- Re-implement batching under new api topology
- Lots of duplicated code between instancing and batching, room for abstraction
- StructTypes now require their type argument to extend InstancedPart
- Fix crash when re-allocating drawbuffers
- Disable crumbling pending refactor
- Separate RenderDispatcher#beginFrame from RenderDispatcher#maintainOriginCoordinate
- Inline OriginShiftListener
- Refactor InstancedRenderDispatcher hooks
- InstancingEngine takes over tracking individual draw calls
- Many draw calls are associated with a single ShaderState
- Each ShaderState will bind one shader program
- Make Material a record
- Inline Renderable and move InstancedModel.Layer to DrawCall
- Rename InstanceData -> InstancedPart, and subclasses
- Burger Fences
- Fix GlStateTracker nuking vao element buffer bindings
- GlVertexArray tracks element buffer bindings
- Use vertexAttribPointer offset instead of ..BaseVertex
- Setup code for better crumbling rendering
- Move some logic into CoreShaderInfoMap
- Simplify VertexWriter/VertexList
- Prefer IEventBus#addListener to @SubscribeEvent
- Stop using persistent buffers... for now
- Fix and improve crumbling rendering
- Rename Batched to BatchedStructType and Instanced to
InstancedStructType
- Normalize flw_vertexNormal in context/common.vert
- Add TODO and FIXME comments
- Check for errors immediately after loading shaders
- Done through FileResolution
- Add class SourceChecks to
- Throw away GatherContextEvent, do everything on client init
- Unify FileResolution and Resolver, PartialModel style
- Pass around error reporter to detect errors during load
- Rename MaterialManager -> InstancerManager and Material ->
InstancerFactory
- Create Material class consisting of RenderType and shader files
- Make ModelSuppliers use Materials as keys instead of RenderTypes
- Move diffuse calculation from context shaders to material/shaded.vert
- Merge used templates directly into compilers and remove template
classes
- Overhaul shader pipeline
- Remove VertexType#getShaderHeader in favor of vertex-only "layout
shaders"
- Remove program specs in favor of vertex-only "instance shaders"
- Add API GLSL headers for vertex and fragment shaders that store
common values
- Separate context shaders into vertex and fragment shader files
- Improve import handling
- Ensure imports are not added more than once
- Change regex for import directives to be more in line with C-style
directives
- Move some classes and GLSL files
- Respect fog shape and type
- Add uFogShape uniform
- Add macro for current FogType
- Respect dimension constant ambient light when calculating diffuse
- Add uConstantAmbientLight uniform
- Redirect diffuse calculation to FLWDiffuse
- Do not automatically include diffuse.glsl
- Fill CoreShaderInfoMap with info for vanilla and Forge core shaders
- Use try-with-resources when querying ResourceManager
- Move WorldProgram uniform uploading from bind to uploadUniforms
- Use correct layer when rendering bells (cutout -> solid)
- Update Starlight
- Remove OnlyIn annotations
- Replace JSR305 Nullable and Nonnull with Jetbrains Nullable and
NotNull
- Fix indentation in flywheel.mixins.json
- Organize imports
- Inline a handful of interfaces that had limited use
- Removed IndexedModel/VBOModel as a first step in refactoring model uploads
- InstancedModels/GPUInstancers support multiple VAOs
- Fix padding issue
- Instancer VAO logic moved to InstancedModel
- Refactor GlVertexArray for easier debugging
- Sort of clean up BufferLayout
- Padding items are broken, needs fix
- Abstract ModelUtil.getBufferBuilder and .getBufferBuilderFromTemplate
- BakedModelBuilder and WorldModelBuilder as parameter objects for getBufferBuilder
- WorldModelBuilder supports IModelData
- Slight simplification to model allocation
- RenderLayers are no more
- Add basically empty RenderTypeRegistry, to be filled out as I go
- Move content of RenderLayerEvent into RenderContext
- BlockModels have PartialModel names now
- Futz around with InstancingEngine's render logic
- Fix invisibility bug caused by not unbinding vaos
- Inline all the things
- Now MaterialManager -> Material -> Instancer
- ModelSuppliers store RenderType
- Currently broken, it only renders chunk layers so vanilla instances are invisible
- User no longer needs to supply a key object to get an instancer
- Move memoization logic outside of MaterialManager
- Provide utility class for memoizing models
- Vanilla instances use static objects for their models
- PartialModels don't actually need reference states, use air instead
- Move RenderLayer to api package
- Kill lazy killable
- Only one task engine for everything now
- Fixes crash on LightUpdater init when Flywheel is off
- Attempt to wrangle InstanceWorld init code
- Misc. cleanup/renaming
- LightUpdates now uses task engine to update listeners in parallel
- Basic workgroup system to run something on the main thread after a group of tasks is complete
- Add ShadeSeparatedBufferBuilder and other code to allow separating all
non-shaded vertices to the end of the buffer
- Add ShadedVertexList to allow defining if a certain vertex is shaded
or not
- Add new methods to ModelUtil with arguments for more flexibility
- Refactor VirtualEmptyBlockGetter to allow defining arbitrary light
values
- Add shaded argument to DiffuseLightCalculator
- Fix entitiesForRendering redirect to be compatible with carpet.
- Use more reflection for dealing with optifine
- Fixes issue where flywheel would still be on immediately after enabling optifine shaders
- The batching engine may run in parallel, meaning the force diffuse
state will not be read correctly
- ModelTransformers only process vertices that will be rendered in the
world
- Add 0.6.1 to the issue template
- Remove BooleanConfig, BooleanConfigCommand, and BooleanDirective
- Make Flywheel.VERSION private so it cannot be changed
- Move createUpdateLimiter from FlwConfig to InstanceManager
- Extract update limiting behavior to interface
- Move original impl to BandedPrimeLimiter
- Add dummy NonLimiter impl
- Add command/config to toggle update limiting
- Refactor InstanceManager to be more consistent between frame updates and tick updates
- Bump version - 0.6.1
- GlCompat now a singleton, doesn't need to be re-created
- Fix crash with F3 open and backend off
- Clear program spec map before loading
- Merge FlywheelClient and Flywheel classes
- ModelPools no longer undbind their buffer during construction, as state should be maintained externally
- Call BufferUploader#reset at the end of renderChunkLayer to partially fix crash rendering contraptions
- Remove Engine#getName, it was replaced by FlwEngine
