1. new tesseletor and stroker are used: less vertexes generated
In general, the previous implementation was based on the path-outline-mesh approach.
It has now been changed to a path-mesh approach, so we skip the path-outline transformation.
For shape fills, a BW-tesselator now used, and all submeshes (moveTo) are stored in a single buffer.
For strokes, all intermediate operations such as trimming and dash use path-path logic instead of outline-outline logic.
In addition, the new stroker generates fewer polygons for joints, especially for Rounds
2. render all sub-shapes by single draw call
https://github.com/thorvg/thorvg/issues/3557https://github.com/thorvg/thorvg/issues/3288https://github.com/thorvg/thorvg/issues/3273
Manage the global buffer memory for vertex and indexed vertex buffers,
increase the memory size incrementally twice by default and reduce
the default buffer size, which is not suitable for typical scenarios.
This could reduce the a bit stack memory usage and improve
the portability across systems where has the stack memory
limitation and potentially gaining performance enhancement
by avoiding brutal stack memory usage at the many function calls.
added the internal functions:
- WgVertexBuffer* mpoolReqVertexBuffer(float scale = 1.0f);
- WgIndexedVertexBuffer* mpoolReqIndexedVertexBuffer(float scale = 1.0f);
- void mpoolRetVertexBuffer(WgVertexBuffer* buffer);
- void mpoolRetIndexedVertexBuffer(WgIndexedVertexBuffer* buffer);
issue: https://github.com/thorvg/thorvg/issues/3159
Streaming model for massive vertex and index creations: minimize memory allocations, range checks and other conditions
Reduce number of segments length calculations (sqrt) and bbox (min and max).
Update distances and bboxes on a whole buffer and only if necessary. For shapes without strokes compute distances not necessary at all. bbox can be updated only on the final stage of geometry workflow, but not on the each stage.
Using stack memory instead of heap. its more cache friendly and did not fragment memory, faster memory allocations (weak place of realization)
Using cache for points distances and whole path length. Updates only if necessary
Validation of geometry consistency executes only on the final stage of path life cicle. It more friendly for data streaming: no any conditions and branches.
Using binary search for strokes trimming
Pre-cached circles geometry for caps and joints
Refactored strokes elements generation functions. Code is more readable and modifiable in general. Can be easily fixed if some geometry issues will be finded
Fixed detection of close vertices using comparison with epsilon and related artifacts in stencil buffer as extra pixels.
Fixed incorrect tessellation of curves using scaling.
Deep shader refactoring for the following purposes:
* used pre-calculated gradient texture instead of per-pixel gradient map computation
* used HW wrap samples for fill spread setting
* unified gradient shader types
* used single shader module for composition instead of signle module per composition type
* used single shader module for blending for each of fill type (solid, gradient, image) instaed of signle module per blend type
* much easier add new composition and blend equations
* get rided std::string uasge
* shaders code is more readable
* bind groups creation in real time removed - performance boost
* blend and composition shaders decomposed - performance boost
* shader modules and pipeline layouts generalized - less memory usage
* shared single stencil buffer used - less memory usage
* bind groups usage simplified
* general context API simplified and generalized
* all rendering logic moved into new composition class
* ready for hardware MSAA (in next steps)
* ready for direct mask applience (in next steps)
[issues 1479: ClipPath](#1479)
Supports ClipPath composition.
Clip path composition is an only composition type who doesn't ignore blend method.
Clip path is a combination of composition approach and blend approach using compute shader
deprecate the `identifier()` APIs by replacing them with `type()`.
ThorVG is going to introduce an instance `id()`,
and this could be confused with the `identifier()` methods.
with this new type() method can reduce the memory size
by removing unncessary type data.
New Experimental C APIs:
- enum Tvg_Type
- Tvg_Result tvg_paint_get_type(const Tvg_Paint* paint, Tvg_Type* type)
- Tvg_Result tvg_gradient_get_type(const Tvg_Gradient* grad, Tvg_Type* type)
New Experimental C++ APIs:
- Type Paint::type() const
- Type Fill::type() const
- Type LinearGradient::type() const
- Type RadialGradient::type() const
- Type Shape::type() const
- Type Scene::type() const
- Type Picture::type() const
- Type Text::type() const
Deprecated C APIs:
- enum Tvg_Identifier
- Tvg_Result tvg_paint_get_identifier(const Tvg_Paint* paint, Tvg_Identifier* identifier)
- Tvg_Result tvg_gradient_get_identifier(const Tvg_Gradient* grad, Tvg_Identifier* identifier)
Deprecated C++ APIs:
- enum class Type
- uint32_t Paint::identifier() const
- uint32_t Fill::identifier() const
- static uint32_t Picture::identifier()
- static uint32_t Scene::identifier()
- static uint32_t Shape::identifier()
- static uint32_t LinearGradient:identifier()
- static uint32_T RadialGradient::identfier()
Removed Experimental APIs:
- static uint32_t Text::identifier()
issue: https://github.com/thorvg/thorvg/issues/1372
Skip shapes rendering, if opacity is 0 and if fill color for shape and strokes also equal to 0
This behavior is used in sw renderer and fix visual artifacts in referenced animations.
Also this rule fix composition results in case of AlphaMask and InvAlphaMask methods
[issues 1479: lottie](#1479)
- optimaze stroking algorithm by prevent vector normalizations
- using render meshes heaps to reduce webgpu instances re-creations
- using single instance of pilylines for standard operations such as trim and split
- "on-the-fly" strokes generations with dash pattern
- "on-the-fly" mesh objects generation in a process of path decoding
- merge strokes into single mesh by each shape
[issues 1479: lottie](#1479)
Vertex, Index and uniform buffers now updates instead of recreate.
Implemented pools form mesh objects and render shapes data
it increase performance in 30-40% in massive animations scenes
Before the current changes, all surfaces were painted using a full-screen overlay, no matter how large the object was rendered. This approach is redundant and required reorganization. At the moment, all objects are rendered using an overlay equal to the box of the object itself, which reduces the cost of filling the surface.
Also surfaces and images were divided into different entities, which reduces the pressure on memory.
Also geometry data for rendering and geometry data for calculations in system memory were logically separated.
