previous logic doesn't work if the clipping shape
has any inner corners. replace the logic
with a intermediate composition approach for
stability.
- performance can be drop at texture clipping by ~11%
- size is reduced by -0.5kb
issue: https://github.com/thorvg/thorvg/issues/3520
The old approach often produced incorrect results,
especially when the fixed pixel had a noticeably
different color from the texture due to the AA
target blending position was fixed.
Although the previous method worked well as an analytical AA
solution with good speed and fair quality, it couldn't
overcome the above limitation.
The new approach still applies AA only to polygon edges
for efficiency. While the quality may be slightly reduced,
it offers greater stability.
- binary size: -1.1kb
- performance diff: ignoreable
issue: https://github.com/thorvg/thorvg/issues/1729
- Replaced `modff()` with a custom method,
boosting texture mapping performance by ~15%.
- Unified opacity/non-opacity logic for improved
binary size efficiency(-0.5kb).
- Implemented minor changes for better cache effectiveness.
This commit has two purposes:
- refactoring to introduce y indexing method for the upcoming partial rendering.
- replaces the RLE-specific memory allocation with a shared array structure,
eliminating potential memory overflows during RLE clipping.
redefiend properties so that min/max are prioritized,
as they are accessed more frequently than pos/size
during rendering calculations.
also introduced miscellaneous functions to improve usability.
In some clipping cases, the memory allocated for storing spans
was too small. As a result, the entire clipped area might not
have been rendered.
This has been resolved by adding an experimental factor to increase
the size of allocated memory.
@issue: https://github.com/thorvg/thorvg/issues/3461
Added drawing exceptions when target is not properly ready.
Now, Canvas::update() Canvas::draw() will return InsufficientCondition
if the target has not been set beforehand.
revert the alpha skip condition introduced in
commit 2990e72b23.
color values with alpha = 0 must be transformed to 0
to ensure correct intermediate composition results.
Simplified parameters and ensured proper backend engine
initialization by using reference checking through canvas
instances.
C++ API Modification:
- Result Initializer::init(uint32_t threads, CanvasEngine engine)
-> Result Initializer::init(uint32_t threads)
- Result Initializer::term(CanvasEngine engine)
-> Result Initializer::term()
C API Modification:
- Tvg_Result tvg_engine_init(Tvg_Engine engine_method, unsigned threads)
-> Tvg_Result tvg_engine_init(unsigned threads);
- Tvg_Result tvg_engine_term(Tvg_Engine engine_method)
-> Tvg_Result tvg_engine_term()
issue: https://github.com/thorvg/thorvg/issues/3116
The following is a redesign that extends the class internally.
The main goal is to preserve the characteristics of the pImpl idiom
for data encapsulation, while simultaneously reducing the memory
allocation overhead typically associated with pImpl.
The stragegy is here:
Rather than alloc the impl memory inside of the thorvg engine,
impl extends the API classes in order to consolidate the memory.
size has been decreased by -4kb with optimization=s
issue: https://github.com/thorvg/thorvg/issues/3214
ThorVG did not consider updates to the clipping path
when a paint object was modified after being added to the scene.
This fix ensures that any updates to clipping are correctly
reflected during rendering.
issue: https://github.com/thorvg/thorvg/issues/3403
- modify the concept of AABB to apply only to transformed shapes.
- transform points before computing the bounding box min/max
to obtain a more compact shape region.
- trimmming memory by removing the cached matrix, about 36kb
of memory has been reduced per paint instance.
Generating the outline may return false, e.g., if the shape
is completely trimmed. If such a shape was used as a clip,
an attempt to generate 'rle' will still be made. In such a case,
a crash will occur because the outline is nullptr.
Until now, only stroke trimming was possible,
but the Lottie standard requires support for
trimming both (if present), stroke and fill.
This commit changes the API behavior.
@Issue: https://github.com/thorvg/thorvg/issues/3118
Before the C++ compiler couldn't decide the template of std::min (that happen if `int` type is not an `int32_t`), so explicitly changing all types to int32_t fixes this.
Support the bindings to be more integrable with a system's coherent memory management.
Pleaes note that thorvg now only allow the desinated memory allocators here:
malloc -> tvg::malloc
calloc -> tvg::calloc
realloc -> tvg::realloc
free -> tvg::free
issue: https://github.com/thorvg/thorvg/issues/2652
In cases where the fill was changed while the stroke existed
but remained unchanged, the stroke would disappear because it
was being reset during the shape preparing (shapeReset).
Fixed by disabling the reset of stroke rle from shape reseting.
Also the shape should be prepared not only when the RenderUpdateFlag
is set to Color, but also when it is set to Gradient.
@Issue: https://github.com/thorvg/thorvg/issues/3237
We have improved the functionality to manage the memory pool safely,
by figuring out the current working threads. Users no longer need to
manually configure memory pool management, and the related APIs
have been removed.
API Removals:
- Result SwCanvas::mempool(MempoolPolicy policy)
- enum SwCanvas::MempoolPolicy
- Tvg_Result tvg_swcanvas_set_mempool(Tvg_Canvas* canvas, Tvg_Mempool_Policy policy)
- enum Tvg_Mempool_Policy
issue: https://github.com/thorvg/thorvg/issues/3116
The trim handling from sw_engine has been removed.
Instead the common logic from the TrimPath structure
is used.
On the sw_engine side, the dashed outline is now
used for both: dashed and/or trimmed strokes.
@Issue: https://github.com/thorvg/thorvg/issues/2854
Dodge d / (1 - s):
- handle d = 0 first to avoid 0/0 when s = 1
- for d != 0 and s = 1, the formula results in d/0 (infinity),
which maps to 1 in the 0-1 range (or 255 in the 0-255 range)
Burn 1 - (1 - d) / s:
- handle d = 1 first to avoid 0/0 when s = 0
- for d != 1 and s = 0, the formula results in 1 - 0/0 -> 1 - inf = -inf
Negative infinity in the 0-1 range corresponds to the minimum possible
value, which is 0 (255 if we operate in 0-255 range).
Applied the compromised approach for the gaussian blur
since the effect is a bit burdensome for cpu processing
as animatable effects.
- Optimized performance and quality with negligible observable differences.
- Disabled the border option until specific use cases are identified.
