wg_engine: clean code++

- initialize stage buffers before drawing
- confirm the surface size in sync stage

Co-authored-by: Jinny You <jinny@lottiefiles.com>

CODEOWNERS

@@ -4,14 +4,8 @@
 
 * @hermet
 /src/renderer/sw_engine @mgrudzinska
-/src/renderer/gl_engine @RuiwenTang @SergeyLebedkin
+/src/renderer/gl_engine @SergeyLebedkin @RuiwenTang
 /src/renderer/wg_engine @SergeyLebedkin
-/src/loaders/external_webp @JSUYA
-/src/loaders/raw @JSUYA
-/src/loaders/svg @JSUYA @mgrudzinska
-/src/loaders/webp @JSUYA
+/src/loaders/svg @mgrudzinska @JSUYA
 /src/loaders/lottie @mgrudzinska
-/src/bindings/capi @mgrudzinska
 /src/bindings/wasm @tinyjin
-/src/savers/gif @JSUYA
-/src/tools/svg2png @JSUYA

CONTRIBUTORS

@@ -44,3 +44,4 @@ Thaddeus Crews <repiteo@outlook.com>
 Benjamin <benjaminhalko@hotmail.com>
 Benson Muite <benson_muite@emailplus.com>
 kkocdko <kkocdko@gmail.com>
+SoonGeon Noh <nors.nsg@gmail.com>

README.md

@@ -1,5 +1,6 @@
 [![Discord](https://img.shields.io/badge/Community-5865f2?style=flat&logo=discord&logoColor=white)](https://discord.gg/n25xj6J6HM)
 [![ThorVGPT](https://img.shields.io/badge/ThorVGPT-76A99C?style=flat&logo=openai&logoColor=white)](https://chat.openai.com/g/g-Ht3dYIwLO-thorvgpt)
+[![DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/thorvg/thorvg)
 [![OpenCollective](https://img.shields.io/badge/OpenCollective-84B5FC?style=flat&logo=opencollective&logoColor=white)](https://opencollective.com/thorvg)
 [![License](https://img.shields.io/badge/licence-MIT-green.svg?style=flat)](LICENSE)
 ![BinarySize](https://img.shields.io/badge/Size->150kb-black)
@@ -98,7 +99,7 @@ ThorVG is actively under development, continuously expanding its support for ess
     - [LVGL](#lvgl)
     - [Tizen](#tizen)
   - [References](#references)
-  - [APIs](#apis)
+  - [Documentation](#documentation)
   - [Examples](#examples)    
   - [Tools](#tools)
     - [ThorVG Viewer](#thorvg-viewer)
@@ -294,7 +295,7 @@ ThorVG facilitates [SVG Tiny Specification](https://www.w3.org/TR/SVGTiny12/) re
 The figure below highlights ThorVG's SVG rendering capabilities:
 
 <p align="center">
-  <img width="780" height="auto" src="https://github.com/thorvg/thorvg/blob/main/res/example_svg.png">
+  <img width="780" height="auto" src="https://github.com/thorvg/thorvg/blob/main/res/example_svg.jpg">
 </p>
 
 The following code snippet shows how to draw SVG image using ThorVG:
@@ -420,8 +421,10 @@ ThorVG has been integrated into the [Tizen](https://www.tizen.org) platform as t
 [Back to contents](#contents)
 <br />
 <br />
-## APIs
-The ThorVG API documentation can be accessed at [thorvg.org/apis](https://www.thorvg.org/apis), and is also available in the [C++ API](https://github.com/thorvg/thorvg/blob/main/inc/thorvg.h), [C API](https://github.com/thorvg/thorvg/blob/main/src/bindings/capi/thorvg_capi.h) within this repository.
+## Documentation
+The ThorVG API documentation is available at [thorvg.org/apis](https://www.thorvg.org/apis), and can also be found directly in this repository via the [C++ API](https://github.com/thorvg/thorvg/blob/main/inc/thorvg.h) and [C API](https://github.com/thorvg/thorvg/blob/main/src/bindings/capi/thorvg_capi.h). 
+
+For comprehensive and well-structured technical information, please visit the [DeepWiki](https://deepwiki.com/thorvg/thorvg), which offers in-depth guidance on ThorVG's architecture, features, and usage.
 
 [Back to contents](#contents)
 <br />
@@ -529,14 +532,16 @@ meson setup builddir -Dbindings="capi"
 <br />
 <br />
 ## Dependencies
-ThorVG offers versatile support for image loading, accommodating both static and external loaders. This flexibility ensures that, even in environments without external libraries, users can still leverage static loaders as a reliable alternative. At its foundation, the ThorVG core library is engineered to function autonomously, free from external dependencies. However, it is important to note that ThorVG also encompasses a range of optional feature extensions, each with its specific set of dependencies. The dependencies associated with these selective features are outlined as follows:
+ThorVG provides flexible image loading capabilities, supporting both static and external loaders. This design ensures that even in environments lacking external libraries, users can rely on built-in static loaders for core functionality. At its core, the ThorVG library is fully self-contained and operates without mandatory external dependencies. However, several optional feature extensions are available, each with its own set of dependencies.
 
-* GL engine: [OpenGL v3.3](https://www.khronos.org/opengl/) or [GLES v3.0](https://www.khronos.org/opengles/)
-* WG engine: [webgpu-native](https://github.com/gfx-rs/wgpu-native)
-* External PNG support: [libpng](https://github.com/glennrp/libpng)
-* External JPG support: [turbojpeg](https://github.com/libjpeg-turbo/libjpeg-turbo)
-* External WebP support: [libwebp](https://developers.google.com/speed/webp/download)
-* Examples: [SDL2](https://www.libsdl.org/)
+The following outlines the dependencies for these optional features:
+
+* **GL Engine**: [OpenGL 3.3](https://www.khronos.org/opengl/), [OpenGL ES 3.0](https://www.khronos.org/opengles/), or a browser with [WebGL2](https://www.khronos.org/webgl/) support.
+* **WG Engine**: [webgpu-native v0.22](https://github.com/gfx-rs/wgpu-native) or a browser with [WebGPU](https://www.w3.org/TR/webgpu/) support.
+* **PNG Loader** (external): [libpng](https://github.com/pnggroup/libpng)
+* **JPEG Loader** (external): [libjpeg-turbo](https://github.com/libjpeg-turbo/libjpeg-turbo)
+* **WebP Loader** (external): [libwebp](https://developers.google.com/speed/webp/download)
+* **Examples**: [SDL2](https://www.libsdl.org/)
 
 [Back to contents](#contents)
 <br />

cross/ios_arm64.txt

@@ -17,6 +17,6 @@ cpp_link_args = ['-miphoneos-version-min=11.0']
 system = 'darwin'
 subsystem = 'ios'
 kernel = 'xnu'
-cpu_family = 'arm64'
-cpu = 'arm64'
+cpu_family = 'aarch64'
+cpu = 'aarch64'
 endian = 'little'

examples/PicturePng.cpp

@@ -65,7 +65,7 @@ struct UserExample : tvgexam::Example
         auto picture = tvg::Picture::gen();
         if (!tvgexam::verify(picture->load(data, size, "png", "", true))) return false;
         free(data);
-        picture->translate(400, 0);
+        picture->translate(380, 0);
         picture->scale(0.8);
         canvas->push(picture);
 

inc/thorvg.h

@@ -65,6 +65,7 @@ namespace tvg
 
 class RenderMethod;
 class Animation;
+class Shape;
 
 /**
  * @defgroup ThorVG ThorVG
@@ -229,9 +230,9 @@ enum class BlendMethod : uint8_t
 enum class SceneEffect : uint8_t
 {
     ClearAll = 0,      ///< Reset all previously applied scene effects, restoring the scene to its original state.
-    GaussianBlur,      ///< Apply a blur effect with a Gaussian filter. Param(3) = {sigma(float)[> 0], direction(int)[both: 0 / horizontal: 1 / vertical: 2], border(int)[duplicate: 0 / wrap: 1], quality(int)[0 - 100]}
+    GaussianBlur,      ///< Apply a blur effect with a Gaussian filter. Param(4) = {sigma(float)[> 0], direction(int)[both: 0 / horizontal: 1 / vertical: 2], border(int)[duplicate: 0 / wrap: 1], quality(int)[0 - 100]}
     DropShadow,        ///< Apply a drop shadow effect with a Gaussian Blur filter. Param(8) = {color_R(int)[0 - 255], color_G(int)[0 - 255], color_B(int)[0 - 255], opacity(int)[0 - 255], angle(double)[0 - 360], distance(double), blur_sigma(double)[> 0], quality(int)[0 - 100]}
-    Fill,              ///< Override the scene content color with a given fill information (Experimental API). Param(5) = {color_R(int)[0 - 255], color_G(int)[0 - 255], color_B(int)[0 - 255], opacity(int)[0 - 255]}
+    Fill,              ///< Override the scene content color with a given fill information (Experimental API). Param(4) = {color_R(int)[0 - 255], color_G(int)[0 - 255], color_B(int)[0 - 255], opacity(int)[0 - 255]}
     Tint,              ///< Tinting the current scene color with a given black, white color paramters (Experimental API). Param(7) = {black_R(int)[0 - 255], black_G(int)[0 - 255], black_B(int)[0 - 255], white_R(int)[0 - 255], white_G(int)[0 - 255], white_B(int)[0 - 255], intensity(float)[0 - 100]}
     Tritone            ///< Apply a tritone color effect to the scene using three color parameters for shadows, midtones, and highlights (Experimental API). Param(9) = {Shadow_R(int)[0 - 255], Shadow_G(int)[0 - 255], Shadow_B(int)[0 - 255], Midtone_R(int)[0 - 255], Midtone_G(int)[0 - 255], Midtone_B(int)[0 - 255], Highlight_R(int)[0 - 255], Highlight_G(int)[0 - 255], Highlight_B(int)[0 - 255]}
 };
@@ -396,13 +397,13 @@ public:
      *
      * @param[in] clipper The shape object as the clipper.
      *
-     * @retval Result::NonSupport If the @p clipper type is not Shape.
-     * @retval Result::InsufficientCondition if the target has already belonged to another paint.
+     * @retval Result::InsufficientCondition if the @p clipper has already belonged to another paint.
+     *
+     * @see Paint::clip()
      *
-     * @note @p clipper only supports the Shape type.
      * @since 1.0
      */
-    Result clip(Paint* clipper) noexcept;
+    Result clip(Shape* clipper) noexcept;
 
     /**
      * @brief Sets the blending method for the paint object.
@@ -478,6 +479,19 @@ public:
      */
     MaskMethod mask(const Paint** target) const noexcept;
 
+    /**
+     * @brief Get the clipper shape of the paint object.
+     *
+     * This function returns the clipper that has been previously set to this paint object.
+     *
+     * @return The shape object used as the clipper, or @c nullptr if no clipper is set.
+     *
+     * @see Paint::clip(Shape* clipper)
+     *
+     * @since 1.0
+     */
+    Shape* clip() const noexcept;
+
     /**
      * @brief Increment the reference count for the Paint instance.
      *

src/bindings/capi/thorvg_capi.h

@@ -972,7 +972,7 @@ TVG_API Tvg_Result tvg_paint_set_mask_method(Tvg_Paint* paint, Tvg_Paint* target
 TVG_API Tvg_Result tvg_paint_get_mask_method(const Tvg_Paint* paint, const Tvg_Paint** target, Tvg_Mask_Method* method);
 
 
-/*!
+/**
 * @brief Clip the drawing region of the paint object.
 *
 * This function restricts the drawing area of the paint object to the specified shape's paths.
@@ -985,10 +985,24 @@ TVG_API Tvg_Result tvg_paint_get_mask_method(const Tvg_Paint* paint, const Tvg_P
 * @retval TVG_RESULT_INSUFFICIENT_CONDITION if the target has already belonged to another paint.
 * @retval TVG_RESULT_NOT_SUPPORTED If the @p clipper type is not Shape.
 *
+* @see tvg_paint_get_clip()
+
 * @since 1.0
 */
-TVG_API Tvg_Result tvg_paint_clip(Tvg_Paint* paint, Tvg_Paint* clipper);
+TVG_API Tvg_Result tvg_paint_set_clip(Tvg_Paint* paint, Tvg_Paint* clipper);
 
+/**
+ * @brief Get the clipper shape of the paint object.
+ *
+ * This function returns the clipper that has been previously set to this paint object.
+ *
+ * @return The shape object used as the clipper, or @c nullptr if no clipper is set.
+ *
+ * @see tvg_paint_set_clip()
+ *
+ * @since 1.0
+ */
+TVG_API Tvg_Paint* tvg_paint_get_clip(const Tvg_Paint* paint);
 
 /**
  * @brief Retrieves the parent paint object.

src/bindings/wasm/tvgWasmLottieAnimation.cpp

@@ -47,6 +47,8 @@ struct TvgEngineMethod
     }
 };
 
+#ifdef THORVG_SW_RASTER_SUPPORT
+
 struct TvgSwEngine : TvgEngineMethod
 {
     uint8_t* buffer = nullptr;
@@ -78,6 +80,8 @@ struct TvgSwEngine : TvgEngineMethod
     }
 };
 
+#endif
+
 
 #ifdef THORVG_WG_RASTER_SUPPORT
 
@@ -243,12 +247,14 @@ public:
     {
         errorMsg = NoError;
 
+#ifdef THORVG_SW_RASTER_SUPPORT
         if (engine == "sw") this->engine = new TvgSwEngine;
+#endif
 #ifdef THORVG_GL_RASTER_SUPPORT
-        else if (engine == "gl") this->engine = new TvgGLEngine;
+        if (engine == "gl") this->engine = new TvgGLEngine;
 #endif
 #ifdef THORVG_WG_RASTER_SUPPORT
-        else if (engine == "wg") this->engine = new TvgWgEngine;
+        if (engine == "wg") this->engine = new TvgWgEngine;
 #endif
 
         if (!this->engine) {

src/common/tvgArray.h

@@ -95,6 +95,24 @@ struct Array
         return data[idx];
     }
 
+    void operator=(const Array& rhs)
+    {
+        reserve(rhs.count);
+        if (rhs.count > 0) memcpy(data, rhs.data, sizeof(T) * rhs.count);
+        count = rhs.count;
+    }
+
+    void move(Array& to)
+    {
+        to.reset();
+        to.data = data;
+        to.count = count;
+        to.reserved = reserved;
+
+        data = nullptr;
+        count = reserved = 0;
+    }
+
     const T* begin() const
     {
         return data;
@@ -130,6 +148,12 @@ struct Array
         return data[count - 1];
     }
 
+    T& next()
+    {
+        if (full()) grow(count + 1);
+        return data[count++];
+    }
+
     T& first()
     {
         return data[0];
@@ -157,17 +181,14 @@ struct Array
         return count == 0;
     }
 
-    template<class COMPARE>
-    void sort()
+    bool full()
     {
-        qsort<COMPARE>(data, 0, static_cast<int32_t>(count) - 1);
+        return count == reserved;
     }
 
-    void operator=(const Array& rhs)
+    template<class COMPARE> void sort()
     {
-        reserve(rhs.count);
-        if (rhs.count > 0) memcpy(data, rhs.data, sizeof(T) * rhs.count);
-        count = rhs.count;
+        qsort<COMPARE>(data, 0, (int32_t)(count - 1));
     }
 
     ~Array()
@@ -180,20 +201,14 @@ private:
     void qsort(T* arr, int32_t low, int32_t high)
     {
         if (low < high) {
-            int32_t i = low;
-            int32_t j = high;
-            T tmp = arr[low];
+            auto i = low;
+            auto j = high;
+            auto tmp = arr[low];
             while (i < j) {
                 while (i < j && !COMPARE{}(arr[j], tmp)) --j;
-                if (i < j) {
-                    arr[i] = arr[j];
-                    ++i;
-                }
+                if (i < j) arr[i++] = arr[j];
                 while (i < j && COMPARE{}(arr[i], tmp)) ++i;
-                if (i < j) {
-                    arr[j] = arr[i];
-                    --j;
-                }
+                if (i < j) arr[j--] = arr[i];
             }
             arr[i] = tmp;
             qsort<COMPARE>(arr, low, i - 1);

src/common/tvgMath.h

@@ -90,7 +90,7 @@ bool operator==(const Matrix& lhs, const Matrix& rhs);
 static inline bool rightAngle(const Matrix& m)
 {
    auto radian = fabsf(tvg::atan2(m.e21, m.e11));
-   if (radian < FLOAT_EPSILON || tvg::equal(radian, MATH_PI2) || tvg::equal(radian, MATH_PI)) return true;
+   if (tvg::zero(radian) || tvg::zero(radian - MATH_PI2) || tvg::zero(radian - MATH_PI)) return true;
    return false;
 }
 

src/loaders/jpg/tvgJpgd.cpp

@@ -33,21 +33,11 @@
 
 #include <stdlib.h>
 #include <stdio.h>
-#include <setjmp.h>
 #include <stdint.h>
 
 #include "tvgCommon.h"
 #include "tvgJpgd.h"
 
-#ifdef _MSC_VER
-  #pragma warning (disable : 4611) // warning C4611: interaction between '_setjmp' and C++ object destruction is non-portable
-  #define JPGD_NORETURN __declspec(noreturn)
-#elif defined(__GNUC__)
-  #define JPGD_NORETURN __attribute__ ((noreturn))
-#else
-  #define JPGD_NORETURN
-#endif
-
 /************************************************************************/
 /* Internal Class Implementation                                        */
 /************************************************************************/
@@ -171,7 +161,7 @@ private:
     jpeg_decoder(const jpeg_decoder &);
     jpeg_decoder &operator =(const jpeg_decoder &);
 
-    typedef void (*pDecode_block_func)(jpeg_decoder *, int, int, int);
+    typedef bool (*pDecode_block_func)(jpeg_decoder*, int, int, int);
 
     struct huff_tables
     {
@@ -198,7 +188,6 @@ private:
       char m_data[1];
     };
 
-    jmp_buf m_jmp_state;
     mem_block *m_pMem_blocks;
     int m_image_x_size;
     int m_image_y_size;
@@ -274,21 +263,21 @@ private:
     int m_total_bytes_read;
 
     void free_all_blocks();
-    JPGD_NORETURN void stop_decoding(jpgd_status status);
+    bool stop_decoding(jpgd_status status);
     void *alloc(size_t n, bool zero = false);
     void word_clear(void *p, uint16_t c, uint32_t n);
-    void prep_in_buffer();
-    void read_dht_marker();
-    void read_dqt_marker();
-    void read_sof_marker();
-    void skip_variable_marker();
-    void read_dri_marker();
-    void read_sos_marker();
+    bool prep_in_buffer();
+    bool read_dht_marker();
+    bool read_dqt_marker();
+    bool read_sof_marker();
+    bool skip_variable_marker();
+    bool read_dri_marker();
+    bool read_sos_marker();
     int next_marker();
     int process_markers();
-    void locate_soi_marker();
-    void locate_sof_marker();
-    int locate_sos_marker();
+    bool locate_soi_marker();
+    bool locate_sof_marker();
+    bool locate_sos_marker();
     void init(jpeg_decoder_stream * pStream);
     void create_look_ups();
     void fix_in_buffer();
@@ -297,18 +286,18 @@ private:
     coeff_buf* coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y);
     inline jpgd_block_t *coeff_buf_getp(coeff_buf *cb, int block_x, int block_y);
     void load_next_row();
-    void decode_next_row();
+    bool decode_next_row();
     void make_huff_table(int index, huff_tables *pH);
-    void check_quant_tables();
-    void check_huff_tables();
+    bool check_quant_tables();
+    bool check_huff_tables();
     void calc_mcu_block_order();
     int init_scan();
-    void init_frame();
-    void process_restart();
-    void decode_scan(pDecode_block_func decode_block_func);
-    void init_progressive();
-    void init_sequential();
-    void decode_start();
+    bool init_frame();
+    bool process_restart();
+    bool decode_scan(pDecode_block_func decode_block_func);
+    bool init_progressive();
+    bool init_sequential();
+    bool decode_start();
     void decode_init(jpeg_decoder_stream * pStream);
     void H2V2Convert();
     void H2V1Convert();
@@ -326,10 +315,10 @@ private:
     inline int huff_decode(huff_tables *pH);
     inline int huff_decode(huff_tables *pH, int& extrabits);
     static inline uint8_t clamp(int i);
-    static void decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
+    static bool decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
+    static bool decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
+    static bool decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
+    static bool decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
 };
 
 
@@ -630,7 +619,7 @@ inline uint32_t jpeg_decoder::get_char()
     // Any bytes remaining in buffer?
     if (!m_in_buf_left) {
         // Try to get more bytes.
-        prep_in_buffer();
+        if (!prep_in_buffer()) return 0xFF;
         // Still nothing to get?
         if (!m_in_buf_left) {
             // Pad the end of the stream with 0xFF 0xD9 (EOI marker)
@@ -650,7 +639,7 @@ inline uint32_t jpeg_decoder::get_char()
 inline uint32_t jpeg_decoder::get_char(bool *pPadding_flag)
 {
     if (!m_in_buf_left) {
-        prep_in_buffer();
+        if (!prep_in_buffer()) return 0xFF;
         if (!m_in_buf_left) {
             *pPadding_flag = true;
             int t = m_tem_flag;
@@ -1093,13 +1082,15 @@ void jpeg_decoder::free_all_blocks()
 }
 
 
-// This method handles all errors. It will never return.
-// It could easily be changed to use C++ exceptions.
-JPGD_NORETURN void jpeg_decoder::stop_decoding(jpgd_status status)
+bool jpeg_decoder::stop_decoding(jpgd_status status)
 {
+#if 0 //for debugging
     m_error_code = status;
+#else
+    m_error_code = JPGD_FAILED;
+#endif
     free_all_blocks();
-    longjmp(m_jmp_state, status);
+    return false;
 }
 
 
@@ -1116,8 +1107,7 @@ void *jpeg_decoder::alloc(size_t nSize, bool zero)
     }
     if (!rv) {
         int capacity = JPGD_MAX(32768 - 256, (nSize + 2047) & ~2047);
-        mem_block *b = tvg::malloc<mem_block*>(sizeof(mem_block) + capacity);
-        if (!b) stop_decoding(JPGD_NOTENOUGHMEM);
+        auto b = tvg::malloc<mem_block*>(sizeof(mem_block) + capacity);
         b->m_pNext = m_pMem_blocks; m_pMem_blocks = b;
         b->m_used_count = nSize;
         b->m_size = capacity;
@@ -1142,16 +1132,15 @@ void jpeg_decoder::word_clear(void *p, uint16_t c, uint32_t n)
 // Refill the input buffer.
 // This method will sit in a loop until (A) the buffer is full or (B)
 // the stream's read() method reports and end of file condition.
-void jpeg_decoder::prep_in_buffer()
+bool jpeg_decoder::prep_in_buffer()
 {
     m_in_buf_left = 0;
     m_pIn_buf_ofs = m_in_buf;
-
-    if (m_eof_flag) return;
+    if (m_eof_flag) return true;
 
     do {
-        int bytes_read = m_pStream->read(m_in_buf + m_in_buf_left, JPGD_IN_BUF_SIZE - m_in_buf_left, &m_eof_flag);
-        if (bytes_read == -1) stop_decoding(JPGD_STREAM_READ);
+        auto bytes_read = m_pStream->read(m_in_buf + m_in_buf_left, JPGD_IN_BUF_SIZE - m_in_buf_left, &m_eof_flag);
+        if (bytes_read == -1) return stop_decoding(JPGD_STREAM_READ);
         m_in_buf_left += bytes_read;
     } while ((m_in_buf_left < JPGD_IN_BUF_SIZE) && (!m_eof_flag));
 
@@ -1160,18 +1149,19 @@ void jpeg_decoder::prep_in_buffer()
     // Pad the end of the block with M_EOI (prevents the decompressor from going off the rails if the stream is invalid).
     // (This dates way back to when this decompressor was written in C/asm, and the all-asm Huffman decoder did some fancy things to increase perf.)
     word_clear(m_pIn_buf_ofs + m_in_buf_left, 0xD9FF, 64);
+    return true;
 }
 
 
 // Read a Huffman code table.
-void jpeg_decoder::read_dht_marker()
+bool jpeg_decoder::read_dht_marker()
 {
     int i, index, count;
     uint8_t huff_num[17];
     uint8_t huff_val[256];
     uint32_t num_left = get_bits(16);
 
-    if (num_left < 2) stop_decoding(JPGD_BAD_DHT_MARKER);
+    if (num_left < 2) return stop_decoding(JPGD_BAD_DHT_MARKER);
     num_left -= 2;
 
     while (num_left) {
@@ -1184,19 +1174,19 @@ void jpeg_decoder::read_dht_marker()
             count += huff_num[i];
         }
 
-        if (count > 255) stop_decoding(JPGD_BAD_DHT_COUNTS);
+        if (count > 255) return stop_decoding(JPGD_BAD_DHT_COUNTS);
 
         for (i = 0; i < count; i++)
             huff_val[i] = static_cast<uint8_t>(get_bits(8));
 
         i = 1 + 16 + count;
 
-        if (num_left < (uint32_t)i) stop_decoding(JPGD_BAD_DHT_MARKER);
+        if (num_left < (uint32_t)i) return stop_decoding(JPGD_BAD_DHT_MARKER);
         num_left -= i;
 
-        if ((index & 0x10) > 0x10) stop_decoding(JPGD_BAD_DHT_INDEX);
+        if ((index & 0x10) > 0x10) return stop_decoding(JPGD_BAD_DHT_INDEX);
         index = (index & 0x0F) + ((index & 0x10) >> 4) * (JPGD_MAX_HUFF_TABLES >> 1);
-        if (index >= JPGD_MAX_HUFF_TABLES) stop_decoding(JPGD_BAD_DHT_INDEX);
+        if (index >= JPGD_MAX_HUFF_TABLES) return stop_decoding(JPGD_BAD_DHT_INDEX);
 
         if (!m_huff_num[index]) m_huff_num[index] = (uint8_t *)alloc(17);
         if (!m_huff_val[index]) m_huff_val[index] = (uint8_t *)alloc(256);
@@ -1205,16 +1195,17 @@ void jpeg_decoder::read_dht_marker()
         memcpy(m_huff_num[index], huff_num, 17);
         memcpy(m_huff_val[index], huff_val, 256);
     }
+    return true;
 }
 
 
 // Read a quantization table.
-void jpeg_decoder::read_dqt_marker()
+bool jpeg_decoder::read_dqt_marker()
 {
     int n, i, prec;
     uint32_t temp;
     uint32_t num_left = get_bits(16);
-    if (num_left < 2) stop_decoding(JPGD_BAD_DQT_MARKER);
+    if (num_left < 2) return stop_decoding(JPGD_BAD_DQT_MARKER);
     num_left -= 2;
 
     while (num_left) {
@@ -1222,7 +1213,7 @@ void jpeg_decoder::read_dqt_marker()
         prec = n >> 4;
         n &= 0x0F;
 
-        if (n >= JPGD_MAX_QUANT_TABLES) stop_decoding(JPGD_BAD_DQT_TABLE);
+        if (n >= JPGD_MAX_QUANT_TABLES) return stop_decoding(JPGD_BAD_DQT_TABLE);
 
         if (!m_quant[n]) m_quant[n] = (jpgd_quant_t *)alloc(64 * sizeof(jpgd_quant_t));
 
@@ -1234,30 +1225,31 @@ void jpeg_decoder::read_dqt_marker()
         }
         i = 64 + 1;
         if (prec) i += 64;
-        if (num_left < (uint32_t)i) stop_decoding(JPGD_BAD_DQT_LENGTH);
+        if (num_left < (uint32_t)i) return stop_decoding(JPGD_BAD_DQT_LENGTH);
         num_left -= i;
     }
+    return true;
 }
 
 
 // Read the start of frame (SOF) marker.
-void jpeg_decoder::read_sof_marker()
+bool jpeg_decoder::read_sof_marker()
 {
     int i;
     uint32_t num_left = get_bits(16);
 
-    if (get_bits(8) != 8) stop_decoding(JPGD_BAD_PRECISION);   /* precision: sorry, only 8-bit precision is supported right now */
+    if (get_bits(8) != 8) return stop_decoding(JPGD_BAD_PRECISION);   /* precision: sorry, only 8-bit precision is supported right now */
 
     m_image_y_size = get_bits(16);
-    if ((m_image_y_size < 1) || (m_image_y_size > JPGD_MAX_HEIGHT)) stop_decoding(JPGD_BAD_HEIGHT);
+    if ((m_image_y_size < 1) || (m_image_y_size > JPGD_MAX_HEIGHT)) return stop_decoding(JPGD_BAD_HEIGHT);
 
     m_image_x_size = get_bits(16);
-    if ((m_image_x_size < 1) || (m_image_x_size > JPGD_MAX_WIDTH)) stop_decoding(JPGD_BAD_WIDTH);
+    if ((m_image_x_size < 1) || (m_image_x_size > JPGD_MAX_WIDTH)) return stop_decoding(JPGD_BAD_WIDTH);
 
     m_comps_in_frame = get_bits(8);
-    if (m_comps_in_frame > JPGD_MAX_COMPONENTS) stop_decoding(JPGD_TOO_MANY_COMPONENTS);
+    if (m_comps_in_frame > JPGD_MAX_COMPONENTS) return stop_decoding(JPGD_TOO_MANY_COMPONENTS);
 
-    if (num_left != (uint32_t)(m_comps_in_frame * 3 + 8)) stop_decoding(JPGD_BAD_SOF_LENGTH);
+    if (num_left != (uint32_t)(m_comps_in_frame * 3 + 8)) return stop_decoding(JPGD_BAD_SOF_LENGTH);
 
     for (i = 0; i < m_comps_in_frame; i++) {
         m_comp_ident[i]  = get_bits(8);
@@ -1265,33 +1257,36 @@ void jpeg_decoder::read_sof_marker()
         m_comp_v_samp[i] = get_bits(4);
         m_comp_quant[i]  = get_bits(8);
     }
+    return true;
 }
 
 
 // Used to skip unrecognized markers.
-void jpeg_decoder::skip_variable_marker()
+bool jpeg_decoder::skip_variable_marker()
 {
     uint32_t num_left = get_bits(16);
-    if (num_left < 2) stop_decoding(JPGD_BAD_VARIABLE_MARKER);
+    if (num_left < 2) return stop_decoding(JPGD_BAD_VARIABLE_MARKER);
     num_left -= 2;
 
     while (num_left) {
         get_bits(8);
         num_left--;
     }
+    return true;
 }
 
 
 // Read a define restart interval (DRI) marker.
-void jpeg_decoder::read_dri_marker()
+bool jpeg_decoder::read_dri_marker()
 {
-    if (get_bits(16) != 4) stop_decoding(JPGD_BAD_DRI_LENGTH);
+    if (get_bits(16) != 4) return stop_decoding(JPGD_BAD_DRI_LENGTH);
     m_restart_interval = get_bits(16);
+    return true;
 }
 
 
 // Read a start of scan (SOS) marker.
-void jpeg_decoder::read_sos_marker()
+bool jpeg_decoder::read_sos_marker()
 {
     int i, ci, c, cc;
     uint32_t num_left = get_bits(16);
@@ -1300,7 +1295,7 @@ void jpeg_decoder::read_sos_marker()
     m_comps_in_scan = n;
     num_left -= 3;
 
-    if ( (num_left != (uint32_t)(n * 2 + 3)) || (n < 1) || (n > JPGD_MAX_COMPS_IN_SCAN) ) stop_decoding(JPGD_BAD_SOS_LENGTH);
+    if ( (num_left != (uint32_t)(n * 2 + 3)) || (n < 1) || (n > JPGD_MAX_COMPS_IN_SCAN) ) return stop_decoding(JPGD_BAD_SOS_LENGTH);
 
     for (i = 0; i < n; i++) {
         cc = get_bits(8);
@@ -1310,7 +1305,7 @@ void jpeg_decoder::read_sos_marker()
         for (ci = 0; ci < m_comps_in_frame; ci++)
           if (cc == m_comp_ident[ci]) break;
 
-        if (ci >= m_comps_in_frame) stop_decoding(JPGD_BAD_SOS_COMP_ID);
+        if (ci >= m_comps_in_frame) return stop_decoding(JPGD_BAD_SOS_COMP_ID);
 
         m_comp_list[i]    = ci;
         m_comp_dc_tab[ci] = (c >> 4) & 15;
@@ -1331,6 +1326,7 @@ void jpeg_decoder::read_sos_marker()
         get_bits(8);
         num_left--;
     }
+    return true;
 }
 
 
@@ -1380,21 +1376,21 @@ int jpeg_decoder::process_markers()
             case M_EOI:
             case M_SOS: return c;
             case M_DHT: {
-                read_dht_marker();
-                break;
+                if (read_dht_marker()) break;
+                else return M_EOI;
             }
             // No arithmetic support - dumb patents!
             case M_DAC: {
                 stop_decoding(JPGD_NO_ARITHMETIC_SUPPORT);
-                break;
+                return M_EOI;
             }
             case M_DQT: {
-                read_dqt_marker();
-                break;
+                if (read_dqt_marker()) break;
+                else return M_EOI;
             }
             case M_DRI: {
-                read_dri_marker();
-                break;
+                if (read_dri_marker()) break;
+                else return M_EOI;
             }
             //case M_APP0:  /* no need to read the JFIF marker */
             case M_JPG:
@@ -1408,11 +1404,11 @@ int jpeg_decoder::process_markers()
             case M_RST7:
             case M_TEM: {
                 stop_decoding(JPGD_UNEXPECTED_MARKER);
-                break;
+                return M_EOI;
             }
             default: {   /* must be DNL, DHP, EXP, APPn, JPGn, COM, or RESn or APP0 */
-                skip_variable_marker();
-                break;
+                if (skip_variable_marker()) break;
+                else return M_EOI;
             }
         }
     }
@@ -1422,71 +1418,62 @@ int jpeg_decoder::process_markers()
 // Finds the start of image (SOI) marker.
 // This code is rather defensive: it only checks the first 512 bytes to avoid
 // false positives.
-void jpeg_decoder::locate_soi_marker()
+bool jpeg_decoder::locate_soi_marker()
 {
     uint32_t lastchar = get_bits(8);
     uint32_t thischar = get_bits(8);
 
     /* ok if it's a normal JPEG file without a special header */
-    if ((lastchar == 0xFF) && (thischar == M_SOI)) return;
+    if ((lastchar == 0xFF) && (thischar == M_SOI)) return true;
 
     uint32_t bytesleft = 4096; //512;
 
     while (true) {
-        if (--bytesleft == 0) stop_decoding(JPGD_NOT_JPEG);
+        if (--bytesleft == 0) return stop_decoding(JPGD_NOT_JPEG);
 
         lastchar = thischar;
         thischar = get_bits(8);
 
         if (lastchar == 0xFF) {
           if (thischar == M_SOI) break;
-          else if (thischar == M_EOI) stop_decoding(JPGD_NOT_JPEG); // get_bits will keep returning M_EOI if we read past the end
+          else if (thischar == M_EOI) return stop_decoding(JPGD_NOT_JPEG); // get_bits will keep returning M_EOI if we read past the end
         }
     }
 
     // Check the next character after marker: if it's not 0xFF, it can't be the start of the next marker, so the file is bad.
     thischar = (m_bit_buf >> 24) & 0xFF;
-    if (thischar != 0xFF) stop_decoding(JPGD_NOT_JPEG);
+    if (thischar != 0xFF) return stop_decoding(JPGD_NOT_JPEG);
+    return true;
 }
 
 
 // Find a start of frame (SOF) marker.
-void jpeg_decoder::locate_sof_marker()
+bool jpeg_decoder::locate_sof_marker()
 {
-    locate_soi_marker();
-    int c = process_markers();
-
-    switch (c) {
+    if (!locate_soi_marker()) return false;
+    switch (process_markers()) {
         case M_SOF2: {
             m_progressive_flag = true;
-            read_sof_marker();
-            break;
+            return read_sof_marker();
         }
         case M_SOF0:  /* baseline DCT */
         case M_SOF1: { /* extended sequential DCT */
-          read_sof_marker();
-          break;
-        }
-        case M_SOF9: {  /* Arithmetic coding */
-          stop_decoding(JPGD_NO_ARITHMETIC_SUPPORT);
-          break;
-        }
-        default: {
-          stop_decoding(JPGD_UNSUPPORTED_MARKER);
-          break;
+            return read_sof_marker();
         }
+        case M_SOF9: return stop_decoding(JPGD_NO_ARITHMETIC_SUPPORT);
+        default: return stop_decoding(JPGD_UNSUPPORTED_MARKER);
     }
+    return true;
 }
 
 
 // Find a start of scan (SOS) marker.
-int jpeg_decoder::locate_sos_marker()
+bool jpeg_decoder::locate_sos_marker()
 {
-    int c = process_markers();
+    auto c = process_markers();
     if (c == M_EOI) return false;
-    else if (c != M_SOS) stop_decoding(JPGD_UNEXPECTED_MARKER);
-    read_sos_marker();
-    return true;
+    else if (c != M_SOS) return stop_decoding(JPGD_UNEXPECTED_MARKER);
+    return read_sos_marker();
 }
 
 
@@ -1828,7 +1815,7 @@ void jpeg_decoder::load_next_row()
 
 
 // Restart interval processing.
-void jpeg_decoder::process_restart()
+bool jpeg_decoder::process_restart()
 {
     int i;
     int c = 0;
@@ -1842,15 +1829,15 @@ void jpeg_decoder::process_restart()
     for (i = 1536; i > 0; i--) {
         if (get_char() == 0xFF) break;
     }
-    if (i == 0) stop_decoding(JPGD_BAD_RESTART_MARKER);
+    if (i == 0) return stop_decoding(JPGD_BAD_RESTART_MARKER);
 
     for ( ; i > 0; i--) {
         if ((c = get_char()) != 0xFF) break;
     }
-    if (i == 0) stop_decoding(JPGD_BAD_RESTART_MARKER);
+    if (i == 0) return stop_decoding(JPGD_BAD_RESTART_MARKER);
 
     // Is it the expected marker? If not, something bad happened.
-    if (c != (m_next_restart_num + M_RST0)) stop_decoding(JPGD_BAD_RESTART_MARKER);
+    if (c != (m_next_restart_num + M_RST0)) return stop_decoding(JPGD_BAD_RESTART_MARKER);
 
     // Reset each component's DC prediction values.
     memset(&m_last_dc_val, 0, m_comps_in_frame * sizeof(uint32_t));
@@ -1863,6 +1850,8 @@ void jpeg_decoder::process_restart()
     m_bits_left = 16;
     get_bits_no_markers(16);
     get_bits_no_markers(16);
+
+    return true;
 }
 
 
@@ -1873,10 +1862,12 @@ static inline int dequantize_ac(int c, int q)
 }
 
 // Decodes and dequantizes the next row of coefficients.
-void jpeg_decoder::decode_next_row()
+bool jpeg_decoder::decode_next_row()
 {
     for (int mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) {
-        if ((m_restart_interval) && (m_restarts_left == 0)) process_restart();
+        if ((m_restart_interval) && (m_restarts_left == 0)) {
+            if (!process_restart()) return false;
+        }
 
         jpgd_block_t* p = m_pMCU_coefficients;
 
@@ -1903,7 +1894,7 @@ void jpeg_decoder::decode_next_row()
 
                 if (s) {
                     if (r) {
-                        if ((k + r) > 63) stop_decoding(JPGD_DECODE_ERROR);
+                        if ((k + r) > 63) return stop_decoding(JPGD_DECODE_ERROR);
                         if (k < prev_num_set) {
                             int n = JPGD_MIN(r, prev_num_set - k);
                             int kt = k;
@@ -1916,7 +1907,7 @@ void jpeg_decoder::decode_next_row()
                     p[g_ZAG[k]] = static_cast<jpgd_block_t>(dequantize_ac(s, q[k])); //s * q[k];
                 } else {
                     if (r == 15) {
-                        if ((k + 16) > 64) stop_decoding(JPGD_DECODE_ERROR);
+                        if ((k + 16) > 64) return stop_decoding(JPGD_DECODE_ERROR);
                         if (k < prev_num_set) {
                             int n = JPGD_MIN(16, prev_num_set - k);
                             int kt = k;
@@ -1942,6 +1933,7 @@ void jpeg_decoder::decode_next_row()
         else transform_mcu(mcu_row);
         m_restarts_left--;
     }
+    return true;
 }
 
 
@@ -2183,9 +2175,8 @@ int jpeg_decoder::decode(const void** pScan_line, uint32_t* pScan_line_len)
     if ((m_error_code) || (!m_ready_flag)) return JPGD_FAILED;
     if (m_total_lines_left == 0) return JPGD_DONE;
     if (m_mcu_lines_left == 0) {
-        if (setjmp(m_jmp_state)) return JPGD_FAILED;
         if (m_progressive_flag) load_next_row();
-        else decode_next_row();
+        else if (!decode_next_row()) return JPGD_FAILED;
         // Find the EOI marker if that was the last row.
         if (m_total_lines_left <= m_max_mcu_y_size) find_eoi();
         m_mcu_lines_left = m_max_mcu_y_size;
@@ -2341,20 +2332,21 @@ void jpeg_decoder::make_huff_table(int index, huff_tables *pH)
 
 
 // Verifies the quantization tables needed for this scan are available.
-void jpeg_decoder::check_quant_tables()
+bool jpeg_decoder::check_quant_tables()
 {
     for (int i = 0; i < m_comps_in_scan; i++) {
-        if (m_quant[m_comp_quant[m_comp_list[i]]] == nullptr) stop_decoding(JPGD_UNDEFINED_QUANT_TABLE);
+        if (!m_quant[m_comp_quant[m_comp_list[i]]]) return stop_decoding(JPGD_UNDEFINED_QUANT_TABLE);
     }
+    return true;
 }
 
 
 // Verifies that all the Huffman tables needed for this scan are available.
-void jpeg_decoder::check_huff_tables()
+bool jpeg_decoder::check_huff_tables()
 {
     for (int i = 0; i < m_comps_in_scan; i++) {
-      if ((m_spectral_start == 0) && (m_huff_num[m_comp_dc_tab[m_comp_list[i]]] == nullptr)) stop_decoding(JPGD_UNDEFINED_HUFF_TABLE);
-      if ((m_spectral_end > 0) && (m_huff_num[m_comp_ac_tab[m_comp_list[i]]] == nullptr)) stop_decoding(JPGD_UNDEFINED_HUFF_TABLE);
+      if ((m_spectral_start == 0) && !m_huff_num[m_comp_dc_tab[m_comp_list[i]]]) return stop_decoding(JPGD_UNDEFINED_HUFF_TABLE);
+      if ((m_spectral_end > 0) && !m_huff_num[m_comp_ac_tab[m_comp_list[i]]]) return stop_decoding(JPGD_UNDEFINED_HUFF_TABLE);
     }
 
     for (int i = 0; i < JPGD_MAX_HUFF_TABLES; i++) {
@@ -2363,6 +2355,7 @@ void jpeg_decoder::check_huff_tables()
             make_huff_table(i, m_pHuff_tabs[i]);
         }
     }
+    return true;
 }
 
 
@@ -2418,7 +2411,7 @@ int jpeg_decoder::init_scan()
 
     calc_mcu_block_order();
     check_huff_tables();
-    check_quant_tables();
+    if (!check_quant_tables()) return false;
 
     memset(m_last_dc_val, 0, m_comps_in_frame * sizeof(uint32_t));
 
@@ -2435,19 +2428,20 @@ int jpeg_decoder::init_scan()
 
 // Starts a frame. Determines if the number of components or sampling factors
 // are supported.
-void jpeg_decoder::init_frame()
+bool jpeg_decoder::init_frame()
 {
     int i;
 
     if (m_comps_in_frame == 1) {
-        if ((m_comp_h_samp[0] != 1) || (m_comp_v_samp[0] != 1)) stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
+        if ((m_comp_h_samp[0] != 1) || (m_comp_v_samp[0] != 1)) return stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
         m_scan_type = JPGD_GRAYSCALE;
         m_max_blocks_per_mcu = 1;
         m_max_mcu_x_size = 8;
         m_max_mcu_y_size = 8;
     } else if (m_comps_in_frame == 3) {
-        if (((m_comp_h_samp[1] != 1) || (m_comp_v_samp[1] != 1)) || ((m_comp_h_samp[2] != 1) || (m_comp_v_samp[2] != 1)))
-            stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
+        if (((m_comp_h_samp[1] != 1) || (m_comp_v_samp[1] != 1)) || ((m_comp_h_samp[2] != 1) || (m_comp_v_samp[2] != 1))) {
+            return stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
+        }
 
         if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 1)) {
             m_scan_type = JPGD_YH1V1;
@@ -2469,8 +2463,8 @@ void jpeg_decoder::init_frame()
             m_max_blocks_per_mcu = 6;
             m_max_mcu_x_size = 16;
             m_max_mcu_y_size = 16;
-        } else stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
-    } else stop_decoding(JPGD_UNSUPPORTED_COLORSPACE);
+        } else return stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS);
+    } else return stop_decoding(JPGD_UNSUPPORTED_COLORSPACE);
 
     m_max_mcus_per_row = (m_image_x_size + (m_max_mcu_x_size - 1)) / m_max_mcu_x_size;
     m_max_mcus_per_col = (m_image_y_size + (m_max_mcu_y_size - 1)) / m_max_mcu_y_size;
@@ -2491,7 +2485,7 @@ void jpeg_decoder::init_frame()
     m_max_blocks_per_row = m_max_mcus_per_row * m_max_blocks_per_mcu;
 
     // Should never happen
-    if (m_max_blocks_per_row > JPGD_MAX_BLOCKS_PER_ROW) stop_decoding(JPGD_ASSERTION_ERROR);
+    if (m_max_blocks_per_row > JPGD_MAX_BLOCKS_PER_ROW) return stop_decoding(JPGD_ASSERTION_ERROR);
 
     // Allocate the coefficient buffer, enough for one MCU
     m_pMCU_coefficients = (jpgd_block_t*)alloc(m_max_blocks_per_mcu * 64 * sizeof(jpgd_block_t));
@@ -2517,6 +2511,8 @@ void jpeg_decoder::init_frame()
     m_total_lines_left = m_image_y_size;
     m_mcu_lines_left = 0;
     create_look_ups();
+
+    return true;
 }
 
 
@@ -2546,7 +2542,7 @@ inline jpgd_block_t *jpeg_decoder::coeff_buf_getp(coeff_buf *cb, int block_x, in
 
 // The following methods decode the various types of m_blocks encountered
 // in progressively encoded images.
-void jpeg_decoder::decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y)
+bool jpeg_decoder::decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y)
 {
     int s, r;
     jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y);
@@ -2557,25 +2553,28 @@ void jpeg_decoder::decode_block_dc_first(jpeg_decoder *pD, int component_id, int
     }
     pD->m_last_dc_val[component_id] = (s += pD->m_last_dc_val[component_id]);
     p[0] = static_cast<jpgd_block_t>(static_cast<unsigned int>(s) << pD->m_successive_low);
+
+    return true;
 }
 
 
-void jpeg_decoder::decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y)
+bool jpeg_decoder::decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y)
 {
     if (pD->get_bits_no_markers(1)) {
         jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y);
         p[0] |= (1 << pD->m_successive_low);
     }
+    return true;
 }
 
 
-void jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y)
+bool jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y)
 {
     int k, s, r;
 
     if (pD->m_eob_run) {
         pD->m_eob_run--;
-        return;
+        return true;
     }
     jpgd_block_t *p = pD->coeff_buf_getp(pD->m_ac_coeffs[component_id], block_x, block_y);
 
@@ -2584,13 +2583,13 @@ void jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int
         r = s >> 4;
         s &= 15;
         if (s) {
-            if ((k += r) > 63) pD->stop_decoding(JPGD_DECODE_ERROR);
+            if ((k += r) > 63) return pD->stop_decoding(JPGD_DECODE_ERROR);
             r = pD->get_bits_no_markers(s);
             s = JPGD_HUFF_EXTEND(r, s);
             p[g_ZAG[k]] = static_cast<jpgd_block_t>(static_cast<unsigned int>(s) << pD->m_successive_low);
         } else {
             if (r == 15) {
-                if ((k += 15) > 63) pD->stop_decoding(JPGD_DECODE_ERROR);
+                if ((k += 15) > 63) return pD->stop_decoding(JPGD_DECODE_ERROR);
             } else {
                 pD->m_eob_run = 1 << r;
                 if (r) pD->m_eob_run += pD->get_bits_no_markers(r);
@@ -2599,10 +2598,11 @@ void jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int
             }
         }
     }
+    return true;
 }
 
 
-void jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y)
+bool jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y)
 {
     int s, k, r;
     int p1 = 1 << pD->m_successive_low;
@@ -2619,7 +2619,7 @@ void jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, in
             r = s >> 4;
             s &= 15;
             if (s) {
-                if (s != 1) pD->stop_decoding(JPGD_DECODE_ERROR);
+                if (s != 1) return pD->stop_decoding(JPGD_DECODE_ERROR);
                 if (pD->get_bits_no_markers(1)) s = p1;
                 else s = m1;
             } else {
@@ -2667,11 +2667,12 @@ void jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, in
         }
         pD->m_eob_run--;
     }
+    return true;
 }
 
 
 // Decode a scan in a progressively encoded image.
-void jpeg_decoder::decode_scan(pDecode_block_func decode_block_func)
+bool jpeg_decoder::decode_scan(pDecode_block_func decode_block_func)
 {
     int mcu_row, mcu_col, mcu_block;
     int block_x_mcu[JPGD_MAX_COMPONENTS], m_block_y_mcu[JPGD_MAX_COMPONENTS];
@@ -2685,11 +2686,13 @@ void jpeg_decoder::decode_scan(pDecode_block_func decode_block_func)
         for (mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) {
             int block_x_mcu_ofs = 0, block_y_mcu_ofs = 0;
 
-            if ((m_restart_interval) && (m_restarts_left == 0)) process_restart();
+            if ((m_restart_interval) && (m_restarts_left == 0)) {
+                if (!process_restart()) return false;
+            }
 
             for (mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) {
                 component_id = m_mcu_org[mcu_block];
-                decode_block_func(this, component_id, block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs);
+                if (!decode_block_func(this, component_id, block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs)) return false;
 
                 if (m_comps_in_scan == 1) block_x_mcu[component_id]++;
                 else {
@@ -2714,15 +2717,16 @@ void jpeg_decoder::decode_scan(pDecode_block_func decode_block_func)
             }
         }
     }
+    return true;
 }
 
 
 // Decode a progressively encoded image.
-void jpeg_decoder::init_progressive()
+bool jpeg_decoder::init_progressive()
 {
     int i;
 
-    if (m_comps_in_frame == 4) stop_decoding(JPGD_UNSUPPORTED_COLORSPACE);
+    if (m_comps_in_frame == 4) return stop_decoding(JPGD_UNSUPPORTED_COLORSPACE);
 
     // Allocate the coefficient buffers.
     for (i = 0; i < m_comps_in_frame; i++) {
@@ -2739,15 +2743,13 @@ void jpeg_decoder::init_progressive()
         dc_only_scan = (m_spectral_start == 0);
         refinement_scan = (m_successive_high != 0);
 
-        if ((m_spectral_start > m_spectral_end) || (m_spectral_end > 63)) stop_decoding(JPGD_BAD_SOS_SPECTRAL);
+        if ((m_spectral_start > m_spectral_end) || (m_spectral_end > 63)) return stop_decoding(JPGD_BAD_SOS_SPECTRAL);
 
         if (dc_only_scan) {
-            if (m_spectral_end) stop_decoding(JPGD_BAD_SOS_SPECTRAL);
-        } else if (m_comps_in_scan != 1) {  /* AC scans can only contain one component */
-            stop_decoding(JPGD_BAD_SOS_SPECTRAL);
-        }
+            if (m_spectral_end) return stop_decoding(JPGD_BAD_SOS_SPECTRAL);
+        } else if (m_comps_in_scan != 1) return stop_decoding(JPGD_BAD_SOS_SPECTRAL);  // AC scans can only contain one component
 
-        if ((refinement_scan) && (m_successive_low != m_successive_high - 1)) stop_decoding(JPGD_BAD_SOS_SUCCESSIVE);
+        if ((refinement_scan) && (m_successive_low != m_successive_high - 1)) return stop_decoding(JPGD_BAD_SOS_SUCCESSIVE);
 
         if (dc_only_scan) {
             if (refinement_scan) decode_block_func = decode_block_dc_refine;
@@ -2756,7 +2758,7 @@ void jpeg_decoder::init_progressive()
             if (refinement_scan) decode_block_func = decode_block_ac_refine;
             else decode_block_func = decode_block_ac_first;
         }
-        decode_scan(decode_block_func);
+        if (!decode_scan(decode_block_func)) return false;
         m_bits_left = 16;
         get_bits(16);
         get_bits(16);
@@ -2769,20 +2771,23 @@ void jpeg_decoder::init_progressive()
     }
 
     calc_mcu_block_order();
+
+    return true;
 }
 
 
-void jpeg_decoder::init_sequential()
+bool jpeg_decoder::init_sequential()
 {
-    if (!init_scan()) stop_decoding(JPGD_UNEXPECTED_MARKER);
+    if (!init_scan()) return stop_decoding(JPGD_UNEXPECTED_MARKER);
+    return true;
 }
 
 
-void jpeg_decoder::decode_start()
+bool jpeg_decoder::decode_start()
 {
-    init_frame();
-    if (m_progressive_flag) init_progressive();
-    else init_sequential();
+    if (!init_frame()) return false;
+    if (m_progressive_flag) return init_progressive();
+    return init_sequential();
 }
 
 
@@ -2795,7 +2800,6 @@ void jpeg_decoder::decode_init(jpeg_decoder_stream *pStream)
 
 jpeg_decoder::jpeg_decoder(jpeg_decoder_stream *pStream)
 {
-    if (setjmp(m_jmp_state)) return;
     decode_init(pStream);
 }
 
@@ -2804,9 +2808,7 @@ int jpeg_decoder::begin_decoding()
 {
     if (m_ready_flag) return JPGD_SUCCESS;
     if (m_error_code) return JPGD_FAILED;
-    if (setjmp(m_jmp_state)) return JPGD_FAILED;
-
-    decode_start();
+    if (!decode_start()) return JPGD_FAILED;
     m_ready_flag = true;
 
     return JPGD_SUCCESS;
@@ -2970,7 +2972,7 @@ unsigned char* jpgdDecompress(jpeg_decoder* decoder)
     if (!pImage_data) return nullptr;
 
     for (int y = 0; y < image_height; y++) {
-        const uint8_t* pScan_line;
+        const uint8_t* pScan_line = nullptr;
         uint32_t scan_line_len;
         if (decoder->decode((const void**)&pScan_line, &scan_line_len) != JPGD_SUCCESS) {
             tvg::free(pImage_data);

src/loaders/lottie/tvgLottieBuilder.cpp

@@ -352,13 +352,11 @@ static void _repeat(LottieGroup* parent, Shape* path, RenderContext* ctx)
 
         for (int i = 0; i < repeater->cnt; ++i) {
             auto multiplier = repeater->offset + static_cast<float>(i);
-
             ARRAY_FOREACH(p, propagators) {
                 auto shape = static_cast<Shape*>((*p)->duplicate());
                 SHAPE(shape)->rs.path = SHAPE(path)->rs.path;
-
-                auto opacity = repeater->interpOpacity ? tvg::lerp<uint8_t>(repeater->startOpacity, repeater->endOpacity, static_cast<float>(i + 1) / repeater->cnt) : repeater->startOpacity;
-                shape->opacity(opacity);
+                auto opacity = tvg::lerp<uint8_t>(repeater->startOpacity, repeater->endOpacity, static_cast<float>(i + 1) / repeater->cnt);
+                shape->opacity(MULTIPLY((*p)->opacity(), opacity));
 
                 Matrix m;
                 tvg::identity(&m);
@@ -366,11 +364,10 @@ static void _repeat(LottieGroup* parent, Shape* path, RenderContext* ctx)
                 scale(&m, {powf(repeater->scale.x * 0.01f, multiplier), powf(repeater->scale.y * 0.01f, multiplier)});
                 rotate(&m, repeater->rotation * multiplier);
                 translateR(&m, -repeater->anchor);
-                m = repeater->transform * m;
 
                 Matrix inv;
                 inverse(&repeater->transform, &inv);
-                shape->transform(m * (inv * shape->transform()));
+                shape->transform((repeater->transform * m) * (inv * shape->transform()));
                 shapes.push(shape);
             }
         }
@@ -613,7 +610,7 @@ void LottieBuilder::updatePolygon(LottieGroup* parent, LottiePolyStar* star, flo
     auto radius = star->outerRadius(frameNo, tween, exps);
     auto outerRoundness = star->outerRoundness(frameNo, tween, exps) * 0.01f;
 
-    auto angle = deg2rad(-90.0f);
+    auto angle = -MATH_PI2;
     auto anglePerPoint = 2.0f * MATH_PI / float(ptsCnt);
     auto direction = star->clockwise ? 1.0f : -1.0f;
     auto hasRoundness = !tvg::zero(outerRoundness);
@@ -641,6 +638,7 @@ void LottieBuilder::updatePolygon(LottieGroup* parent, LottiePolyStar* star, flo
     auto in = Point{x, y} * transform;
     shape->moveTo(in.x, in.y);
 
+    auto coeff = anglePerPoint * radius * outerRoundness * POLYGON_MAGIC_NUMBER;
     for (size_t i = 0; i < ptsCnt; i++) {
         auto previousX = x;
         auto previousY = y;
@@ -649,16 +647,11 @@ void LottieBuilder::updatePolygon(LottieGroup* parent, LottiePolyStar* star, flo
 
         if (hasRoundness) {
             auto cp1Theta = tvg::atan2(previousY, previousX) - MATH_PI2 * direction;
-            auto cp1Dx = cosf(cp1Theta);
-            auto cp1Dy = sinf(cp1Theta);
+            auto cp1x = coeff * cosf(cp1Theta);
+            auto cp1y = coeff * sinf(cp1Theta);
             auto cp2Theta = tvg::atan2(y, x) - MATH_PI2 * direction;
-            auto cp2Dx = cosf(cp2Theta);
-            auto cp2Dy = sinf(cp2Theta);
-
-            auto cp1x = radius * outerRoundness * POLYGON_MAGIC_NUMBER * cp1Dx;
-            auto cp1y = radius * outerRoundness * POLYGON_MAGIC_NUMBER * cp1Dy;
-            auto cp2x = radius * outerRoundness * POLYGON_MAGIC_NUMBER * cp2Dx;
-            auto cp2y = radius * outerRoundness * POLYGON_MAGIC_NUMBER * cp2Dy;
+            auto cp2x = coeff * cosf(cp2Theta);
+            auto cp2y = coeff * sinf(cp2Theta);
 
             auto in2 = Point{previousX - cp1x, previousY - cp1y} * transform;
             auto in3 = Point{x + cp2x, y + cp2y} * transform;
@@ -743,7 +736,6 @@ void LottieBuilder::updateRepeater(TVG_UNUSED LottieGroup* parent, LottieObject*
     r.startOpacity = repeater->startOpacity(frameNo, tween, exps);
     r.endOpacity = repeater->endOpacity(frameNo, tween, exps);
     r.inorder = repeater->inorder;
-    r.interpOpacity = (r.startOpacity == r.endOpacity) ? false : true;
     ctx->repeaters.push(r);
 
     ctx->merging = nullptr;
@@ -938,7 +930,7 @@ void LottieBuilder::updateText(LottieLayer* layer, float frameNo)
 
     if (!p || !text->font) return;
 
-    if (text->font->origin != LottieFont::Origin::Embedded) {
+    if (text->font->origin != LottieFont::Origin::Local || text->font->chars.empty()) {
         _fontText(doc, layer->scene);
         return;
     }
@@ -1178,8 +1170,8 @@ void LottieBuilder::updateMasks(LottieLayer* layer, float frameNo)
 {
     if (layer->masks.count == 0) return;
 
-    //Introduce an intermediate scene for embracing the matte + masking
-    if (layer->matteTarget) {
+    //Introduce an intermediate scene for embracing matte + masking or precomp clipping + masking replaced by clipping
+    if (layer->matteTarget || layer->type == LottieLayer::Precomp) {
         auto scene = Scene::gen();
         scene->push(layer->scene);
         layer->scene = scene;
@@ -1196,7 +1188,6 @@ void LottieBuilder::updateMasks(LottieLayer* layer, float frameNo)
         auto method = mask->method;
         auto opacity = mask->opacity(frameNo);
         auto expand = mask->expand(frameNo);
-        auto fastTrack = false;  //single clipping
 
         //the first mask
         if (!pShape) {
@@ -1207,7 +1198,6 @@ void LottieBuilder::updateMasks(LottieLayer* layer, float frameNo)
             if (layer->masks.count == 1 && compMethod == MaskMethod::Alpha) {
                 layer->scene->opacity(MULTIPLY(layer->scene->opacity(), opacity));
                 layer->scene->clip(pShape);
-                fastTrack = true;
             } else {
                 layer->scene->mask(pShape, compMethod);
             }
@@ -1233,9 +1223,6 @@ void LottieBuilder::updateMasks(LottieLayer* layer, float frameNo)
             auto offset = LottieOffsetModifier(expand);
             mask->pathset(frameNo, SHAPE(pShape)->rs.path, nullptr, tween, exps, &offset);
         }
-
-        if (fastTrack) return;
-
         pOpacity = opacity;
         pMethod = method;
     }
@@ -1245,7 +1232,7 @@ void LottieBuilder::updateMasks(LottieLayer* layer, float frameNo)
 bool LottieBuilder::updateMatte(LottieComposition* comp, float frameNo, Scene* scene, LottieLayer* layer)
 {
     auto target = layer->matteTarget;
-    if (!target) return true;
+    if (!target || target->type == LottieLayer::Null) return true;
 
     updateLayer(comp, scene, target, frameNo);
 

src/loaders/lottie/tvgLottieBuilder.h

@@ -42,7 +42,6 @@ struct RenderRepeater
     float rotation;
     uint8_t startOpacity;
     uint8_t endOpacity;
-    bool interpOpacity;
     bool inorder;
 };
 

src/loaders/lottie/tvgLottieModel.cpp

@@ -80,36 +80,47 @@ float LottieTextFollowPath::prepare(LottieMask* mask, float frameNo, float scale
 
 Point LottieTextFollowPath::position(float lenSearched, float& angle)
 {
+    //position before the start of the curve
+    if (lenSearched <= 0.0f) {
+        //shape is closed -> wrapping
+        if (path.cmds.last() == PathCommand::Close) {
+            while (lenSearched < 0.0f) lenSearched += totalLen;
+            pts = path.pts.data;
+            cmds = path.cmds.data;
+            cmdsCnt = path.cmds.count;
+            currentLen = 0.0f;
+        //linear interpolation
+        } else {
+            if (cmds >= path.cmds.data + path.cmds.count - 1) return *start;
+            switch (*(cmds + 1)) {
+                case PathCommand::LineTo: {
+                    auto dp = *(pts + 1) - *pts;
+                    angle = tvg::atan2(dp.y, dp.x);
+                    return {pts->x + lenSearched * cos(angle), pts->y + lenSearched * sin(angle)};
+                }
+                case PathCommand::CubicTo: {
+                    angle = deg2rad(Bezier{*pts, *(pts + 1), *(pts + 2), *(pts + 3)}.angle(0.0001f));
+                    return {pts->x + lenSearched * cos(angle), pts->y + lenSearched * sin(angle)};
+                }
+                default:
+                    angle = 0.0f;
+                return *start;
+            }
+        }
+    }
+
     auto shift = [&]() -> void {
         switch (*cmds) {
-            case PathCommand::MoveTo:
-                start = pts;
-                ++pts;
-                break;
-            case PathCommand::LineTo:
-                ++pts;
-                break;
-            case PathCommand::CubicTo:
-                pts += 3;
-                break;
-            case PathCommand::Close:
-                break;
+            case PathCommand::MoveTo: start = pts; ++pts; break;
+            case PathCommand::LineTo: ++pts; break;
+            case PathCommand::CubicTo: pts += 3; break;
+            case PathCommand::Close: break;
         }
         ++cmds;
         --cmdsCnt;
     };
 
-    auto length = [&]() -> float {
-        switch (*cmds) {
-            case PathCommand::MoveTo: return 0.0f;
-            case PathCommand::LineTo: return tvg::length(pts - 1, pts);
-            case PathCommand::CubicTo: return Bezier{*(pts - 1), *pts, *(pts + 1), *(pts + 2)}.length();
-            case PathCommand::Close: return tvg::length(pts - 1, start);
-        }
-        return 0.0f;
-    };
-
-    //beyond the curve
+    //position beyond the end of the curve
     if (lenSearched >= totalLen) {
         //shape is closed -> wrapping
         if (path.cmds.last() == PathCommand::Close) {
@@ -146,6 +157,24 @@ Point LottieTextFollowPath::position(float lenSearched, float& angle)
         }
     }
 
+    //reset required if text partially crosses curve start
+    if (lenSearched < currentLen) {
+        pts = path.pts.data;
+        cmds = path.cmds.data;
+        cmdsCnt = path.cmds.count;
+        currentLen = 0.0f;
+    }
+
+    auto length = [&]() -> float {
+        switch (*cmds) {
+            case PathCommand::MoveTo: return 0.0f;
+            case PathCommand::LineTo: return tvg::length(pts - 1, pts);
+            case PathCommand::CubicTo: return Bezier{*(pts - 1), *pts, *(pts + 1), *(pts + 2)}.length();
+            case PathCommand::Close: return tvg::length(pts - 1, start);
+            default: return 0.0f;
+        }
+    };
+
     while (cmdsCnt > 0) {
         auto dLen = length();
         if (currentLen + dLen < lenSearched) {
@@ -664,7 +693,7 @@ void LottieLayer::prepare(RGB24* color)
 
 float LottieLayer::remap(LottieComposition* comp, float frameNo, LottieExpressions* exp)
 {
-    if (timeRemap.frames || timeRemap.value) {
+    if (timeRemap.frames || timeRemap.value >= 0.0f) {
         frameNo = comp->frameAtTime(timeRemap(frameNo, exp));
     } else {
         frameNo -= startFrame;

src/loaders/lottie/tvgLottieModel.h

@@ -386,7 +386,7 @@ struct LottieTextRange
 
 struct LottieFont
 {
-    enum Origin : uint8_t { Local = 0, CssURL, ScriptURL, FontURL, Embedded };
+    enum Origin : uint8_t {Local = 0, CssURL, ScriptURL, FontURL};
 
     ~LottieFont()
     {
@@ -408,7 +408,7 @@ struct LottieFont
     char* style = nullptr;
     size_t dataSize = 0;
     float ascent = 0.0f;
-    Origin origin = Embedded;
+    Origin origin = Local;
 
     void prepare();
 };
@@ -980,7 +980,7 @@ struct LottieLayer : LottieGroup
 
     char* name = nullptr;
     LottieLayer* parent = nullptr;
-    LottieFloat timeRemap = 0.0f;
+    LottieFloat timeRemap = -1.0f;
     LottieLayer* comp = nullptr;  //Precompositor, current layer is belonges.
     LottieTransform* transform = nullptr;
     Array<LottieMask*> masks;

src/loaders/lottie/tvgLottieModifier.cpp

@@ -27,6 +27,12 @@
 /* Internal Class Implementation                                        */
 /************************************************************************/
 
+static bool _colinear(const Point* p)
+{
+    return tvg::zero(*p - *(p + 1)) && tvg::zero(*(p + 2) - *(p + 3));
+}
+
+
 static void _roundCorner(Array<PathCommand>& cmds, Array<Point>& pts, Point& prev, Point& curr, Point& next, float r)
 {
     auto lenPrev = length(prev - curr);
@@ -115,9 +121,12 @@ void LottieOffsetModifier::corner(RenderPath& out, Line& line, Line& nextLine, u
                 auto norm = normal(line.pt1, line.pt2);
                 auto nextNorm = normal(nextLine.pt1, nextLine.pt2);
                 auto miterDirection = (norm + nextNorm) / length(norm + nextNorm);
+                if (1.0f <= miterLimit * fabsf(miterDirection.x * norm.x + miterDirection.y * norm.y)) {
                     out.cmds.push(PathCommand::LineTo);
-                if (1.0f <= miterLimit * fabsf(miterDirection.x * norm.x + miterDirection.y * norm.y)) out.pts.push(intersect);
-                else out.pts.push(nextLine.pt1);
+                    out.pts.push(intersect);
+                }
+                out.cmds.push(PathCommand::LineTo);
+                out.pts.push(nextLine.pt1);
             } else {
                 out.cmds.push(PathCommand::LineTo);
                 out.pts.push(nextLine.pt1);
@@ -192,14 +201,10 @@ bool LottieRoundnessModifier::modifyPath(PathCommand* inCmds, uint32_t inCmdsCnt
                 break;
             }
             case PathCommand::CubicTo: {
+                if (iCmds < inCmdsCnt - 1 && _colinear(inPts + iPts - 1)) {
                     auto& prev = inPts[iPts - 1];
                     auto& curr = inPts[iPts + 2];
-                if (iCmds < inCmdsCnt - 1 &&
-                    tvg::zero(inPts[iPts - 1] - inPts[iPts]) &&
-                    tvg::zero(inPts[iPts + 1] - inPts[iPts + 2])) {
-                    if (inCmds[iCmds + 1] == PathCommand::CubicTo &&
-                        tvg::zero(inPts[iPts + 2] - inPts[iPts + 3]) &&
-                        tvg::zero(inPts[iPts + 4] - inPts[iPts + 5])) {
+                    if (inCmds[iCmds + 1] == PathCommand::CubicTo && _colinear(inPts + iPts + 2)) {
                         _roundCorner(path.cmds, path.pts, prev, curr, inPts[iPts + 5], r);
                         iPts += 3;
                         break;

src/loaders/svg/tvgSvgLoader.cpp

@@ -3401,6 +3401,13 @@ static void _svgLoaderParserXmlClose(SvgLoaderData* loader, const char* content,
         }
     }
 
+    for (unsigned int i = 0; i < sizeof(gradientTags) / sizeof(gradientTags[0]); i++) {
+        if (!strncmp(tagName, gradientTags[i].tag, sz)) {
+            loader->gradientStack.pop();
+            break;
+        }
+    }
+
     for (unsigned int i = 0; i < sizeof(graphicsTags) / sizeof(graphicsTags[0]); i++) {
         if (!strncmp(tagName, graphicsTags[i].tag, sz)) {
             loader->currentGraphicsNode = nullptr;
@@ -3472,7 +3479,6 @@ static void _svgLoaderParserXmlOpen(SvgLoaderData* loader, const char* content,
         if (node->type != SvgNodeType::Defs || !empty) {
             loader->stack.push(node);
         }
-        loader->latestGradient = nullptr;
     } else if ((method = _findGraphicsFactory(tagName))) {
         if (loader->stack.count > 0) parent = loader->stack.last();
         else parent = loader->doc;
@@ -3483,10 +3489,11 @@ static void _svgLoaderParserXmlOpen(SvgLoaderData* loader, const char* content,
             loader->stack.push(defs);
             loader->currentGraphicsNode = node;
         }
-        loader->latestGradient = nullptr;
     } else if ((gradientMethod = _findGradientFactory(tagName))) {
         SvgStyleGradient* gradient;
         gradient = gradientMethod(loader, attrs, attrsLength);
+        //Gradients do not allow nested declarations, so only the earliest declared Gradient is valid.
+        if (loader->gradientStack.count == 0) {
             //FIXME: The current parsing structure does not distinguish end tags.
             //       There is no way to know if the currently parsed gradient is in defs.
             //       If a gradient is declared outside of defs after defs is set, it is included in the gradients of defs.
@@ -3498,9 +3505,10 @@ static void _svgLoaderParserXmlOpen(SvgLoaderData* loader, const char* content,
             } else {
                 loader->gradients.push(gradient);
             }
-        loader->latestGradient = gradient;
+        }
+        if (!empty) loader->gradientStack.push(gradient);
     } else if (STR_AS(tagName, "stop")) {
-        if (!loader->latestGradient) {
+        if (loader->gradientStack.count == 0) {
             TVGLOG("SVG", "Stop element is used outside of the Gradient element");
             return;
         }
@@ -3508,9 +3516,8 @@ static void _svgLoaderParserXmlOpen(SvgLoaderData* loader, const char* content,
         loader->svgParse->gradStop = {0.0f, 0, 0, 0, 255};
         loader->svgParse->flags = SvgStopStyleFlags::StopDefault;
         xmlParseAttributes(attrs, attrsLength, _attrParseStops, loader);
-        loader->latestGradient->stops.push(loader->svgParse->gradStop);
+        loader->gradientStack.last()->stops.push(loader->svgParse->gradStop);
     } else {
-        loader->latestGradient = nullptr;
         if (!isIgnoreUnsupportedLogElements(tagName)) TVGLOG("SVG", "Unsupported elements used [Elements: %s]", tagName);
     }
 }
@@ -3831,6 +3838,7 @@ void SvgLoader::clear(bool all)
         tvg::free(*p);
     }
     loaderData.gradients.reset();
+    loaderData.gradientStack.reset();
 
     _freeNode(loaderData.doc);
     loaderData.doc = nullptr;

src/loaders/svg/tvgSvgLoaderCommon.h

@@ -590,7 +590,7 @@ struct SvgLoaderData
     SvgNode* def = nullptr; //also used to store nested graphic nodes
     SvgNode* cssStyle = nullptr;
     Array<SvgStyleGradient*> gradients;
-    SvgStyleGradient* latestGradient = nullptr; //For stops
+    Array<SvgStyleGradient*> gradientStack; //For stops
     SvgParser* svgParse = nullptr;
     Array<SvgNodeIdPair> cloneNodes;
     Array<SvgNodeIdPair> nodesToStyle;

src/renderer/gl_engine/tvgGlGeometry.cpp

@@ -57,16 +57,17 @@ bool GlGeometry::tesselate(const RenderShape& rshape, RenderUpdateFlag flag)
 
 bool GlGeometry::tesselate(const RenderSurface* image, RenderUpdateFlag flag)
 {
-    if (flag & RenderUpdateFlag::Image) {
+    if (!(flag & RenderUpdateFlag::Image)) return true;
+
     fill.clear();
 
     fill.vertex.reserve(5 * 4);
     fill.index.reserve(6);
 
-        float left = 0.f;
-        float top = 0.f;
-        float right = image->w;
-        float bottom = image->h;
+    auto left = 0.f;
+    auto top = 0.f;
+    auto right = float(image->w);
+    auto bottom = float(image->h);
 
     // left top point
     fill.vertex.push(left);
@@ -101,11 +102,7 @@ bool GlGeometry::tesselate(const RenderSurface* image, RenderUpdateFlag flag)
     fill.index.push(1);
     fill.index.push(3);
 
-        bounds.x = 0;
-        bounds.y = 0;
-        bounds.w = image->w;
-        bounds.h = image->h;
-    }
+    bounds = {{0, 0}, {int32_t(image->w), int32_t(image->h)}};
 
     return true;
 }
@@ -155,31 +152,20 @@ GlStencilMode GlGeometry::getStencilMode(RenderUpdateFlag flag)
 
 RenderRegion GlGeometry::getBounds() const
 {
-    if (tvg::identity(&matrix)) {
-        return bounds;
-    } else {
-        Point lt{static_cast<float>(bounds.x), static_cast<float>(bounds.y)};
-        Point lb{static_cast<float>(bounds.x), static_cast<float>(bounds.y + bounds.h)};
-        Point rt{static_cast<float>(bounds.x + bounds.w), static_cast<float>(bounds.y)};
-        Point rb{static_cast<float>(bounds.x + bounds.w), static_cast<float>(bounds.y + bounds.h)};
+    if (tvg::identity(&matrix)) return bounds;
 
-        lt *= matrix;
-        lb *= matrix;
-        rt *= matrix;
-        rb *= matrix;
+    auto lt = Point{float(bounds.min.x), float(bounds.min.y)} * matrix;
+    auto lb = Point{float(bounds.min.x), float(bounds.max.y)} * matrix;
+    auto rt = Point{float(bounds.max.x), float(bounds.min.y)} * matrix;
+    auto rb = Point{float(bounds.max.x), float(bounds.max.y)} * matrix;
 
-        float left = min(min(lt.x, lb.x), min(rt.x, rb.x));
-        float top = min(min(lt.y, lb.y), min(rt.y, rb.y));
-        float right = max(max(lt.x, lb.x), max(rt.x, rb.x));
-        float bottom = max(max(lt.y, lb.y), max(rt.y, rb.y));
+    auto left = min(min(lt.x, lb.x), min(rt.x, rb.x));
+    auto top = min(min(lt.y, lb.y), min(rt.y, rb.y));
+    auto right = max(max(lt.x, lb.x), max(rt.x, rb.x));
+    auto bottom = max(max(lt.y, lb.y), max(rt.y, rb.y));
+
+    auto bounds = RenderRegion {{int32_t(floor(left)), int32_t(floor(top))}, {int32_t(ceil(right)), int32_t(ceil(bottom))}};
+    if (bounds.valid()) return bounds;
+    return this->bounds;
 
-        auto bounds = RenderRegion {
-            static_cast<int32_t>(floor(left)),
-            static_cast<int32_t>(floor(top)),
-            static_cast<int32_t>(ceil(right - floor(left))),
-            static_cast<int32_t>(ceil(bottom - floor(top))),
-        };
-        if (bounds.w < 0 || bounds.h < 0) return this->bounds;
-        else return bounds;
-    }
 }

src/renderer/gl_engine/tvgGlRenderPass.cpp

@@ -52,17 +52,17 @@ void GlRenderPass::addRenderTask(GlRenderTask* task)
 
 void GlRenderPass::getMatrix(float *dst, const Matrix &matrix) const
 {
-    const auto& vp = getViewport();
-
     Matrix postMatrix{};
     tvg::identity(&postMatrix);
-    translate(&postMatrix, {(float)-vp.x, (float)-vp.y});
+
+    const auto& vp = getViewport();
+    translate(&postMatrix, {(float)-vp.sx(), (float)-vp.sy()});
 
     auto m = postMatrix * matrix;
 
     float modelMatrix[16];
     GET_MATRIX44(m, modelMatrix);
-    MVP_MATRIX(vp.w, vp.h);
+    MVP_MATRIX(vp.w(), vp.h());
 
     MULTIPLY_MATRIX(mvp, modelMatrix, dst);
 }

src/renderer/gl_engine/tvgGlRenderPass.h

@@ -62,10 +62,7 @@ public:
         }
 
         auto task = new T(program, targetFbo, mFbo, std::move(mTasks));
-
-        const auto& vp = mFbo->getViewport();
-
-        task->setRenderSize(static_cast<uint32_t>(vp.w), static_cast<uint32_t>(vp.h));
+        task->setRenderSize(mFbo->getViewport().w(),  mFbo->getViewport().h());
 
         return task;
     }

src/renderer/gl_engine/tvgGlRenderTarget.cpp

@@ -22,26 +22,20 @@
 
 #include "tvgGlRenderTarget.h"
 
-GlRenderTarget::GlRenderTarget(uint32_t width, uint32_t height): mWidth(width), mHeight(height) {}
+GlRenderTarget::GlRenderTarget() {}
 
 GlRenderTarget::~GlRenderTarget()
 {
-    if (mFbo == 0) return;
-    GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, 0));
-    GL_CHECK(glDeleteFramebuffers(1, &mFbo));
-
-    if (mColorTex != 0) {
-        GL_CHECK(glDeleteTextures(1, &mColorTex));
-    }
-    if (mDepthStencilBuffer != 0) {
-        GL_CHECK(glDeleteRenderbuffers(1, &mDepthStencilBuffer));
-    }
+    reset();
 }
 
-void GlRenderTarget::init(GLint resolveId)
+void GlRenderTarget::init(uint32_t width, uint32_t height, GLint resolveId)
 {
-    if (mFbo != 0 || mWidth == 0 || mHeight == 0) return;
+    if (mFbo != GL_INVALID_VALUE || width == 0 || height == 0) return;
+    mWidth = width;
+    mHeight = height;
 
+    //TODO: fbo is used. maybe we can consider the direct rendering with resolveId as well.
     GL_CHECK(glGenFramebuffers(1, &mFbo));
 
     GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, mFbo));
@@ -81,6 +75,18 @@ void GlRenderTarget::init(GLint resolveId)
     GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, resolveId));
 }
 
+void GlRenderTarget::reset()
+{
+    if (mFbo == 0) return;
+    GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, 0));
+    GL_CHECK(glDeleteFramebuffers(1, &mFbo));
+    GL_CHECK(glDeleteRenderbuffers(1, &mColorBuffer));
+    GL_CHECK(glDeleteRenderbuffers(1, &mDepthStencilBuffer));
+    GL_CHECK(glDeleteFramebuffers(1, &mResolveFbo));
+    GL_CHECK(glDeleteTextures(1, &mColorTex));
+    mFbo = GL_INVALID_VALUE;
+}
+
 GlRenderTargetPool::GlRenderTargetPool(uint32_t maxWidth, uint32_t maxHeight): mMaxWidth(maxWidth), mMaxHeight(maxHeight), mPool() {}
 
 GlRenderTargetPool::~GlRenderTargetPool()
@@ -101,31 +107,28 @@ uint32_t alignPow2(uint32_t value)
 
 GlRenderTarget* GlRenderTargetPool::getRenderTarget(const RenderRegion& vp, GLuint resolveId)
 {
-    uint32_t width = static_cast<uint32_t>(vp.w);
-    uint32_t height = static_cast<uint32_t>(vp.h);
+    auto width = vp.w();
+    auto height = vp.h();
 
     // pow2 align width and height
     if (width >= mMaxWidth) width = mMaxWidth;
     else width = alignPow2(width);
-
     if (width >= mMaxWidth) width = mMaxWidth;
 
     if (height >= mMaxHeight) height = mMaxHeight;
     else height = alignPow2(height);
-
     if (height >= mMaxHeight) height = mMaxHeight;
 
     for (uint32_t i = 0; i < mPool.count; i++) {
         auto rt = mPool[i];
-
         if (rt->getWidth() == width && rt->getHeight() == height) {
             rt->setViewport(vp);
             return rt;
         }
     }
 
-    auto rt = new GlRenderTarget(width, height);
-    rt->init(resolveId);
+    auto rt = new GlRenderTarget();
+    rt->init(width, height, resolveId);
     rt->setViewport(vp);
     mPool.push(rt);
     return rt;

src/renderer/gl_engine/tvgGlRenderTarget.h

@@ -28,11 +28,11 @@
 class GlRenderTarget
 {
 public:
-    GlRenderTarget() = default;
-    GlRenderTarget(uint32_t width, uint32_t height);
+    GlRenderTarget();
     ~GlRenderTarget();
 
-    void init(GLint resolveId);
+    void init(uint32_t width, uint32_t height, GLint resolveId);
+    void reset();
 
     GLuint getFboId() { return mFbo; }
     GLuint getResolveFboId() { return mResolveFbo; }
@@ -44,12 +44,13 @@ public:
     void setViewport(const RenderRegion& vp) { mViewport = vp; }
     const RenderRegion& getViewport() const { return mViewport; }
 
+    bool invalid() const { return mFbo == GL_INVALID_VALUE; }
 
 private:
     uint32_t mWidth = 0;
     uint32_t mHeight = 0;
     RenderRegion mViewport{};
-    GLuint mFbo = 0;
+    GLuint mFbo = GL_INVALID_VALUE;
     GLuint mColorBuffer = 0;
     GLuint mDepthStencilBuffer = 0;
     GLuint mResolveFbo = 0;

src/renderer/gl_engine/tvgGlRenderTask.cpp

@@ -24,6 +24,9 @@
 #include "tvgGlProgram.h"
 #include "tvgGlRenderPass.h"
 
+/************************************************************************/
+/* GlRenderTask Class Implementation                                    */
+/************************************************************************/
 
 GlRenderTask::GlRenderTask(GlProgram* program, GlRenderTask* other): mProgram(program)
 {
@@ -33,6 +36,7 @@ GlRenderTask::GlRenderTask(GlProgram* program, GlRenderTask* other): mProgram(pr
     mIndexCount = other->mIndexCount;
 }
 
+
 void GlRenderTask::run()
 {
     // bind shader
@@ -45,7 +49,7 @@ void GlRenderTask::run()
     }
 
     // setup scissor rect
-    GL_CHECK(glScissor(mViewport.x, mViewport.y, mViewport.w, mViewport.h));
+    GL_CHECK(glScissor(mViewport.sx(), mViewport.sy(), mViewport.sw(), mViewport.sh()));
 
     // setup attribute layout
     for (uint32_t i = 0; i < mVertexLayout.count; i++) {
@@ -81,36 +85,44 @@ void GlRenderTask::run()
     }
 }
 
+
 void GlRenderTask::addVertexLayout(const GlVertexLayout &layout)
 {
     mVertexLayout.push(layout);
 }
 
+
 void GlRenderTask::addBindResource(const GlBindingResource &binding)
 {
     mBindingResources.push(binding);
 }
 
+
 void GlRenderTask::setDrawRange(uint32_t offset, uint32_t count)
 {
     mIndexOffset = offset;
     mIndexCount = count;
 }
 
+
 void GlRenderTask::setViewport(const RenderRegion &viewport)
 {
     mViewport = viewport;
-    if (mViewport.w < 0) {
-        mViewport.w = 0;
-    }
-
-    if (mViewport.h < 0) {
-        mViewport.h = 0;
-    }
+    if (mViewport.max.x < mViewport.min.x) mViewport.max.x = mViewport.min.x;
+    if (mViewport.max.y < mViewport.min.y) mViewport.max.y = mViewport.min.y;
 }
 
+
+/************************************************************************/
+/* GlStencilCoverTask Class Implementation                              */
+/************************************************************************/
+
 GlStencilCoverTask::GlStencilCoverTask(GlRenderTask* stencil, GlRenderTask* cover, GlStencilMode mode)
- :GlRenderTask(nullptr), mStencilTask(stencil), mCoverTask(cover), mStencilMode(mode) {}
+ :GlRenderTask(nullptr), mStencilTask(stencil), mCoverTask(cover), mStencilMode(mode)
+ {
+
+ }
+
 
 GlStencilCoverTask::~GlStencilCoverTask()
 {
@@ -118,6 +130,7 @@ GlStencilCoverTask::~GlStencilCoverTask()
     delete mCoverTask;
 }
 
+
 void GlStencilCoverTask::run()
 {
     GL_CHECK(glEnable(GL_STENCIL_TEST));
@@ -151,12 +164,18 @@ void GlStencilCoverTask::run()
     GL_CHECK(glDisable(GL_STENCIL_TEST));
 }
 
+
 void GlStencilCoverTask::normalizeDrawDepth(int32_t maxDepth)
 {
     mCoverTask->normalizeDrawDepth(maxDepth);
     mStencilTask->normalizeDrawDepth(maxDepth);
 }
 
+
+/************************************************************************/
+/* GlComposeTask Class Implementation                                   */
+/************************************************************************/
+
 GlComposeTask::GlComposeTask(GlProgram* program, GLuint target, GlRenderTarget* fbo, Array<GlRenderTask*>&& tasks)
  :GlRenderTask(program) ,mTargetFbo(target), mFbo(fbo), mTasks()
 {
@@ -164,33 +183,32 @@ GlComposeTask::GlComposeTask(GlProgram* program, GLuint target, GlRenderTarget*
     tasks.clear();
 }
 
+
 GlComposeTask::~GlComposeTask()
 {
     ARRAY_FOREACH(p, mTasks) delete(*p);
     mTasks.clear();
 }
 
+
 void GlComposeTask::run()
 {
     GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, getSelfFbo()));
 
-    // clear this fbo
-    if (mClearBuffer) {
     // we must clear all area of fbo
     GL_CHECK(glViewport(0, 0, mFbo->getWidth(), mFbo->getHeight()));
     GL_CHECK(glScissor(0, 0, mFbo->getWidth(), mFbo->getHeight()));
     GL_CHECK(glClearColor(0, 0, 0, 0));
     GL_CHECK(glClearStencil(0));
-    #ifdef THORVG_GL_TARGET_GLES
+#ifdef THORVG_GL_TARGET_GLES
     GL_CHECK(glClearDepthf(0.0));
-    #else
+#else
     GL_CHECK(glClearDepth(0.0));
-    #endif
+#endif
     GL_CHECK(glDepthMask(1));
 
     GL_CHECK(glClear(GL_COLOR_BUFFER_BIT | GL_STENCIL_BUFFER_BIT | GL_DEPTH_BUFFER_BIT));
     GL_CHECK(glDepthMask(0));
-    }
 
     GL_CHECK(glViewport(0, 0, mRenderWidth, mRenderHeight));
     GL_CHECK(glScissor(0, 0, mRenderWidth, mRenderHeight));
@@ -209,29 +227,43 @@ void GlComposeTask::run()
     onResolve();
 }
 
-GLuint GlComposeTask::getSelfFbo() { return mFbo->getFboId(); }
 
-GLuint GlComposeTask::getResolveFboId() { return mFbo->getResolveFboId(); }
+GLuint GlComposeTask::getSelfFbo()
+{
+    return mFbo->getFboId();
+}
 
-void GlComposeTask::onResolve() {
+
+GLuint GlComposeTask::getResolveFboId()
+{
+    return mFbo->getResolveFboId();
+}
+
+
+void GlComposeTask::onResolve()
+{
     GL_CHECK(glBindFramebuffer(GL_READ_FRAMEBUFFER, getSelfFbo()));
     GL_CHECK(glBindFramebuffer(GL_DRAW_FRAMEBUFFER, getResolveFboId()));
-
-
     GL_CHECK(glBlitFramebuffer(0, 0, mRenderWidth, mRenderHeight, 0, 0, mRenderWidth, mRenderHeight, GL_COLOR_BUFFER_BIT, GL_NEAREST));
 }
 
+
+/************************************************************************/
+/* GlBlitTask Class Implementation                                      */
+/************************************************************************/
+
 GlBlitTask::GlBlitTask(GlProgram* program, GLuint target, GlRenderTarget* fbo, Array<GlRenderTask*>&& tasks)
  : GlComposeTask(program, target, fbo, std::move(tasks)), mColorTex(fbo->getColorTexture())
 {
 }
 
+
 void GlBlitTask::run()
 {
     GlComposeTask::run();
 
     GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, getTargetFbo()));
-    GL_CHECK(glViewport(mTargetViewport.x, mTargetViewport.y, mTargetViewport.w, mTargetViewport.h));
+    GL_CHECK(glViewport(mTargetViewport.x(), mTargetViewport.y(), mTargetViewport.w(), mTargetViewport.h()));
 
     if (mClearBuffer) {
         GL_CHECK(glClearColor(0, 0, 0, 0));
@@ -246,16 +278,24 @@ void GlBlitTask::run()
     GlRenderTask::run();
 }
 
+
+/************************************************************************/
+/* GlDrawBlitTask Class Implementation                                  */
+/************************************************************************/
+
+
 GlDrawBlitTask::GlDrawBlitTask(GlProgram* program, GLuint target, GlRenderTarget* fbo, Array<GlRenderTask*>&& tasks)
  : GlComposeTask(program, target, fbo, std::move(tasks))
 {
 }
 
+
 GlDrawBlitTask::~GlDrawBlitTask()
 {
     if (mPrevTask) delete mPrevTask;
 }
 
+
 void GlDrawBlitTask::run()
 {
     if (mPrevTask) mPrevTask->run();
@@ -269,15 +309,22 @@ void GlDrawBlitTask::run()
     GlRenderTask::run();
 }
 
+
+/************************************************************************/
+/* GlClipTask Class Implementation                                      */
+/************************************************************************/
+
 GlClipTask::GlClipTask(GlRenderTask* clip, GlRenderTask* mask)
  :GlRenderTask(nullptr), mClipTask(clip), mMaskTask(mask) {}
 
+
 GlClipTask::~GlClipTask()
 {
     delete mClipTask;
     delete mMaskTask;
 }
 
+
 void GlClipTask::run()
 {
     GL_CHECK(glEnable(GL_STENCIL_TEST));
@@ -304,14 +351,22 @@ void GlClipTask::run()
     GL_CHECK(glDisable(GL_STENCIL_TEST));
 }
 
+
 void GlClipTask::normalizeDrawDepth(int32_t maxDepth)
 {
     mClipTask->normalizeDrawDepth(maxDepth);
     mMaskTask->normalizeDrawDepth(maxDepth);
 }
 
+/************************************************************************/
+/* GlSimpleBlendTask Class Implementation                               */
+/************************************************************************/
+
 GlSimpleBlendTask::GlSimpleBlendTask(BlendMethod method, GlProgram* program)
- : GlRenderTask(program), mBlendMethod(method) {}
+ : GlRenderTask(program), mBlendMethod(method)
+ {
+ }
+
 
 void GlSimpleBlendTask::run()
 {
@@ -332,8 +387,17 @@ void GlSimpleBlendTask::run()
     glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
 }
 
+
+/************************************************************************/
+/* GlComplexBlendTask Class Implementation                              */
+/************************************************************************/
+
+
 GlComplexBlendTask::GlComplexBlendTask(GlProgram* program, GlRenderTarget* dstFbo, GlRenderTarget* dstCopyFbo, GlRenderTask* stencilTask, GlComposeTask* composeTask)
- : GlRenderTask(program), mDstFbo(dstFbo), mDstCopyFbo(dstCopyFbo), mStencilTask(stencilTask), mComposeTask(composeTask) {}
+ : GlRenderTask(program), mDstFbo(dstFbo), mDstCopyFbo(dstCopyFbo), mStencilTask(stencilTask), mComposeTask(composeTask)
+ {
+ }
+
 
 GlComplexBlendTask::~GlComplexBlendTask()
 {
@@ -341,6 +405,7 @@ GlComplexBlendTask::~GlComplexBlendTask()
     delete mComposeTask;
 }
 
+
 void GlComplexBlendTask::run()
 {
     mComposeTask->run();
@@ -349,12 +414,11 @@ void GlComplexBlendTask::run()
     GL_CHECK(glBindFramebuffer(GL_READ_FRAMEBUFFER, mDstFbo->getFboId()));
     GL_CHECK(glBindFramebuffer(GL_DRAW_FRAMEBUFFER, mDstCopyFbo->getResolveFboId()));
 
-    GL_CHECK(glViewport(0, 0, mDstFbo->getViewport().w, mDstFbo->getViewport().h));
-    GL_CHECK(glScissor(0, 0, mDstFbo->getViewport().w, mDstFbo->getViewport().h));
+    GL_CHECK(glViewport(0, 0, mDstFbo->getViewport().w(), mDstFbo->getViewport().h()));
+    GL_CHECK(glScissor(0, 0, mDstFbo->getViewport().w(), mDstFbo->getViewport().h()));
     
     const auto& vp = getViewport();
-
-    GL_CHECK(glBlitFramebuffer(vp.x, vp.y, vp.x + vp.w, vp.y + vp.h, 0, 0, vp.w, vp.h, GL_COLOR_BUFFER_BIT, GL_LINEAR));
+    GL_CHECK(glBlitFramebuffer(vp.min.x, vp.min.y, vp.max.x, vp.max.y, 0, 0, vp.w(), vp.h(), GL_COLOR_BUFFER_BIT, GL_LINEAR));
 
     GL_CHECK(glBindFramebuffer(GL_FRAMEBUFFER, mDstFbo->getFboId()));
 
@@ -381,12 +445,17 @@ void GlComplexBlendTask::run()
     GL_CHECK(glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA));
 }
 
+
 void GlComplexBlendTask::normalizeDrawDepth(int32_t maxDepth)
 {
     mStencilTask->normalizeDrawDepth(maxDepth);
     GlRenderTask::normalizeDrawDepth(maxDepth);
 }
 
+/************************************************************************/
+/* GlGaussianBlurTask Class Implementation                              */
+/************************************************************************/
+
 void GlGaussianBlurTask::run()
 {
     const auto vp = getViewport();
@@ -439,6 +508,9 @@ void GlGaussianBlurTask::run()
     GL_CHECK(glEnable(GL_BLEND));
 }
 
+/************************************************************************/
+/* GlEffectDropShadowTask Class Implementation                          */
+/************************************************************************/
 
 void GlEffectDropShadowTask::run()
 {
@@ -492,6 +564,9 @@ void GlEffectDropShadowTask::run()
     GL_CHECK(glEnable(GL_BLEND));
 }
 
+/************************************************************************/
+/* GlEffectColorTransformTask Class Implementation                      */
+/************************************************************************/
 
 void GlEffectColorTransformTask::run()
 {

src/renderer/gl_engine/tvgGlRenderTask.h

@@ -130,12 +130,10 @@ public:
 
 protected:
     GLuint getTargetFbo() { return mTargetFbo; }
-
     GLuint getSelfFbo();
-
     GLuint getResolveFboId();
-
     void onResolve();
+
 private:
     GLuint mTargetFbo;
     GlRenderTarget* mFbo;

src/renderer/gl_engine/tvgGlRenderer.cpp

@@ -52,6 +52,8 @@ void GlRenderer::flush()
 {
     clearDisposes();
 
+    mRootTarget.reset();
+
     ARRAY_FOREACH(p, mComposePool) delete(*p);
     mComposePool.clear();
 
@@ -181,10 +183,10 @@ void GlRenderer::drawPrimitive(GlShape& sdata, const RenderColor& c, RenderUpdat
         bbox.intersect(vp);
     }
 
-    auto x = bbox.x - vp.x;
-    auto y = bbox.y - vp.y;
-    auto w = bbox.w;
-    auto h = bbox.h;
+    auto x = bbox.sx() - vp.sx();
+    auto y = bbox.sy() - vp.sy();
+    auto w = bbox.sw();
+    auto h = bbox.sh();
 
     GlRenderTask* task = nullptr;
     if (mBlendMethod != BlendMethod::Normal && !complexBlend) task = new GlSimpleBlendTask(mBlendMethod, mPrograms[RT_Color]);
@@ -197,7 +199,8 @@ void GlRenderer::drawPrimitive(GlShape& sdata, const RenderColor& c, RenderUpdat
         return;
     }
 
-    task->setViewport({x, vp.h - y - h, w, h});
+    y = vp.sh() - y - h;
+    task->setViewport({{x, y}, {x + w, y + h}});
 
     GlRenderTask* stencilTask = nullptr;
 
@@ -266,7 +269,6 @@ void GlRenderer::drawPrimitive(GlShape& sdata, const Fill* fill, RenderUpdateFla
 {
     auto vp = currentPass()->getViewport();
     auto bbox = sdata.geometry.viewport;
-
     bbox.intersect(vp);
 
     const Fill::ColorStop* stops = nullptr;
@@ -275,13 +277,9 @@ void GlRenderer::drawPrimitive(GlShape& sdata, const Fill* fill, RenderUpdateFla
 
     GlRenderTask* task = nullptr;
 
-    if (fill->type() == Type::LinearGradient) {
-        task = new GlRenderTask(mPrograms[RT_LinGradient]);
-    } else if (fill->type() == Type::RadialGradient) {
-        task = new GlRenderTask(mPrograms[RT_RadGradient]);
-    } else {
-        return;
-    }
+    if (fill->type() == Type::LinearGradient) task = new GlRenderTask(mPrograms[RT_LinGradient]);
+    else if (fill->type() == Type::RadialGradient) task = new GlRenderTask(mPrograms[RT_RadGradient]);
+    else return;
 
     task->setDrawDepth(depth);
 
@@ -291,13 +289,11 @@ void GlRenderer::drawPrimitive(GlShape& sdata, const Fill* fill, RenderUpdateFla
     }
 
     auto complexBlend = beginComplexBlending(bbox, sdata.geometry.getBounds());
-
     if (complexBlend) vp = currentPass()->getViewport();
 
-    auto x = bbox.x - vp.x;
-    auto y = bbox.y - vp.y;
-
-    task->setViewport({x, vp.h - y - bbox.h, bbox.w, bbox.h});
+    auto x = bbox.sx() - vp.sx();
+    auto y = vp.sh() - (bbox.sy() - vp.sy()) - bbox.sh();
+    task->setViewport({{x, y}, {x + bbox.sw(), y + bbox.sh()}});
 
     GlRenderTask* stencilTask = nullptr;
     GlStencilMode stencilMode = sdata.geometry.getStencilMode(flag);
@@ -496,22 +492,18 @@ void GlRenderer::drawClip(Array<RenderData>& clips)
 
     for (uint32_t i = 0; i < clips.count; ++i) {
         auto sdata = static_cast<GlShape*>(clips[i]);
-
         auto clipTask = new GlRenderTask(mPrograms[RT_Stencil]);
-
         clipTask->setDrawDepth(clipDepths[i]);
 
         auto flag = (sdata->geometry.stroke.vertex.count > 0) ? RenderUpdateFlag::Stroke : RenderUpdateFlag::Path;
         sdata->geometry.draw(clipTask, &mGpuBuffer, flag);
 
         auto bbox = sdata->geometry.viewport;
-
         bbox.intersect(vp);
 
-        auto x = bbox.x - vp.x;
-        auto y = bbox.y - vp.y;
-
-        clipTask->setViewport({x, vp.h - y - bbox.h, bbox.w, bbox.h});
+        auto x = bbox.sx() - vp.sx();
+        auto y = vp.sh() - (bbox.sy() - vp.sy()) - bbox.sh();
+        clipTask->setViewport({{x, y}, {x + bbox.sw(), y + bbox.sh()}});
 
         float matrix44[16];
         currentPass()->getMatrix(matrix44, sdata->geometry.matrix);
@@ -527,7 +519,7 @@ void GlRenderer::drawClip(Array<RenderData>& clips)
         maskTask->addVertexLayout(GlVertexLayout{0, 2, 2 * sizeof(float), identityVertexOffset});
         maskTask->addBindResource(GlBindingResource{0, loc, mGpuBuffer.getBufferId(), mat4Offset, 16 * sizeof(float), });
         maskTask->setDrawRange(identityIndexOffset, 6);
-        maskTask->setViewport({0, 0, static_cast<int32_t>(vp.w), static_cast<int32_t>(vp.h)});
+        maskTask->setViewport({{0, 0}, {vp.sw(), vp.sh()}});
 
         currentPass()->addRenderTask(new GlClipTask(clipTask, maskTask));
     }
@@ -541,11 +533,10 @@ GlRenderPass* GlRenderer::currentPass()
 
 bool GlRenderer::beginComplexBlending(const RenderRegion& vp, RenderRegion bounds)
 {
-    if (vp.w == 0 || vp.h == 0) return false;
+    if (vp.invalid()) return false;
 
     bounds.intersect(vp);
-
-    if (bounds.w == 0 || bounds.h == 0) return false;
+    if (bounds.invalid()) return false;
 
     if (mBlendMethod == BlendMethod::Normal || mBlendMethod == BlendMethod::Add || mBlendMethod == BlendMethod::Darken || mBlendMethod == BlendMethod::Lighten) return false;
 
@@ -571,20 +562,12 @@ void GlRenderer::endBlendingCompose(GlRenderTask* stencilTask, const Matrix& mat
     if (mBlendPool.count < 2) mBlendPool.push(new GlRenderTargetPool(surface.w, surface.h));
     auto dstCopyFbo = mBlendPool[1]->getRenderTarget(vp);
 
-    {
-        const auto& passVp = currentPass()->getViewport();
-
-        auto x = vp.x;
-        auto y = vp.y;
-        auto w = vp.w;
-        auto h = vp.h;
-
-        stencilTask->setViewport({x, passVp.h - y - h, w, h});
-    }
+    auto x = vp.sx();
+    auto y = currentPass()->getViewport().sh() - vp.sy() - vp.sh();
+    stencilTask->setViewport({{x, y}, {x + vp.sw(), y + vp.sh()}});
 
     stencilTask->setDrawDepth(currentPass()->nextDrawDepth());
 
-    {
     // set view matrix
     float matrix44[16];
     currentPass()->getMatrix(matrix44, matrix);
@@ -596,12 +579,9 @@ void GlRenderer::endBlendingCompose(GlRenderTask* stencilTask, const Matrix& mat
         viewOffset,
         16 * sizeof(float),
     });
-    }
 
     auto task = new GlComplexBlendTask(getBlendProgram(), currentPass()->getFbo(), dstCopyFbo, stencilTask, composeTask);
-
     prepareCmpTask(task, vp, blendPass->getFboWidth(), blendPass->getFboHeight());
-
     task->setDrawDepth(currentPass()->nextDrawDepth());
 
     // src and dst texture
@@ -632,8 +612,7 @@ GlProgram* GlRenderer::getBlendProgram()
 
 void GlRenderer::prepareBlitTask(GlBlitTask* task)
 {
-    RenderRegion region{0, 0, static_cast<int32_t>(surface.w), static_cast<int32_t>(surface.h)};
-    prepareCmpTask(task, region, surface.w, surface.h);
+    prepareCmpTask(task, {{0, 0}, {int32_t(surface.w), int32_t(surface.h)}}, surface.w, surface.h);
     task->addBindResource(GlBindingResource{0, task->getColorTexture(), task->getProgram()->getUniformLocation("uSrcTexture")});
 }
 
@@ -648,13 +627,13 @@ void GlRenderer::prepareCmpTask(GlRenderTask* task, const RenderRegion& vp, uint
     auto taskVp = vp;
     taskVp.intersect(passVp);
 
-    auto x = taskVp.x - passVp.x;
-    auto y = taskVp.y - passVp.y;
-    auto w = taskVp.w;
-    auto h = taskVp.h;
+    auto x = taskVp.sx() - passVp.sx();
+    auto y = taskVp.sy() - passVp.sy();
+    auto w = taskVp.sw();
+    auto h = taskVp.sh();
 
-    float rw = static_cast<float>(passVp.w);
-    float rh = static_cast<float>(passVp.h);
+    float rw = static_cast<float>(passVp.w());
+    float rh = static_cast<float>(passVp.h());
 
     float l = static_cast<float>(x);
     float t = static_cast<float>(rh - y);
@@ -705,10 +684,9 @@ void GlRenderer::prepareCmpTask(GlRenderTask* task, const RenderRegion& vp, uint
 
     task->addVertexLayout(GlVertexLayout{0, 2, 4 * sizeof(float), vertexOffset});
     task->addVertexLayout(GlVertexLayout{1, 2, 4 * sizeof(float), vertexOffset + 2 * sizeof(float)});
-
     task->setDrawRange(indexOffset, indices.count);
-
-    task->setViewport({x, static_cast<int32_t>((passVp.h - y - h)), w, h});
+    y = (passVp.sh() - y - h);
+    task->setViewport({{x, y}, {x + w, y + h}});
 }
 
 
@@ -741,11 +719,11 @@ void GlRenderer::endRenderPass(RenderCompositor* cmp)
         if (program && !selfPass->isEmpty() && !maskPass->isEmpty()) {
             auto prev_task = maskPass->endRenderPass<GlComposeTask>(nullptr, currentPass()->getFboId());
             prev_task->setDrawDepth(currentPass()->nextDrawDepth());
-            prev_task->setRenderSize(static_cast<uint32_t>(glCmp->bbox.w), static_cast<uint32_t>(glCmp->bbox.h));
+            prev_task->setRenderSize(glCmp->bbox.w(), glCmp->bbox.h());
             prev_task->setViewport(glCmp->bbox);
 
             auto compose_task = selfPass->endRenderPass<GlDrawBlitTask>(program, currentPass()->getFboId());
-            compose_task->setRenderSize(static_cast<uint32_t>(glCmp->bbox.w), static_cast<uint32_t>(glCmp->bbox.h));
+            compose_task->setRenderSize(glCmp->bbox.w(), glCmp->bbox.h());
             compose_task->setPrevTask(prev_task);
 
             prepareCmpTask(compose_task, glCmp->bbox, selfPass->getFboWidth(), selfPass->getFboHeight());
@@ -754,7 +732,7 @@ void GlRenderer::endRenderPass(RenderCompositor* cmp)
             compose_task->addBindResource(GlBindingResource{1, maskPass->getTextureId(), program->getUniformLocation("uMaskTexture")});
 
             compose_task->setDrawDepth(currentPass()->nextDrawDepth());
-            compose_task->setParentSize(static_cast<uint32_t>(currentPass()->getViewport().w), static_cast<uint32_t>(currentPass()->getViewport().h));
+            compose_task->setParentSize(currentPass()->getViewport().w(), currentPass()->getViewport().h());
             currentPass()->addRenderTask(compose_task);
         }
 
@@ -767,7 +745,7 @@ void GlRenderer::endRenderPass(RenderCompositor* cmp)
 
         if (!renderPass->isEmpty()) {
             auto task = renderPass->endRenderPass<GlDrawBlitTask>(mPrograms[RT_Image], currentPass()->getFboId());
-            task->setRenderSize(static_cast<uint32_t>(glCmp->bbox.w), static_cast<uint32_t>(glCmp->bbox.h));
+            task->setRenderSize(glCmp->bbox.w(), glCmp->bbox.h());
             prepareCmpTask(task, glCmp->bbox, renderPass->getFboWidth(), renderPass->getFboHeight());
             task->setDrawDepth(currentPass()->nextDrawDepth());
 
@@ -795,7 +773,7 @@ void GlRenderer::endRenderPass(RenderCompositor* cmp)
 
             // texture id
             task->addBindResource(GlBindingResource{0, renderPass->getTextureId(), task->getProgram()->getUniformLocation("uTexture")});
-            task->setParentSize(static_cast<uint32_t>(currentPass()->getViewport().w), static_cast<uint32_t>(currentPass()->getViewport().h));
+            task->setParentSize(currentPass()->getViewport().w(), currentPass()->getViewport().h());
             currentPass()->addRenderTask(std::move(task));
         }
         delete(renderPass);
@@ -809,6 +787,8 @@ void GlRenderer::endRenderPass(RenderCompositor* cmp)
 
 bool GlRenderer::clear()
 {
+    if (mRootTarget.invalid()) return false;
+
     mClearBuffer = true;
     return true;
 }
@@ -832,9 +812,8 @@ bool GlRenderer::target(void* context, int32_t id, uint32_t w, uint32_t h)
 
     currentContext();
 
-    mRootTarget = GlRenderTarget(surface.w, surface.h);
-    mRootTarget.setViewport({0, 0, static_cast<int32_t>(surface.w), static_cast<int32_t>(surface.h)});
-    mRootTarget.init(mTargetFboId);
+    mRootTarget.setViewport({{0, 0}, {int32_t(surface.w), int32_t(surface.h)}});
+    mRootTarget.init(surface.w, surface.h, mTargetFboId);
 
     return true;
 }
@@ -861,7 +840,7 @@ bool GlRenderer::sync()
     prepareBlitTask(task);
 
     task->mClearBuffer = mClearBuffer;
-    task->setTargetViewport({0, 0, static_cast<int32_t>(surface.w), static_cast<int32_t>(surface.h)});
+    task->setTargetViewport({{0, 0}, {int32_t(surface.w), int32_t(surface.h)}});
 
     if (mGpuBuffer.flushToGPU()) {
         mGpuBuffer.bind();
@@ -874,6 +853,9 @@ bool GlRenderer::sync()
 
     clearDisposes();
 
+    // Reset clear buffer flag to default (false) after use.    
+    mClearBuffer = false; 
+
     delete task;
 
     return true;
@@ -882,7 +864,7 @@ bool GlRenderer::sync()
 
 RenderRegion GlRenderer::region(RenderData data)
 {
-    if (currentPass()->isEmpty()) return {0, 0, 0, 0};
+    if (currentPass()->isEmpty()) return {};
 
     auto shape = reinterpret_cast<GlShape*>(data);
     auto bounds = shape->geometry.getBounds();
@@ -896,6 +878,8 @@ RenderRegion GlRenderer::region(RenderData data)
 
 bool GlRenderer::preRender()
 {
+    if (mRootTarget.invalid()) return false;
+
     currentContext();
     if (mPrograms.empty()) initShaders();
     mRenderPassStack.push(new GlRenderPass(&mRootTarget));
@@ -930,20 +914,11 @@ bool GlRenderer::beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_
     cmp->opacity = opacity;
 
     uint32_t index = mRenderPassStack.count - 1;
-
-    if (index >= mComposePool.count) {
-        mComposePool.push( new GlRenderTargetPool(surface.w, surface.h));
-    }
+    if (index >= mComposePool.count) mComposePool.push( new GlRenderTargetPool(surface.w, surface.h));
 
     auto glCmp = static_cast<GlCompositor*>(cmp);
-
-    if (glCmp->bbox.w > 0 && glCmp->bbox.h > 0) {
-        auto renderTarget = mComposePool[index]->getRenderTarget(glCmp->bbox);
-        mRenderPassStack.push(new GlRenderPass(renderTarget));
-    } else {
-        // empty render pass
-        mRenderPassStack.push(new GlRenderPass(nullptr));
-    }
+    if (glCmp->bbox.valid()) mRenderPassStack.push(new GlRenderPass(mComposePool[index]->getRenderTarget(glCmp->bbox)));
+    else mRenderPassStack.push(new GlRenderPass(nullptr));
 
     return true;
 }
@@ -1064,12 +1039,12 @@ bool GlRenderer::effectGaussianBlurRegion(RenderEffectGaussianBlur* effect)
 {
     auto gaussianBlur = (GlGaussianBlur*)effect->rd;
     if (effect->direction != 2) {
-        effect->extend.x = -gaussianBlur->extend;
-        effect->extend.w = +gaussianBlur->extend * 2;
+        effect->extend.min.x = -gaussianBlur->extend;
+        effect->extend.max.x = +gaussianBlur->extend;
     }
     if (effect->direction != 1) {
-        effect->extend.y = -gaussianBlur->extend;
-        effect->extend.h = +gaussianBlur->extend * 2;
+        effect->extend.min.y = -gaussianBlur->extend;
+        effect->extend.max.y = +gaussianBlur->extend;
     }
     return true;
 };
@@ -1078,10 +1053,10 @@ bool GlRenderer::effectGaussianBlurRegion(RenderEffectGaussianBlur* effect)
 bool GlRenderer::effectDropShadowRegion(RenderEffectDropShadow* effect)
 {
     auto gaussianBlur = (GlDropShadow*)effect->rd;
-    effect->extend.x = -gaussianBlur->extend;
-    effect->extend.w = +gaussianBlur->extend * 2;
-    effect->extend.y = -gaussianBlur->extend;
-    effect->extend.h = +gaussianBlur->extend * 2;
+    effect->extend.min.x = -gaussianBlur->extend;
+    effect->extend.max.x = +gaussianBlur->extend;
+    effect->extend.min.y = -gaussianBlur->extend;
+    effect->extend.max.y = +gaussianBlur->extend;
     return true;
 };
 
@@ -1131,7 +1106,6 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
     auto voffset = mGpuBuffer.push((void*)vdata, sizeof(vdata));
     auto ioffset = mGpuBuffer.pushIndex((void*)idata, sizeof(idata));
 
-    // effect gaussian blur
     if (effect->type == SceneEffect::GaussianBlur) {
         // get gaussian programs
         GlProgram* programHorz = mPrograms[RT_GaussianHorz];
@@ -1147,7 +1121,7 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
         // create gaussian blur tasks
         auto gaussianTask = new GlGaussianBlurTask(dstFbo, dstCopyFbo0, dstCopyFbo1);
         gaussianTask->effect = (RenderEffectGaussianBlur*)effect;
-        gaussianTask->setViewport({0, 0, vp.w, vp.h});
+        gaussianTask->setViewport({{0, 0}, {vp.sw(), vp.sh()}});
         // horizontal blur task and geometry
         gaussianTask->horzTask = new GlRenderTask(programHorz);
         gaussianTask->horzTask->addBindResource(GlBindingResource{0, programHorz->getUniformBlockIndex("Gaussian"), mGpuBuffer.getBufferId(), blurOffset, sizeof(GlGaussianBlur)});
@@ -1160,8 +1134,7 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
         gaussianTask->vertTask->setDrawRange(ioffset, 6);
         // add task to render pipeline
         pass->addRenderTask(gaussianTask);
-    } // effect drop shadow
-    else if (effect->type == SceneEffect::DropShadow) {
+    } else if (effect->type == SceneEffect::DropShadow) {
         // get programs
         GlProgram* program = mPrograms[RT_DropShadow];
         GlProgram* programHorz = mPrograms[RT_GaussianHorz];
@@ -1177,7 +1150,7 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
         // create gaussian blur tasks
         auto task = new GlEffectDropShadowTask(program, dstFbo, dstCopyFbo0, dstCopyFbo1);
         task->effect = (RenderEffectDropShadow*)effect;
-        task->setViewport({0, 0, vp.w, vp.h});
+        task->setViewport({{0, 0}, {vp.sw(), vp.sh()}});
         task->addBindResource(GlBindingResource{0, program->getUniformBlockIndex("DropShadow"), mGpuBuffer.getBufferId(), paramsOffset, sizeof(GlDropShadow)});
         task->addVertexLayout(GlVertexLayout{0, 2, 2 * sizeof(float), voffset});
         task->setDrawRange(ioffset, 6);
@@ -1193,8 +1166,7 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
         task->vertTask->setDrawRange(ioffset, 6);
         // add task to render pipeline
         pass->addRenderTask(task);
-    } // effect fill, tint, tritone
-    else if ((effect->type == SceneEffect::Fill) || (effect->type == SceneEffect::Tint) || (effect->type == SceneEffect::Tritone)) {
+    } else if ((effect->type == SceneEffect::Fill) || (effect->type == SceneEffect::Tint) || (effect->type == SceneEffect::Tritone)) {
         GlProgram* program{};
         if (effect->type == SceneEffect::Fill) program = mPrograms[RT_EffectFill];
         else if (effect->type == SceneEffect::Tint) program = mPrograms[RT_EffectTint];
@@ -1208,7 +1180,7 @@ bool GlRenderer::render(TVG_UNUSED RenderCompositor* cmp, const RenderEffect* ef
         auto paramsOffset = mGpuBuffer.push(params, sizeof(GlEffectParams), true);
         // create and setup task
         auto task = new GlEffectColorTransformTask(program, dstFbo, dstCopyFbo);
-        task->setViewport({0, 0, vp.w, vp.h});
+        task->setViewport({{0, 0}, {vp.sw(), vp.sh()}});
         task->addBindResource(GlBindingResource{0, program->getUniformBlockIndex("Params"), mGpuBuffer.getBufferId(), paramsOffset, sizeof(GlEffectParams)});
         task->addVertexLayout(GlVertexLayout{0, 2, 2 * sizeof(float), voffset});
         task->setDrawRange(ioffset, 6);
@@ -1251,25 +1223,18 @@ bool GlRenderer::blend(BlendMethod method)
 bool GlRenderer::renderImage(void* data)
 {
     auto sdata = static_cast<GlShape*>(data);
-
     if (!sdata) return false;
 
-    if (currentPass()->isEmpty()) return true;
+    if (currentPass()->isEmpty() || !(sdata->updateFlag & RenderUpdateFlag::Image)) return true;
 
-    if ((sdata->updateFlag & RenderUpdateFlag::Image) == 0) return true;
     auto vp = currentPass()->getViewport();
-
     auto bbox = sdata->geometry.viewport;
-
     bbox.intersect(vp);
+    if (bbox.invalid()) return true;
 
-    if (bbox.w <= 0 || bbox.h <= 0) return true;
-
-    auto x = bbox.x - vp.x;
-    auto y = bbox.y - vp.y;
-
-
-    int32_t drawDepth = currentPass()->nextDrawDepth();
+    auto x = bbox.sx() - vp.sx();
+    auto y = bbox.sy() - vp.sy();
+    auto drawDepth = currentPass()->nextDrawDepth();
 
     if (!sdata->clips.empty()) drawClip(sdata->clips);
 
@@ -1282,7 +1247,6 @@ bool GlRenderer::renderImage(void* data)
     }
 
     bool complexBlend = beginComplexBlending(bbox, sdata->geometry.getBounds());
-
     if (complexBlend) vp = currentPass()->getViewport();
 
     // matrix buffer
@@ -1311,7 +1275,11 @@ bool GlRenderer::renderImage(void* data)
     // texture id
     task->addBindResource(GlBindingResource{0, sdata->texId, task->getProgram()->getUniformLocation("uTexture")});
 
-    task->setViewport({x, vp.h - y - bbox.h, bbox.w, bbox.h});
+    y = vp.sh() - y - bbox.sh();
+    auto x2 = x + bbox.sw();
+    auto y2 = y + bbox.sh();
+
+    task->setViewport({{x, y}, {x2, y2}});
 
     currentPass()->addRenderTask(task);
 
@@ -1327,33 +1295,25 @@ bool GlRenderer::renderImage(void* data)
 
 bool GlRenderer::renderShape(RenderData data)
 {
-    auto sdata = static_cast<GlShape*>(data);
-
     if (currentPass()->isEmpty()) return true;
 
+    auto sdata = static_cast<GlShape*>(data);
     if (sdata->updateFlag == RenderUpdateFlag::None) return true;
 
-    const auto& vp = currentPass()->getViewport();
-
     auto bbox = sdata->geometry.viewport;
-    bbox.intersect(vp);
-
-    if (bbox.w <= 0 || bbox.h <= 0) return true;
+    bbox.intersect(currentPass()->getViewport());
+    if (bbox.invalid()) return true;
 
     int32_t drawDepth1 = 0, drawDepth2 = 0;
 
-    size_t flags = static_cast<size_t>(sdata->updateFlag);
-
-    if (flags == 0) return false;
-
-    if (flags & (RenderUpdateFlag::Gradient | RenderUpdateFlag::Color)) drawDepth1 = currentPass()->nextDrawDepth();
-
-    if (flags & (RenderUpdateFlag::Stroke | RenderUpdateFlag::GradientStroke)) drawDepth2 = currentPass()->nextDrawDepth();
+    if (sdata->updateFlag == RenderUpdateFlag::None) return false;
+    if (sdata->updateFlag & (RenderUpdateFlag::Gradient | RenderUpdateFlag::Color)) drawDepth1 = currentPass()->nextDrawDepth();
+    if (sdata->updateFlag & (RenderUpdateFlag::Stroke | RenderUpdateFlag::GradientStroke)) drawDepth2 = currentPass()->nextDrawDepth();
 
     if (!sdata->clips.empty()) drawClip(sdata->clips);
 
     auto processFill = [&]() {
-        if (flags & (RenderUpdateFlag::Color | RenderUpdateFlag::Gradient)) {
+        if (sdata->updateFlag & (RenderUpdateFlag::Color | RenderUpdateFlag::Gradient)) {
             if (const auto& gradient = sdata->rshape->fill) {
                 drawPrimitive(*sdata, gradient, RenderUpdateFlag::Gradient, drawDepth1);
             } else if (sdata->rshape->color.a > 0) {
@@ -1364,7 +1324,7 @@ bool GlRenderer::renderShape(RenderData data)
 
     auto processStroke = [&]() {
         if (!sdata->rshape->stroke) return;
-        if (flags & (RenderUpdateFlag::Stroke | RenderUpdateFlag::GradientStroke)) {
+        if (sdata->updateFlag & (RenderUpdateFlag::Stroke | RenderUpdateFlag::GradientStroke)) {
             if (const auto& gradient = sdata->rshape->strokeFill()) {
                 drawPrimitive(*sdata, gradient, RenderUpdateFlag::GradientStroke, drawDepth2);
             } else if (sdata->rshape->stroke->color.a > 0) {
@@ -1525,6 +1485,8 @@ bool GlRenderer::viewport(const RenderRegion& vp)
 
 bool GlRenderer::preUpdate()
 {
+    if (mRootTarget.invalid()) return false;
+
     currentContext();
     return true;
 }

src/renderer/gl_engine/tvgGlRenderer.h

@@ -153,7 +153,7 @@ private:
     } mDisposed;
 
     BlendMethod mBlendMethod = BlendMethod::Normal;
-    bool mClearBuffer = true;  //FIXME: clear buffer should be optional (default is false)
+    bool mClearBuffer = false;
 };
 
 #endif /* _TVG_GL_RENDERER_H_ */

src/renderer/gl_engine/tvgGlTessellator.cpp

@@ -1516,12 +1516,7 @@ void Stroker::stroke(const RenderShape *rshape, const RenderPath& path)
 
 RenderRegion Stroker::bounds() const
 {
-    return RenderRegion {
-        static_cast<int32_t>(floor(mLeftTop.x)),
-        static_cast<int32_t>(floor(mLeftTop.y)),
-        static_cast<int32_t>(ceil(mRightBottom.x - floor(mLeftTop.x))),
-        static_cast<int32_t>(ceil(mRightBottom.y - floor(mLeftTop.y))),
-    };
+    return {{int32_t(floor(mLeftTop.x)), int32_t(floor(mLeftTop.y))}, {int32_t(ceil(mRightBottom.x)), int32_t(ceil(mRightBottom.y))}};
 }
 
 
@@ -2197,26 +2192,15 @@ void BWTessellator::tessellate(const RenderPath& path, const Matrix& matrix)
 
 RenderRegion BWTessellator::bounds() const
 {
-    return RenderRegion {
-        static_cast<int32_t>(floor(bbox.min.x)),
-        static_cast<int32_t>(floor(bbox.min.y)),
-        static_cast<int32_t>(ceil(bbox.max.x - floor(bbox.min.x))),
-        static_cast<int32_t>(ceil(bbox.max.y - floor(bbox.min.y))),
-    };
+    return {{int32_t(floor(bbox.min.x)), int32_t(floor(bbox.min.y))}, {int32_t(ceil(bbox.max.x)), int32_t(ceil(bbox.max.y))}};
 }
 
 
 uint32_t BWTessellator::pushVertex(float x, float y)
 {
     auto index = _pushVertex(mBuffer->vertex, x, y);
-
-    if (index == 0) {
-        bbox.max = bbox.min = {x, y};
-    } else {
-        bbox.min = {std::min(bbox.min.x, x), std::min(bbox.min.y, y)};
-        bbox.max = {std::max(bbox.max.x, x), std::max(bbox.max.y, y)};
-    }
-
+    if (index == 0) bbox.max = bbox.min = {x, y};
+    else bbox = {{std::min(bbox.min.x, x), std::min(bbox.min.y, y)}, {std::max(bbox.max.x, x), std::max(bbox.max.y, y)}};
     return index;
 }
 

src/renderer/gl_engine/tvgGlTessellator.h

@@ -156,7 +156,7 @@ private:
     void pushTriangle(uint32_t a, uint32_t b, uint32_t c);
 
     GlGeometryBuffer* mBuffer;
-    BBox bbox = {{}, {}};
+    BBox bbox = {};
 };
 
 }  // namespace tvg

src/renderer/sw_engine/tvgSwCommon.h

@@ -121,19 +121,20 @@ struct SwSpan
 
 struct SwRle
 {
-    SwSpan *spans;
-    uint32_t alloc;
-    uint32_t size;
-};
+    Array<SwSpan> spans;
 
-struct SwBBox
-{
-    SwPoint min, max;
-
-    void reset()
+    bool invalid() const
     {
-        min.x = min.y = max.x = max.y = 0;
+        return spans.empty();
     }
+
+    bool valid() const
+    {
+        return !invalid();
+    }
+
+    uint32_t size() const { return spans.count; }
+    SwSpan* data() const { return spans.data; }
 };
 
 struct SwFill
@@ -215,7 +216,7 @@ struct SwShape
     SwFill* fill = nullptr;
     SwRle* rle = nullptr;
     SwRle* strokeRle = nullptr;
-    SwBBox       bbox;           //Keep it boundary without stroke region. Using for optimal filling.
+    RenderRegion bbox;           //Keep it boundary without stroke region. Using for optimal filling.
 
     bool fastTrack = false;   //Fast Track: axis-aligned rectangle without any clips?
 };
@@ -279,7 +280,7 @@ struct SwCompositor : RenderCompositor
     SwSurface* recoverSfc;                  //Recover surface when composition is started
     SwCompositor* recoverCmp;               //Recover compositor when composition is done
     SwImage image;
-    SwBBox bbox;
+    RenderRegion bbox;
     bool valid;
 };
 
@@ -499,16 +500,15 @@ SwFixed mathLength(const SwPoint& pt);
 int mathCubicAngle(const SwPoint* base, SwFixed& angleIn, SwFixed& angleMid, SwFixed& angleOut);
 SwFixed mathMean(SwFixed angle1, SwFixed angle2);
 SwPoint mathTransform(const Point* to, const Matrix& transform);
-bool mathUpdateOutlineBBox(const SwOutline* outline, const SwBBox& clipRegion, SwBBox& renderRegion, bool fastTrack);
-bool mathClipBBox(const SwBBox& clipper, SwBBox& clippee);
+bool mathUpdateOutlineBBox(const SwOutline* outline, const RenderRegion& clipBox, RenderRegion& renderBox, bool fastTrack);
 
 void shapeReset(SwShape* shape);
-bool shapePrepare(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid, bool hasComposite);
+bool shapePrepare(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid, bool hasComposite);
 bool shapePrepared(const SwShape* shape);
 bool shapeGenRle(SwShape* shape, const RenderShape* rshape, bool antiAlias);
 void shapeDelOutline(SwShape* shape, SwMpool* mpool, uint32_t tid);
 void shapeResetStroke(SwShape* shape, const RenderShape* rshape, const Matrix& transform);
-bool shapeGenStrokeRle(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid);
+bool shapeGenStrokeRle(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid);
 void shapeFree(SwShape* shape);
 void shapeDelStroke(SwShape* shape);
 bool shapeGenFillColors(SwShape* shape, const Fill* fill, const Matrix& transform, SwSurface* surface, uint8_t opacity, bool ctable);
@@ -523,8 +523,8 @@ bool strokeParseOutline(SwStroke* stroke, const SwOutline& outline);
 SwOutline* strokeExportOutline(SwStroke* stroke, SwMpool* mpool, unsigned tid);
 void strokeFree(SwStroke* stroke);
 
-bool imagePrepare(SwImage* image, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid);
-bool imageGenRle(SwImage* image, const SwBBox& renderRegion, bool antiAlias);
+bool imagePrepare(SwImage* image, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid);
+bool imageGenRle(SwImage* image, const RenderRegion& bbox, bool antiAlias);
 void imageDelOutline(SwImage* image, SwMpool* mpool, uint32_t tid);
 void imageReset(SwImage* image);
 void imageFree(SwImage* image);
@@ -547,13 +547,13 @@ void fillRadial(const SwFill* fill, uint32_t* dst, uint32_t y, uint32_t x, uint3
 void fillRadial(const SwFill* fill, uint32_t* dst, uint32_t y, uint32_t x, uint32_t len, SwBlender op, SwBlender op2, uint8_t a);                          //blending + BlendingMethod(op2) ver.
 void fillRadial(const SwFill* fill, uint32_t* dst, uint32_t y, uint32_t x, uint32_t len, uint8_t* cmp, SwAlpha alpha, uint8_t csize, uint8_t opacity);     //matting ver.
 
-SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegion, bool antiAlias);
-SwRle* rleRender(const SwBBox* bbox);
+SwRle* rleRender(SwRle* rle, const SwOutline* outline, const RenderRegion& bbox, bool antiAlias);
+SwRle* rleRender(const RenderRegion* bbox);
 void rleFree(SwRle* rle);
 void rleReset(SwRle* rle);
 void rleMerge(SwRle* rle, SwRle* clip1, SwRle* clip2);
 bool rleClip(SwRle* rle, const SwRle* clip);
-bool rleClip(SwRle* rle, const SwBBox* clip);
+bool rleClip(SwRle* rle, const RenderRegion* clip);
 
 SwMpool* mpoolInit(uint32_t threads);
 bool mpoolTerm(SwMpool* mpool);
@@ -568,7 +568,11 @@ void mpoolRetDashOutline(SwMpool* mpool, unsigned idx);
 bool rasterCompositor(SwSurface* surface);
 bool rasterGradientShape(SwSurface* surface, SwShape* shape, const Fill* fdata, uint8_t opacity);
 bool rasterShape(SwSurface* surface, SwShape* shape, RenderColor& c);
-bool rasterImage(SwSurface* surface, SwImage* image, const Matrix& transform, const SwBBox& bbox, uint8_t opacity);
+bool rasterTexmapPolygon(SwSurface* surface, const SwImage& image, const Matrix& transform, const RenderRegion& bbox, uint8_t opacity);
+bool rasterScaledImage(SwSurface* surface, const SwImage& image, const Matrix& transform, const RenderRegion& bbox, uint8_t opacity);
+bool rasterDirectImage(SwSurface* surface, const SwImage& image, const RenderRegion& bbox, uint8_t opacity);
+bool rasterScaledRleImage(SwSurface* surface, const SwImage& image, const Matrix& transform, const RenderRegion& bbox, uint8_t opacity);
+bool rasterDirectRleImage(SwSurface* surface, const SwImage& image, uint8_t opacity);
 bool rasterStroke(SwSurface* surface, SwShape* shape, RenderColor& c);
 bool rasterGradientStroke(SwSurface* surface, SwShape* shape, const Fill* fdata, uint8_t opacity);
 bool rasterClear(SwSurface* surface, uint32_t x, uint32_t y, uint32_t w, uint32_t h, pixel_t val = 0);
@@ -576,7 +580,7 @@ void rasterPixel32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len);
 void rasterTranslucentPixel32(uint32_t* dst, uint32_t* src, uint32_t len, uint8_t opacity);
 void rasterPixel32(uint32_t* dst, uint32_t* src, uint32_t len, uint8_t opacity);
 void rasterGrayscale8(uint8_t *dst, uint8_t val, uint32_t offset, int32_t len);
-void rasterXYFlip(uint32_t* src, uint32_t* dst, int32_t stride, int32_t w, int32_t h, const SwBBox& bbox, bool flipped);
+void rasterXYFlip(uint32_t* src, uint32_t* dst, int32_t stride, int32_t w, int32_t h, const RenderRegion& bbox, bool flipped);
 void rasterUnpremultiply(RenderSurface* surface);
 void rasterPremultiply(RenderSurface* surface);
 bool rasterConvertCS(RenderSurface* surface, ColorSpace to);

src/renderer/sw_engine/tvgSwImage.cpp

@@ -72,7 +72,7 @@ static bool _genOutline(SwImage* image, const Matrix& transform, SwMpool* mpool,
 /* External Class Implementation                                        */
 /************************************************************************/
 
-bool imagePrepare(SwImage* image, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid)
+bool imagePrepare(SwImage* image, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid)
 {
     image->direct = _onlyShifted(transform);
 
@@ -91,13 +91,13 @@ bool imagePrepare(SwImage* image, const Matrix& transform, const SwBBox& clipReg
     }
 
     if (!_genOutline(image, transform, mpool, tid)) return false;
-    return mathUpdateOutlineBBox(image->outline, clipRegion, renderRegion, image->direct);
+    return mathUpdateOutlineBBox(image->outline, clipBox, renderBox, image->direct);
 }
 
 
-bool imageGenRle(SwImage* image, const SwBBox& renderRegion, bool antiAlias)
+bool imageGenRle(SwImage* image, const RenderRegion& renderBox, bool antiAlias)
 {
-    if ((image->rle = rleRender(image->rle, image->outline, renderRegion, antiAlias))) return true;
+    if ((image->rle = rleRender(image->rle, image->outline, renderBox, antiAlias))) return true;
 
     return false;
 }

src/renderer/sw_engine/tvgSwMath.cpp

@@ -271,27 +271,12 @@ SwPoint mathTransform(const Point* to, const Matrix& transform)
 }
 
 
-bool mathClipBBox(const SwBBox& clipper, SwBBox& clippee)
-{
-    clippee.max.x = (clippee.max.x < clipper.max.x) ? clippee.max.x : clipper.max.x;
-    clippee.max.y = (clippee.max.y < clipper.max.y) ? clippee.max.y : clipper.max.y;
-    clippee.min.x = (clippee.min.x > clipper.min.x) ? clippee.min.x : clipper.min.x;
-    clippee.min.y = (clippee.min.y > clipper.min.y) ? clippee.min.y : clipper.min.y;
-
-    //Check boundary
-    if (clippee.min.x >= clipper.max.x || clippee.min.y >= clipper.max.y ||
-        clippee.max.x <= clipper.min.x || clippee.max.y <= clipper.min.y) return false;
-
-    return true;
-}
-
-
-bool mathUpdateOutlineBBox(const SwOutline* outline, const SwBBox& clipRegion, SwBBox& renderRegion, bool fastTrack)
+bool mathUpdateOutlineBBox(const SwOutline* outline, const RenderRegion& clipBox, RenderRegion& renderBox, bool fastTrack)
 {
     if (!outline) return false;
 
     if (outline->pts.empty() || outline->cntrs.empty()) {
-        renderRegion.reset();
+        renderBox.reset();
         return false;
     }
 
@@ -310,16 +295,13 @@ bool mathUpdateOutlineBBox(const SwOutline* outline, const SwBBox& clipRegion, S
     }
 
     if (fastTrack) {
-        renderRegion.min.x = static_cast<SwCoord>(round(xMin / 64.0f));
-        renderRegion.max.x = static_cast<SwCoord>(round(xMax / 64.0f));
-        renderRegion.min.y = static_cast<SwCoord>(round(yMin / 64.0f));
-        renderRegion.max.y = static_cast<SwCoord>(round(yMax / 64.0f));
+        renderBox.min = {int32_t(round(xMin / 64.0f)), int32_t(round(yMin / 64.0f))};
+        renderBox.max = {int32_t(round(xMax / 64.0f)), int32_t(round(yMax / 64.0f))};
     } else {
-        renderRegion.min.x = xMin >> 6;
-        renderRegion.max.x = (xMax + 63) >> 6;
-        renderRegion.min.y = yMin >> 6;
-        renderRegion.max.y = (yMax + 63) >> 6;
+        renderBox.min = {xMin >> 6, yMin >> 6};
+        renderBox.max = {(xMax + 63) >> 6, (yMax + 63) >> 6};
     }
 
-    return mathClipBBox(clipRegion, renderRegion);
+    renderBox.intersect(clipBox);
+    return renderBox.valid();
 }

src/renderer/sw_engine/tvgSwPostEffect.cpp

@@ -61,7 +61,7 @@ static inline int _gaussianRemap(int end, int idx)
 
 //TODO: SIMD OPTIMIZATION?
 template<int border = 0>
-static void _gaussianFilter(uint8_t* dst, uint8_t* src, int32_t stride, int32_t w, int32_t h, const SwBBox& bbox, int32_t dimension, bool flipped)
+static void _gaussianFilter(uint8_t* dst, uint8_t* src, int32_t stride, int32_t w, int32_t h, const RenderRegion& bbox, int32_t dimension, bool flipped)
 {
     if (flipped) {
         src += (bbox.min.x * stride + bbox.min.y) << 2;
@@ -136,17 +136,17 @@ static int _gaussianInit(SwGaussianBlur* data, float sigma, int quality)
 
 bool effectGaussianBlurRegion(RenderEffectGaussianBlur* params)
 {
-    //bbox region expansion for feathering
-    auto& region = params->extend;
+    //region expansion for feathering
+    auto& bbox = params->extend;
     auto extra = static_cast<SwGaussianBlur*>(params->rd)->extends;
 
     if (params->direction != 2) {
-        region.x = -extra;
-        region.w = extra * 2;
+        bbox.min.x = -extra;
+        bbox.max.x = extra;
     }
     if (params->direction != 1) {
-        region.y = -extra;
-        region.h = extra * 2;
+        bbox.min.y = -extra;
+        bbox.max.y = extra;
     }
 
     return true;
@@ -230,7 +230,7 @@ struct SwDropShadow : SwGaussianBlur
 
 
 //TODO: SIMD OPTIMIZATION?
-static void _dropShadowFilter(uint32_t* dst, uint32_t* src, int stride, int w, int h, const SwBBox& bbox, int32_t dimension, uint32_t color, bool flipped)
+static void _dropShadowFilter(uint32_t* dst, uint32_t* src, int stride, int w, int h, const RenderRegion& bbox, int32_t dimension, uint32_t color, bool flipped)
 {
     if (flipped) {
         src += (bbox.min.x * stride + bbox.min.y);
@@ -267,20 +267,20 @@ static void _dropShadowFilter(uint32_t* dst, uint32_t* src, int stride, int w, i
 }
 
 
-static void _dropShadowShift(uint32_t* dst, uint32_t* src, int dstride, int sstride, SwBBox& region, SwPoint& offset, uint8_t opacity, bool direct)
+static void _dropShadowShift(uint32_t* dst, uint32_t* src, int dstride, int sstride, RenderRegion& bbox, SwPoint& offset, uint8_t opacity, bool direct)
 {
-    src += (region.min.y * sstride + region.min.x);
-    dst += (region.min.y * dstride + region.min.x);
+    src += (bbox.min.y * sstride + bbox.min.x);
+    dst += (bbox.min.y * dstride + bbox.min.x);
 
-    auto w = region.max.x - region.min.x;
-    auto h = region.max.y - region.min.y;
+    auto w = bbox.max.x - bbox.min.x;
+    auto h = bbox.max.y - bbox.min.y;
     auto translucent = (direct || opacity < 255);
 
     //shift offset
-    if (region.min.x + offset.x < 0) src -= offset.x;
+    if (bbox.min.x + offset.x < 0) src -= offset.x;
     else dst += offset.x;
 
-    if (region.min.y + offset.y < 0) src -= (offset.y * sstride);
+    if (bbox.min.y + offset.y < 0) src -= (offset.y * sstride);
     else dst += (offset.y * dstride);
 
     for (auto y = 0; y < h; ++y) {
@@ -294,20 +294,19 @@ static void _dropShadowShift(uint32_t* dst, uint32_t* src, int dstride, int sstr
 
 bool effectDropShadowRegion(RenderEffectDropShadow* params)
 {
-    //bbox region expansion for feathering
-    auto& region = params->extend;
+    //region expansion for feathering
+    auto& bbox = params->extend;
     auto& offset = static_cast<SwDropShadow*>(params->rd)->offset;
     auto extra = static_cast<SwDropShadow*>(params->rd)->extends;
 
-    region.x = -extra;
-    region.w = extra * 2;
-    region.y = -extra;
-    region.h = extra * 2;
+    bbox.min = {-extra, -extra};
+    bbox.max = {extra, extra};
 
-    region.x = std::min(region.x + (int32_t)offset.x, region.x);
-    region.y = std::min(region.y + (int32_t)offset.y, region.y);
-    region.w += abs(offset.x);
-    region.h += abs(offset.y);
+    if (offset.x < 0) bbox.min.x += (int32_t) offset.x;
+    else bbox.max.x += offset.x;
+
+    if (offset.y < 0) bbox.min.y += (int32_t) offset.y;
+    else bbox.max.y += offset.y;
 
     return true;
 }

src/renderer/sw_engine/tvgSwRasterAvx.h

@@ -99,15 +99,15 @@ static void avxRasterPixel32(uint32_t *dst, uint32_t val, uint32_t offset, int32
 }
 
 
-static bool avxRasterTranslucentRect(SwSurface* surface, const SwBBox& region, const RenderColor& c)
+static bool avxRasterTranslucentRect(SwSurface* surface, const RenderRegion& bbox, const RenderColor& c)
 {
-    auto h = static_cast<uint32_t>(region.max.y - region.min.y);
-    auto w = static_cast<uint32_t>(region.max.x - region.min.x);
+    auto h = bbox.h();
+    auto w = bbox.w();
 
     //32bits channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
-        auto buffer = surface->buf32 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf32 + (bbox.min.y * surface->stride) + bbox.min.x;
 
         uint32_t ialpha = 255 - c.a;
 
@@ -145,7 +145,7 @@ static bool avxRasterTranslucentRect(SwSurface* surface, const SwBBox& region, c
     //8bit grayscale
     } else if (surface->channelSize == sizeof(uint8_t)) {
         TVGLOG("SW_ENGINE", "Require AVX Optimization, Channel Size = %d", surface->channelSize);
-        auto buffer = surface->buf8 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf8 + (bbox.min.y * surface->stride) + bbox.min.x;
         auto ialpha = ~c.a;
         for (uint32_t y = 0; y < h; ++y) {
             auto dst = &buffer[y * surface->stride];
@@ -160,14 +160,12 @@ static bool avxRasterTranslucentRect(SwSurface* surface, const SwBBox& region, c
 
 static bool avxRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const RenderColor& c)
 {
-    auto span = rle->spans;
-
     //32bit channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
         uint32_t src;
 
-        for (uint32_t i = 0; i < rle->size; ++i) {
+        ARRAY_FOREACH(span, rle->spans) {
             auto dst = &surface->buf32[span->y * surface->stride + span->x];
 
             if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
@@ -213,7 +211,7 @@ static bool avxRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const
     } else if (surface->channelSize == sizeof(uint8_t)) {
         TVGLOG("SW_ENGINE", "Require AVX Optimization, Channel Size = %d", surface->channelSize);
         uint8_t src;
-        for (uint32_t i = 0; i < rle->size; ++i, ++span) {
+        ARRAY_FOREACH(span, rle->spans) {
             auto dst = &surface->buf8[span->y * surface->stride + span->x];
             if (span->coverage < 255) src = MULTIPLY(span->coverage, c.a);
             else src = c.a;

src/renderer/sw_engine/tvgSwRasterC.h

@@ -94,13 +94,11 @@ static void inline cRasterPixels(PIXEL_T* dst, PIXEL_T val, uint32_t offset, int
 
 static bool inline cRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const RenderColor& c)
 {
-    auto span = rle->spans;
-
     //32bit channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
         uint32_t src;
-        for (uint32_t i = 0; i < rle->size; ++i, ++span) {
+        ARRAY_FOREACH(span, rle->spans) {
             auto dst = &surface->buf32[span->y * surface->stride + span->x];
             if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
             else src = color;
@@ -112,7 +110,7 @@ static bool inline cRasterTranslucentRle(SwSurface* surface, const SwRle* rle, c
     //8bit grayscale
     } else if (surface->channelSize == sizeof(uint8_t)) {
         uint8_t src;
-        for (uint32_t i = 0; i < rle->size; ++i, ++span) {
+        ARRAY_FOREACH(span, rle->spans) {
             auto dst = &surface->buf8[span->y * surface->stride + span->x];
             if (span->coverage < 255) src = MULTIPLY(span->coverage, c.a);
             else src = c.a;
@@ -126,29 +124,26 @@ static bool inline cRasterTranslucentRle(SwSurface* surface, const SwRle* rle, c
 }
 
 
-static bool inline cRasterTranslucentRect(SwSurface* surface, const SwBBox& region, const RenderColor& c)
+static bool inline cRasterTranslucentRect(SwSurface* surface, const RenderRegion& bbox, const RenderColor& c)
 {
-    auto h = static_cast<uint32_t>(region.max.y - region.min.y);
-    auto w = static_cast<uint32_t>(region.max.x - region.min.x);
-
     //32bits channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
-        auto buffer = surface->buf32 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf32 + (bbox.min.y * surface->stride) + bbox.min.x;
         auto ialpha = 255 - c.a;
-        for (uint32_t y = 0; y < h; ++y) {
+        for (uint32_t y = 0; y < bbox.h(); ++y) {
             auto dst = &buffer[y * surface->stride];
-            for (uint32_t x = 0; x < w; ++x, ++dst) {
+            for (uint32_t x = 0; x < bbox.w(); ++x, ++dst) {
                 *dst = color + ALPHA_BLEND(*dst, ialpha);
             }
         }
     //8bit grayscale
     } else if (surface->channelSize == sizeof(uint8_t)) {
-        auto buffer = surface->buf8 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf8 + (bbox.min.y * surface->stride) + bbox.min.x;
         auto ialpha = ~c.a;
-        for (uint32_t y = 0; y < h; ++y) {
+        for (uint32_t y = 0; y < bbox.h(); ++y) {
             auto dst = &buffer[y * surface->stride];
-            for (uint32_t x = 0; x < w; ++x, ++dst) {
+            for (uint32_t x = 0; x < bbox.w(); ++x, ++dst) {
                 *dst = c.a + MULTIPLY(*dst, ialpha);
             }
         }

src/renderer/sw_engine/tvgSwRasterNeon.h

@@ -91,8 +91,6 @@ static void neonRasterPixel32(uint32_t *dst, uint32_t val, uint32_t offset, int3
 
 static bool neonRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const RenderColor& c)
 {
-    auto span = rle->spans;
-
     //32bit channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
@@ -100,7 +98,7 @@ static bool neonRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const
         uint8x8_t *vDst = nullptr;
         uint16_t align;
 
-        for (uint32_t i = 0; i < rle->size; ++i) {
+        ARRAY_FOREACH(span, rle->spans) {
             if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
             else src = color;
 
@@ -132,7 +130,7 @@ static bool neonRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const
     } else if (surface->channelSize == sizeof(uint8_t)) {
         TVGLOG("SW_ENGINE", "Require Neon Optimization, Channel Size = %d", surface->channelSize);
         uint8_t src;
-        for (uint32_t i = 0; i < rle->size; ++i, ++span) {
+        ARRAY_FOREACH(span, rle->spans) {
             auto dst = &surface->buf8[span->y * surface->stride + span->x];
             if (span->coverage < 255) src = MULTIPLY(span->coverage, c.a);
             else src = c.a;
@@ -146,15 +144,15 @@ static bool neonRasterTranslucentRle(SwSurface* surface, const SwRle* rle, const
 }
 
 
-static bool neonRasterTranslucentRect(SwSurface* surface, const SwBBox& region, const RenderColor& c)
+static bool neonRasterTranslucentRect(SwSurface* surface, const RenderRegion& bbox, const RenderColor& c)
 {
-    auto h = static_cast<uint32_t>(region.max.y - region.min.y);
-    auto w = static_cast<uint32_t>(region.max.x - region.min.x);
+    auto h = bbox.h();
+    auto w = bbox.w();
 
     //32bits channels
     if (surface->channelSize == sizeof(uint32_t)) {
         auto color = surface->join(c.r, c.g, c.b, c.a);
-        auto buffer = surface->buf32 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf32 + (bbox.min.y * surface->stride) + bbox.min.x;
         auto ialpha = 255 - c.a;
 
         auto vColor = vdup_n_u32(color);
@@ -185,7 +183,7 @@ static bool neonRasterTranslucentRect(SwSurface* surface, const SwBBox& region,
     //8bit grayscale
     } else if (surface->channelSize == sizeof(uint8_t)) {
         TVGLOG("SW_ENGINE", "Require Neon Optimization, Channel Size = %d", surface->channelSize);
-        auto buffer = surface->buf8 + (region.min.y * surface->stride) + region.min.x;
+        auto buffer = surface->buf8 + (bbox.min.y * surface->stride) + bbox.min.x;
         auto ialpha = ~c.a;
         for (uint32_t y = 0; y < h; ++y) {
             auto dst = &buffer[y * surface->stride];

src/renderer/sw_engine/tvgSwRasterTexmap.h

@@ -51,100 +51,57 @@ static float dxdya, dxdyb, dudya, dvdya;
 static float xa, xb, ua, va;
 
 
-//Y Range exception handling
-static bool _arrange(const SwImage* image, const SwBBox* region, int& yStart, int& yEnd)
+static inline int32_t _modf(float v)
 {
-    int32_t regionTop, regionBottom;
-
-    if (region) {
-        regionTop = region->min.y;
-        regionBottom = region->max.y;
-    } else {
-        regionTop = image->rle->spans->y;
-        regionBottom = image->rle->spans[image->rle->size - 1].y;
-    }
-
-    if (yStart < regionTop) yStart = regionTop;
-    if (yEnd > regionBottom) yEnd = regionBottom;
-
-    return yEnd > yStart;
+    return 255 - ((int(v * 256.0f)) & 255);
 }
 
 
-static bool _rasterMaskedPolygonImageSegment(SwSurface* surface, const SwImage* image, const SwBBox* region, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity, uint8_t dirFlag = 0)
+static bool _rasterMaskedPolygonImageSegment(SwSurface* surface, const SwImage& image, const RenderRegion& bbox, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity, uint8_t dirFlag = 0)
 {
     TVGERR("SW_ENGINE", "TODO: _rasterMaskedPolygonImageSegment()");
     return false;
 }
 
 
-static void _rasterBlendingPolygonImageSegment(SwSurface* surface, const SwImage* image, const SwBBox* region, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity)
+static void _rasterBlendingPolygonImageSegment(SwSurface* surface, const SwImage& image, const RenderRegion& bbox, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity)
 {
     float _dudx = dudx, _dvdx = dvdx;
     float _dxdya = dxdya, _dxdyb = dxdyb, _dudya = dudya, _dvdya = dvdya;
     float _xa = xa, _xb = xb, _ua = ua, _va = va;
-    auto sbuf = image->buf32;
+    auto sbuf = image.buf32;
     auto dbuf = surface->buf32;
-    int32_t sw = static_cast<int32_t>(image->w);
-    int32_t sh = static_cast<int32_t>(image->h);
+    int32_t sw = static_cast<int32_t>(image.w);
+    int32_t sh = static_cast<int32_t>(image.h);
     int32_t x1, x2, x, y, ar, ab, iru, irv, px, ay;
     int32_t vv = 0, uu = 0;
-    int32_t minx = INT32_MAX, maxx = 0;
-    float dx, u, v, iptr;
+    float dx, u, v;
     uint32_t* buf;
-    SwSpan* span = nullptr;         //used only when rle based.
 
-    if (!_arrange(image, region, yStart, yEnd)) return;
-
-    //Loop through all lines in the segment
-    uint32_t spanIdx = 0;
-
-    if (region) {
-        minx = region->min.x;
-        maxx = region->max.x;
-    } else {
-        span = image->rle->spans;
-        while (span->y < yStart) {
-            ++span;
-            ++spanIdx;
-        }
-    }
+    if (yStart < bbox.min.y) yStart = bbox.min.y;
+    if (yEnd > bbox.max.y) yEnd = bbox.max.y;
 
     y = yStart;
 
     while (y < yEnd) {
-        x1 = (int32_t)_xa;
-        x2 = (int32_t)_xb;
-
-        if (!region) {
-            minx = INT32_MAX;
-            maxx = 0;
-            //one single row, could be consisted of multiple spans.
-            while (span->y == y && spanIdx < image->rle->size) {
-                if (minx > span->x) minx = span->x;
-                if (maxx < span->x + span->len) maxx = span->x + span->len;
-                ++span;
-                ++spanIdx;
-            }
-        }
-        if (x1 < minx) x1 = minx;
-        if (x2 > maxx) x2 = maxx;
+        x1 = std::max((int32_t)_xa, bbox.min.x);
+        x2 = std::min((int32_t)_xb, bbox.max.x);
 
         //Anti-Aliasing frames
+        if (aaSpans) {
             ay = y - aaSpans->yStart;
             if (aaSpans->lines[ay].x[0] > x1) aaSpans->lines[ay].x[0] = x1;
             if (aaSpans->lines[ay].x[1] < x2) aaSpans->lines[ay].x[1] = x2;
+        }
 
         //Range allowed
-        if ((x2 - x1) >= 1 && (x1 < maxx) && (x2 > minx)) {
+        if ((x2 - x1) >= 1 && (x1 < bbox.max.x) && (x2 > bbox.min.x)) {
 
             //Perform subtexel pre-stepping on UV
             dx = 1 - (_xa - x1);
             u = _ua + dx * _dudx;
             v = _va + dx * _dvdx;
-
             buf = dbuf + ((y * surface->stride) + x1);
-
             x = x1;
 
             //Draw horizontal line
@@ -152,29 +109,29 @@ static void _rasterBlendingPolygonImageSegment(SwSurface* surface, const SwImage
                 uu = (int) u;
                 vv = (int) v;
 
-                if ((uint32_t) uu >= image->w || (uint32_t) vv >= image->h) continue;
+                if ((uint32_t) uu >= image.w || (uint32_t) vv >= image.h) continue;
 
-                ar = (int)(255 * (1 - modff(u, &iptr)));
-                ab = (int)(255 * (1 - modff(v, &iptr)));
+                ar = _modf(u);
+                ab = _modf(v);
                 iru = uu + 1;
                 irv = vv + 1;
 
-                px = *(sbuf + (vv * image->stride) + uu);
+                px = *(sbuf + (vv * image.stride) + uu);
 
                 /* horizontal interpolate */
                 if (iru < sw) {
                     /* right pixel */
-                    int px2 = *(sbuf + (vv * image->stride) + iru);
+                    int px2 = *(sbuf + (vv * image.stride) + iru);
                     px = INTERPOLATE(px, px2, ar);
                 }
                 /* vertical interpolate */
                 if (irv < sh) {
                     /* bottom pixel */
-                    int px2 = *(sbuf + (irv * image->stride) + uu);
+                    int px2 = *(sbuf + (irv * image.stride) + uu);
                     /* horizontal interpolate */
                     if (iru < sw) {
                         /* bottom right pixel */
-                        int px3 = *(sbuf + (irv * image->stride) + iru);
+                        int px3 = *(sbuf + (irv * image.stride) + iru);
                         px2 = INTERPOLATE(px2, px3, ar);
                     }
                     px = INTERPOLATE(px, px2, ab);
@@ -195,8 +152,6 @@ static void _rasterBlendingPolygonImageSegment(SwSurface* surface, const SwImage
         _ua += _dudya;
         _va += _dvdya;
 
-        if (!region && spanIdx >= image->rle->size) break;
-
         ++y;
     }
     xa = _xa;
@@ -206,122 +161,95 @@ static void _rasterBlendingPolygonImageSegment(SwSurface* surface, const SwImage
 }
 
 
-static void _rasterPolygonImageSegment(SwSurface* surface, const SwImage* image, const SwBBox* region, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity, bool matting)
+static void _rasterPolygonImageSegment(SwSurface* surface, const SwImage& image, const RenderRegion& bbox, int yStart, int yEnd, AASpans* aaSpans, uint8_t opacity, bool matting)
 {
     float _dudx = dudx, _dvdx = dvdx;
     float _dxdya = dxdya, _dxdyb = dxdyb, _dudya = dudya, _dvdya = dvdya;
     float _xa = xa, _xb = xb, _ua = ua, _va = va;
-    auto sbuf = image->buf32;
+    auto sbuf = image.buf32;
     auto dbuf = surface->buf32;
-    int32_t sw = static_cast<int32_t>(image->w);
-    int32_t sh = static_cast<int32_t>(image->h);
+    int32_t sw = static_cast<int32_t>(image.w);
+    int32_t sh = static_cast<int32_t>(image.h);
     int32_t x1, x2, x, y, ar, ab, iru, irv, px, ay;
     int32_t vv = 0, uu = 0;
-    int32_t minx = INT32_MAX, maxx = 0;
-    float dx, u, v, iptr;
+    float dx, u, v;
     uint32_t* buf;
-    SwSpan* span = nullptr;         //used only when rle based.
 
     //for matting(composition)
     auto csize = matting ? surface->compositor->image.channelSize: 0;
     auto alpha = matting ? surface->alpha(surface->compositor->method) : nullptr;
     uint8_t* cmp = nullptr;
 
-    if (!_arrange(image, region, yStart, yEnd)) return;
-
-    //Loop through all lines in the segment
-    uint32_t spanIdx = 0;
-
-    if (region) {
-        minx = region->min.x;
-        maxx = region->max.x;
-    } else {
-        span = image->rle->spans;
-        while (span->y < yStart) {
-            ++span;
-            ++spanIdx;
-        }
-    }
+    if (yStart < bbox.min.y) yStart = bbox.min.y;
+    if (yEnd > bbox.max.y) yEnd = bbox.max.y;
 
     y = yStart;
 
     while (y < yEnd) {
-        x1 = (int32_t)_xa;
-        x2 = (int32_t)_xb;
-
-        if (!region) {
-            minx = INT32_MAX;
-            maxx = 0;
-            //one single row, could be consisted of multiple spans.
-            while (span->y == y && spanIdx < image->rle->size) {
-                if (minx > span->x) minx = span->x;
-                if (maxx < span->x + span->len) maxx = span->x + span->len;
-                ++span;
-                ++spanIdx;
-            }
-        }
-        if (x1 < minx) x1 = minx;
-        if (x2 > maxx) x2 = maxx;
+        x1 = std::max((int32_t)_xa, bbox.min.x);
+        x2 = std::min((int32_t)_xb, bbox.max.x);
 
         //Anti-Aliasing frames
+        if (aaSpans) {
             ay = y - aaSpans->yStart;
             if (aaSpans->lines[ay].x[0] > x1) aaSpans->lines[ay].x[0] = x1;
             if (aaSpans->lines[ay].x[1] < x2) aaSpans->lines[ay].x[1] = x2;
+        }
 
         //Range allowed
-        if ((x2 - x1) >= 1 && (x1 < maxx) && (x2 > minx)) {
+        if ((x2 - x1) >= 1 && (x1 < bbox.max.x) && (x2 > bbox.min.x)) {
 
             //Perform subtexel pre-stepping on UV
             dx = 1 - (_xa - x1);
             u = _ua + dx * _dudx;
             v = _va + dx * _dvdx;
-
             buf = dbuf + ((y * surface->stride) + x1);
-
             x = x1;
 
             if (matting) cmp = &surface->compositor->image.buf8[(y * surface->compositor->image.stride + x1) * csize];
 
-            if (opacity == 255) {
+            const auto fullOpacity = (opacity == 255);
+
             //Draw horizontal line
             while (x++ < x2) {
                 uu = (int) u;
                 vv = (int) v;
 
-                    if ((uint32_t) uu >= image->w || (uint32_t) vv >= image->h) continue;
+                if ((uint32_t) uu >= image.w || (uint32_t) vv >= image.h) continue;
 
-                    ar = (int)(255.0f * (1.0f - modff(u, &iptr)));
-                    ab = (int)(255.0f * (1.0f - modff(v, &iptr)));
+                ar = _modf(u);
+                ab = _modf(v);
                 iru = uu + 1;
                 irv = vv + 1;
 
-                    px = *(sbuf + (vv * image->stride) + uu);
+                px = *(sbuf + (vv * image.stride) + uu);
 
                 /* horizontal interpolate */
                 if (iru < sw) {
                     /* right pixel */
-                        int px2 = *(sbuf + (vv * image->stride) + iru);
+                    int px2 = *(sbuf + (vv * image.stride) + iru);
                     px = INTERPOLATE(px, px2, ar);
                 }
                 /* vertical interpolate */
                 if (irv < sh) {
                     /* bottom pixel */
-                        int px2 = *(sbuf + (irv * image->stride) + uu);
+                    int px2 = *(sbuf + (irv * image.stride) + uu);
 
                     /* horizontal interpolate */
                     if (iru < sw) {
                         /* bottom right pixel */
-                            int px3 = *(sbuf + (irv * image->stride) + iru);
+                        int px3 = *(sbuf + (irv * image.stride) + iru);
                         px2 = INTERPOLATE(px2, px3, ar);
                     }
                     px = INTERPOLATE(px, px2, ab);
                 }
                 uint32_t src;
                 if (matting) {
-                        src = ALPHA_BLEND(px, alpha(cmp));
+                    auto a = alpha(cmp);
+                    src = fullOpacity ? ALPHA_BLEND(px, a) : ALPHA_BLEND(px, MULTIPLY(opacity, a));
                     cmp += csize;
                 } else {
-                        src = px;
+                    src = fullOpacity ? px : ALPHA_BLEND(px, opacity);
                 }
                 *buf = src + ALPHA_BLEND(*buf, IA(src));
                 ++buf;
@@ -330,55 +258,6 @@ static void _rasterPolygonImageSegment(SwSurface* surface, const SwImage* image,
                 u += _dudx;
                 v += _dvdx;
             }
-            } else {
-                //Draw horizontal line
-                while (x++ < x2) {
-                    uu = (int) u;
-                    vv = (int) v;
-
-                    if ((uint32_t) uu >= image->w || (uint32_t) vv >= image->h) continue;
-
-                    ar = (int)(255.0f * (1.0f - modff(u, &iptr)));
-                    ab = (int)(255.0f * (1.0f - modff(v, &iptr)));
-                    iru = uu + 1;
-                    irv = vv + 1;
-
-                    px = *(sbuf + (vv * sw) + uu);
-
-                    /* horizontal interpolate */
-                    if (iru < sw) {
-                        /* right pixel */
-                        int px2 = *(sbuf + (vv * image->stride) + iru);
-                        px = INTERPOLATE(px, px2, ar);
-                    }
-                    /* vertical interpolate */
-                    if (irv < sh) {
-                        /* bottom pixel */
-                        int px2 = *(sbuf + (irv * image->stride) + uu);
-
-                        /* horizontal interpolate */
-                        if (iru < sw) {
-                            /* bottom right pixel */
-                            int px3 = *(sbuf + (irv * image->stride) + iru);
-                            px2 = INTERPOLATE(px2, px3, ar);
-                        }
-                        px = INTERPOLATE(px, px2, ab);
-                    }
-                    uint32_t src;
-                    if (matting) {
-                        src = ALPHA_BLEND(px, MULTIPLY(opacity, alpha(cmp)));
-                        cmp += csize;
-                    } else {
-                        src = ALPHA_BLEND(px, opacity);
-                    }
-                    *buf = src + ALPHA_BLEND(*buf, IA(src));
-                    ++buf;
-
-                    //Step UV horizontally
-                    u += _dudx;
-                    v += _dvdx;
-                }
-            }
         }
 
         //Step along both edges
@@ -387,8 +266,6 @@ static void _rasterPolygonImageSegment(SwSurface* surface, const SwImage* image,
         _ua += _dudya;
         _va += _dvdya;
 
-        if (!region && spanIdx >= image->rle->size) break;
-
         ++y;
     }
     xa = _xa;
@@ -399,7 +276,7 @@ static void _rasterPolygonImageSegment(SwSurface* surface, const SwImage* image,
 
 
 /* This mapping algorithm is based on Mikael Kalms's. */
-static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const SwBBox* region, Polygon& polygon, AASpans* aaSpans, uint8_t opacity)
+static void _rasterPolygonImage(SwSurface* surface, const SwImage& image, const RenderRegion& bbox, Polygon& polygon, AASpans* aaSpans, uint8_t opacity)
 {
     float x[3] = {polygon.vertex[0].pt.x, polygon.vertex[1].pt.x, polygon.vertex[2].pt.x};
     float y[3] = {polygon.vertex[0].pt.y, polygon.vertex[1].pt.y, polygon.vertex[2].pt.y};
@@ -460,7 +337,6 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
     if (tvg::equal(y[0], y[1])) side = x[0] > x[1];
     if (tvg::equal(y[1], y[2])) side = x[2] > x[1];
 
-    auto regionTop = region ? region->min.y : image->rle->spans->y;  //Normal Image or Rle Image?
     auto compositing = _compositing(surface);   //Composition required
     auto blending = _blending(surface);         //Blending required
 
@@ -479,7 +355,7 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
 
         //Draw upper segment if possibly visible
         if (yi[0] < yi[1]) {
-            off_y = y[0] < regionTop ? (regionTop - y[0]) : 0;
+            off_y = y[0] < bbox.min.y ? (bbox.min.y - y[0]) : 0;
             xa += (off_y * dxdya);
             ua += (off_y * dudya);
             va += (off_y * dvdya);
@@ -489,18 +365,18 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
             xb = x[0] + dy * dxdyb + (off_y * dxdyb);
 
             if (compositing) {
-                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, true);
-                else _rasterMaskedPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, 1);
+                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, true);
+                else _rasterMaskedPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, 1);
             } else if (blending) {
-                _rasterBlendingPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity);
+                _rasterBlendingPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity);
             } else {
-                _rasterPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, false);
+                _rasterPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, false);
             }
             upper = true;
         }
         //Draw lower segment if possibly visible
         if (yi[1] < yi[2]) {
-            off_y = y[1] < regionTop ? (regionTop - y[1]) : 0;
+            off_y = y[1] < bbox.min.y ? (bbox.min.y - y[1]) : 0;
             if (!upper) {
                 xa += (off_y * dxdya);
                 ua += (off_y * dudya);
@@ -510,12 +386,12 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
             dxdyb = dxdy[2];
             xb = x[1] + (1 - (y[1] - yi[1])) * dxdyb + (off_y * dxdyb);
             if (compositing) {
-                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, true);
-                else _rasterMaskedPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, 2);
+                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, true);
+                else _rasterMaskedPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, 2);
             } else if (blending) {
-                 _rasterBlendingPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity);
+                 _rasterBlendingPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity);
             } else {
-                _rasterPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, false);
+                _rasterPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, false);
             }
         }
     //Longer edge is on the right side
@@ -527,7 +403,7 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
 
         //Draw upper segment if possibly visible
         if (yi[0] < yi[1]) {
-            off_y = y[0] < regionTop ? (regionTop - y[0]) : 0;
+            off_y = y[0] < bbox.min.y ? (bbox.min.y - y[0]) : 0;
             xb += (off_y *dxdyb);
 
             // Set slopes along left edge and perform subpixel pre-stepping
@@ -540,18 +416,18 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
             va = v[0] + dy * dvdya + (off_y * dvdya);
 
             if (compositing) {
-                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, true);
-                else _rasterMaskedPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, 3);
+                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, true);
+                else _rasterMaskedPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, 3);
             } else if (blending) {
-                _rasterBlendingPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity);
+                _rasterBlendingPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity);
             } else {
-                _rasterPolygonImageSegment(surface, image, region, yi[0], yi[1], aaSpans, opacity, false);
+                _rasterPolygonImageSegment(surface, image, bbox, yi[0], yi[1], aaSpans, opacity, false);
             }
             upper = true;
         }
         //Draw lower segment if possibly visible
         if (yi[1] < yi[2]) {
-            off_y = y[1] < regionTop ? (regionTop - y[1]) : 0;
+            off_y = y[1] < bbox.min.y ? (bbox.min.y - y[1]) : 0;
             if (!upper) xb += (off_y *dxdyb);
 
             // Set slopes along left edge and perform subpixel pre-stepping
@@ -564,25 +440,20 @@ static void _rasterPolygonImage(SwSurface* surface, const SwImage* image, const
             va = v[1] + dy * dvdya + (off_y * dvdya);
 
             if (compositing) {
-                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, true);
-                else _rasterMaskedPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, 4);
+                if (_matting(surface)) _rasterPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, true);
+                else _rasterMaskedPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, 4);
             } else if (blending) {
-                _rasterBlendingPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity);
+                _rasterBlendingPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity);
             } else {
-                _rasterPolygonImageSegment(surface, image, region, yi[1], yi[2], aaSpans, opacity, false);
+                _rasterPolygonImageSegment(surface, image, bbox, yi[1], yi[2], aaSpans, opacity, false);
             }
         }
     }
 }
 
 
-static AASpans* _AASpans(float ymin, float ymax, const SwImage* image, const SwBBox* region)
+static AASpans* _AASpans(int yStart, int yEnd)
 {
-    auto yStart = static_cast<int>(ymin);
-    auto yEnd = static_cast<int>(ymax);
-
-    if (!_arrange(image, region, yStart, yEnd)) return nullptr;
-
     auto aaSpans = tvg::malloc<AASpans*>(sizeof(AASpans));
     aaSpans->yStart = yStart;
     aaSpans->yEnd = yEnd;
@@ -661,14 +532,9 @@ static void _calcAAEdge(AASpans *aaSpans, int32_t eidx)
         ptx[1] = tx[1]; \
     } while (0)
 
-    struct Point
-    {
-        int32_t x, y;
-    };
-
     int32_t y = 0;
-    Point pEdge = {-1, -1};       //previous edge point
-    Point edgeDiff = {0, 0};      //temporary used for point distance
+    SwPoint pEdge = {-1, -1};       //previous edge point
+    SwPoint edgeDiff = {0, 0};      //temporary used for point distance
 
     /* store bigger to tx[0] between prev and current edge's x positions. */
     int32_t tx[2] = {0, 0};
@@ -791,29 +657,24 @@ static void _calcAAEdge(AASpans *aaSpans, int32_t eidx)
 }
 
 
-static bool _apply(SwSurface* surface, AASpans* aaSpans)
+static void _apply(SwSurface* surface, AASpans* aaSpans)
 {
     auto end = surface->buf32 + surface->h * surface->stride;
+    auto buf = surface->buf32 + surface->stride * aaSpans->yStart;
     auto y = aaSpans->yStart;
-    uint32_t pixel;
+    auto line = aaSpans->lines;
+    uint32_t pix;
     uint32_t* dst;
     int32_t pos;
 
-   //left side
-   _calcAAEdge(aaSpans, 0);
-   //right side
-   _calcAAEdge(aaSpans, 1);
+   _calcAAEdge(aaSpans, 0);    //left side
+   _calcAAEdge(aaSpans, 1);    //right side
 
     while (y < aaSpans->yEnd) {
-        auto line = &aaSpans->lines[y - aaSpans->yStart];
-        auto width = line->x[1] - line->x[0];
-        if (width > 0) {
-            auto offset = y * surface->stride;
-
+        if (line->x[1] - line->x[0] > 0) {
             //Left edge
-            dst = surface->buf32 + (offset + line->x[0]);
-            if (line->x[0] > 1) pixel = *(dst - 1);
-            else pixel = *dst;
+            dst = buf + line->x[0];
+            pix = *(dst - ((line->x[0] > 1) ? 1 : 0));
             pos = 1;
 
             //exceptional handling. out of memory bound.
@@ -822,34 +683,31 @@ static bool _apply(SwSurface* surface, AASpans* aaSpans)
             }
 
             while (pos <= line->length[0]) {
-                *dst = INTERPOLATE(*dst, pixel, line->coverage[0] * pos);
+                *dst = INTERPOLATE(*dst, pix, line->coverage[0] * pos);
                 ++dst;
                 ++pos;
             }
 
             //Right edge
-            dst = surface->buf32 + offset + line->x[1] - 1;
-
-            if (line->x[1] < (int32_t)(surface->w - 1)) pixel = *(dst + 1);
-            else pixel = *dst;
+            dst = buf + line->x[1] - 1;
+            pix = *(dst + (line->x[1] < (int32_t)(surface->w - 1) ? 1 : 0));
             pos = line->length[1];
 
             //exceptional handling. out of memory bound.
             if (dst - pos < surface->buf32) --pos;
 
             while (pos > 0) {
-                *dst = INTERPOLATE(*dst, pixel, 255 - (line->coverage[1] * pos));
+                *dst = INTERPOLATE(*dst, pix, 255 - (line->coverage[1] * pos));
                 --dst;
                 --pos;
             }
         }
-        y++;
+        buf += surface->stride;
+        ++line;
+        ++y;
     }
-
     tvg::free(aaSpans->lines);
     tvg::free(aaSpans);
-
-    return true;
 }
 
 
@@ -863,23 +721,19 @@ static bool _apply(SwSurface* surface, AASpans* aaSpans)
     | /  |
     3 -- 2
 */
-static bool _rasterTexmapPolygon(SwSurface* surface, const SwImage* image, const Matrix& transform, const SwBBox* region, uint8_t opacity)
+bool rasterTexmapPolygon(SwSurface* surface, const SwImage& image, const Matrix& transform, const RenderRegion& bbox, uint8_t opacity)
 {
     if (surface->channelSize == sizeof(uint8_t)) {
         TVGERR("SW_ENGINE", "Not supported grayscale Textmap polygon!");
         return false;
     }
 
-    //Exceptions: No dedicated drawing area?
-    if ((!image->rle && !region) || (image->rle && image->rle->size == 0)) return true;
-
-   /* Prepare vertices.
-      shift XY coordinates to match the sub-pixeling technique. */
+    //Prepare vertices. Shift XY coordinates to match the sub-pixeling technique.
     Vertex vertices[4];
     vertices[0] = {{0.0f, 0.0f}, {0.0f, 0.0f}};
-    vertices[1] = {{float(image->w), 0.0f}, {float(image->w), 0.0f}};
-    vertices[2] = {{float(image->w), float(image->h)}, {float(image->w), float(image->h)}};
-    vertices[3] = {{0.0f, float(image->h)}, {0.0f, float(image->h)}};
+    vertices[1] = {{float(image.w), 0.0f}, {float(image.w), 0.0f}};
+    vertices[2] = {{float(image.w), float(image.h)}, {float(image.w), float(image.h)}};
+    vertices[3] = {{0.0f, float(image.h)}, {0.0f, float(image.h)}};
 
     float ys = FLT_MAX, ye = -1.0f;
     for (int i = 0; i < 4; i++) {
@@ -888,8 +742,9 @@ static bool _rasterTexmapPolygon(SwSurface* surface, const SwImage* image, const
         if (vertices[i].pt.y > ye) ye = vertices[i].pt.y;
     }
 
-    auto aaSpans = _AASpans(ys, ye, image, region);
-    if (!aaSpans) return true;
+    auto yStart = std::max(static_cast<int>(ys), bbox.min.y);
+    auto yEnd = std::min(static_cast<int>(ye), bbox.max.y);
+    auto aaSpans = rightAngle(transform) ?  nullptr : _AASpans(yStart, yEnd);
 
     Polygon polygon;
 
@@ -898,19 +753,20 @@ static bool _rasterTexmapPolygon(SwSurface* surface, const SwImage* image, const
     polygon.vertex[1] = vertices[1];
     polygon.vertex[2] = vertices[3];
 
-    _rasterPolygonImage(surface, image, region, polygon, aaSpans, opacity);
+    _rasterPolygonImage(surface, image, bbox, polygon, aaSpans, opacity);
 
     //Draw the second polygon
     polygon.vertex[0] = vertices[1];
     polygon.vertex[1] = vertices[2];
     polygon.vertex[2] = vertices[3];
 
-    _rasterPolygonImage(surface, image, region, polygon, aaSpans, opacity);
+    _rasterPolygonImage(surface, image, bbox, polygon, aaSpans, opacity);
 
 #if 0
     if (_compositing(surface) && _masking(surface) && !_direct(surface->compositor->method)) {
         _compositeMaskImage(surface, &surface->compositor->image, surface->compositor->bbox);
     }
 #endif
-    return _apply(surface, aaSpans);
+    if (aaSpans) _apply(surface, aaSpans);
+    return true;
 }

src/renderer/sw_engine/tvgSwRenderer.cpp

@@ -40,7 +40,7 @@ struct SwTask : Task
 {
     SwSurface* surface = nullptr;
     SwMpool* mpool = nullptr;
-    SwBBox bbox;                          //Rendering Region
+    RenderRegion bbox;                          //Rendering Region
     Matrix transform;
     Array<RenderData> clips;
     RenderUpdateFlag flags = RenderUpdateFlag::None;
@@ -48,22 +48,11 @@ struct SwTask : Task
     bool pushed = false;                  //Pushed into task list?
     bool disposed = false;                //Disposed task?
 
-    RenderRegion bounds()
+    const RenderRegion& bounds()
     {
         //Can we skip the synchronization?
         done();
-
-        RenderRegion region;
-
-        //Range over?
-        region.x = bbox.min.x > 0 ? bbox.min.x : 0;
-        region.y = bbox.min.y > 0 ? bbox.min.y : 0;
-        region.w = bbox.max.x - region.x;
-        region.h = bbox.max.y - region.y;
-        if (region.w < 0) region.w = 0;
-        if (region.h < 0) region.h = 0;
-
-        return region;
+        return bbox;
     }
 
     virtual void dispose() = 0;
@@ -117,7 +106,7 @@ struct SwShapeTask : SwTask
         }
 
         auto strokeWidth = validStrokeWidth(clipper);
-        SwBBox renderRegion{};
+        RenderRegion renderBox{};
         auto updateShape = flags & (RenderUpdateFlag::Path | RenderUpdateFlag::Transform | RenderUpdateFlag::Clip);
         auto updateFill = false;
 
@@ -126,11 +115,11 @@ struct SwShapeTask : SwTask
             updateFill = (MULTIPLY(rshape->color.a, opacity) || rshape->fill);
             if (updateShape) shapeReset(&shape);
             if (updateFill || clipper) {
-                if (shapePrepare(&shape, rshape, transform, bbox, renderRegion, mpool, tid, clips.count > 0 ? true : false)) {
+                if (shapePrepare(&shape, rshape, transform, bbox, renderBox, mpool, tid, clips.count > 0 ? true : false)) {
                     if (!shapeGenRle(&shape, rshape, antialiasing(strokeWidth))) goto err;
                 } else {
                     updateFill = false;
-                    renderRegion.reset();
+                    renderBox.reset();
                 }
             }
         }
@@ -146,7 +135,7 @@ struct SwShapeTask : SwTask
         if (updateShape || flags & RenderUpdateFlag::Stroke) {
             if (strokeWidth > 0.0f) {
                 shapeResetStroke(&shape, rshape, transform);
-                if (!shapeGenStrokeRle(&shape, rshape, transform, bbox, renderRegion, mpool, tid)) goto err;
+                if (!shapeGenStrokeRle(&shape, rshape, transform, bbox, renderBox, mpool, tid)) goto err;
                 if (auto fill = rshape->strokeFill()) {
                     auto ctable = (flags & RenderUpdateFlag::GradientStroke) ? true : false;
                     if (ctable) shapeResetStrokeFill(&shape);
@@ -168,7 +157,7 @@ struct SwShapeTask : SwTask
             if (!clipShapeRle && !clipStrokeRle) goto err;
         }
 
-        bbox = renderRegion; //sync
+        bbox = renderBox; //sync
 
         return;
 
@@ -199,7 +188,7 @@ struct SwImageTask : SwTask
 
     void run(unsigned tid) override
     {
-        auto clipRegion = bbox;
+        auto clipBox = bbox;
 
         //Convert colorspace if it's not aligned.
         rasterConvertCS(source, surface->cs);
@@ -216,7 +205,7 @@ struct SwImageTask : SwTask
             imageReset(&image);
             if (!image.data || image.w == 0 || image.h == 0) goto end;
 
-            if (!imagePrepare(&image, transform, clipRegion, bbox, mpool, tid)) goto end;
+            if (!imagePrepare(&image, transform, clipBox, bbox, mpool, tid)) goto end;
 
             if (clips.count > 0) {
                 if (!imageGenRle(&image, bbox, false)) goto end;
@@ -245,26 +234,26 @@ struct SwImageTask : SwTask
 };
 
 
-static void _renderFill(SwShapeTask* task, SwSurface* surface, uint8_t opacity)
+static void _renderFill(SwShapeTask* task, SwSurface* surface)
 {
     if (auto fill = task->rshape->fill) {
-        rasterGradientShape(surface, &task->shape, fill, opacity);
+        rasterGradientShape(surface, &task->shape, fill, task->opacity);
     } else {
         RenderColor c;
         task->rshape->fillColor(&c.r, &c.g, &c.b, &c.a);
-        c.a = MULTIPLY(opacity, c.a);
+        c.a = MULTIPLY(task->opacity, c.a);
         if (c.a > 0) rasterShape(surface, &task->shape, c);
     }
 }
 
-static void _renderStroke(SwShapeTask* task, SwSurface* surface, uint8_t opacity)
+static void _renderStroke(SwShapeTask* task, SwSurface* surface)
 {
     if (auto strokeFill = task->rshape->strokeFill()) {
-        rasterGradientStroke(surface, &task->shape, strokeFill, opacity);
+        rasterGradientStroke(surface, &task->shape, strokeFill, task->opacity);
     } else {
         RenderColor c;
         if (task->rshape->strokeFill(&c.r, &c.g, &c.b, &c.a)) {
-            c.a = MULTIPLY(opacity, c.a);
+            c.a = MULTIPLY(task->opacity, c.a);
             if (c.a > 0) rasterStroke(surface, &task->shape, c);
         }
     }
@@ -359,7 +348,7 @@ bool SwRenderer::target(pixel_t* data, uint32_t stride, uint32_t w, uint32_t h,
 
 bool SwRenderer::preUpdate()
 {
-    return true;
+    return surface != nullptr;
 }
 
 
@@ -371,7 +360,7 @@ bool SwRenderer::postUpdate()
 
 bool SwRenderer::preRender()
 {
-    return true;
+    return surface != nullptr;
 }
 
 
@@ -394,12 +383,6 @@ bool SwRenderer::postRender()
         rasterUnpremultiply(surface);
     }
 
-    ARRAY_FOREACH(p, tasks) {
-        if ((*p)->disposed) delete(*p);
-        else (*p)->pushed = false;
-    }
-    tasks.clear();
-
     return true;
 }
 
@@ -411,7 +394,32 @@ bool SwRenderer::renderImage(RenderData data)
 
     if (task->opacity == 0) return true;
 
-    return rasterImage(surface, &task->image, task->transform, task->bbox, task->opacity);
+    //Outside of the viewport, skip the rendering
+    auto& bbox = task->bbox;
+    if (bbox.invalid() || bbox.x() >= surface->w || bbox.y() >= surface->h) return true;
+
+    auto& image = task->image;
+
+    //RLE Image
+    if (image.rle) {
+        if (image.direct) return rasterDirectRleImage(surface, image, task->opacity);
+        else if (image.scaled) return rasterScaledRleImage(surface, image, task->transform, bbox, task->opacity);
+        else {
+            //create a intermediate buffer for rle clipping
+            auto cmp = request(sizeof(pixel_t), false);
+            cmp->compositor->method = MaskMethod::None;
+            cmp->compositor->valid = true;
+            cmp->compositor->image.rle = image.rle;
+            rasterClear(cmp, bbox.x(), bbox.y(), bbox.w(), bbox.h(), 0);
+            rasterTexmapPolygon(cmp, image, task->transform, bbox, 255);
+            return rasterDirectRleImage(surface, cmp->compositor->image, task->opacity);
+        }
+    //Whole Image
+    } else {
+        if (image.direct) return rasterDirectImage(surface, image, bbox, task->opacity);
+        else if (image.scaled) return rasterScaledImage(surface, image, task->transform, bbox, task->opacity);
+        else return rasterTexmapPolygon(surface, image, task->transform, bbox, task->opacity);
+    }
 }
 
 
@@ -426,11 +434,11 @@ bool SwRenderer::renderShape(RenderData data)
 
     //Main raster stage
     if (task->rshape->strokeFirst()) {
-        _renderStroke(task, surface, task->opacity);
-        _renderFill(task, surface, task->opacity);
+        _renderStroke(task, surface);
+        _renderFill(task, surface);
     } else {
-        _renderFill(task, surface, task->opacity);
-        _renderStroke(task, surface, task->opacity);
+        _renderFill(task, surface);
+        _renderStroke(task, surface);
     }
 
     return true;
@@ -570,38 +578,19 @@ SwSurface* SwRenderer::request(int channelSize, bool square)
 
 RenderCompositor* SwRenderer::target(const RenderRegion& region, ColorSpace cs, CompositionFlag flags)
 {
-    auto x = region.x;
-    auto y = region.y;
-    auto w = region.w;
-    auto h = region.h;
-    auto sw = static_cast<int32_t>(surface->w);
-    auto sh = static_cast<int32_t>(surface->h);
-
-    //Out of boundary
-    if (x >= sw || y >= sh || x + w < 0 || y + h < 0) return nullptr;
+    auto bbox = RenderRegion::intersect(region, {{0, 0}, {int32_t(surface->w), int32_t(surface->h)}});
+    if (bbox.invalid()) return nullptr;
 
     auto cmp = request(CHANNEL_SIZE(cs), (flags & CompositionFlag::PostProcessing));
-
-    //Boundary Check
-    if (x < 0) x = 0;
-    if (y < 0) y = 0;
-    if (x + w > sw) w = (sw - x);
-    if (y + h > sh) h = (sh - y);
-
-    if (w == 0 || h == 0) return nullptr;
-
     cmp->compositor->recoverSfc = surface;
     cmp->compositor->recoverCmp = surface->compositor;
     cmp->compositor->valid = false;
-    cmp->compositor->bbox.min.x = x;
-    cmp->compositor->bbox.min.y = y;
-    cmp->compositor->bbox.max.x = x + w;
-    cmp->compositor->bbox.max.y = y + h;
+    cmp->compositor->bbox = bbox;
 
     /* TODO: Currently, only blending might work.
        Blending and composition must be handled together. */
     auto color = (surface->blender && !surface->compositor) ? 0x00ffffff : 0x00000000;
-    rasterClear(cmp, x, y, w, h, color);
+    rasterClear(cmp, bbox.x(), bbox.y(), bbox.w(), bbox.h(), color);
 
     //Switch render target
     surface = cmp;
@@ -626,8 +615,7 @@ bool SwRenderer::endComposite(RenderCompositor* cmp)
 
     //Default is alpha blending
     if (p->method == MaskMethod::None) {
-        auto m = tvg::identity();
-        return rasterImage(surface, &p->image, m, p->bbox, p->opacity);
+        return rasterDirectImage(surface, p->image, p->bbox, p->opacity);
     }
 
     return true;
@@ -722,6 +710,20 @@ void* SwRenderer::prepareCommon(SwTask* task, const Matrix& transform, const Arr
 {
     if (!surface) return task;
     if (flags == RenderUpdateFlag::None) return task;
+    if ((transform.e11 == 0.0f && transform.e12 == 0.0f) || (transform.e21 == 0.0f && transform.e22 == 0.0f)) return task;  //zero size?
+
+    task->surface = surface;
+    task->mpool = mpool;
+    task->bbox = RenderRegion::intersect(vport, {{0, 0}, {int32_t(surface->w), int32_t(surface->h)}});
+    task->transform = transform;
+    task->clips = clips;
+    task->opacity = opacity;
+    task->flags = flags;
+
+    if (!task->pushed) {
+        task->pushed = true;
+        tasks.push(task);
+    }
 
     //TODO: Failed threading them. It would be better if it's possible.
     //See: https://github.com/thorvg/thorvg/issues/1409
@@ -730,27 +732,6 @@ void* SwRenderer::prepareCommon(SwTask* task, const Matrix& transform, const Arr
         static_cast<SwTask*>(*p)->done();
     }
 
-    task->clips = clips;
-    task->transform = transform;
-    
-    //zero size?
-    if (task->transform.e11 == 0.0f && task->transform.e12 == 0.0f) return task; //zero width
-    if (task->transform.e21 == 0.0f && task->transform.e22 == 0.0f) return task; //zero height
-
-    task->opacity = opacity;
-    task->surface = surface;
-    task->mpool = mpool;
-    task->flags = flags;
-    task->bbox.min.x = std::max(static_cast<SwCoord>(0), static_cast<SwCoord>(vport.x));
-    task->bbox.min.y = std::max(static_cast<SwCoord>(0), static_cast<SwCoord>(vport.y));
-    task->bbox.max.x = std::min(static_cast<SwCoord>(surface->w), static_cast<SwCoord>(vport.x + vport.w));
-    task->bbox.max.y = std::min(static_cast<SwCoord>(surface->h), static_cast<SwCoord>(vport.y + vport.h));
-
-    if (!task->pushed) {
-        task->pushed = true;
-        tasks.push(task);
-    }
-
     TaskScheduler::request(task);
 
     return task;

src/renderer/sw_engine/tvgSwRenderer.h

@@ -56,6 +56,8 @@ public:
     bool sync() override;
     bool target(pixel_t* data, uint32_t stride, uint32_t w, uint32_t h, ColorSpace cs);
 
+    SwSurface* request(int channelSize, bool square);
+
     RenderCompositor* target(const RenderRegion& region, ColorSpace cs, CompositionFlag flags) override;
     bool beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_t opacity) override;
     bool endComposite(RenderCompositor* cmp) override;
@@ -80,7 +82,6 @@ private:
     SwRenderer();
     ~SwRenderer();
 
-    SwSurface* request(int channelSize, bool square);
     RenderData prepareCommon(SwTask* task, const Matrix& transform, const Array<RenderData>& clips, uint8_t opacity, RenderUpdateFlag flags);
 };
 

src/renderer/sw_engine/tvgSwRle.cpp

@@ -324,27 +324,18 @@ static void _horizLine(RleWorker& rw, SwCoord x, SwCoord y, SwCoord area, SwCoor
     if (!rw.antiAlias) coverage = 255;
 
     //see whether we can add this span to the current list
-    if (rle->size > 0) {
-        auto span = rle->spans + rle->size - 1;
-        if ((span->coverage == coverage) && (span->y == y) && (span->x + span->len == x)) {
+    if (!rle->spans.empty()) {
+        auto& span = rle->spans.last();
+        if ((span.coverage == coverage) && (span.y == y) && (span.x + span.len == x)) {
             //Clip x range
             SwCoord xOver = 0;
             if (x + aCount >= rw.cellMax.x) xOver -= (x + aCount - rw.cellMax.x);
             if (x < rw.cellMin.x) xOver -= (rw.cellMin.x - x);
-            span->len += (aCount + xOver);
+            span.len += (aCount + xOver);
             return;
         }
     }
 
-    //span pool is full, grow it.
-    if (rle->size >= rle->alloc) {
-        auto newSize = (rle->size > 0) ? (rle->size * 2) : 256;
-        if (rle->alloc < newSize) {
-            rle->alloc = newSize;
-            rle->spans = tvg::realloc<SwSpan*>(rle->spans, rle->alloc * sizeof(SwSpan));
-        }
-    }
-
     //Clip x range
     SwCoord xOver = 0;
     if (x + aCount >= rw.cellMax.x) xOver -= (x + aCount - rw.cellMax.x);
@@ -357,12 +348,7 @@ static void _horizLine(RleWorker& rw, SwCoord x, SwCoord y, SwCoord area, SwCoor
     if (aCount + xOver <= 0) return;
 
     //add a span to the current list
-    auto span = rle->spans + rle->size;
-    span->x = x;
-    span->y = y;
-    span->len = (aCount + xOver);
-    span->coverage = coverage;
-    rle->size++;
+    rle->spans.next() = {(uint16_t)x, (uint16_t)y, uint16_t(aCount + xOver), (uint8_t)coverage};
 }
 
 
@@ -737,110 +723,11 @@ static bool _genRle(RleWorker& rw)
 }
 
 
-static SwSpan* _intersectSpansRegion(const SwRle *clip, const SwRle *target, SwSpan *outSpans, uint32_t outSpansCnt)
-{
-    auto out = outSpans;
-    auto spans = target->spans;
-    auto end = target->spans + target->size;
-    auto clipSpans = clip->spans;
-    auto clipEnd = clip->spans + clip->size;
-
-    while (spans < end && clipSpans < clipEnd) {
-        //align y-coordinates.
-        if (clipSpans->y > spans->y) {
-            ++spans;
-            continue;
-        }
-        if (spans->y > clipSpans->y) {
-            ++clipSpans;
-            continue;
-        }
-
-        //Try clipping with all clip spans which have a same y-coordinate.
-        auto temp = clipSpans;
-        while(temp < clipEnd && outSpansCnt > 0 && temp->y == clipSpans->y) {
-            auto sx1 = spans->x;
-            auto sx2 = sx1 + spans->len;
-            auto cx1 = temp->x;
-            auto cx2 = cx1 + temp->len;
-
-            //The span must be left(x1) to right(x2) direction. Not intersected.
-            if (cx2 < sx1 || sx2 < cx1) {
-                ++temp;
-                continue;
-            }
-
-            //clip span region.
-            auto x = sx1 > cx1 ? sx1 : cx1;
-            auto len = (sx2 < cx2 ? sx2 : cx2) - x;
-            if (len > 0) {
-                out->x = x;
-                out->y = temp->y;
-                out->len = len;
-                out->coverage = (uint8_t)(((spans->coverage * temp->coverage) + 0xff) >> 8);
-                ++out;
-                --outSpansCnt;
-            }
-            ++temp;
-        }
-        ++spans;
-    }
-    return out;
-}
-
-
-static SwSpan* _intersectSpansRect(const SwBBox *bbox, const SwRle *targetRle, SwSpan *outSpans, uint32_t outSpansCnt)
-{
-    auto out = outSpans;
-    auto spans = targetRle->spans;
-    auto end = targetRle->spans + targetRle->size;
-    auto minx = static_cast<int16_t>(bbox->min.x);
-    auto miny = static_cast<int16_t>(bbox->min.y);
-    auto maxx = minx + static_cast<int16_t>(bbox->max.x - bbox->min.x) - 1;
-    auto maxy = miny + static_cast<int16_t>(bbox->max.y - bbox->min.y) - 1;
-
-    while (outSpansCnt > 0 && spans < end) {
-        if (spans->y > maxy) {
-            spans = end;
-            break;
-        }
-        if (spans->y < miny || spans->x > maxx || spans->x + spans->len <= minx) {
-            ++spans;
-            continue;
-        }
-        if (spans->x < minx) {
-            out->len = (spans->len - (minx - spans->x)) < (maxx - minx + 1) ? (spans->len - (minx - spans->x)) : (maxx - minx + 1);
-            out->x = minx;
-        }
-        else {
-            out->x = spans->x;
-            out->len = spans->len < (maxx - spans->x + 1) ? spans->len : (maxx - spans->x + 1);
-        }
-        if (out->len > 0) {
-            out->y = spans->y;
-            out->coverage = spans->coverage;
-            ++out;
-            --outSpansCnt;
-        }
-        ++spans;
-    }
-    return out;
-}
-
-
-void _replaceClipSpan(SwRle *rle, SwSpan* clippedSpans, uint32_t size)
-{
-    tvg::free(rle->spans);
-    rle->spans = clippedSpans;
-    rle->size = rle->alloc = size;
-}
-
-
 /************************************************************************/
 /* External Class Implementation                                        */
 /************************************************************************/
 
-SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegion, bool antiAlias)
+SwRle* rleRender(SwRle* rle, const SwOutline* outline, const RenderRegion& bbox, bool antiAlias)
 {
     if (!outline) return nullptr;
 
@@ -861,8 +748,8 @@ SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegio
     rw.area = 0;
     rw.cover = 0;
     rw.invalid = true;
-    rw.cellMin = renderRegion.min;
-    rw.cellMax = renderRegion.max;
+    rw.cellMin = {bbox.min.x, bbox.min.y};
+    rw.cellMax = {bbox.max.x, bbox.max.y};
     rw.cellXCnt = rw.cellMax.x - rw.cellMin.x;
     rw.cellYCnt = rw.cellMax.y - rw.cellMin.y;
     rw.outline = const_cast<SwOutline*>(outline);
@@ -870,8 +757,9 @@ SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegio
     rw.bandShoot = 0;
     rw.antiAlias = antiAlias;
 
-    if (!rle) rw.rle = tvg::calloc<SwRle*>(1, sizeof(SwRle));
+    if (!rle) rw.rle = new SwRle;
     else rw.rle = rle;
+    rw.rle->spans.reserve(256);
 
     //Generate RLE
     Band bands[BAND_SIZE];
@@ -934,7 +822,10 @@ SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegio
 
             /* This is too complex for a single scanline; there must
                be some problems */
-            if (middle == bottom) goto error;
+            if (middle == bottom) {
+                rleFree(rw.rle);
+                return nullptr;
+            }
 
             if (bottom - top >= rw.bandSize) ++rw.bandShoot;
 
@@ -945,34 +836,26 @@ SwRle* rleRender(SwRle* rle, const SwOutline* outline, const SwBBox& renderRegio
             ++band;
         }
     }
-
-    if (rw.bandShoot > 8 && rw.bandSize > 16)
+    if (rw.bandShoot > 8 && rw.bandSize > 16) {
         rw.bandSize = (rw.bandSize >> 1);
-
+    }
     return rw.rle;
-
-error:
-    tvg::free(rw.rle);
-    return nullptr;
 }
 
 
-SwRle* rleRender(const SwBBox* bbox)
+SwRle* rleRender(const RenderRegion* bbox)
 {
-    auto width = static_cast<uint16_t>(bbox->max.x - bbox->min.x);
-    auto height = static_cast<uint16_t>(bbox->max.y - bbox->min.y);
+    auto rle = tvg::calloc<SwRle*>(sizeof(SwRle), 1);
+    rle->spans.reserve(bbox->h());
+    rle->spans.count = bbox->h();
 
-    auto rle = tvg::malloc<SwRle*>(sizeof(SwRle));
-    rle->spans = tvg::malloc<SwSpan*>(sizeof(SwSpan) * height);
-    rle->size = height;
-    rle->alloc = height;
+    //cheaper without push()
+    auto x = uint16_t(bbox->min.x);
+    auto y = uint16_t(bbox->min.y);
+    auto len = uint16_t(bbox->w());
 
-    auto span = rle->spans;
-    for (uint16_t i = 0; i < height; ++i, ++span) {
-        span->x = bbox->min.x;
-        span->y = bbox->min.y + i;
-        span->len = width;
-        span->coverage = 255;
+    ARRAY_FOREACH(p, rle->spans) {
+        *p = {x, y++, len, 255};
     }
 
     return rle;
@@ -981,44 +864,88 @@ SwRle* rleRender(const SwBBox* bbox)
 
 void rleReset(SwRle* rle)
 {
-    if (!rle) return;
-    rle->size = 0;
+    if (rle) rle->spans.clear();
 }
 
 
 void rleFree(SwRle* rle)
 {
-    if (!rle) return;
-    if (rle->spans) tvg::free(rle->spans);
-    tvg::free(rle);
+    delete(rle);
 }
 
 
 bool rleClip(SwRle *rle, const SwRle *clip)
 {
-    if (rle->size == 0 || clip->size == 0) return false;
-    auto spanCnt = rle->size > clip->size ? rle->size : clip->size;
-    auto spans = tvg::malloc<SwSpan*>(sizeof(SwSpan) * (spanCnt));
-    auto spansEnd = _intersectSpansRegion(clip, rle, spans, spanCnt);
+    if (rle->spans.empty() || clip->spans.empty()) return false;
 
-    _replaceClipSpan(rle, spans, spansEnd - spans);
+    Array<SwSpan> out;
+    out.reserve(std::max(rle->spans.count, clip->spans.count));
 
-    TVGLOG("SW_ENGINE", "Using Path Clipping!");
+    auto spans = rle->data();
+    auto end = rle->spans.end();
+    auto cspans = clip->data();
+    auto cend = clip->spans.end();
 
+    while(spans < end && cspans < cend) {
+        //align y-coordinates.
+        if (cspans->y > spans->y) {
+            ++spans;
+            continue;
+        }
+        if (spans->y > cspans->y) {
+            ++cspans;
+            continue;
+        }
+        //try clipping with all clip spans which have a same y-coordinate.
+        auto temp = cspans;
+        while(temp < cend && temp->y == cspans->y) {
+            //span must be left(x1) to right(x2) direction. Not intersected.
+            if ((spans->x + spans->len) < spans->x || (temp->x + temp->len) < temp->x) {
+                ++temp;
+                continue;
+            }
+            //clip span region
+            auto x = std::max(spans->x, temp->x);
+            auto len = std::min((spans->x + spans->len), (temp->x + temp->len)) - x;
+            if (len > 0) out.next() = {uint16_t(x), temp->y, uint16_t(len), (uint8_t)(((spans->coverage * temp->coverage) + 0xff) >> 8)};
+            ++temp;
+        }
+        ++spans;
+    }
+    out.move(rle->spans);
     return true;
 }
 
 
-bool rleClip(SwRle *rle, const SwBBox* clip)
+//Need to confirm: dead code?
+bool rleClip(SwRle *rle, const RenderRegion* clip)
 {
-    if (rle->size == 0) return false;
-    auto spans = tvg::malloc<SwSpan*>(sizeof(SwSpan) * (rle->size));
-    auto spansEnd = _intersectSpansRect(clip, rle, spans, rle->size);
+    if (rle->spans.empty() || clip->invalid()) return false;
 
-    _replaceClipSpan(rle, spans, spansEnd - spans);
+    auto& min = clip->min;
+    auto& max = clip->max;
 
-    TVGLOG("SW_ENGINE", "Using Box Clipping!");
+    Array<SwSpan> out;
+    out.reserve(rle->spans.count);
+    auto data = out.data;
+    uint16_t x, len;
 
+    ARRAY_FOREACH(p, rle->spans) {
+        if (p->y >= max.y) break;
+        if (p->y < min.y || p->x >= max.x || (p->x + p->len) <= min.x) continue;
+        if (p->x < min.x) {
+            x = min.x;
+            len = std::min((p->len - (x - p->x)), (max.x - x));
+        } else {
+            x = p->x;
+            len = std::min(p->len, uint16_t(max.x - x));
+        }
+        if (len > 0) {
+            *data = {x, p->y, len, p->coverage};
+            ++data;
+            ++out.count;
+        }
+    }
+    out.move(rle->spans);
     return true;
 }
-

src/renderer/sw_engine/tvgSwShape.cpp

@@ -419,13 +419,13 @@ static SwOutline* _genOutline(SwShape* shape, const RenderShape* rshape, const M
 /* External Class Implementation                                        */
 /************************************************************************/
 
-bool shapePrepare(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid, bool hasComposite)
+bool shapePrepare(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid, bool hasComposite)
 {
     if (auto out = _genOutline(shape, rshape, transform, mpool, tid, hasComposite, rshape->trimpath())) shape->outline = out;
     else return false;
-    if (!mathUpdateOutlineBBox(shape->outline, clipRegion, renderRegion, shape->fastTrack)) return false;
+    if (!mathUpdateOutlineBBox(shape->outline, clipBox, renderBox, shape->fastTrack)) return false;
 
-    shape->bbox = renderRegion;
+    shape->bbox = renderBox;
     return true;
 }
 
@@ -500,7 +500,7 @@ void shapeResetStroke(SwShape* shape, const RenderShape* rshape, const Matrix& t
 }
 
 
-bool shapeGenStrokeRle(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const SwBBox& clipRegion, SwBBox& renderRegion, SwMpool* mpool, unsigned tid)
+bool shapeGenStrokeRle(SwShape* shape, const RenderShape* rshape, const Matrix& transform, const RenderRegion& clipBox, RenderRegion& renderBox, SwMpool* mpool, unsigned tid)
 {
     SwOutline* shapeOutline = nullptr;
     SwOutline* strokeOutline = nullptr;
@@ -528,12 +528,12 @@ bool shapeGenStrokeRle(SwShape* shape, const RenderShape* rshape, const Matrix&
 
     strokeOutline = strokeExportOutline(shape->stroke, mpool, tid);
 
-    if (!mathUpdateOutlineBBox(strokeOutline, clipRegion, renderRegion, false)) {
+    if (!mathUpdateOutlineBBox(strokeOutline, clipBox, renderBox, false)) {
         ret = false;
         goto clear;
     }
 
-    shape->strokeRle = rleRender(shape->strokeRle, strokeOutline, renderRegion, true);
+    shape->strokeRle = rleRender(shape->strokeRle, strokeOutline, renderBox, true);
 
 clear:
     if (dashStroking) mpoolRetDashOutline(mpool, tid);

src/renderer/tvgCanvas.h

@@ -31,7 +31,7 @@ struct Canvas::Impl
 {
     Scene* scene;
     RenderMethod* renderer;
-    RenderRegion vport = {0, 0, INT32_MAX, INT32_MAX};
+    RenderRegion vport = {{0, 0}, {INT32_MAX, INT32_MAX}};
     Status status = Status::Synced;
 
     Impl() : scene(Scene::gen())
@@ -119,11 +119,11 @@ struct Canvas::Impl
     {
         if (status != Status::Damaged && status != Status::Synced) return Result::InsufficientCondition;
 
-        RenderRegion val = {x, y, w, h};
+        RenderRegion val = {{x, y}, {x + w, y + h}};
         //intersect if the target buffer is already set.
         auto surface = renderer->mainSurface();
         if (surface && surface->w > 0 && surface->h > 0) {
-            val.intersect({0, 0, (int32_t)surface->w, (int32_t)surface->h});
+            val.intersect({{0, 0}, {(int32_t)surface->w, (int32_t)surface->h}});
         }
         if (vport == val) return Result::Success;
         renderer->viewport(val);

src/renderer/tvgGlCanvas.cpp

@@ -51,7 +51,7 @@ Result GlCanvas::target(void* context, int32_t id, uint32_t w, uint32_t h, Color
     if (!renderer) return Result::MemoryCorruption;
 
     if (!renderer->target(context, id, w, h)) return Result::Unknown;
-    pImpl->vport = {0, 0, (int32_t)w, (int32_t)h};
+    pImpl->vport = {{0, 0}, {(int32_t)w, (int32_t)h}};
     renderer->viewport(pImpl->vport);
 
     //Paints must be updated again with this new target.

src/renderer/tvgPaint.cpp

@@ -58,7 +58,7 @@ static Result _clipRect(RenderMethod* renderer, const Point* pts, const Matrix&
         if (tmp[i].y > max.y) max.y = tmp[i].y;
     }
 
-    float region[4] = {float(before.x), float(before.x + before.w), float(before.y), float(before.y + before.h)};
+    float region[4] = {float(before.min.x), float(before.max.x), float(before.min.y), float(before.max.y)};
 
     //figure out if the clipper is a superset of the current viewport(before) region
     if (min.x <= region[0] && max.x >= region[1] && min.y <= region[2] && max.y >= region[3]) {
@@ -66,7 +66,7 @@ static Result _clipRect(RenderMethod* renderer, const Point* pts, const Matrix&
         return Result::Success;
     //figure out if the clipper is totally outside of the viewport
     } else if (max.x <= region[0] || min.x >= region[1] || max.y <= region[2] || min.y >= region[3]) {
-        renderer->viewport({0, 0, 0, 0});
+        renderer->viewport({});
         return Result::Success;
     }
     return Result::InsufficientCondition;
@@ -122,13 +122,13 @@ static Result _compFastTrack(RenderMethod* renderer, Paint* cmpTarget, const Mat
         if (v1.x > v2.x) std::swap(v1.x, v2.x);
         if (v1.y > v2.y) std::swap(v1.y, v2.y);
 
-        after.x = static_cast<int32_t>(nearbyint(v1.x));
-        after.y = static_cast<int32_t>(nearbyint(v1.y));
-        after.w = static_cast<int32_t>(nearbyint(v2.x)) - after.x;
-        after.h = static_cast<int32_t>(nearbyint(v2.y)) - after.y;
+        after.min.x = static_cast<int32_t>(nearbyint(v1.x));
+        after.min.y = static_cast<int32_t>(nearbyint(v1.y));
+        after.max.x = static_cast<int32_t>(nearbyint(v2.x));
+        after.max.y = static_cast<int32_t>(nearbyint(v2.y));
 
-        if (after.w < 0) after.w = 0;
-        if (after.h < 0) after.h = 0;
+        if (after.max.x < after.min.x) after.max.x = after.min.x;
+        if (after.max.y < after.min.y) after.max.y = after.min.y;
 
         after.intersect(before);
         renderer->viewport(after);
@@ -168,7 +168,7 @@ Paint* Paint::Impl::duplicate(Paint* ret)
     ret->pImpl->opacity = opacity;
 
     if (maskData) ret->mask(maskData->target->duplicate(), maskData->method);
-    if (clipper) ret->clip(clipper->duplicate());
+    if (clipper) ret->clip(static_cast<Shape*>(clipper->duplicate()));
 
     return ret;
 }
@@ -185,7 +185,7 @@ bool Paint::Impl::render(RenderMethod* renderer)
         PAINT_METHOD(region, bounds(renderer));
 
         if (MASK_REGION_MERGING(maskData->method)) region.add(PAINT(maskData->target)->bounds(renderer));
-        if (region.w == 0 || region.h == 0) return true;
+        if (region.invalid()) return true;
         cmp = renderer->target(region, MASK_TO_COLORSPACE(renderer, maskData->method), CompositionFlag::Masking);
         if (renderer->beginComposite(cmp, MaskMethod::None, 255)) {
             maskData->target->pImpl->render(renderer);
@@ -376,16 +376,18 @@ Paint* Paint::duplicate() const noexcept
 }
 
 
-Result Paint::clip(Paint* clipper) noexcept
+Result Paint::clip(Shape* clipper) noexcept
 {
-    if (clipper && clipper->type() != Type::Shape) {
-        TVGERR("RENDERER", "Clipping only supports the Shape!");
-        return Result::NonSupport;
-    }
     return pImpl->clip(clipper);
 }
 
 
+Shape* Paint::clip() const noexcept
+{
+    return pImpl->clipper;
+}
+
+
 Result Paint::mask(Paint* target, MaskMethod method) noexcept
 {
     return pImpl->mask(target, method);

src/renderer/tvgPaint.h

@@ -54,7 +54,7 @@ namespace tvg
         Paint* paint = nullptr;
         Paint* parent = nullptr;
         Mask* maskData = nullptr;
-        Paint* clipper = nullptr;
+        Shape* clipper = nullptr;
         RenderMethod* renderer = nullptr;
         RenderData rd = nullptr;
 
@@ -80,9 +80,9 @@ namespace tvg
         } tr;
         RenderUpdateFlag renderFlag = RenderUpdateFlag::None;
         BlendMethod blendMethod;
+        uint16_t refCnt = 0;       //reference count
         uint8_t ctxFlag;
         uint8_t opacity;
-        uint8_t refCnt = 0;       //reference count
 
         Impl(Paint* pnt) : paint(pnt)
         {
@@ -107,9 +107,7 @@ namespace tvg
 
         uint8_t ref()
         {
-            if (refCnt == UINT8_MAX) TVGERR("RENDERER", "Reference Count Overflow!");
-            else ++refCnt;
-            return refCnt;
+            return ++refCnt;
         }
 
         uint8_t unref(bool free = true)
@@ -163,7 +161,7 @@ namespace tvg
             return tm;
         }
 
-        Result clip(Paint* clp)
+        Result clip(Shape* clp)
         {
             if (clp && PAINT(clp)->parent) return Result::InsufficientCondition;
             if (clipper) PAINT(clipper)->unref(clipper != clp);

src/renderer/tvgPicture.h

@@ -253,7 +253,7 @@ struct PictureImpl : Picture
 
     bool render(RenderMethod* renderer)
     {
-        bool ret = false;
+        auto ret = true;
         renderer->blend(impl.blendMethod);
 
         if (bitmap) return renderer->renderImage(impl.rd);
@@ -273,7 +273,7 @@ struct PictureImpl : Picture
     {
         if (impl.rd) return renderer->region(impl.rd);
         if (vector) return vector->pImpl->bounds(renderer);
-        return {0, 0, 0, 0};
+        return {};
     }
 
     Result load(ImageLoader* loader)

src/renderer/tvgRender.cpp

@@ -106,33 +106,14 @@ bool RenderPath::bounds(Matrix* m, float* x, float* y, float* w, float* h)
 
 void RenderRegion::intersect(const RenderRegion& rhs)
 {
-    auto x1 = x + w;
-    auto y1 = y + h;
-    auto x2 = rhs.x + rhs.w;
-    auto y2 = rhs.y + rhs.h;
+    if (min.x < rhs.min.x) min.x = rhs.min.x;
+    if (min.y < rhs.min.y) min.y = rhs.min.y;
+    if (max.x > rhs.max.x) max.x = rhs.max.x;
+    if (max.y > rhs.max.y) max.y = rhs.max.y;
 
-    x = (x > rhs.x) ? x : rhs.x;
-    y = (y > rhs.y) ? y : rhs.y;
-    w = ((x1 < x2) ? x1 : x2) - x;
-    h = ((y1 < y2) ? y1 : y2) - y;
-
-    if (w < 0) w = 0;
-    if (h < 0) h = 0;
-}
-
-
-void RenderRegion::add(const RenderRegion& rhs)
-{
-    if (rhs.x < x) {
-        w += (x - rhs.x);
-        x = rhs.x;
-    }
-    if (rhs.y < y) {
-        h += (y - rhs.y);
-        y = rhs.y;
-    }
-    if (rhs.x + rhs.w > x + w) w = (rhs.x + rhs.w) - x;
-    if (rhs.y + rhs.h > y + h) h = (rhs.y + rhs.h) - y;
+    // Not intersected: collapse to zero-area region
+    if (max.x < min.x) max.x = min.x;
+    if (max.y < min.y) max.y = min.y;
 }
 
 /************************************************************************/

src/renderer/tvgRender.h

@@ -94,16 +94,51 @@ struct RenderCompositor
 
 struct RenderRegion
 {
-    int32_t x, y, w, h;
+    struct {
+        int32_t x, y;
+    } min;
+
+    struct {
+        int32_t x, y;
+    } max;
+
+    static constexpr RenderRegion intersect(const RenderRegion& lhs, const RenderRegion& rhs)
+    {
+        RenderRegion ret = {{std::max(lhs.min.x, rhs.min.x), std::max(lhs.min.y, rhs.min.y)}, {std::min(lhs.max.x, rhs.max.x), std::min(lhs.max.y, rhs.max.y)}};
+        // Not intersected: collapse to zero-area region
+        if (ret.min.x > ret.max.x) ret.max.x = ret.min.x;
+        if (ret.min.y > ret.max.y) ret.max.y = ret.min.y;
+        return ret;
+    }
 
     void intersect(const RenderRegion& rhs);
-    void add(const RenderRegion& rhs);
+
+    void add(const RenderRegion& rhs)
+    {
+        if (rhs.min.x < min.x) min.x = rhs.min.x;
+        if (rhs.min.y < min.y) min.y = rhs.min.y;
+        if (rhs.max.x > max.x) max.x = rhs.max.x;
+        if (rhs.max.y > max.y) max.y = rhs.max.y;
+    }
 
     bool operator==(const RenderRegion& rhs) const
     {
-        if (x == rhs.x && y == rhs.y && w == rhs.w && h == rhs.h) return true;
-        return false;
+        return (min.x == rhs.min.x && min.y == rhs.min.y && max.x == rhs.max.x && max.y == rhs.max.y);
     }
+
+    void reset() { min.x = min.y = max.x = max.y = 0; }
+    bool valid() const { return (max.x > min.x && max.y > min.y); }
+    bool invalid() const { return !valid(); }
+
+    int32_t sx() const { return min.x; }
+    int32_t sy() const { return min.y; }
+    int32_t sw() const { return max.x - min.x; }
+    int32_t sh() const { return max.y - min.y; }
+
+    uint32_t x() const { return (uint32_t) sx(); }
+    uint32_t y() const { return (uint32_t) sy(); }
+    uint32_t w() const { return (uint32_t) sw(); }
+    uint32_t h() const { return (uint32_t) sh(); }
 };
 
 struct RenderPath
@@ -272,7 +307,7 @@ struct RenderShape
 struct RenderEffect
 {
     RenderData rd = nullptr;
-    RenderRegion extend = {0, 0, 0, 0};
+    RenderRegion extend{};
     SceneEffect type;
     bool valid = false;
 
@@ -353,7 +388,7 @@ struct RenderEffectTint : RenderEffect
         inst->white[0] = va_arg(args, int);
         inst->white[1] = va_arg(args, int);
         inst->white[2] = va_arg(args, int);
-        inst->intensity = (uint8_t)(va_arg(args, double) * 2.55);
+        inst->intensity = (uint8_t)(static_cast<float>(va_arg(args, double)) * 2.55f);
         inst->type = SceneEffect::Tint;
         return inst;
     }

src/renderer/tvgScene.h

@@ -64,10 +64,11 @@ struct SceneImpl : Scene
 {
     Paint::Impl impl;
     list<Paint*> paints;     //children list
-    RenderRegion vport = {0, 0, INT32_MAX, INT32_MAX};
+    RenderRegion vport = {};
     Array<RenderEffect*>* effects = nullptr;
     uint8_t compFlag = CompositionFlag::Invalid;
     uint8_t opacity;         //for composition
+    bool vdirty = false;
 
     SceneImpl() : impl(Paint::Impl(this))
     {
@@ -93,10 +94,11 @@ struct SceneImpl : Scene
         //Half translucent requires intermediate composition.
         if (opacity == 255) return compFlag;
 
-        //If scene has several children or only scene, it may require composition.
-        //OPTIMIZE: the bitmap type of the picture would not need the composition.
-        //OPTIMIZE: a single paint of a scene would not need the composition.
-        if (paints.size() == 1 && paints.front()->type() == Type::Shape) return compFlag;
+        //Only shape or picture may not require composition.
+        if (paints.size() == 1) {
+            auto type = paints.front()->type();
+            if (type == Type::Shape || type == Type::Picture) return compFlag;
+        }
 
         compFlag |= CompositionFlag::Opacity;
 
@@ -105,7 +107,7 @@ struct SceneImpl : Scene
 
     RenderData update(RenderMethod* renderer, const Matrix& transform, Array<RenderData>& clips, uint8_t opacity, RenderUpdateFlag flag, TVG_UNUSED bool clipper)
     {
-        this->vport = renderer->viewport();
+        if (paints.empty()) return nullptr;
 
         if (needComposition(opacity)) {
             /* Overriding opacity value. If this scene is half-translucent,
@@ -123,11 +125,17 @@ struct SceneImpl : Scene
             }
         }
 
+        //this viewport update is more performant than in bounds()?
+        vport = renderer->viewport();
+        vdirty = true;
+
         return nullptr;
     }
 
     bool render(RenderMethod* renderer)
     {
+        if (paints.empty()) return true;
+
         RenderCompositor* cmp = nullptr;
         auto ret = true;
 
@@ -157,42 +165,38 @@ struct SceneImpl : Scene
         return ret;
     }
 
-    RenderRegion bounds(RenderMethod* renderer) const
+    RenderRegion bounds(RenderMethod* renderer)
     {
-        if (paints.empty()) return {0, 0, 0, 0};
-
-        int32_t x1 = INT32_MAX;
-        int32_t y1 = INT32_MAX;
-        int32_t x2 = 0;
-        int32_t y2 = 0;
-
-        for (auto paint : paints) {
-            auto region = paint->pImpl->bounds(renderer);
+        if (paints.empty()) return {};
+        if (!vdirty) return vport;
+        vdirty = false;
 
         //Merge regions
-            if (region.x < x1) x1 = region.x;
-            if (x2 < region.x + region.w) x2 = (region.x + region.w);
-            if (region.y < y1) y1 = region.y;
-            if (y2 < region.y + region.h) y2 = (region.y + region.h);
+        RenderRegion pRegion = {{INT32_MAX, INT32_MAX}, {0, 0}};
+        for (auto paint : paints) {
+            auto region = paint->pImpl->bounds(renderer);
+            if (region.min.x < pRegion.min.x) pRegion.min.x = region.min.x;
+            if (pRegion.max.x < region.max.x) pRegion.max.x = region.max.x;
+            if (region.min.y < pRegion.min.y) pRegion.min.y = region.min.y;
+            if (pRegion.max.y < region.max.y) pRegion.max.y = region.max.y;
         }
 
         //Extends the render region if post effects require
-        int32_t ex = 0, ey = 0, ew = 0, eh = 0;
+        RenderRegion eRegion{};
         if (effects) {
             ARRAY_FOREACH(p, *effects) {
                 auto effect = *p;
-                if (effect->valid && renderer->region(effect)) {
-                    ex = std::min(ex, effect->extend.x);
-                    ey = std::min(ey, effect->extend.y);
-                    ew = std::max(ew, effect->extend.w);
-                    eh = std::max(eh, effect->extend.h);
-                }
+                if (effect->valid && renderer->region(effect)) eRegion.add(effect->extend);
             }
         }
 
-        auto ret = RenderRegion{x1 + ex, y1 + ey, (x2 - x1) + ew, (y2 - y1) + eh};
-        ret.intersect(this->vport);
-        return ret;
+        pRegion.min.x += eRegion.min.x;
+        pRegion.min.y += eRegion.min.y;
+        pRegion.max.x += eRegion.max.x;
+        pRegion.max.y += eRegion.max.y;
+
+        vport = RenderRegion::intersect(vport, pRegion);
+        return vport;
     }
 
     Result bounds(Point* pt4, Matrix& m, bool obb, bool stroking)

src/renderer/tvgShape.h

@@ -114,7 +114,7 @@ struct ShapeImpl : Shape
 
     RenderRegion bounds(RenderMethod* renderer)
     {
-        if (!impl.rd) return {0, 0, 0, 0};
+        if (!impl.rd) return {};
         return renderer->region(impl.rd);
     }
 

src/renderer/tvgSwCanvas.cpp

@@ -54,7 +54,7 @@ Result SwCanvas::target(uint32_t* buffer, uint32_t stride, uint32_t w, uint32_t
     if (!renderer) return Result::MemoryCorruption;
 
     if (!renderer->target(buffer, stride, w, h, cs)) return Result::InvalidArguments;
-    pImpl->vport = {0, 0, (int32_t)w, (int32_t)h};
+    pImpl->vport = {{0, 0}, {(int32_t)w, (int32_t)h}};
     renderer->viewport(pImpl->vport);
 
     //FIXME: The value must be associated with an individual canvas instance.

src/renderer/tvgWgCanvas.cpp

@@ -55,7 +55,7 @@ Result WgCanvas::target(void* device, void* instance, void* target, uint32_t w,
     if (!renderer) return Result::MemoryCorruption;
 
     if (!renderer->target((WGPUDevice)device, (WGPUInstance)instance, target, w, h, type)) return Result::Unknown;
-    pImpl->vport = {0, 0, (int32_t)w, (int32_t)h};
+    pImpl->vport = {{0, 0}, {(int32_t)w, (int32_t)h}};
     renderer->viewport(pImpl->vport);
 
     //Paints must be updated again with this new target.

src/renderer/wg_engine/meson.build

@@ -7,6 +7,7 @@ source_file = [
     'tvgWgRenderData.h',
     'tvgWgRenderer.h',
     'tvgWgRenderTarget.h',
+    'tvgWgRenderTask.h',
     'tvgWgShaderSrc.h',
     'tvgWgShaderTypes.h',
     'tvgWgBindGroups.cpp',
@@ -17,6 +18,7 @@ source_file = [
     'tvgWgRenderData.cpp',
     'tvgWgRenderer.cpp',
     'tvgWgRenderTarget.cpp',
+    'tvgWgRenderTask.cpp',
     'tvgWgShaderSrc.cpp',
     'tvgWgShaderTypes.cpp'
 ]

src/renderer/wg_engine/tvgWgBindGroups.cpp

@@ -103,9 +103,15 @@ WGPUBindGroup WgBindGroupLayouts::createBindGroupStrorage3RO(WGPUTextureView tex
 
 
 WGPUBindGroup WgBindGroupLayouts::createBindGroupBuffer1Un(WGPUBuffer buff)
+{
+    return createBindGroupBuffer1Un(buff, 0, wgpuBufferGetSize(buff));
+}
+
+
+WGPUBindGroup WgBindGroupLayouts::createBindGroupBuffer1Un(WGPUBuffer buff, uint64_t offset, uint64_t size)
 {
     const WGPUBindGroupEntry bindGroupEntrys[] = {
-        { .binding = 0, .buffer = buff, .size = wgpuBufferGetSize(buff) }
+        { .binding = 0, .buffer = buff, .offset = offset, .size = size }
     };
     const WGPUBindGroupDescriptor bindGroupDesc { .layout = layoutBuffer1Un, .entryCount = 1, .entries = bindGroupEntrys };
     return wgpuDeviceCreateBindGroup(device, &bindGroupDesc);

src/renderer/wg_engine/tvgWgBindGroups.h

@@ -49,6 +49,7 @@ public:
     WGPUBindGroup createBindGroupStrorage2RO(WGPUTextureView texView0, WGPUTextureView texView1);
     WGPUBindGroup createBindGroupStrorage3RO(WGPUTextureView texView0, WGPUTextureView texView1, WGPUTextureView texView2);
     WGPUBindGroup createBindGroupBuffer1Un(WGPUBuffer buff);
+    WGPUBindGroup createBindGroupBuffer1Un(WGPUBuffer buff, uint64_t offset, uint64_t size);
     WGPUBindGroup createBindGroupBuffer2Un(WGPUBuffer buff0, WGPUBuffer buff1);
     WGPUBindGroup createBindGroupBuffer3Un(WGPUBuffer buff0, WGPUBuffer buff1, WGPUBuffer buff2);
     void releaseBindGroup(WGPUBindGroup& bindGroup);

src/renderer/wg_engine/tvgWgCommon.cpp

@@ -200,10 +200,7 @@ bool WgContext::allocateBufferVertex(WGPUBuffer& buffer, const float* data, uint
         wgpuQueueWriteBuffer(queue, buffer, 0, data, size);
     else {
         releaseBuffer(buffer);
-        const WGPUBufferDescriptor bufferDesc {
-            .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Vertex,
-            .size = size > WG_VERTEX_BUFFER_MIN_SIZE ? size : WG_VERTEX_BUFFER_MIN_SIZE
-        };
+        const WGPUBufferDescriptor bufferDesc { .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Vertex, .size = size };
         buffer = wgpuDeviceCreateBuffer(device, &bufferDesc);
         wgpuQueueWriteBuffer(queue, buffer, 0, data, size);
         return true;
@@ -218,10 +215,7 @@ bool WgContext::allocateBufferIndex(WGPUBuffer& buffer, const uint32_t* data, ui
         wgpuQueueWriteBuffer(queue, buffer, 0, data, size);
     else {
         releaseBuffer(buffer);
-        const WGPUBufferDescriptor bufferDesc {
-            .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Index,
-            .size = size > WG_INDEX_BUFFER_MIN_SIZE ? size : WG_INDEX_BUFFER_MIN_SIZE
-        };
+        const WGPUBufferDescriptor bufferDesc { .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Index, .size = size };
         buffer = wgpuDeviceCreateBuffer(device, &bufferDesc);
         wgpuQueueWriteBuffer(queue, buffer, 0, data, size);
         return true;
@@ -241,10 +235,7 @@ bool WgContext::allocateBufferIndexFan(uint64_t vertexCount)
             indexes.push(i + 2);
         }
         releaseBuffer(bufferIndexFan);
-        WGPUBufferDescriptor bufferDesc{
-            .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Index,
-            .size = indexCount * sizeof(uint32_t)
-        };
+        WGPUBufferDescriptor bufferDesc{ .usage = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Index, .size = indexCount * sizeof(uint32_t) };
         bufferIndexFan = wgpuDeviceCreateBuffer(device, &bufferDesc);
         wgpuQueueWriteBuffer(queue, bufferIndexFan, 0, &indexes[0], indexCount * sizeof(uint32_t));
         return true;
@@ -269,3 +260,34 @@ void WgContext::releaseQueue(WGPUQueue& queue)
         queue = nullptr;
     }
 }
+
+
+WGPUCommandEncoder WgContext::createCommandEncoder()
+{
+    WGPUCommandEncoderDescriptor commandEncoderDesc{};
+    return wgpuDeviceCreateCommandEncoder(device, &commandEncoderDesc);
+}
+
+
+void WgContext::submitCommandEncoder(WGPUCommandEncoder commandEncoder)
+{
+    const WGPUCommandBufferDescriptor commandBufferDesc{};
+    WGPUCommandBuffer commandsBuffer = wgpuCommandEncoderFinish(commandEncoder, &commandBufferDesc);
+    wgpuQueueSubmit(queue, 1, &commandsBuffer);
+    wgpuCommandBufferRelease(commandsBuffer);
+}
+
+
+void WgContext::releaseCommandEncoder(WGPUCommandEncoder& commandEncoder)
+{
+    if (commandEncoder) {
+        wgpuCommandEncoderRelease(commandEncoder);
+        commandEncoder = nullptr;
+    }
+}
+
+
+bool WgContext::invalid()
+{
+    return !instance || !device;
+}

src/renderer/wg_engine/tvgWgCommon.h

@@ -25,9 +25,6 @@
 
 #include "tvgWgBindGroups.h"
 
-#define WG_VERTEX_BUFFER_MIN_SIZE 2048
-#define WG_INDEX_BUFFER_MIN_SIZE 2048
-
 struct WgContext {
     // external webgpu handles
     WGPUInstance instance{};
@@ -70,6 +67,13 @@ struct WgContext {
 
     // release buffer objects
     void releaseBuffer(WGPUBuffer& buffer);
+
+    // command encoder
+    WGPUCommandEncoder createCommandEncoder();
+    void submitCommandEncoder(WGPUCommandEncoder encoder);
+    void releaseCommandEncoder(WGPUCommandEncoder& encoder);
+
+    bool invalid();
 };
 
 #endif // _TVG_WG_COMMON_H_

src/renderer/wg_engine/tvgWgCompositor.h

@@ -38,6 +38,9 @@ class WgCompositor
 private:
     // pipelines
     WgPipelines pipelines{};
+    // stage buffers
+    WgStageBufferGeometry stageBufferGeometry{};
+    WgStageBufferUniform<WgShaderTypePaintSettings> stageBufferPaint;
     // global stencil/depth buffer handles
     WGPUTexture texDepthStencil{};
     WGPUTextureView texViewDepthStencil{};
@@ -52,21 +55,26 @@ private:
     // current render pass handles
     WGPURenderPassEncoder renderPassEncoder{};
     WGPUCommandEncoder commandEncoder{};
-    WgRenderStorage* currentTarget{};
-    // intermediate render storages
-    WgRenderStorage storageTemp0;
-    WgRenderStorage storageTemp1;
+    WgRenderTarget* currentTarget{};
+    // intermediate render targets
+    WgRenderTarget targetTemp0;
+    WgRenderTarget targetTemp1;
     WGPUBindGroup bindGroupStorageTemp{};
     // composition and blend geometries
-    WgMeshData meshData;
+    WgMeshData meshDataBlit;
     // render target dimensions
     uint32_t width{};
     uint32_t height{};
     
     // viewport utilities
     RenderRegion shrinkRenderRegion(RenderRegion& rect);
-    void copyTexture(const WgRenderStorage* dst, const WgRenderStorage* src);
-    void copyTexture(const WgRenderStorage* dst, const WgRenderStorage* src, const RenderRegion& region);
+    void copyTexture(const WgRenderTarget* dst, const WgRenderTarget* src);
+    void copyTexture(const WgRenderTarget* dst, const WgRenderTarget* src, const RenderRegion& region);
+
+    // base meshes draw
+    void drawMesh(WgContext& context, WgMeshData* meshData);
+    void drawMeshFan(WgContext& context, WgMeshData* meshData);
+    void drawMeshImage(WgContext& context, WgMeshData* meshData);
 
     // shapes
     void drawShape(WgContext& context, WgRenderDataShape* renderData);
@@ -84,8 +92,8 @@ private:
     void clipImage(WgContext& context, WgRenderDataPicture* renderData);
 
     // scenes
-    void drawScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose);
-    void blendScene(WgContext& context, WgRenderStorage* src, WgCompose* compose);
+    void drawScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose);
+    void blendScene(WgContext& context, WgRenderTarget* src, WgCompose* compose);
 
     // the renderer prioritizes clipping with the stroke over the shape's fill
     void markupClipPath(WgContext& context, WgRenderDataShape* renderData);
@@ -100,24 +108,32 @@ public:
     void resize(WgContext& context, uint32_t width, uint32_t height);
 
     // render passes workflow
-    void beginRenderPass(WGPUCommandEncoder encoder, WgRenderStorage* target, bool clear, WGPUColor clearColor = { 0.0, 0.0, 0.0, 0.0 });
+    void beginRenderPass(WGPUCommandEncoder encoder, WgRenderTarget* target, bool clear, WGPUColor clearColor = { 0.0, 0.0, 0.0, 0.0 });
     void endRenderPass();
 
+    // stage buffers operations
+    void reset(WgContext& context);
+    void flush(WgContext& context);
+
+    // request shapes for drawing (staging)
+    void requestShape(WgRenderDataShape* renderData);
+    void requestImage(WgRenderDataPicture* renderData);
+
     // render shapes, images and scenes
     void renderShape(WgContext& context, WgRenderDataShape* renderData, BlendMethod blendMethod);
     void renderImage(WgContext& context, WgRenderDataPicture* renderData, BlendMethod blendMethod);
-    void renderScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose);
-    void composeScene(WgContext& context, WgRenderStorage* src, WgRenderStorage* mask, WgCompose* compose);
+    void renderScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose);
+    void composeScene(WgContext& context, WgRenderTarget* src, WgRenderTarget* mask, WgCompose* compose);
 
-    // blit render storage to texture view (f.e. screen buffer)
-    void blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderStorage* src, WGPUTextureView dstView);
+    // blit render target to texture view (f.e. screen buffer)
+    void blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderTarget* src, WGPUTextureView dstView);
 
     // effects
-    bool gaussianBlur(WgContext& context, WgRenderStorage* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose);
-    bool dropShadow(WgContext& context, WgRenderStorage* dst, const RenderEffectDropShadow* params, const WgCompose* compose);
-    bool fillEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectFill* params, const WgCompose* compose);
-    bool tintEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTint* params, const WgCompose* compose);
-    bool tritoneEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTritone* params, const WgCompose* compose);
+    bool gaussianBlur(WgContext& context, WgRenderTarget* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose);
+    bool dropShadow(WgContext& context, WgRenderTarget* dst, const RenderEffectDropShadow* params, const WgCompose* compose);
+    bool fillEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectFill* params, const WgCompose* compose);
+    bool tintEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectTint* params, const WgCompose* compose);
+    bool tritoneEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectTritone* params, const WgCompose* compose);
 };
 
 #endif // _TVG_WG_COMPOSITOR_H_

src/renderer/wg_engine/tvgWgPipelines.cpp

@@ -167,17 +167,17 @@ void WgPipelines::initialize(WgContext& context)
 
     const WgBindGroupLayouts& layouts = context.layouts;
     // bind group layouts helpers
-    const WGPUBindGroupLayout bindGroupLayoutsStencil[] { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un };
-    const WGPUBindGroupLayout bindGroupLayoutsDepth[]   { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutBuffer1Un };
+    const WGPUBindGroupLayout bindGroupLayoutsStencil[] { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un };
+    const WGPUBindGroupLayout bindGroupLayoutsDepth[]   { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutBuffer1Un };
     // bind group layouts normal blend
-    const WGPUBindGroupLayout bindGroupLayoutsSolid[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutBuffer1Un };
-    const WGPUBindGroupLayout bindGroupLayoutsGradient[] { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutTexSampledBuff2Un };
-    const WGPUBindGroupLayout bindGroupLayoutsImage[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutTexSampled };
+    const WGPUBindGroupLayout bindGroupLayoutsSolid[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un };
+    const WGPUBindGroupLayout bindGroupLayoutsGradient[] { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled };
+    const WGPUBindGroupLayout bindGroupLayoutsImage[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled };
     const WGPUBindGroupLayout bindGroupLayoutsScene[]    { layouts.layoutTexSampled, layouts.layoutBuffer1Un };
     // bind group layouts custom blend
-    const WGPUBindGroupLayout bindGroupLayoutsSolidBlend[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled };
-    const WGPUBindGroupLayout bindGroupLayoutsGradientBlend[] { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutTexSampledBuff2Un, layouts.layoutTexSampled };
-    const WGPUBindGroupLayout bindGroupLayoutsImageBlend[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer2Un, layouts.layoutTexSampled, layouts.layoutTexSampled };
+    const WGPUBindGroupLayout bindGroupLayoutsSolidBlend[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled };
+    const WGPUBindGroupLayout bindGroupLayoutsGradientBlend[] { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled, layouts.layoutTexSampled };
+    const WGPUBindGroupLayout bindGroupLayoutsImageBlend[]    { layouts.layoutBuffer1Un, layouts.layoutBuffer1Un, layouts.layoutTexSampled, layouts.layoutTexSampled };
     const WGPUBindGroupLayout bindGroupLayoutsSceneBlend[]    { layouts.layoutTexSampled, layouts.layoutTexSampled, layouts.layoutBuffer1Un };
     // bind group layouts scene compose
     const WGPUBindGroupLayout bindGroupLayoutsSceneCompose[] { layouts.layoutTexSampled, layouts.layoutTexSampled };
@@ -229,7 +229,7 @@ void WgPipelines::initialize(WgContext& context)
     layout_stencil = createPipelineLayout(context.device, bindGroupLayoutsStencil, 2);
     layout_depth = createPipelineLayout(context.device, bindGroupLayoutsDepth, 3);
     // layouts normal blend
-    layout_solid    = createPipelineLayout(context.device, bindGroupLayoutsSolid, 3);
+    layout_solid    = createPipelineLayout(context.device, bindGroupLayoutsSolid, 2);
     layout_gradient = createPipelineLayout(context.device, bindGroupLayoutsGradient, 3);
     layout_image    = createPipelineLayout(context.device, bindGroupLayoutsImage, 3);
     layout_scene    = createPipelineLayout(context.device, bindGroupLayoutsScene, 2);

src/renderer/wg_engine/tvgWgRenderData.cpp

@@ -30,163 +30,115 @@
 // WgMeshData
 //***********************************************************************
 
-void WgMeshData::draw(WgContext& context, WGPURenderPassEncoder renderPassEncoder)
-{
-    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, bufferPosition, 0, vertexCount * sizeof(float) * 2);
-    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, bufferIndex, WGPUIndexFormat_Uint32, 0, indexCount * sizeof(uint32_t));
-    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, indexCount, 1, 0, 0, 0);
-}
-
-
-void WgMeshData::drawFan(WgContext& context, WGPURenderPassEncoder renderPassEncoder)
-{
-    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, bufferPosition, 0, vertexCount * sizeof(float) * 2);
-    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, context.bufferIndexFan, WGPUIndexFormat_Uint32, 0, indexCount * sizeof(uint32_t));
-    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, indexCount, 1, 0, 0, 0);
-}
-
-
-void WgMeshData::drawImage(WgContext& context, WGPURenderPassEncoder renderPassEncoder)
-{
-    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, bufferPosition, 0, vertexCount * sizeof(float) * 2);
-    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 1, bufferTexCoord, 0, vertexCount * sizeof(float) * 2);
-    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, bufferIndex, WGPUIndexFormat_Uint32, 0, indexCount * sizeof(uint32_t));
-    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, indexCount, 1, 0, 0, 0);
-};
-
-
-void WgMeshData::update(WgContext& context, const WgVertexBuffer& vertexBuffer)
+void WgMeshData::update(const WgVertexBuffer& vertexBuffer)
 {
     assert(vertexBuffer.count > 2);
-    vertexCount = vertexBuffer.count;
-    indexCount = (vertexBuffer.count - 2) * 3;
-    context.allocateBufferVertex(bufferPosition, (float*)vertexBuffer.data, vertexCount * sizeof(float) * 2);
-    context.allocateBufferIndexFan(vertexCount);
+    // setup vertex data
+    vbuffer.reserve(vertexBuffer.count);
+    vbuffer.count = vertexBuffer.count;
+    memcpy(vbuffer.data, vertexBuffer.data, sizeof(vertexBuffer.data[0])*vertexBuffer.count);
+    // setup tex coords data
+    tbuffer.clear();
 }
 
 
-void WgMeshData::update(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd)
+void WgMeshData::update(const WgIndexedVertexBuffer& vertexBufferInd)
 {
     assert(vertexBufferInd.vcount > 2);
-    vertexCount = vertexBufferInd.vcount;
-    indexCount = vertexBufferInd.icount;
-    if (vertexCount > 0) context.allocateBufferVertex(bufferPosition, (float*)vertexBufferInd.vbuff, vertexCount * sizeof(float) * 2);
-    if (indexCount > 0) context.allocateBufferIndex(bufferIndex, vertexBufferInd.ibuff, indexCount * sizeof(uint32_t));
+    // setup vertex data
+    vbuffer.reserve(vertexBufferInd.vcount);
+    vbuffer.count = vertexBufferInd.vcount;
+    memcpy(vbuffer.data, vertexBufferInd.vbuff, sizeof(vertexBufferInd.vbuff[0])*vertexBufferInd.vcount);
+    // setup tex coords data
+    tbuffer.clear();
+    // copy index data
+    ibuffer.reserve(vertexBufferInd.icount);
+    ibuffer.count = vertexBufferInd.icount;
+    memcpy(ibuffer.data, vertexBufferInd.ibuff, sizeof(vertexBufferInd.ibuff[0])*vertexBufferInd.icount);
 };
 
 
-void WgMeshData::bbox(WgContext& context, const Point pmin, const Point pmax)
+void WgMeshData::bbox(const Point pmin, const Point pmax)
 {
-    vertexCount = 4;
-    indexCount = 6;
     const float data[] = {pmin.x, pmin.y, pmax.x, pmin.y, pmax.x, pmax.y, pmin.x, pmax.y};
-    context.allocateBufferVertex(bufferPosition, data, sizeof(data));
-    context.allocateBufferIndexFan(vertexCount);
+    // setup vertex data
+    vbuffer.reserve(4);
+    vbuffer.count = 4;
+    memcpy(vbuffer.data, data, sizeof(data));
+    // setup tex coords data
+    tbuffer.clear();
 }
 
 
-void WgMeshData::imageBox(WgContext& context, float w, float h)
+void WgMeshData::imageBox(float w, float h)
 {
-    vertexCount = 4;
-    indexCount = 6;
     const float vdata[] = {0.0f, 0.0f, w, 0.0f, w, h, 0.0f, h};
     const float tdata[] = {0.0f, 0.0f, 1.0f, 0.0f, 1.0f, 1.0f, 0.0f, 1.0f};
     const uint32_t idata[] = {0, 1, 2, 0, 2, 3};
-    context.allocateBufferVertex(bufferPosition, vdata, sizeof(vdata));
-    context.allocateBufferVertex(bufferTexCoord, tdata, sizeof(tdata));
-    context.allocateBufferIndex(bufferIndex, idata, sizeof(idata));
+    // setup vertex data
+    vbuffer.reserve(4);
+    vbuffer.count = 4;
+    memcpy(vbuffer.data, vdata, sizeof(vdata));
+    // setup tex coords data
+    tbuffer.reserve(4);
+    tbuffer.count = 4;
+    memcpy(tbuffer.data, tdata, sizeof(tdata));
+    // setup indexes data
+    ibuffer.reserve(6);
+    ibuffer.count = 6;
+    memcpy(ibuffer.data, idata, sizeof(idata));
 }
 
 
-void WgMeshData::blitBox(WgContext& context)
+void WgMeshData::blitBox()
 {
-    vertexCount = 4;
-    indexCount = 6;
     const float vdata[] = {-1.0f, +1.0f, +1.0f, +1.0f, +1.0f, -1.0f, -1.0f, -1.0f};
     const float tdata[] = {+0.0f, +0.0f, +1.0f, +0.0f, +1.0f, +1.0f, +0.0f, +1.0f};
     const uint32_t idata[] = { 0, 1, 2, 0, 2, 3 };
-    context.allocateBufferVertex(bufferPosition, vdata, sizeof(vdata));
-    context.allocateBufferVertex(bufferTexCoord, tdata, sizeof(tdata));
-    context.allocateBufferIndex(bufferIndex, idata, sizeof(idata));
+    // setup vertex data
+    vbuffer.reserve(4);
+    vbuffer.count = 4;
+    memcpy(vbuffer.data, vdata, sizeof(vdata));
+    // setup tex coords data
+    tbuffer.reserve(4);
+    tbuffer.count = 4;
+    memcpy(tbuffer.data, tdata, sizeof(tdata));
+    // setup indexes data
+    ibuffer.reserve(6);
+    ibuffer.count = 6;
+    memcpy(ibuffer.data, idata, sizeof(idata));
 }
 
-
-void WgMeshData::release(WgContext& context)
-{
-    context.releaseBuffer(bufferIndex);
-    context.releaseBuffer(bufferTexCoord);
-    context.releaseBuffer(bufferPosition);
-};
-
-
-//***********************************************************************
-// WgMeshDataPool
-//***********************************************************************
-
-WgMeshData* WgMeshDataPool::allocate(WgContext& context)
-{
-    WgMeshData* meshData{};
-    if (mPool.count > 0) {
-        meshData = mPool.last();
-        mPool.pop();
-    } else {
-        meshData = new WgMeshData();
-        mList.push(meshData);
-    }
-    return meshData;
-}
-
-
-void WgMeshDataPool::free(WgContext& context, WgMeshData* meshData)
-{
-    mPool.push(meshData);
-}
-
-
-void WgMeshDataPool::release(WgContext& context)
-{
-    ARRAY_FOREACH(p, mList) {
-        (*p)->release(context);
-        delete(*p);
-    }
-    mPool.clear();
-    mList.clear();
-}
-
-WgMeshDataPool gMeshDataPoolInstance;
-WgMeshDataPool* WgMeshDataPool::gMeshDataPool = &gMeshDataPoolInstance;
-
 //***********************************************************************
 // WgMeshDataGroup
 //***********************************************************************
 
-void WgMeshDataGroup::append(WgContext& context, const WgVertexBuffer& vertexBuffer)
+void WgMeshDataGroup::append(const WgVertexBuffer& vertexBuffer)
 {
     assert(vertexBuffer.count >= 3);
-    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
-    meshes.last()->update(context, vertexBuffer);
+    meshes.push(new WgMeshData());
+    meshes.last()->update(vertexBuffer);
 }
 
 
-void WgMeshDataGroup::append(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd)
+void WgMeshDataGroup::append(const WgIndexedVertexBuffer& vertexBufferInd)
 {
     assert(vertexBufferInd.vcount >= 3);
-    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
-    meshes.last()->update(context, vertexBufferInd);
+    meshes.push(new WgMeshData());
+    meshes.last()->update(vertexBufferInd);
 }
 
 
-void WgMeshDataGroup::append(WgContext& context, const Point pmin, const Point pmax)
+void WgMeshDataGroup::append(const Point pmin, const Point pmax)
 {
-    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
-    meshes.last()->bbox(context, pmin, pmax);
+    meshes.push(new WgMeshData());
+    meshes.last()->bbox(pmin, pmax);
 }
 
 
-void WgMeshDataGroup::release(WgContext& context)
+void WgMeshDataGroup::release()
 {
-    ARRAY_FOREACH(p, meshes)
-        WgMeshDataPool::gMeshDataPool->free(context, *p);
+    ARRAY_FOREACH(p, meshes) delete *p;
     meshes.clear();
 };
 
@@ -209,12 +161,38 @@ void WgImageData::update(WgContext& context, const RenderSurface* surface)
     if (texHandleChanged) {
         context.releaseTextureView(textureView);
         textureView = context.createTextureView(texture);
+        // update bind group
+        context.layouts.releaseBindGroup(bindGroup);
+        bindGroup = context.layouts.createBindGroupTexSampled(context.samplerLinearRepeat, textureView);
+    }
+};
+
+
+void WgImageData::update(WgContext& context, const Fill* fill)
+{
+    // compute gradient data
+    WgShaderTypeGradientData gradientData;
+    gradientData.update(fill);
+    // allocate new texture handle
+    bool texHandleChanged = context.allocateTexture(texture, WG_TEXTURE_GRADIENT_SIZE, 1, WGPUTextureFormat_RGBA8Unorm, gradientData.data);
+    // update texture view of texture handle was changed
+    if (texHandleChanged) {
+        context.releaseTextureView(textureView);
+        textureView = context.createTextureView(texture);
+        // get sampler by spread type
+        WGPUSampler sampler = context.samplerLinearClamp;
+        if (fill->spread() == FillSpread::Reflect) sampler = context.samplerLinearMirror;
+        if (fill->spread() == FillSpread::Repeat) sampler = context.samplerLinearRepeat;
+        // update bind group
+        context.layouts.releaseBindGroup(bindGroup);
+        bindGroup = context.layouts.createBindGroupTexSampled(sampler, textureView);
     }
 };
 
 
 void WgImageData::release(WgContext& context)
 {
+    context.layouts.releaseBindGroup(bindGroup);
     context.releaseTextureView(textureView);
     context.releaseTexture(texture);
 };
@@ -223,52 +201,31 @@ void WgImageData::release(WgContext& context)
 // WgRenderSettings
 //***********************************************************************
 
-void WgRenderSettings::updateFill(WgContext& context, const Fill* fill)
+void WgRenderSettings::update(WgContext& context, const tvg::Matrix& transform, tvg::ColorSpace cs, uint8_t opacity)
 {
-    rasterType = WgRenderRasterType::Gradient;
-    // get gradient transfrom matrix
-    Matrix invFillTransform;
-    WgShaderTypeMat4x4f gradientTrans; // identity by default
-    if (inverse(&fill->transform(), &invFillTransform))
-        gradientTrans.update(invFillTransform);
+    settings.transform.update(transform);
+    settings.options.update(cs, opacity);
+}
+
+void WgRenderSettings::update(WgContext& context, const Fill* fill)
+{
+    assert(fill);
+    settings.gradient.update(fill);
+    gradientData.update(context, fill);
     // get gradient rasterisation settings
-    WgShaderTypeGradient gradient;
-    if (fill->type() == Type::LinearGradient) {
-        gradient.update((LinearGradient*)fill);
+    rasterType = WgRenderRasterType::Gradient;
+    if (fill->type() == Type::LinearGradient)
         fillType = WgRenderSettingsType::Linear;
-    } else if (fill->type() == Type::RadialGradient) {
-        gradient.update((RadialGradient*)fill);
+    else if (fill->type() == Type::RadialGradient)
         fillType = WgRenderSettingsType::Radial;
-    }
-    // update gpu assets
-    bool bufferGradientSettingsChanged = context.allocateBufferUniform(bufferGroupGradient, &gradient.settings, sizeof(gradient.settings));
-    bool bufferGradientTransformChanged = context.allocateBufferUniform(bufferGroupTransfromGrad, &gradientTrans.mat, sizeof(gradientTrans.mat));
-    bool textureGradientChanged = context.allocateTexture(texGradient, WG_TEXTURE_GRADIENT_SIZE, 1, WGPUTextureFormat_RGBA8Unorm, gradient.texData);
-    if (bufferGradientSettingsChanged || textureGradientChanged || bufferGradientTransformChanged) {
-        // update texture view
-        context.releaseTextureView(texViewGradient);
-        texViewGradient = context.createTextureView(texGradient);
-        // get sampler by spread type
-        WGPUSampler sampler = context.samplerLinearClamp;
-        if (fill->spread() == FillSpread::Reflect) sampler = context.samplerLinearMirror;
-        if (fill->spread() == FillSpread::Repeat) sampler = context.samplerLinearRepeat;
-        // update bind group
-        context.layouts.releaseBindGroup(bindGroupGradient);
-        bindGroupGradient = context.layouts.createBindGroupTexSampledBuff2Un(
-            sampler, texViewGradient, bufferGroupGradient, bufferGroupTransfromGrad);
-    }
     skip = false;
 };
 
 
-void WgRenderSettings::updateColor(WgContext& context, const RenderColor& c)
+void WgRenderSettings::update(WgContext& context, const RenderColor& c)
 {
+    settings.color.update(c);
     rasterType = WgRenderRasterType::Solid;
-    WgShaderTypeVec4f solidColor(c);
-    if (context.allocateBufferUniform(bufferGroupSolid, &solidColor, sizeof(solidColor))) {
-        context.layouts.releaseBindGroup(bindGroupSolid);
-        bindGroupSolid = context.layouts.createBindGroupBuffer1Un(bufferGroupSolid);
-    }
     fillType = WgRenderSettingsType::Solid;
     skip = (c.a == 0);
 };
@@ -276,13 +233,7 @@ void WgRenderSettings::updateColor(WgContext& context, const RenderColor& c)
 
 void WgRenderSettings::release(WgContext& context)
 {
-    context.layouts.releaseBindGroup(bindGroupSolid);
-    context.layouts.releaseBindGroup(bindGroupGradient);
-    context.releaseBuffer(bufferGroupSolid);
-    context.releaseBuffer(bufferGroupGradient);
-    context.releaseBuffer(bufferGroupTransfromGrad);
-    context.releaseTexture(texGradient);
-    context.releaseTextureView(texViewGradient);
+    gradientData.release(context);
 };
 
 //***********************************************************************
@@ -291,26 +242,10 @@ void WgRenderSettings::release(WgContext& context)
 
 void WgRenderDataPaint::release(WgContext& context)
 {
-    context.layouts.releaseBindGroup(bindGroupPaint);
-    context.releaseBuffer(bufferModelMat);
-    context.releaseBuffer(bufferBlendSettings);
     clips.clear();
 };
 
 
-void WgRenderDataPaint::update(WgContext& context, const tvg::Matrix& transform, tvg::ColorSpace cs, uint8_t opacity)
-{
-    WgShaderTypeMat4x4f modelMat(transform);
-    WgShaderTypeVec4f blendSettings(cs, opacity);
-    bool bufferModelMatChanged = context.allocateBufferUniform(bufferModelMat, &modelMat, sizeof(modelMat));
-    bool bufferBlendSettingsChanged = context.allocateBufferUniform(bufferBlendSettings, &blendSettings, sizeof(blendSettings));
-    if (bufferModelMatChanged || bufferBlendSettingsChanged) {
-        context.layouts.releaseBindGroup(bindGroupPaint);
-        bindGroupPaint = context.layouts.createBindGroupBuffer2Un(bufferModelMat, bufferBlendSettings);
-    }
-}
-
-
 void WgRenderDataPaint::updateClips(tvg::Array<tvg::RenderData> &clips) {
     this->clips.clear();
     ARRAY_FOREACH(p, clips) {
@@ -322,24 +257,24 @@ void WgRenderDataPaint::updateClips(tvg::Array<tvg::RenderData> &clips) {
 // WgRenderDataShape
 //***********************************************************************
 
-void WgRenderDataShape::appendShape(WgContext& context, const WgVertexBuffer& vertexBuffer)
+void WgRenderDataShape::appendShape(const WgVertexBuffer& vertexBuffer)
 {
     if (vertexBuffer.count < 3) return;
     Point pmin{}, pmax{};
     vertexBuffer.getMinMax(pmin, pmax);
-    meshGroupShapes.append(context, vertexBuffer);
-    meshGroupShapesBBox.append(context, pmin, pmax);
+    meshGroupShapes.append(vertexBuffer);
+    meshGroupShapesBBox.append(pmin, pmax);
     updateBBox(pmin, pmax);
 }
 
 
-void WgRenderDataShape::appendStroke(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd)
+void WgRenderDataShape::appendStroke(const WgIndexedVertexBuffer& vertexBufferInd)
 {
     if (vertexBufferInd.vcount < 3) return;
     Point pmin{}, pmax{};
     vertexBufferInd.getMinMax(pmin, pmax);
-    meshGroupStrokes.append(context, vertexBufferInd);
-    meshGroupStrokesBBox.append(context, pmin, pmax);
+    meshGroupStrokes.append(vertexBufferInd);
+    meshGroupStrokesBBox.append(pmin, pmax);
     updateBBox(pmin, pmax);
 }
 
@@ -358,14 +293,14 @@ void WgRenderDataShape::updateAABB(const Matrix& tr) {
     auto p1 = Point{pMax.x, pMin.y} * tr;
     auto p2 = Point{pMin.x, pMax.y} * tr;
     auto p3 = Point{pMax.x, pMax.y} * tr;
-    aabb.min = {std::min({p0.x, p1.x, p2.x, p3.x}), std::min({p0.y, p1.y, p2.y, p3.y})};
-    aabb.max = {std::max({p0.x, p1.x, p2.x, p3.x}), std::max({p0.y, p1.y, p2.y, p3.y})};
+    aabb.min = {std::min(std::min(p0.x, p1.x), std::min(p2.x, p3.x)), std::min(std::min(p0.y, p1.y), std::min(p2.y, p3.y))};
+    aabb.max = {std::max(std::max(p0.x, p1.x), std::max(p2.x, p3.x)), std::max(std::max(p0.y, p1.y), std::max(p2.y, p3.y))};
 }
 
 
-void WgRenderDataShape::updateMeshes(WgContext& context, const RenderShape &rshape, const Matrix& tr, WgGeometryBufferPool* pool)
+void WgRenderDataShape::updateMeshes(const RenderShape &rshape, const Matrix& tr, WgGeometryBufferPool* pool)
 {
-    releaseMeshes(context);
+    releaseMeshes();
     strokeFirst = rshape.strokeFirst();
 
     // get object scale
@@ -375,8 +310,8 @@ void WgRenderDataShape::updateMeshes(WgContext& context, const RenderShape &rsha
     auto pbuff = pool->reqVertexBuffer(scale);
 
     pbuff->decodePath(rshape, true, [&](const WgVertexBuffer& path_buff) {
-        appendShape(context, path_buff);
-        if ((rshape.stroke) && (rshape.stroke->width > 0)) proceedStrokes(context, rshape.stroke, path_buff, pool);
+        appendShape(path_buff);
+        if ((rshape.stroke) && (rshape.stroke->width > 0)) proceedStrokes(rshape.stroke, path_buff, pool);
     }, rshape.trimpath());
 
     // update shapes bbox (with empty path handling)
@@ -384,31 +319,31 @@ void WgRenderDataShape::updateMeshes(WgContext& context, const RenderShape &rsha
         (this->meshGroupStrokesBBox.meshes.count > 0)) {
         updateAABB(tr);
     } else aabb = {{0, 0}, {0, 0}};
-    meshDataBBox.bbox(context, pMin, pMax);
+    meshDataBBox.bbox(pMin, pMax);
 
     pool->retVertexBuffer(pbuff);
 }
 
 
-void WgRenderDataShape::proceedStrokes(WgContext& context, const RenderStroke* rstroke, const WgVertexBuffer& buff, WgGeometryBufferPool* pool)
+void WgRenderDataShape::proceedStrokes(const RenderStroke* rstroke, const WgVertexBuffer& buff, WgGeometryBufferPool* pool)
 {
     assert(rstroke);
     auto strokesGenerator = pool->reqIndexedVertexBuffer(buff.scale);
     if (rstroke->dash.count == 0) strokesGenerator->appendStrokes(buff, rstroke);
     else strokesGenerator->appendStrokesDashed(buff, rstroke);
 
-    appendStroke(context, *strokesGenerator);
+    appendStroke(*strokesGenerator);
 
     pool->retIndexedVertexBuffer(strokesGenerator);
 }
 
 
-void WgRenderDataShape::releaseMeshes(WgContext& context)
+void WgRenderDataShape::releaseMeshes()
 {
-    meshGroupStrokesBBox.release(context);
-    meshGroupStrokes.release(context);
-    meshGroupShapesBBox.release(context);
-    meshGroupShapes.release(context);
+    meshGroupStrokesBBox.release();
+    meshGroupStrokes.release();
+    meshGroupShapesBBox.release();
+    meshGroupShapes.release();
     pMin = {FLT_MAX, FLT_MAX};
     pMax = {0.0f, 0.0f};
     aabb = {{0, 0}, {0, 0}};
@@ -418,8 +353,7 @@ void WgRenderDataShape::releaseMeshes(WgContext& context)
 
 void WgRenderDataShape::release(WgContext& context)
 {
-    releaseMeshes(context);
-    meshDataBBox.release(context);
+    releaseMeshes();
     renderSettingsStroke.release(context);
     renderSettingsShape.release(context);
     WgRenderDataPaint::release(context);
@@ -445,10 +379,10 @@ WgRenderDataShape* WgRenderDataShapePool::allocate(WgContext& context)
 
 void WgRenderDataShapePool::free(WgContext& context, WgRenderDataShape* renderData)
 {
-    renderData->meshGroupShapes.release(context);
-    renderData->meshGroupShapesBBox.release(context);
-    renderData->meshGroupStrokes.release(context);
-    renderData->meshGroupStrokesBBox.release(context);
+    renderData->meshGroupShapes.release();
+    renderData->meshGroupShapesBBox.release();
+    renderData->meshGroupStrokes.release();
+    renderData->meshGroupStrokesBBox.release();
     renderData->clips.clear();
     mPool.push(renderData);
 }
@@ -471,22 +405,16 @@ void WgRenderDataShapePool::release(WgContext& context)
 void WgRenderDataPicture::updateSurface(WgContext& context, const RenderSurface* surface)
 {
     // upoate mesh data
-    meshData.imageBox(context, surface->w, surface->h);
+    meshData.imageBox(surface->w, surface->h);
     // update texture data
     imageData.update(context, surface);
-    // update texture bind group
-    context.layouts.releaseBindGroup(bindGroupPicture);
-    bindGroupPicture = context.layouts.createBindGroupTexSampled(
-        context.samplerLinearRepeat, imageData.textureView
-    );
 }
 
 
 void WgRenderDataPicture::release(WgContext& context)
 {
-    context.layouts.releaseBindGroup(bindGroupPicture);
+    renderSettings.release(context);
     imageData.release(context);
-    meshData.release(context);
     WgRenderDataPaint::release(context);
 }
 
@@ -681,3 +609,85 @@ void WgRenderDataEffectParamsPool::release(WgContext& context)
     mPool.clear();
     mList.clear();
 }
+
+//***********************************************************************
+// WgStageBufferGeometry
+//***********************************************************************
+
+void WgStageBufferGeometry::append(WgMeshData* meshData)
+{
+    assert(meshData);
+    uint32_t vsize = meshData->vbuffer.count * sizeof(meshData->vbuffer[0]);
+    uint32_t tsize = meshData->tbuffer.count * sizeof(meshData->tbuffer[0]);
+    uint32_t isize = meshData->ibuffer.count * sizeof(meshData->ibuffer[0]);
+    vmaxcount = std::max(vmaxcount, meshData->vbuffer.count);
+    // append vertex data
+    if (vbuffer.reserved < vbuffer.count + vsize)
+        vbuffer.grow(std::max(vsize, vbuffer.reserved));
+    if (meshData->vbuffer.count > 0) {
+        meshData->voffset = vbuffer.count;
+        memcpy(vbuffer.data + vbuffer.count, meshData->vbuffer.data, vsize);
+        vbuffer.count += vsize;
+    }
+    // append tex coords data
+    if (vbuffer.reserved < vbuffer.count + tsize)
+        vbuffer.grow(std::max(tsize, vbuffer.reserved));
+    if (meshData->tbuffer.count > 0) {
+        meshData->toffset = vbuffer.count;
+        memcpy(vbuffer.data + vbuffer.count, meshData->tbuffer.data, tsize);
+        vbuffer.count += tsize;
+    }
+    // append index data
+    if (ibuffer.reserved < ibuffer.count + isize)
+        ibuffer.grow(std::max(isize, ibuffer.reserved));
+    if (meshData->ibuffer.count > 0) {
+        meshData->ioffset = ibuffer.count;
+        memcpy(ibuffer.data + ibuffer.count, meshData->ibuffer.data, isize);
+        ibuffer.count += isize;
+    }
+}
+
+
+void WgStageBufferGeometry::append(WgMeshDataGroup* meshDataGroup)
+{
+    ARRAY_FOREACH(p, meshDataGroup->meshes) append(*p);
+}
+
+
+void WgStageBufferGeometry::append(WgRenderDataShape* renderDataShape)
+{
+    append(&renderDataShape->meshGroupShapes);
+    append(&renderDataShape->meshGroupShapesBBox);
+    append(&renderDataShape->meshGroupStrokes);
+    append(&renderDataShape->meshGroupStrokesBBox);
+    append(&renderDataShape->meshDataBBox);
+}
+
+
+void WgStageBufferGeometry::append(WgRenderDataPicture* renderDataPicture)
+{
+    append(&renderDataPicture->meshData);
+}
+
+
+void WgStageBufferGeometry::release(WgContext& context)
+{
+    context.releaseBuffer(vbuffer_gpu);
+    context.releaseBuffer(ibuffer_gpu);
+}
+
+
+void WgStageBufferGeometry::clear()
+{
+    vbuffer.clear();
+    ibuffer.clear();
+    vmaxcount = 0;
+}
+
+
+void WgStageBufferGeometry::flush(WgContext& context) 
+{
+    context.allocateBufferVertex(vbuffer_gpu, (float *)vbuffer.data, vbuffer.count);
+    context.allocateBufferIndex(ibuffer_gpu, (uint32_t *)ibuffer.data, ibuffer.count);
+    context.allocateBufferIndexFan(vmaxcount);
+}

src/renderer/wg_engine/tvgWgRenderData.h

@@ -28,49 +28,36 @@
 #include "tvgWgShaderTypes.h"
 
 struct WgMeshData {
-    WGPUBuffer bufferPosition{};
-    WGPUBuffer bufferTexCoord{};
-    WGPUBuffer bufferIndex{};
-    size_t vertexCount{};
-    size_t indexCount{};
+    Array<Point> vbuffer;
+    Array<Point> tbuffer;
+    Array<uint32_t> ibuffer;
+    size_t voffset{};
+    size_t toffset{};
+    size_t ioffset{};
 
-    void draw(WgContext& context, WGPURenderPassEncoder renderPassEncoder);
-    void drawFan(WgContext& context, WGPURenderPassEncoder renderPassEncoder);
-    void drawImage(WgContext& context, WGPURenderPassEncoder renderPassEncoder);
-
-    void update(WgContext& context, const WgVertexBuffer& vertexBuffer);
-    void update(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd);
-    void bbox(WgContext& context, const Point pmin, const Point pmax);
-    void imageBox(WgContext& context, float w, float h);
-    void blitBox(WgContext& context);
-    void release(WgContext& context);
-};
-
-class WgMeshDataPool {
-private:
-    Array<WgMeshData*> mPool;
-    Array<WgMeshData*> mList;
-public:
-    static WgMeshDataPool* gMeshDataPool;
-    WgMeshData* allocate(WgContext& context);
-    void free(WgContext& context, WgMeshData* meshData);
-    void release(WgContext& context);
+    void update(const WgVertexBuffer& vertexBuffer);
+    void update(const WgIndexedVertexBuffer& vertexBufferInd);
+    void bbox(const Point pmin, const Point pmax);
+    void imageBox(float w, float h);
+    void blitBox();
 };
 
 struct WgMeshDataGroup {
     Array<WgMeshData*> meshes{};
     
-    void append(WgContext& context, const WgVertexBuffer& vertexBuffer);
-    void append(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd);
-    void append(WgContext& context, const Point pmin, const Point pmax);
-    void release(WgContext& context);
+    void append(const WgVertexBuffer& vertexBuffer);
+    void append(const WgIndexedVertexBuffer& vertexBufferInd);
+    void append(const Point pmin, const Point pmax);
+    void release();
 };
 
 struct WgImageData {
     WGPUTexture texture{};
     WGPUTextureView textureView{};
+    WGPUBindGroup bindGroup{};
 
     void update(WgContext& context, const RenderSurface* surface);
+    void update(WgContext& context, const Fill* fill);
     void release(WgContext& context);
 };
 
@@ -79,37 +66,29 @@ enum class WgRenderRasterType { Solid = 0, Gradient, Image };
 
 struct WgRenderSettings
 {
-    WGPUBuffer bufferGroupSolid{};
-    WGPUBindGroup bindGroupSolid{};
-    WGPUTexture texGradient{};
-    WGPUTextureView texViewGradient{};
-    WGPUBuffer bufferGroupGradient{};
-    WGPUBuffer bufferGroupTransfromGrad{};
-    WGPUBindGroup bindGroupGradient{};
+    uint32_t bindGroupInd{};
+    WgShaderTypePaintSettings settings;
+    WgImageData gradientData;
     WgRenderSettingsType fillType{};
     WgRenderRasterType rasterType{};
     bool skip{};
 
-    void updateFill(WgContext& context, const Fill* fill);
-    void updateColor(WgContext& context, const RenderColor& c);
+    void update(WgContext& context, const tvg::Matrix& transform, tvg::ColorSpace cs, uint8_t opacity);
+    void update(WgContext& context, const Fill* fill);
+    void update(WgContext& context, const RenderColor& c);
     void release(WgContext& context);
 };
 
 struct WgRenderDataPaint
 {
-    WGPUBuffer bufferModelMat{};
-    WGPUBuffer bufferBlendSettings{};
-    WGPUBindGroup bindGroupPaint{};
-    RenderRegion viewport{};
     BBox aabb{{},{}};
-    float opacity{};
+    RenderRegion viewport{};
     Array<WgRenderDataPaint*> clips;
 
     virtual ~WgRenderDataPaint() {};
     virtual void release(WgContext& context);
     virtual Type type() { return Type::Undefined; };
 
-    void update(WgContext& context, const tvg::Matrix& transform, tvg::ColorSpace cs, uint8_t opacity);
     void updateClips(tvg::Array<tvg::RenderData> &clips);
 };
 
@@ -127,13 +106,13 @@ struct WgRenderDataShape: public WgRenderDataPaint
     bool strokeFirst{};
     FillRule fillRule{};
 
-    void appendShape(WgContext& context, const WgVertexBuffer& vertexBuffer);
-    void appendStroke(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd);
+    void appendShape(const WgVertexBuffer& vertexBuffer);
+    void appendStroke(const WgIndexedVertexBuffer& vertexBufferInd);
     void updateBBox(Point pmin, Point pmax);
     void updateAABB(const Matrix& tr);
-    void updateMeshes(WgContext& context, const RenderShape& rshape, const Matrix& tr, WgGeometryBufferPool* pool);
-    void proceedStrokes(WgContext& context, const RenderStroke* rstroke, const WgVertexBuffer& buff, WgGeometryBufferPool* pool);
-    void releaseMeshes(WgContext& context);
+    void updateMeshes(const RenderShape& rshape, const Matrix& tr, WgGeometryBufferPool* pool);
+    void proceedStrokes(const RenderStroke* rstroke, const WgVertexBuffer& buff, WgGeometryBufferPool* pool);
+    void releaseMeshes();
     void release(WgContext& context) override;
     Type type() override { return Type::Shape; };
 };
@@ -150,7 +129,7 @@ public:
 
 struct WgRenderDataPicture: public WgRenderDataPaint
 {
-    WGPUBindGroup bindGroupPicture{};
+    WgRenderSettings renderSettings{};
     WgImageData imageData{};
     WgMeshData meshData{};
 
@@ -224,4 +203,69 @@ public:
     void release(WgContext& context);
 };
 
+class WgStageBufferGeometry {
+private:
+    Array<uint8_t> vbuffer;
+    Array<uint8_t> ibuffer;
+    uint32_t vmaxcount{};
+public:
+    WGPUBuffer vbuffer_gpu{};
+    WGPUBuffer ibuffer_gpu{};
+
+    void append(WgMeshData* meshData);
+    void append(WgMeshDataGroup* meshDataGroup);
+    void append(WgRenderDataShape* renderDataShape);
+    void append(WgRenderDataPicture* renderDataPicture);
+    void initialize(WgContext& context){};
+    void release(WgContext& context);
+    void clear();
+    void flush(WgContext& context);
+};
+
+// typed uniform stage buffer with related bind groups handling
+template<typename T>
+class WgStageBufferUniform {
+private:
+    Array<T> ubuffer;
+    WGPUBuffer ubuffer_gpu{};
+    Array<WGPUBindGroup> bbuffer;
+public:
+    // append uniform data to cpu staged buffer and return related bind group index
+    uint32_t append(const T& value) {
+        ubuffer.push(value);
+        return ubuffer.count - 1;
+    }
+
+    void flush(WgContext& context) {
+        // flush data to gpu buffer from cpu memory including reserved data to prevent future gpu buffer reallocations
+        bool bufferChanged = context.allocateBufferUniform(ubuffer_gpu, (void*)ubuffer.data, ubuffer.reserved*sizeof(T));
+        // if gpu buffer handle was changed we must to remove all created binding groups
+        if (bufferChanged) releaseBindGroups(context);
+        // allocate bind groups for all new data items
+        for (uint32_t i = bbuffer.count; i < ubuffer.count; i++)
+            bbuffer.push(context.layouts.createBindGroupBuffer1Un(ubuffer_gpu, i*sizeof(T), sizeof(T)));
+        assert(bbuffer.count >= ubuffer.count);
+    }
+
+    // please, use index that was returned from append method
+    WGPUBindGroup operator[](const uint32_t index) const {
+        return bbuffer[index];
+    }
+
+    void clear() {
+        ubuffer.clear();
+    }
+
+    void release(WgContext& context) {
+        context.releaseBuffer(ubuffer_gpu);
+        releaseBindGroups(context);
+    }
+
+    void releaseBindGroups(WgContext& context) {
+        ARRAY_FOREACH(p, bbuffer)
+            context.layouts.releaseBindGroup(*p);
+        bbuffer.clear();
+    }
+};
+
 #endif // _TVG_WG_RENDER_DATA_H_

src/renderer/wg_engine/tvgWgRenderTarget.cpp

@@ -22,7 +22,7 @@
 
 #include "tvgWgRenderTarget.h"
 
-void WgRenderStorage::initialize(WgContext& context, uint32_t width, uint32_t height)
+void WgRenderTarget::initialize(WgContext& context, uint32_t width, uint32_t height)
 {
     this->width = width;
     this->height = height;
@@ -36,7 +36,7 @@ void WgRenderStorage::initialize(WgContext& context, uint32_t width, uint32_t he
 }
 
 
-void WgRenderStorage::release(WgContext& context)
+void WgRenderTarget::release(WgContext& context)
 {
     context.layouts.releaseBindGroup(bindGroupTexure);
     context.layouts.releaseBindGroup(bindGroupWrite);
@@ -50,38 +50,38 @@ void WgRenderStorage::release(WgContext& context)
 }
 
 //*****************************************************************************
-// render storage pool
+// render target pool
 //*****************************************************************************
 
-WgRenderStorage* WgRenderStoragePool::allocate(WgContext& context)
+WgRenderTarget* WgRenderTargetPool::allocate(WgContext& context)
 {
-    WgRenderStorage* renderStorage{};
+    WgRenderTarget* renderTarget{};
     if (pool.count > 0) {
-        renderStorage = pool.last();
+        renderTarget = pool.last();
         pool.pop();
     } else {
-        renderStorage = new WgRenderStorage;
-        renderStorage->initialize(context, width, height);
-        list.push(renderStorage);
+        renderTarget = new WgRenderTarget;
+        renderTarget->initialize(context, width, height);
+        list.push(renderTarget);
     }
-    return renderStorage;
+    return renderTarget;
 };
 
 
-void WgRenderStoragePool::free(WgContext& context, WgRenderStorage* renderStorage)
+void WgRenderTargetPool::free(WgContext& context, WgRenderTarget* renderTarget)
 {
-    pool.push(renderStorage);
+    pool.push(renderTarget);
 };
 
 
-void WgRenderStoragePool::initialize(WgContext& context, uint32_t width, uint32_t height)
+void WgRenderTargetPool::initialize(WgContext& context, uint32_t width, uint32_t height)
 {
     this->width = width;
     this->height = height;
 }
 
 
-void WgRenderStoragePool::release(WgContext& context)
+void WgRenderTargetPool::release(WgContext& context)
 {
     ARRAY_FOREACH(p, list) {
        (*p)->release(context);

src/renderer/wg_engine/tvgWgRenderTarget.h

@@ -26,7 +26,7 @@
 #include "tvgWgPipelines.h"
 #include "tvgRender.h"
 
-struct WgRenderStorage {
+struct WgRenderTarget {
     WGPUTexture texture{};
     WGPUTexture textureMS{};
     WGPUTextureView texView{};
@@ -42,15 +42,15 @@ struct WgRenderStorage {
 };
 
 
-class WgRenderStoragePool {
+class WgRenderTargetPool {
 private:
-    Array<WgRenderStorage*> list;
-    Array<WgRenderStorage*> pool;
+    Array<WgRenderTarget*> list;
+    Array<WgRenderTarget*> pool;
     uint32_t width{};
     uint32_t height{};
 public:
-    WgRenderStorage* allocate(WgContext& context);
-    void free(WgContext& context, WgRenderStorage* renderTarget);
+    WgRenderTarget* allocate(WgContext& context);
+    void free(WgContext& context, WgRenderTarget* renderTarget);
 
     void initialize(WgContext& context, uint32_t width, uint32_t height);
     void release(WgContext& context);

src/renderer/wg_engine/tvgWgRenderTask.cpp

@@ -0,0 +1,93 @@
+/*
+ * Copyright (c) 2025 the ThorVG project. All rights reserved.
+
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "tvgWgRenderTask.h"
+#include <iostream>
+
+//***********************************************************************
+// WgPaintTask
+//***********************************************************************
+
+void WgPaintTask::run(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder)
+{
+    if (renderData->type() == tvg::Type::Shape)
+        compositor.renderShape(context, (WgRenderDataShape*)renderData, blendMethod);
+    if (renderData->type() == tvg::Type::Picture)
+        compositor.renderImage(context, (WgRenderDataPicture*)renderData, blendMethod);
+    else assert(true);
+}
+
+//***********************************************************************
+// WgSceneTask
+//***********************************************************************
+
+void WgSceneTask::run(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder)
+{
+    // begin the render pass for the current scene and clear the target content
+    WGPUColor color{};
+    if ((compose->method == MaskMethod::None) && (compose->blend != BlendMethod::Normal)) color = { 1.0, 1.0, 1.0, 0.0 };
+    compositor.beginRenderPass(encoder, renderTarget, true, color);
+    // run all childs (scenes and shapes)
+    runChildren(context, compositor, encoder);
+    // we must to end current render pass for current scene
+    compositor.endRenderPass();
+    // we must to apply effect for current scene
+    if (effect)
+        runEffect(context, compositor, encoder);
+    // there's no point in continuing if the scene has no destination target (e.g., the root scene)
+    if (!renderTargetDst) return;
+    // apply scene blending
+    if (compose->method == MaskMethod::None) {
+        compositor.beginRenderPass(encoder, renderTargetDst, false);
+        compositor.renderScene(context, renderTarget, compose);
+    // apply scene composition (for scenes, that have a handle to mask)
+    } else if (renderTargetMsk) {
+        compositor.beginRenderPass(encoder, renderTargetDst, false);
+        compositor.composeScene(context, renderTarget, renderTargetMsk, compose);
+    }
+}
+
+
+void WgSceneTask::runChildren(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder)
+{
+    ARRAY_FOREACH(task, children) {
+        WgRenderTask* renderTask = *task;
+        // we need to restore current render pass without clear
+        compositor.beginRenderPass(encoder, renderTarget, false);
+        // run children (shape or scene)
+        renderTask->run(context, compositor, encoder);
+    }
+}
+
+
+void WgSceneTask::runEffect(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder)
+{
+    assert(effect);
+    switch (effect->type) {
+        case SceneEffect::GaussianBlur: compositor.gaussianBlur(context, renderTarget, (RenderEffectGaussianBlur*)effect, compose); break;
+        case SceneEffect::DropShadow: compositor.dropShadow(context, renderTarget, (RenderEffectDropShadow*)effect, compose); break;
+        case SceneEffect::Fill: compositor.fillEffect(context, renderTarget, (RenderEffectFill*)effect, compose); break;
+        case SceneEffect::Tint: compositor.tintEffect(context, renderTarget, (RenderEffectTint*)effect, compose); break;
+        case SceneEffect::Tritone : compositor.tritoneEffect(context, renderTarget, (RenderEffectTritone*)effect, compose); break;
+        default: break;
+    }
+}

src/renderer/wg_engine/tvgWgRenderTask.h

@@ -0,0 +1,72 @@
+/*
+ * Copyright (c) 2025 the ThorVG project. All rights reserved.
+
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _TVG_WG_RENDER_TASK_H_
+#define _TVG_WG_RENDER_TASK_H_
+ 
+#include "tvgWgCompositor.h"
+
+// base class for any renderable objects 
+struct WgRenderTask {
+    virtual ~WgRenderTask() {}
+    virtual void run(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder) = 0;
+};
+
+// task for sinlge shape rendering
+struct WgPaintTask: public WgRenderTask {
+    // shape render properties
+    WgRenderDataPaint* renderData{};
+    BlendMethod blendMethod{};
+
+    WgPaintTask(WgRenderDataPaint* renderData, BlendMethod blendMethod) : 
+        renderData(renderData), blendMethod(blendMethod) {}
+    // apply shape execution, including custom blending and clipping
+    void run(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder) override;
+};
+
+// task for scene rendering with blending, composition and effect
+struct WgSceneTask: public WgRenderTask {
+public:
+    // parent scene (nullptr for root scene)
+    WgSceneTask* parent{};
+    // childs can be shapes or scenes tesks
+    Array<WgRenderTask*> children;
+    // scene blend/compose targets
+    WgRenderTarget* renderTarget{};
+    WgRenderTarget* renderTargetMsk{};
+    WgRenderTarget* renderTargetDst{};
+    // scene blend/compose properties
+    WgCompose* compose{};
+    // scene effect properties
+    const RenderEffect* effect{};
+
+    WgSceneTask(WgRenderTarget* renderTarget, WgCompose* compose, WgSceneTask* parent) :
+        parent(parent), renderTarget(renderTarget), compose(compose) {}
+    // run all, including all shapes drawing, blending, composition and effect
+    void run(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder) override;
+private:
+    void runChildren(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder);
+    void runEffect(WgContext& context, WgCompositor& compositor, WGPUCommandEncoder encoder);
+};
+ 
+ #endif // _TVG_WG_RENDER_TASK_H_
+ 

src/renderer/wg_engine/tvgWgRenderer.cpp

@@ -34,10 +34,8 @@ static atomic<int32_t> rendererCnt{-1};
 
 void WgRenderer::release()
 {
-    // check for general context availability
     if (!mContext.queue) return;
 
-    // dispose stored objects
     disposeObjects();
 
     // clear render data paint pools
@@ -45,15 +43,14 @@ void WgRenderer::release()
     mRenderDataPicturePool.release(mContext);
     mRenderDataViewportPool.release(mContext);
     mRenderDataEffectParamsPool.release(mContext);
-    WgMeshDataPool::gMeshDataPool->release(mContext);
 
-    // clear render storage pool
-    mRenderStoragePool.release(mContext);
+    // clear render  pool
+    mRenderTargetPool.release(mContext);
 
     // clear rendering tree stacks
-    mCompositorStack.clear();
-    mRenderStorageStack.clear();
-    mRenderStorageRoot.release(mContext);
+    mCompositorList.clear();
+    mRenderTargetStack.clear();
+    mRenderTargetRoot.release(mContext);
 
     // release context handles
     mCompositor.release(mContext);
@@ -102,7 +99,6 @@ void WgRenderer::clearTargets()
 
 bool WgRenderer::surfaceConfigure(WGPUSurface surface, WgContext& context, uint32_t width, uint32_t height)
 {
-    // store target surface properties
     this->surface = surface;
     if (width == 0 || height == 0 || !surface) return false;
 
@@ -135,33 +131,29 @@ bool WgRenderer::surfaceConfigure(WGPUSurface surface, WgContext& context, uint3
 
 RenderData WgRenderer::prepare(const RenderShape& rshape, RenderData data, const Matrix& transform, Array<RenderData>& clips, uint8_t opacity, RenderUpdateFlag flags, bool clipper)
 {
-    // get or create render data shape
-    auto renderDataShape = (WgRenderDataShape*)data;
-    if (!renderDataShape)
-        renderDataShape = mRenderDataShapePool.allocate(mContext);
+    auto renderDataShape = data ? (WgRenderDataShape*)data : mRenderDataShapePool.allocate(mContext);
 
     // update geometry
     if ((!data) || (flags & (RenderUpdateFlag::Path | RenderUpdateFlag::Stroke))) {
-        renderDataShape->updateMeshes(mContext, rshape, transform, mBufferPool.pool);
+        renderDataShape->updateMeshes(rshape, transform, mBufferPool.pool);
     }
 
     // update paint settings
     if ((!data) || (flags & (RenderUpdateFlag::Transform | RenderUpdateFlag::Blend))) {
-        renderDataShape->update(mContext, transform, mTargetSurface.cs, opacity);
+        renderDataShape->renderSettingsShape.update(mContext, transform, mTargetSurface.cs, opacity);
+        renderDataShape->renderSettingsStroke.update(mContext, transform, mTargetSurface.cs, opacity);
         renderDataShape->fillRule = rshape.rule;
     }
 
     // setup fill settings
     renderDataShape->viewport = mViewport;
-    renderDataShape->opacity = opacity;
-    if (flags & RenderUpdateFlag::Gradient && rshape.fill) renderDataShape->renderSettingsShape.updateFill(mContext, rshape.fill);
-    else if (flags & RenderUpdateFlag::Color) renderDataShape->renderSettingsShape.updateColor(mContext, rshape.color);
+    if (flags & RenderUpdateFlag::Gradient && rshape.fill) renderDataShape->renderSettingsShape.update(mContext, rshape.fill);
+    else if (flags & RenderUpdateFlag::Color) renderDataShape->renderSettingsShape.update(mContext, rshape.color);
     if (rshape.stroke) {
-        if (flags & RenderUpdateFlag::GradientStroke && rshape.stroke->fill) renderDataShape->renderSettingsStroke.updateFill(mContext, rshape.stroke->fill);
-        else if (flags & RenderUpdateFlag::Stroke) renderDataShape->renderSettingsStroke.updateColor(mContext, rshape.stroke->color);
+        if (flags & RenderUpdateFlag::GradientStroke && rshape.stroke->fill) renderDataShape->renderSettingsStroke.update(mContext, rshape.stroke->fill);
+        else if (flags & RenderUpdateFlag::Stroke) renderDataShape->renderSettingsStroke.update(mContext, rshape.stroke->color);
     }
 
-    // store clips data
     renderDataShape->updateClips(clips);
 
     return renderDataShape;
@@ -170,16 +162,12 @@ RenderData WgRenderer::prepare(const RenderShape& rshape, RenderData data, const
 
 RenderData WgRenderer::prepare(RenderSurface* surface, RenderData data, const Matrix& transform, Array<RenderData>& clips, uint8_t opacity, RenderUpdateFlag flags)
 {
-    // get or create render data shape
-    auto renderDataPicture = (WgRenderDataPicture*)data;
-    if (!renderDataPicture)
-        renderDataPicture = mRenderDataPicturePool.allocate(mContext);
+    auto renderDataPicture = data ? (WgRenderDataPicture*)data : mRenderDataPicturePool.allocate(mContext);
 
     // update paint settings
     renderDataPicture->viewport = mViewport;
-    renderDataPicture->opacity = opacity;
     if (flags & (RenderUpdateFlag::Transform | RenderUpdateFlag::Blend)) {
-        renderDataPicture->update(mContext, transform, surface->cs, opacity);
+        renderDataPicture->renderSettings.update(mContext, transform, surface->cs, opacity);
     }
 
     // update image data
@@ -187,58 +175,87 @@ RenderData WgRenderer::prepare(RenderSurface* surface, RenderData data, const Ma
         renderDataPicture->updateSurface(mContext, surface);
     }
 
-    // store clips data
     renderDataPicture->updateClips(clips);
 
     return renderDataPicture;
 }
 
+
 bool WgRenderer::preRender()
 {
-    // push rot render storage to the render tree stack
-    assert(mRenderStorageStack.count == 0);
-    mRenderStorageStack.push(&mRenderStorageRoot);
-    // create command encoder for drawing
-    WGPUCommandEncoderDescriptor commandEncoderDesc{};
-    mCommandEncoder = wgpuDeviceCreateCommandEncoder(mContext.device, &commandEncoderDesc);
-    // start root render pass
-    mCompositor.beginRenderPass(mCommandEncoder, mRenderStorageStack.last(), true);
+    if (mContext.invalid()) return false;
+
+    mCompositor.reset(mContext);
+
+    assert(mRenderTargetStack.count == 0);
+    mRenderTargetStack.push(&mRenderTargetRoot);
+
+    // create root compose settings
+    WgCompose* compose = new WgCompose();
+    compose->aabb = { { 0, 0 }, { (int32_t)mTargetSurface.w, (int32_t)mTargetSurface.h } };
+    compose->blend = BlendMethod::Normal;
+    compose->method = MaskMethod::None;
+    compose->opacity = 255;
+    mCompositorList.push(compose);
+
+    // create root scene
+    WgSceneTask* sceneTask = new WgSceneTask(&mRenderTargetRoot, compose, nullptr);
+    mRenderTaskList.push(sceneTask);
+    mSceneTaskStack.push(sceneTask);
     return true;
 }
 
 
 bool WgRenderer::renderShape(RenderData data)
 {
-    // temporary simple render data to the current render target
-    mCompositor.renderShape(mContext, (WgRenderDataShape*)data, mBlendMethod);
+    WgPaintTask* paintTask = new WgPaintTask((WgRenderDataPaint*)data, mBlendMethod);
+    WgSceneTask* sceneTask = mSceneTaskStack.last();
+    sceneTask->children.push(paintTask);
+    mRenderTaskList.push(paintTask);
+    mCompositor.requestShape((WgRenderDataShape*)data);
     return true;
 }
 
 
 bool WgRenderer::renderImage(RenderData data)
 {
-    // temporary simple render data to the current render target
-    mCompositor.renderImage(mContext, (WgRenderDataPicture*)data, mBlendMethod);
+    WgPaintTask* paintTask = new WgPaintTask((WgRenderDataPaint*)data, mBlendMethod);
+    WgSceneTask* sceneTask = mSceneTaskStack.last();
+    sceneTask->children.push(paintTask);
+    mRenderTaskList.push(paintTask);
+    mCompositor.requestImage((WgRenderDataPicture*)data);
     return true;
 }
 
 
 bool WgRenderer::postRender()
 {
-    // end root render pass
-    mCompositor.endRenderPass();
-    // release command encoder
-    const WGPUCommandBufferDescriptor commandBufferDesc{};
-    WGPUCommandBuffer commandsBuffer = wgpuCommandEncoderFinish(mCommandEncoder, &commandBufferDesc);
-    wgpuQueueSubmit(mContext.queue, 1, &commandsBuffer);
-    wgpuCommandBufferRelease(commandsBuffer);
-    wgpuCommandEncoderRelease(mCommandEncoder);
-    // pop root render storage to the render tree stack
-    mRenderStorageStack.pop();
-    assert(mRenderStorageStack.count == 0);
-    // clear viewport list and store allocated handles to pool
-    ARRAY_FOREACH(p, mRenderDataViewportList)
+    // flush stage data to gpu
+    mCompositor.flush(mContext);
+
+    // create command encoder for drawing
+    WGPUCommandEncoder commandEncoder = mContext.createCommandEncoder();
+
+    // run rendering (all the fun is here)
+    WgSceneTask* sceneTaskRoot = mSceneTaskStack.last();
+    sceneTaskRoot->run(mContext, mCompositor, commandEncoder);
+
+    // execute and release command encoder
+    mContext.submitCommandEncoder(commandEncoder);
+    mContext.releaseCommandEncoder(commandEncoder);
+
+    // clear the render tasks tree
+    mSceneTaskStack.pop();
+    assert(mSceneTaskStack.count == 0);
+    mRenderTargetStack.pop();
+    assert(mRenderTargetStack.count == 0);
+    ARRAY_FOREACH(p, mRenderTaskList) { delete (*p); };
+    mRenderTaskList.clear();
+    ARRAY_FOREACH(p, mCompositorList) { delete (*p); };
+    mCompositorList.clear();
+    ARRAY_FOREACH(p, mRenderDataViewportList) {
         mRenderDataViewportPool.free(mContext, *p);
+    }
     mRenderDataViewportList.clear();
     return true;
 }
@@ -257,14 +274,9 @@ RenderRegion WgRenderer::region(RenderData data)
     if (renderData->type() == Type::Shape) {
         auto& v1 = renderData->aabb.min;
         auto& v2 = renderData->aabb.max;
-        RenderRegion renderRegion;
-        renderRegion.x = static_cast<int32_t>(nearbyint(v1.x));
-        renderRegion.y = static_cast<int32_t>(nearbyint(v1.y));
-        renderRegion.w = static_cast<int32_t>(nearbyint(v2.x)) - renderRegion.x;
-        renderRegion.h = static_cast<int32_t>(nearbyint(v2.y)) - renderRegion.y;
-        return renderRegion;
+        return {{int32_t(nearbyint(v1.x)), int32_t(nearbyint(v1.y))}, {int32_t(nearbyint(v2.x)), int32_t(nearbyint(v2.y))}};
     }
-    return { 0, 0, (int32_t)mTargetSurface.w, (int32_t)mTargetSurface.h };
+    return {{0, 0}, {(int32_t)mTargetSurface.w, (int32_t)mTargetSurface.h}};
 }
 
 
@@ -301,6 +313,8 @@ const RenderSurface* WgRenderer::mainSurface()
 
 bool WgRenderer::clear()
 {
+    if (mContext.invalid()) return false;
+
     //TODO: clear the current target buffer only if clear() is called
     return true;
 }
@@ -308,6 +322,8 @@ bool WgRenderer::clear()
 
 bool WgRenderer::sync()
 {
+    if (mContext.invalid()) return false;
+
     disposeObjects();
 
     // if texture buffer used
@@ -317,57 +333,45 @@ bool WgRenderer::sync()
         wgpuSurfaceGetCurrentTexture(surface, &surfaceTexture);
         dstTexture = surfaceTexture.texture;
     }
-    // there is no external dest buffer
+
     if (!dstTexture) return false;
 
-    // get external dest buffer
+    // insure that surface and offscreen target have the same size
+    if ((wgpuTextureGetWidth(dstTexture) == mRenderTargetRoot.width) && 
+        (wgpuTextureGetHeight(dstTexture) == mRenderTargetRoot.height)) {
         WGPUTextureView dstTextureView = mContext.createTextureView(dstTexture);
-
-    // create command encoder
-    const WGPUCommandEncoderDescriptor commandEncoderDesc{};
-    WGPUCommandEncoder commandEncoder = wgpuDeviceCreateCommandEncoder(mContext.device, &commandEncoderDesc);
-
+        WGPUCommandEncoder commandEncoder = mContext.createCommandEncoder();
         // show root offscreen buffer
-    mCompositor.blit(mContext, commandEncoder, &mRenderStorageRoot, dstTextureView);
-
-    // release command encoder
-    const WGPUCommandBufferDescriptor commandBufferDesc{};
-    WGPUCommandBuffer commandsBuffer = wgpuCommandEncoderFinish(commandEncoder, &commandBufferDesc);
-    wgpuQueueSubmit(mContext.queue, 1, &commandsBuffer);
-    wgpuCommandBufferRelease(commandsBuffer);
-    wgpuCommandEncoderRelease(commandEncoder);
-
-    // release dest buffer view
+        mCompositor.blit(mContext, commandEncoder, &mRenderTargetRoot, dstTextureView);
+        mContext.submitCommandEncoder(commandEncoder);
+        mContext.releaseCommandEncoder(commandEncoder);
         mContext.releaseTextureView(dstTextureView);
+    }
+
     return true;
 }
 
 
-// render target handle
 bool WgRenderer::target(WGPUDevice device, WGPUInstance instance, void* target, uint32_t width, uint32_t height, int type)
 {
     // release all existing handles
     if (!instance || !device || !target) {
-        // release all handles
         release();
         return true;
     }
 
-    // can not initialize renderer, give up
-    if (!instance || !device || !target || !width || !height)
-        return false;
+    if (!width || !height) return false;
 
     // device or instance was changed, need to recreate all instances
     if ((mContext.device != device) || (mContext.instance != instance)) {
-        // release all handles
         release();
 
         // initialize base rendering handles
         mContext.initialize(instance, device);
 
         // initialize render tree instances
-        mRenderStoragePool.initialize(mContext, width, height);
-        mRenderStorageRoot.initialize(mContext, width, height);
+        mRenderTargetPool.initialize(mContext, width, height);
+        mRenderTargetRoot.initialize(mContext, width, height);
         mCompositor.initialize(mContext, width, height);
 
         // store target properties
@@ -386,12 +390,12 @@ bool WgRenderer::target(WGPUDevice device, WGPUInstance instance, void* target,
     // update render targets dimentions
     if ((mTargetSurface.w != width) || (mTargetSurface.h != height) || (type == 0 ? (surface != (WGPUSurface)target) : (targetTexture != (WGPUTexture)target))) {
         // release render tagets
-        mRenderStoragePool.release(mContext);
-        mRenderStorageRoot.release(mContext);
+        mRenderTargetPool.release(mContext);
+        mRenderTargetRoot.release(mContext);
         clearTargets();
 
-        mRenderStoragePool.initialize(mContext, width, height);
-        mRenderStorageRoot.initialize(mContext, width, height);
+        mRenderTargetPool.initialize(mContext, width, height);
+        mRenderTargetRoot.initialize(mContext, width, height);
         mCompositor.resize(mContext, width, height);
 
         // store target properties
@@ -446,7 +450,7 @@ RenderCompositor* WgRenderer::target(const RenderRegion& region, TVG_UNUSED Colo
         compose->rdViewport->update(mContext, region);
         mRenderDataViewportList.push(compose->rdViewport);
     }
-    mCompositorStack.push(compose);
+    mCompositorList.push(compose);
     return compose;
 }
 
@@ -458,145 +462,101 @@ bool WgRenderer::beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_
     compose->method = method;
     compose->opacity = opacity;
     compose->blend = mBlendMethod;
-    // end current render pass
-    mCompositor.endRenderPass();
-    // allocate new render storage and push to the render tree stack
-    WgRenderStorage* storage = mRenderStoragePool.allocate(mContext);
-    mRenderStorageStack.push(storage);
-    // begin newly added render pass
-    WGPUColor color{};
-    if ((compose->method == MaskMethod::None) && (compose->blend != BlendMethod::Normal)) color = { 1.0, 1.0, 1.0, 0.0 };
-    mCompositor.beginRenderPass(mCommandEncoder, mRenderStorageStack.last(), true, color);
+    WgSceneTask* sceneTaskCurrent = mSceneTaskStack.last();
+    // allocate new render target and push to the render tree stack
+    WgRenderTarget* renderTarget = mRenderTargetPool.allocate(mContext);
+    mRenderTargetStack.push(renderTarget);
+    // create and setup new scene task
+    WgSceneTask* sceneTask = new WgSceneTask(renderTarget, compose, sceneTaskCurrent);
+    sceneTaskCurrent->children.push(sceneTask);
+    mRenderTaskList.push(sceneTask);
+    mSceneTaskStack.push(sceneTask);
     return true;
 }
 
 
 bool WgRenderer::endComposite(RenderCompositor* cmp)
 {
-    // get current composition settings
-    WgCompose* comp = (WgCompose*)cmp;
-    // we must to end current render pass to run blend/composition mechanics
-    mCompositor.endRenderPass();
     // finish scene blending
-    if (comp->method == MaskMethod::None) {
-        // get source and destination render storages
-        WgRenderStorage* src = mRenderStorageStack.last();
-        mRenderStorageStack.pop();
-        WgRenderStorage* dst = mRenderStorageStack.last();
-        // begin previous render pass
-        mCompositor.beginRenderPass(mCommandEncoder, dst, false);
-        // apply composition
-        mCompositor.renderScene(mContext, src, comp);
+    if (cmp->method == MaskMethod::None) {
+        // get source and destination render targets
+        WgRenderTarget* src = mRenderTargetStack.last();
+        mRenderTargetStack.pop();
+
+        WgSceneTask* srcScene = mSceneTaskStack.last();
+        mSceneTaskStack.pop();
+        // setup render target compose destitations
+        srcScene->renderTargetDst = mSceneTaskStack.last()->renderTarget;
+        srcScene->renderTargetMsk =  nullptr;
         // back render targets to the pool
-        mRenderStoragePool.free(mContext, src);
-    } else { // finish composition
-        // get source, mask and destination render storages
-        WgRenderStorage* src = mRenderStorageStack.last();
-        mRenderStorageStack.pop();
-        WgRenderStorage* msk = mRenderStorageStack.last();
-        mRenderStorageStack.pop();
-        WgRenderStorage* dst = mRenderStorageStack.last();
-        // begin previous render pass
-        mCompositor.beginRenderPass(mCommandEncoder, dst, false);
-        // apply composition
-        mCompositor.composeScene(mContext, src, msk, comp);
+        mRenderTargetPool.free(mContext, src);
+    } else { // finish scene composition
+        // get source, mask and destination render targets
+        WgRenderTarget* src = mRenderTargetStack.last();
+        mRenderTargetStack.pop();
+        WgRenderTarget* msk = mRenderTargetStack.last();
+        mRenderTargetStack.pop();
+        // get source and mask scenes
+        WgSceneTask* srcScene = mSceneTaskStack.last();
+        mSceneTaskStack.pop();
+        WgSceneTask* mskScene = mSceneTaskStack.last();
+        mSceneTaskStack.pop();
+        // setup render target compose destitations
+        srcScene->renderTargetDst = mSceneTaskStack.last()->renderTarget;
+        srcScene->renderTargetMsk = mskScene->renderTarget;
         // back render targets to the pool
-        mRenderStoragePool.free(mContext, src);
-        mRenderStoragePool.free(mContext, msk);
+        mRenderTargetPool.free(mContext, src);
+        mRenderTargetPool.free(mContext, msk);
     }
-
-    // delete current compositor settings
-    delete mCompositorStack.last();
-    mCompositorStack.pop();
-
     return true;
 }
 
 
 void WgRenderer::prepare(RenderEffect* effect, const Matrix& transform)
 {
-    // prepare gaussian blur data
+    if (!effect->rd) effect->rd = mRenderDataEffectParamsPool.allocate(mContext);
+    auto renderData = (WgRenderDataEffectParams*)effect->rd;
+
     if (effect->type == SceneEffect::GaussianBlur) {
-        auto renderEffect = (RenderEffectGaussianBlur*)effect;
-        auto renderData = (WgRenderDataEffectParams*)renderEffect->rd;
-        if (!renderData) {
-            renderData = mRenderDataEffectParamsPool.allocate(mContext);
-            renderEffect->rd = renderData;
+        renderData->update(mContext, (RenderEffectGaussianBlur*)effect, transform);
+    } else if (effect->type == SceneEffect::DropShadow) {
+        renderData->update(mContext, (RenderEffectDropShadow*)effect, transform);
+    } else if (effect->type == SceneEffect::Fill) {
+        renderData->update(mContext, (RenderEffectFill*)effect);
+    } else if (effect->type == SceneEffect::Tint) {
+        renderData->update(mContext, (RenderEffectTint*)effect);
+    } else if (effect->type == SceneEffect::Tritone) {
+        renderData->update(mContext, (RenderEffectTritone*)effect);
+    } else {
+        TVGERR("WG_ENGINE", "Missing effect type? = %d", (int) effect->type);
+        return;
     }
-        renderData->update(mContext, renderEffect, transform);
+
     effect->valid = true;
-    } else
-    // prepare drop shadow data
-    if (effect->type == SceneEffect::DropShadow) {
-        auto renderEffect = (RenderEffectDropShadow*)effect;
-        auto renderData = (WgRenderDataEffectParams*)renderEffect->rd;
-        if (!renderData) {
-            renderData = mRenderDataEffectParamsPool.allocate(mContext);
-            renderEffect->rd = renderData;
-        }
-        renderData->update(mContext, renderEffect, transform);
-        effect->valid = true;
-    } else
-    // prepare fill
-    if (effect->type == SceneEffect::Fill) {
-        auto renderEffect = (RenderEffectFill*)effect;
-        auto renderData = (WgRenderDataEffectParams*)renderEffect->rd;
-        if (!renderData) {
-            renderData = mRenderDataEffectParamsPool.allocate(mContext);
-            renderEffect->rd = renderData;
-        }
-        renderData->update(mContext, renderEffect);
-        effect->valid = true;
-    } else
-    // prepare tint
-    if (effect->type == SceneEffect::Tint) {
-        auto renderEffect = (RenderEffectTint*)effect;
-        auto renderData = (WgRenderDataEffectParams*)renderEffect->rd;
-        if (!renderData) {
-            renderData = mRenderDataEffectParamsPool.allocate(mContext);
-            renderEffect->rd = renderData;
-        }
-        renderData->update(mContext, renderEffect);
-        effect->valid = true;
-    } else
-    // prepare tritone
-    if (effect->type == SceneEffect::Tritone) {
-        auto renderEffect = (RenderEffectTritone*)effect;
-        auto renderData = (WgRenderDataEffectParams*)renderEffect->rd;
-        if (!renderData) {
-            renderData = mRenderDataEffectParamsPool.allocate(mContext);
-            renderEffect->rd = renderData;
-        }
-        renderData->update(mContext, renderEffect);
-        effect->valid = true;
-    }
 }
 
 
 bool WgRenderer::region(RenderEffect* effect)
 {
-    // update gaussian blur region
     if (effect->type == SceneEffect::GaussianBlur) {
         auto gaussian = (RenderEffectGaussianBlur*)effect;
         auto renderData = (WgRenderDataEffectParams*)gaussian->rd;
         if (gaussian->direction != 2) {
-            gaussian->extend.x = -renderData->extend;
-            gaussian->extend.w = +renderData->extend * 2;
+            gaussian->extend.min.x = -renderData->extend;
+            gaussian->extend.max.x = +renderData->extend;
         }
         if (gaussian->direction != 1) {
-            gaussian->extend.y = -renderData->extend;
-            gaussian->extend.h = +renderData->extend * 2;
+            gaussian->extend.min.y = -renderData->extend;
+            gaussian->extend.max.y = +renderData->extend;
         }
         return true;
-    } else
-    // update drop shadow region
-    if (effect->type == SceneEffect::DropShadow) {
+    } else if (effect->type == SceneEffect::DropShadow) {
         auto dropShadow = (RenderEffectDropShadow*)effect;
         auto renderData = (WgRenderDataEffectParams*)dropShadow->rd;
-        dropShadow->extend.x = -(renderData->extend + std::abs(renderData->offset.x));
-        dropShadow->extend.w = +(renderData->extend + std::abs(renderData->offset.x)) * 2;
-        dropShadow->extend.y = -(renderData->extend + std::abs(renderData->offset.y));
-        dropShadow->extend.h = +(renderData->extend + std::abs(renderData->offset.y)) * 2;
+        dropShadow->extend.min.x = -(renderData->extend + std::abs(renderData->offset.x));
+        dropShadow->extend.max.x = +(renderData->extend + std::abs(renderData->offset.x));
+        dropShadow->extend.min.y = -(renderData->extend + std::abs(renderData->offset.y));
+        dropShadow->extend.max.y = +(renderData->extend + std::abs(renderData->offset.y));
         return true;
     }
     return false;
@@ -605,20 +565,8 @@ bool WgRenderer::region(RenderEffect* effect)
 
 bool WgRenderer::render(RenderCompositor* cmp, const RenderEffect* effect, TVG_UNUSED bool direct)
 {
-    // we must to end current render pass to resolve ms texture before effect
-    mCompositor.endRenderPass();
-    WgCompose* comp = (WgCompose*)cmp;
-    WgRenderStorage* dst = mRenderStorageStack.last();
-
-    switch (effect->type) {
-        case SceneEffect::GaussianBlur: return mCompositor.gaussianBlur(mContext, dst, (RenderEffectGaussianBlur*)effect, comp);
-        case SceneEffect::DropShadow: return mCompositor.dropShadow(mContext, dst, (RenderEffectDropShadow*)effect, comp);
-        case SceneEffect::Fill: return mCompositor.fillEffect(mContext, dst, (RenderEffectFill*)effect, comp);
-        case SceneEffect::Tint: return mCompositor.tintEffect(mContext, dst, (RenderEffectTint*)effect, comp);
-        case SceneEffect::Tritone : return mCompositor.tritoneEffect(mContext, dst, (RenderEffectTritone*)effect, comp);
-        default: return false;
-    }
-    return false;
+    mSceneTaskStack.last()->effect = effect;
+    return true;
 }
 
 
@@ -632,6 +580,8 @@ void WgRenderer::dispose(RenderEffect* effect)
 
 bool WgRenderer::preUpdate()
 {
+    if (mContext.invalid()) return false;
+
     return true;
 }
 

src/renderer/wg_engine/tvgWgRenderer.h

@@ -23,7 +23,7 @@
 #ifndef _TVG_WG_RENDERER_H_
 #define _TVG_WG_RENDERER_H_
 
-#include "tvgWgCompositor.h"
+#include "tvgWgRenderTask.h"
 
 class WgRenderer : public RenderMethod
 {
@@ -72,13 +72,15 @@ private:
     bool surfaceConfigure(WGPUSurface surface, WgContext& context, uint32_t width, uint32_t height);
 
     // render tree stacks
-    WgRenderStorage mRenderStorageRoot;
-    Array<WgCompose*> mCompositorStack;
-    Array<WgRenderStorage*> mRenderStorageStack;
+    WgRenderTarget mRenderTargetRoot;
+    Array<WgCompose*> mCompositorList;
+    Array<WgRenderTarget*> mRenderTargetStack;
     Array<WgRenderDataViewport*> mRenderDataViewportList;
+    Array<WgSceneTask*> mSceneTaskStack;
+    Array<WgRenderTask*> mRenderTaskList;
 
-    // render storage pool
-    WgRenderStoragePool mRenderStoragePool;
+    // render target pool
+    WgRenderTargetPool mRenderTargetPool;
 
     // render data paint pools
     WgRenderDataShapePool mRenderDataShapePool;
@@ -100,7 +102,6 @@ private:
     Key mDisposeKey{};
 
     // gpu handles
-    WGPUCommandEncoder mCommandEncoder{};
     WGPUTexture targetTexture{}; // external handle
     WGPUSurfaceTexture surfaceTexture{};
     WGPUSurface surface{};  // external handle

src/renderer/wg_engine/tvgWgShaderSrc.cpp

@@ -23,8 +23,6 @@
 #include "tvgWgShaderSrc.h"
 #include <string>
 
-#define WG_SHADER_SOURCE(...) #__VA_ARGS__
-
 //************************************************************************
 // graphics shader source: stencil
 //************************************************************************
@@ -32,14 +30,15 @@
 const char* cShaderSrc_Stencil = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f };
+struct PaintSettings { transform: mat4x4f };
 
 @group(0) @binding(0) var<uniform> uViewMat  : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     return out;
 }
 
@@ -56,15 +55,16 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Depth = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f };
+struct PaintSettings { transform: mat4x4f };
 
 @group(0) @binding(0) var<uniform> uViewMat  : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var<uniform> uDepth : f32;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.position.z = uDepth;
     return out;
 }
@@ -82,23 +82,22 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Solid = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat      : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
-@group(2) @binding(0) var<uniform> uSolidColor : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     return out;
 }
 
 @fragment
 fn fs_main(in: VertexOutput) -> @location(0) vec4f {
-    let Sc = uSolidColor;
-    let So = uBlendSettings.a;
+    let Sc = uPaintSettings.color;
+    let So = uPaintSettings.options.a;
     return vec4f(Sc.rgb * Sc.a * So, Sc.a * So);
 }
 )";
@@ -110,34 +109,32 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Linear = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position : vec4f, @location(0) vGradCoord : vec4f };
-struct GradSettings { settings: vec4f, focal: vec4f };
+struct GradSettings  { transform: mat4x4f, coords: vec4f, focal: vec4f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f, gradient: GradSettings };
 
 // uniforms
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSamplerGrad : sampler;
 @group(2) @binding(1) var uTextureGrad : texture_2d<f32>;
-@group(2) @binding(2) var<uniform> uSettingGrad : GradSettings;
-@group(2) @binding(3) var<uniform> uTransformGrad : mat4x4f;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
-    out.vGradCoord = uTransformGrad * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
+    out.vGradCoord = uPaintSettings.gradient.transform * vec4f(in.position.xy, 0.0, 1.0);
     return out;
 }
 
 @fragment
 fn fs_main(in: VertexOutput) -> @location(0) vec4f {
     let pos = in.vGradCoord.xy;
-    let st = uSettingGrad.settings.xy;
-    let ed = uSettingGrad.settings.zw;
+    let st = uPaintSettings.gradient.coords.xy;
+    let ed = uPaintSettings.gradient.coords.zw;
     let ba = ed - st;
     let t = dot(pos - st, ba) / dot(ba, ba);
     let Sc = textureSample(uTextureGrad, uSamplerGrad, vec2f(t, 0.5));
-    let So = uBlendSettings.a;
+    let So = uPaintSettings.options.a;
     return vec4f(Sc.rgb * Sc.a * So, Sc.a * So);
 }
 )";
@@ -149,37 +146,35 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Radial = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position : vec4f, @location(0) vGradCoord : vec4f };
-struct GradSettings { settings: vec4f, focal: vec4f };
+struct GradSettings  { transform: mat4x4f, coords: vec4f, focal: vec4f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f, gradient: GradSettings };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSamplerGrad : sampler;
 @group(2) @binding(1) var uTextureGrad : texture_2d<f32>;
-@group(2) @binding(2) var<uniform> uSettingGrad : GradSettings;
-@group(2) @binding(3) var<uniform> uTransformGrad : mat4x4f;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
-    out.vGradCoord = uTransformGrad * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
+    out.vGradCoord = uPaintSettings.gradient.transform * vec4f(in.position.xy, 0.0, 1.0);
     return out;
 }
 
 @fragment
 fn fs_main(in: VertexOutput) -> @location(0) vec4f {
     // orignal data
-    let d0 = in.vGradCoord.xy - uSettingGrad.settings.xy;
-    let d1 = uSettingGrad.settings.xy - uSettingGrad.focal.xy;
-    let r0 = uSettingGrad.settings.z;
-    let rd = uSettingGrad.focal.z - uSettingGrad.settings.z;
+    let d0 = in.vGradCoord.xy - uPaintSettings.gradient.coords.xy;
+    let d1 = uPaintSettings.gradient.coords.xy - uPaintSettings.gradient.focal.xy;
+    let r0 = uPaintSettings.gradient.coords.z;
+    let rd = uPaintSettings.gradient.focal.z - uPaintSettings.gradient.coords.z;
     let a = 1.0*dot(d1, d1) - 1.0*rd*rd;
     let b = 2.0*dot(d0, d1) - 2.0*r0*rd;
     let c = 1.0*dot(d0, d0) - 1.0*r0*r0;
     let t = (-b + sqrt(b*b - 4*a*c))/(2*a);
     let Sc = textureSample(uTextureGrad, uSamplerGrad, vec2f(1.0 - t, 0.5));
-    let So = uBlendSettings.a;
+    let So = uPaintSettings.options.a;
     return vec4f(Sc.rgb * Sc.a * So, Sc.a * So);
 }
 )";
@@ -191,17 +186,17 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Image = R"(
 struct VertexInput { @location(0) position: vec2f, @location(1) texCoord: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f, @location(0) vTexCoord: vec2f };
+struct PaintSettings { transform: mat4x4f, options: vec4f };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat      : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSampler     : sampler;
 @group(2) @binding(1) var uTextureView : texture_2d<f32>;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    out.position = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    out.position = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.vTexCoord = in.texCoord;
     return out;
 }
@@ -209,7 +204,7 @@ fn vs_main(in: VertexInput) -> VertexOutput {
 @fragment
 fn fs_main(in: VertexOutput) -> @location(0) vec4f {
     var Sc: vec4f = textureSample(uTextureView, uSampler, in.vTexCoord.xy);
-    let So: f32 = uBlendSettings.a;
+    let So: f32 = uPaintSettings.options.a;
     return vec4f(Sc.rgb * Sc.a * So, Sc.a * So);
 };
 )";
@@ -248,18 +243,18 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4f {
 const char* cShaderSrc_Solid_Blend = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f, @location(1) vScrCoord: vec2f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
-@group(2) @binding(0) var<uniform> uSolidColor : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
+// @group(2) - empty
 @group(3) @binding(0) var uSamplerDst : sampler;
 @group(3) @binding(1) var uTextureDst : texture_2d<f32>;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    let pos = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    let pos = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.position = pos;
     out.vScrCoord = vec2f(0.5 + pos.x * 0.5, 0.5 - pos.y * 0.5);
     return out;
@@ -268,13 +263,13 @@ fn vs_main(in: VertexInput) -> VertexOutput {
 struct FragData { Sc: vec3f, Sa: f32, So: f32, Dc: vec3f, Da: f32 };
 fn getFragData(in: VertexOutput) -> FragData {
     // get source data
-    let colorSrc = uSolidColor;
+    let colorSrc = uPaintSettings.color;
     let colorDst = textureSample(uTextureDst, uSamplerDst, in.vScrCoord.xy);
     // fill fragment data
     var data: FragData;
     data.Sc = colorSrc.rgb;
     data.Sa = colorSrc.a;
-    data.So = uBlendSettings.a;
+    data.So = uPaintSettings.options.a;
     data.Dc = colorDst.rgb;
     data.Da = colorDst.a;
     data.Sc = data.Sa * data.So * data.Sc;
@@ -288,24 +283,22 @@ fn postProcess(d: FragData, R: vec4f) -> vec4f { return R; };
 const char* cShaderSrc_Linear_Blend = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f, @location(0) vGradCoord : vec4f, @location(1) vScrCoord: vec2f };
-struct GradSettings { settings: vec4f, focal: vec4f };
+struct GradSettings  { transform: mat4x4f, coords: vec4f, focal: vec4f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f, gradient: GradSettings };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSamplerGrad : sampler;
 @group(2) @binding(1) var uTextureGrad : texture_2d<f32>;
-@group(2) @binding(2) var<uniform> uSettingGrad : vec4f;
-@group(2) @binding(3) var<uniform> uTransformGrad : mat4x4f;
 @group(3) @binding(0) var uSamplerDst : sampler;
 @group(3) @binding(1) var uTextureDst : texture_2d<f32>;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    let pos = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    let pos = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.position = pos;
-    out.vGradCoord = uTransformGrad * vec4f(in.position.xy, 0.0, 1.0);
+    out.vGradCoord = uPaintSettings.gradient.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.vScrCoord = vec2f(0.5 + pos.x * 0.5, 0.5 - pos.y * 0.5);
     return out;
 }
@@ -314,8 +307,8 @@ struct FragData { Sc: vec3f, Sa: f32, So: f32, Dc: vec3f, Da: f32 };
 fn getFragData(in: VertexOutput) -> FragData {
     // get source data
     let pos = in.vGradCoord.xy;
-    let st = uSettingGrad.xy;
-    let ed = uSettingGrad.zw;
+    let st = uPaintSettings.gradient.coords.xy;
+    let ed = uPaintSettings.gradient.coords.zw;
     let ba = ed - st;
     let t = dot(pos - st, ba) / dot(ba, ba);
     let colorSrc = textureSample(uTextureGrad, uSamplerGrad, vec2f(t, 0.5));
@@ -324,7 +317,7 @@ fn getFragData(in: VertexOutput) -> FragData {
     var data: FragData;
     data.Sc = colorSrc.rgb;
     data.Sa = colorSrc.a;
-    data.So = uBlendSettings.a;
+    data.So = uPaintSettings.options.a;
     data.Dc = colorDst.rgb;
     data.Da = colorDst.a;
     data.Sc = mix(data.Dc, data.Sc, data.Sa * data.So);
@@ -338,34 +331,32 @@ fn postProcess(d: FragData, R: vec4f) -> vec4f { return R; };
 const char* cShaderSrc_Radial_Blend = R"(
 struct VertexInput { @location(0) position: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f, @location(0) vGradCoord : vec4f, @location(1) vScrCoord: vec2f };
-struct GradSettings { settings: vec4f, focal: vec4f };
+struct GradSettings  { transform: mat4x4f, coords: vec4f, focal: vec4f };
+struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f, gradient: GradSettings };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSamplerGrad : sampler;
 @group(2) @binding(1) var uTextureGrad : texture_2d<f32>;
-@group(2) @binding(2) var<uniform> uSettingGrad : GradSettings;
-@group(2) @binding(3) var<uniform> uTransformGrad : mat4x4f;
 @group(3) @binding(0) var uSamplerDst : sampler;
 @group(3) @binding(1) var uTextureDst : texture_2d<f32>;
 
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    let pos = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    let pos = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.position = pos;
-    out.vGradCoord = uTransformGrad * vec4f(in.position.xy, 0.0, 1.0);
+    out.vGradCoord = uPaintSettings.gradient.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.vScrCoord = vec2f(0.5 + pos.x * 0.5, 0.5 - pos.y * 0.5);
     return out;
 }
 
 struct FragData { Sc: vec3f, Sa: f32, So: f32, Dc: vec3f, Da: f32 };
 fn getFragData(in: VertexOutput) -> FragData {
-    let d0 = in.vGradCoord.xy - uSettingGrad.settings.xy;
-    let d1 = uSettingGrad.settings.xy - uSettingGrad.focal.xy;
-    let r0 = uSettingGrad.settings.z;
-    let rd = uSettingGrad.focal.z - uSettingGrad.settings.z;
+    let d0 = in.vGradCoord.xy - uPaintSettings.gradient.coords.xy;
+    let d1 = uPaintSettings.gradient.coords.xy - uPaintSettings.gradient.focal.xy;
+    let r0 = uPaintSettings.gradient.coords.z;
+    let rd = uPaintSettings.gradient.focal.z - uPaintSettings.gradient.coords.z;
     let a = 1.0*dot(d1, d1) - 1.0*rd*rd;
     let b = 2.0*dot(d0, d1) - 2.0*r0*rd;
     let c = 1.0*dot(d0, d0) - 1.0*r0*r0;
@@ -376,7 +367,7 @@ fn getFragData(in: VertexOutput) -> FragData {
     var data: FragData;
     data.Sc = colorSrc.rgb;
     data.Sa = colorSrc.a;
-    data.So = uBlendSettings.a;
+    data.So = uPaintSettings.options.a;
     data.Dc = colorDst.rgb;
     data.Da = colorDst.a;
     data.Sc = mix(data.Dc, data.Sc, data.Sa * data.So);
@@ -390,10 +381,10 @@ fn postProcess(d: FragData, R: vec4f) -> vec4f { return R; };
 const char* cShaderSrc_Image_Blend = R"(
 struct VertexInput { @location(0) position: vec2f, @location(1) texCoord: vec2f };
 struct VertexOutput { @builtin(position) position: vec4f, @location(0) vTexCoord : vec2f, @location(1) vScrCoord: vec2f };
+struct PaintSettings { transform: mat4x4f, options: vec4f };
 
 @group(0) @binding(0) var<uniform> uViewMat : mat4x4f;
-@group(1) @binding(0) var<uniform> uModelMat      : mat4x4f;
-@group(1) @binding(1) var<uniform> uBlendSettings : vec4f;
+@group(1) @binding(0) var<uniform> uPaintSettings : PaintSettings;
 @group(2) @binding(0) var uSamplerSrc : sampler;
 @group(2) @binding(1) var uTextureSrc : texture_2d<f32>;
 @group(3) @binding(0) var uSamplerDst : sampler;
@@ -402,7 +393,7 @@ struct VertexOutput { @builtin(position) position: vec4f, @location(0) vTexCoord
 @vertex
 fn vs_main(in: VertexInput) -> VertexOutput {
     var out: VertexOutput;
-    let pos = uViewMat * uModelMat * vec4f(in.position.xy, 0.0, 1.0);
+    let pos = uViewMat * uPaintSettings.transform * vec4f(in.position.xy, 0.0, 1.0);
     out.position = pos;
     out.vTexCoord = in.texCoord;
     out.vScrCoord = vec2f(0.5 + pos.x * 0.5, 0.5 - pos.y * 0.5);
@@ -418,7 +409,7 @@ fn getFragData(in: VertexOutput) -> FragData {
     var data: FragData;
     data.Sc = colorSrc.rgb;
     data.Sa = colorSrc.a;
-    data.So = uBlendSettings.a;
+    data.So = uPaintSettings.options.a;
     data.Dc = colorDst.rgb;
     data.Da = colorDst.a;
     data.Sc = data.Sc * data.So;

src/renderer/wg_engine/tvgWgShaderTypes.cpp

@@ -118,42 +118,41 @@ void WgShaderTypeVec4f::update(const RenderColor& c)
 
 void WgShaderTypeVec4f::update(const RenderRegion& r)
 {
-    vec[0] = r.x; // left
-    vec[1] = r.y; // top
-    vec[2] = r.x + r.w - 1; // right
-    vec[3] = r.y + r.h - 1; // bottom
+    vec[0] = r.min.x;
+    vec[1] = r.min.y;
+    vec[2] = r.max.x - 1;
+    vec[3] = r.max.y - 1;
 }
 
 //************************************************************************
-// WgShaderTypeGradient
+// WgShaderTypeGradSettings
 //************************************************************************
 
-void WgShaderTypeGradient::update(const LinearGradient* linearGradient)
+void WgShaderTypeGradSettings::update(const Fill* fill)
 {
-    // update gradient data
+    assert(fill);
+    // update transform matrix
+    Matrix invTransform;
+    if (inverse(&fill->transform(), &invTransform))
+        transform.update(invTransform);
+    else transform.identity();
+    // update gradient base points
+    if (fill->type() == Type::LinearGradient)
+        ((LinearGradient*)fill)->linear(&coords.vec[0], &coords.vec[1], &coords.vec[2], &coords.vec[3]);
+    else if (fill->type() == Type::RadialGradient)
+        ((RadialGradient*)fill)->radial(&coords.vec[0], &coords.vec[1], &coords.vec[2], &focal.vec[0], &focal.vec[1], &focal.vec[2]);
+}
+
+//************************************************************************
+// WgShaderTypeGradientData
+//************************************************************************
+
+void WgShaderTypeGradientData::update(const Fill* fill)
+{
+    if (!fill) return;
     const Fill::ColorStop* stops = nullptr;
-    auto stopCnt = linearGradient->colorStops(&stops);
-    updateTexData(stops, stopCnt);
-    // update base points
-    linearGradient->linear(&settings[0], &settings[1], &settings[2], &settings[3]);
-};
-
-
-void WgShaderTypeGradient::update(const RadialGradient* radialGradient)
-{
-    // update gradient data
-    const Fill::ColorStop* stops = nullptr;
-    auto stopCnt = radialGradient->colorStops(&stops);
-    updateTexData(stops, stopCnt);
-    // update base points
-    radialGradient->radial(&settings[0], &settings[1], &settings[2], &settings[4], &settings[5], &settings[6]);
-};
-
-
-void WgShaderTypeGradient::updateTexData(const Fill::ColorStop* stops, uint32_t stopCnt)
-{
+    auto stopCnt = fill->colorStops(&stops);
     if (stopCnt == 0) return;
-
     static Array<Fill::ColorStop> sstops(stopCnt);
     sstops.clear();
     sstops.push(stops[0]);
@@ -167,10 +166,10 @@ void WgShaderTypeGradient::updateTexData(const Fill::ColorStop* stops, uint32_t
     uint32_t range_s = 0;
     uint32_t range_e = uint32_t(sstops[0].offset * (WG_TEXTURE_GRADIENT_SIZE-1));
     for (uint32_t ti = range_s; (ti < range_e) && (ti < WG_TEXTURE_GRADIENT_SIZE); ti++) {
-        texData[ti * 4 + 0] = sstops[0].r;
-        texData[ti * 4 + 1] = sstops[0].g;
-        texData[ti * 4 + 2] = sstops[0].b;
-        texData[ti * 4 + 3] = sstops[0].a;
+        data[ti * 4 + 0] = sstops[0].r;
+        data[ti * 4 + 1] = sstops[0].g;
+        data[ti * 4 + 2] = sstops[0].b;
+        data[ti * 4 + 3] = sstops[0].a;
     }
     // body
     for (uint32_t di = 1; di < sstops.count; di++) {
@@ -179,10 +178,10 @@ void WgShaderTypeGradient::updateTexData(const Fill::ColorStop* stops, uint32_t
         float delta = 1.0f/(range_e - range_s);
         for (uint32_t ti = range_s; (ti < range_e) && (ti < WG_TEXTURE_GRADIENT_SIZE); ti++) {
             float t = (ti - range_s) * delta;
-            texData[ti * 4 + 0] = tvg::lerp(sstops[di-1].r, sstops[di].r, t);
-            texData[ti * 4 + 1] = tvg::lerp(sstops[di-1].g, sstops[di].g, t);
-            texData[ti * 4 + 2] = tvg::lerp(sstops[di-1].b, sstops[di].b, t);
-            texData[ti * 4 + 3] = tvg::lerp(sstops[di-1].a, sstops[di].a, t);
+            data[ti * 4 + 0] = tvg::lerp(sstops[di-1].r, sstops[di].r, t);
+            data[ti * 4 + 1] = tvg::lerp(sstops[di-1].g, sstops[di].g, t);
+            data[ti * 4 + 2] = tvg::lerp(sstops[di-1].b, sstops[di].b, t);
+            data[ti * 4 + 3] = tvg::lerp(sstops[di-1].a, sstops[di].a, t);
         }
     }
     // tail
@@ -190,10 +189,10 @@ void WgShaderTypeGradient::updateTexData(const Fill::ColorStop* stops, uint32_t
     range_s = uint32_t(colorStopLast.offset * (WG_TEXTURE_GRADIENT_SIZE-1));
     range_e = WG_TEXTURE_GRADIENT_SIZE;
     for (uint32_t ti = range_s; ti < range_e; ti++) {
-        texData[ti * 4 + 0] = colorStopLast.r;
-        texData[ti * 4 + 1] = colorStopLast.g;
-        texData[ti * 4 + 2] = colorStopLast.b;
-        texData[ti * 4 + 3] = colorStopLast.a;
+        data[ti * 4 + 0] = colorStopLast.r;
+        data[ti * 4 + 1] = colorStopLast.g;
+        data[ti * 4 + 2] = colorStopLast.b;
+        data[ti * 4 + 3] = colorStopLast.a;
     }
 }
 

src/renderer/wg_engine/tvgWgShaderTypes.h

@@ -55,16 +55,44 @@ struct WgShaderTypeVec4f
     void update(const RenderRegion& r);
 };
 
-// sampler, texture, vec4f
-#define WG_TEXTURE_GRADIENT_SIZE 512
-struct WgShaderTypeGradient
+// WGSL: struct GradSettings  { transform: mat4x4f, coords: vec4f, focal: vec4f };
+struct WgShaderTypeGradSettings
 {
-    float settings[4+4]{}; // WGSL: struct GradSettings { settings: vec4f, focal: vec4f; transform: mat4f };
-    uint8_t texData[WG_TEXTURE_GRADIENT_SIZE * 4];
+    // gradient transform matrix
+    WgShaderTypeMat4x4f transform;
+    // linear: [0] - x1, [1] - y1, [2] - x2, [3] - y2
+    // radial: [0] - cx, [1] - cy, [2] - cr
+    WgShaderTypeVec4f coords;
+    // radial: [0] - fx, [1] - fy, [2] - fr
+    WgShaderTypeVec4f focal;
     
-    void update(const LinearGradient* linearGradient);
-    void update(const RadialGradient* radialGradient);
-    void updateTexData(const Fill::ColorStop* stops, uint32_t stopCnt);
+    void update(const Fill* fill);
+};
+
+// WGSL: struct PaintSettings { transform: mat4x4f, options: vec4f, color: vec4f, gradient: GradSettings };
+struct WgShaderTypePaintSettings
+{
+    // paint transform matrix (must be at offset 0)
+    WgShaderTypeMat4x4f transform;
+    // [0] - color space, [3] - opacity
+    WgShaderTypeVec4f options;
+    // solid color
+    WgShaderTypeVec4f color;
+    // gradient settings (linear/radial)
+    WgShaderTypeGradSettings gradient;
+    // align to 256 bytes (see webgpu spec: minUniformBufferOffsetAlignment)
+    uint8_t _padding[256 - sizeof(transform) - sizeof(options) - sizeof(color) - sizeof(gradient)];
+};
+// see webgpu spec: 3.6.2. Limits - minUniformBufferOffsetAlignment (256)
+static_assert(sizeof(WgShaderTypePaintSettings) == 256, "Uniform shader data type size must be aligned to 256 bytes");
+
+// gradient color map
+#define WG_TEXTURE_GRADIENT_SIZE 512
+struct WgShaderTypeGradientData
+{
+    uint8_t data[WG_TEXTURE_GRADIENT_SIZE * 4];
+
+    void update(const Fill* fill);
 };
 
 // gaussian params: sigma, scale, extend

test/testPaint.cpp

@@ -121,6 +121,9 @@ TEST_CASE("Bounding Box", "[tvgPaint]")
     Initializer::init(0);
 
     auto canvas = unique_ptr<SwCanvas>(SwCanvas::gen());
+    uint32_t buffer[100*100];
+    canvas->target(buffer, 100, 100, 100, ColorSpace::ABGR8888);
+
     auto shape = Shape::gen();
     canvas->push(shape);
     canvas->sync();
@@ -220,7 +223,9 @@ TEST_CASE("Composition", "[tvgPaint]")
 
     //Clipping
     auto comp = Shape::gen();
+    REQUIRE(shape->clip() == nullptr);
     REQUIRE(shape->clip(comp) == Result::Success);
+    REQUIRE(shape->clip() == comp);
 
     //AlphaMask
     comp = Shape::gen();

test/testScene.cpp

@@ -80,7 +80,8 @@ TEST_CASE("Scene Clear And Reuse Shape", "[tvgScene]")
     REQUIRE(Initializer::init(0) == Result::Success);
     {
         auto canvas = unique_ptr<SwCanvas>(SwCanvas::gen());
-        REQUIRE(canvas);
+        uint32_t buffer[100*100];
+        canvas->target(buffer, 100, 100, 100, ColorSpace::ABGR8888);
 
         auto scene = Scene::gen();
         REQUIRE(scene);

test/testText.cpp

@@ -120,6 +120,8 @@ TEST_CASE("Text Basic", "[tvgText]")
     Initializer::init(0);
 
     auto canvas = unique_ptr<SwCanvas>(SwCanvas::gen());
+    uint32_t buffer[100*100];
+    canvas->target(buffer, 100, 100, 100, ColorSpace::ABGR8888);
 
     auto text = Text::gen();
     REQUIRE(text);
@@ -144,6 +146,8 @@ TEST_CASE("Text with composite glyphs", "[tvgText]")
     Initializer::init(0);
 
     auto canvas = unique_ptr<SwCanvas>(SwCanvas::gen());
+    uint32_t buffer[100*100];
+    canvas->target(buffer, 100, 100, 100, ColorSpace::ABGR8888);
 
     auto text = Text::gen();
     REQUIRE(text);

tools/lottie2gif/lottie2gif.cpp

@@ -54,7 +54,7 @@ private:
 
    void helpMsg()
    {
-      cout << "Usage: \n   tvg-lotie2gif [Lottie file] or [Lottie folder] [-r resolution] [-f fps] [-b background color]\n\nExamples: \n    $ tvg-lotie2gif input.json\n    $ tvg-lotie2gif input.json -r 600x600\n    $ tvg-lotie2gif input.json -f 30\n    $ tvg-lotie2gif input.json -r 600x600 -f 30\n    $ tvg-lotie2gif lottiefolder\n    $ tvg-lotie2gif lottiefolder -r 600x600 -f 30 -b fa7410\n\n";
+      cout << "Usage: \n   tvg-lottie2gif [Lottie file] or [Lottie folder] [-r resolution] [-f fps] [-b background color]\n\nExamples: \n    $ tvg-lottie2gif input.json\n    $ tvg-lottie2gif input.json -r 600x600\n    $ tvg-lottie2gif input.json -f 30\n    $ tvg-lottie2gif input.json -r 600x600 -f 30\n    $ tvg-lottie2gif lottiefolder\n    $ tvg-lottie2gif lottiefolder -r 600x600 -f 30 -b fa7410\n\n";
    }
 
    bool validate(string& lottieName)