick misc cleanups from last-frame-visibility
- Smarter multibind logic
- Make offsets in IndirectBuffers dependent on BufferBindings
- Organize buffer bindings based on where they're used to allow each
pass to bind exactly which buffers it needs
- Use DSA for the depth pyramid
- Pass the map of util programs to IndirectPrograms rather than
unpacking them individually
- Actually delete all the indirect utils
- Fix mip levels being half the size they should be
- Use the next lowest po2 from the main render target size for mip 0
- Map from dst texel to src texel rather than naively multiply by 2
- Clamp the estimated mip level in the cull shader
- Use texel fetches in the cull shader (not sure if necessary?)
- Implement hi-z occlusion culling
- Generate depth pyramid just before issuing cull dispatches
- Currently use raw texel fetches but this may be causing loss
- Add _flw_cullData to frame uniforms
- Only update the page table when an allocation is resized
- Only upload the page table after it's uploaded
- Combine various setters for InstancePager.Allocation and
IndirectInstancer
- Free pages when an allocation is deleted
- Fix bit logic on the GPU
- Manually manage the size of the storage and pageTable buffers
- Make object2Page and page2Object static
- Fix instance writing loop
- Fix page table always having full pages
- Fix allocations not shrinking
- Goal: avoid needing to re-upload everything when instance count for
one instancer changes
- Solution: store instances in pages of 32
- Allocate pages in a GPU arena
- Store one uint per page to indicate which model the instances in the
page belong to, and how many instances are actually stored in the page
- Instancers eagerly allocate and free pages as their instance count
changes
- Instancers will not necessarily store instances contiguously anymore,
but that's okay because any given cull workgroup will only reference a
single page
- Culling threads *will* write instances contiguously however, and so we
still need to keep track of a base instance per instancer, and the
target buffer logic does not change
