wg_engine: geometry stage buffers implementation

Implemented task-based rendering and geometry stage buffers:
1. Get information about current frame objects
2. Accumulate geometry data into a stage buffer during frame rendering
3. Flush it to the GPU in single call
4. Run rendering process in post render stage

https://github.com/thorvg/thorvg/issues/3489
https://github.com/thorvg/thorvg/issues/3455

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/tvgWgCommon.cpp

@@ -269,3 +269,34 @@ 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

@@ -71,10 +71,12 @@ struct WgContext {
     // release buffer objects
     void releaseBuffer(WGPUBuffer& buffer);
 
-    bool invalid()
+    // command encoder
-    {
+    WGPUCommandEncoder createCommandEncoder();
-        return !instance || !device;
+    void submitCommandEncoder(WGPUCommandEncoder encoder);
-    }
+    void releaseCommandEncoder(WGPUCommandEncoder& encoder);
+
+    bool invalid();
 };
 
 #endif // _TVG_WG_COMMON_H_

src/renderer/wg_engine/tvgWgCompositor.cpp

@@ -28,6 +28,7 @@ void WgCompositor::initialize(WgContext& context, uint32_t width, uint32_t heigh
 {
     // pipelines (external handle, do not release)
     pipelines.initialize(context);
+    stageBuffer.initialize(context);
     // initialize opacity pool
     initPools(context);
     // allocate global view matrix handles
@@ -37,7 +38,7 @@ void WgCompositor::initialize(WgContext& context, uint32_t width, uint32_t heigh
     // create render targets handles
     resize(context, width, height);
     // composition and blend geometries
-    meshData.blitBox(context);
+    meshDataBlit.blitBox(context);
 }
 
 
@@ -54,7 +55,7 @@ void WgCompositor::initPools(WgContext& context)
 void WgCompositor::release(WgContext& context)
 {
     // composition and blend geometries
-    meshData.release(context);
+    meshDataBlit.release(context);
     // release render targets habdles
     resize(context, 0, 0);
     // release opacity pool
@@ -62,6 +63,8 @@ void WgCompositor::release(WgContext& context)
     // release global view matrix handles
     context.layouts.releaseBindGroup(bindGroupViewMat);
     context.releaseBuffer(bufferViewMat);
+    // release stage buffer
+    stageBuffer.release(context);
     // release pipelines
     pipelines.release(context);
 }
@@ -81,9 +84,9 @@ void WgCompositor::resize(WgContext& context, uint32_t width, uint32_t height) {
     // release existig handles
     if ((this->width != width) || (this->height != height)) {
         context.layouts.releaseBindGroup(bindGroupStorageTemp);
-        // release intermediate render storages
+        // release intermediate render target
-        storageTemp1.release(context);
+        targetTemp1.release(context);
-        storageTemp0.release(context);
+        targetTemp0.release(context);
         // release global stencil buffer handles
         context.releaseTextureView(texViewDepthStencilMS);
         context.releaseTexture(texDepthStencilMS);
@@ -107,10 +110,10 @@ void WgCompositor::resize(WgContext& context, uint32_t width, uint32_t height) {
         texViewDepthStencil = context.createTextureView(texDepthStencil);
         texDepthStencilMS = context.createTexAttachement(width, height, WGPUTextureFormat_Depth24PlusStencil8, 4);
         texViewDepthStencilMS = context.createTextureView(texDepthStencilMS);
-        // initialize intermediate render storages
+        // initialize intermediate render targets
-        storageTemp0.initialize(context, width, height);
+        targetTemp0.initialize(context, width, height);
-        storageTemp1.initialize(context, width, height);
+        targetTemp1.initialize(context, width, height);
-        bindGroupStorageTemp = context.layouts.createBindGroupStrorage2RO(storageTemp0.texView, storageTemp1.texView);
+        bindGroupStorageTemp = context.layouts.createBindGroupStrorage2RO(targetTemp0.texView, targetTemp1.texView);
     }
 }
 
@@ -121,14 +124,14 @@ RenderRegion WgCompositor::shrinkRenderRegion(RenderRegion& rect)
 }
 
 
-void WgCompositor::copyTexture(const WgRenderStorage* dst, const WgRenderStorage* src)
+void WgCompositor::copyTexture(const WgRenderTarget* dst, const WgRenderTarget* src)
 {
     const RenderRegion region = {{0, 0}, {(int32_t)src->width, (int32_t)src->height}};
     copyTexture(dst, src, region);
 }
 
 
-void WgCompositor::copyTexture(const WgRenderStorage* dst, const WgRenderStorage* src, const RenderRegion& region)
+void WgCompositor::copyTexture(const WgRenderTarget* dst, const WgRenderTarget* src, const RenderRegion& region)
 {
     assert(dst);
     assert(src);
@@ -140,16 +143,25 @@ void WgCompositor::copyTexture(const WgRenderStorage* dst, const WgRenderStorage
 }
 
 
-void WgCompositor::beginRenderPass(WGPUCommandEncoder commandEncoder, WgRenderStorage* target, bool clear, WGPUColor clearColor)
+void WgCompositor::beginRenderPass(WGPUCommandEncoder commandEncoder, WgRenderTarget* target, bool clear, WGPUColor clearColor)
 {
-    assert(commandEncoder);
     assert(target);
+    assert(commandEncoder);
+    // do not start same render bass
+    if (target == currentTarget) return;
+    // we must to end render pass first
+    endRenderPass();
     this->currentTarget = target;
+    // start new render pass
     this->commandEncoder = commandEncoder;
     const WGPURenderPassDepthStencilAttachment depthStencilAttachment{ 
         .view = texViewDepthStencilMS,
-        .depthLoadOp = WGPULoadOp_Clear, .depthStoreOp = WGPUStoreOp_Discard, .depthClearValue = 1.0f,
+        .depthLoadOp = WGPULoadOp_Clear,
-        .stencilLoadOp = WGPULoadOp_Clear, .stencilStoreOp = WGPUStoreOp_Discard, .stencilClearValue = 0
+        .depthStoreOp = WGPUStoreOp_Discard,
+        .depthClearValue = 1.0f,
+        .stencilLoadOp = WGPULoadOp_Clear,
+        .stencilStoreOp = WGPUStoreOp_Discard,
+        .stencilClearValue = 0
     };
     const WGPURenderPassColorAttachment colorAttachment{
         .view = target->texViewMS,
@@ -158,7 +170,6 @@ void WgCompositor::beginRenderPass(WGPUCommandEncoder commandEncoder, WgRenderSt
         .loadOp = clear ? WGPULoadOp_Clear : WGPULoadOp_Load,
         .storeOp = WGPUStoreOp_Store,
         .clearValue = clearColor
-        
     };
     WGPURenderPassDescriptor renderPassDesc{ .colorAttachmentCount = 1, .colorAttachments = &colorAttachment, .depthStencilAttachment = &depthStencilAttachment };
     renderPassEncoder = wgpuCommandEncoderBeginRenderPass(commandEncoder, &renderPassDesc);
@@ -172,11 +183,37 @@ void WgCompositor::endRenderPass()
         assert(renderPassEncoder);
         wgpuRenderPassEncoderEnd(renderPassEncoder);
         wgpuRenderPassEncoderRelease(renderPassEncoder);
-        this->renderPassEncoder = nullptr;
+        renderPassEncoder = nullptr;
-        this->currentTarget = nullptr;
+        currentTarget = nullptr;
     }
 }
 
+void WgCompositor::reset(WgContext& context)
+{
+    stageBuffer.clear();
+}
+
+
+void WgCompositor::flush(WgContext& context)
+{
+    stageBuffer.append(&meshDataBlit);
+    stageBuffer.flush(context);
+}
+
+
+void WgCompositor::requestShape(WgRenderDataShape* renderData)
+{
+    stageBuffer.append(renderData);
+    // TODO: expand for fill settings
+}
+
+
+void WgCompositor::requestImage(WgRenderDataPicture* renderData)
+{
+    stageBuffer.append(renderData);
+    // TODO: expand for fill settings
+}
+
 
 void WgCompositor::renderShape(WgContext& context, WgRenderDataShape* renderData, BlendMethod blendMethod)
 {
@@ -232,7 +269,7 @@ void WgCompositor::renderImage(WgContext& context, WgRenderDataPicture* renderDa
 }
 
 
-void WgCompositor::renderScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose)
+void WgCompositor::renderScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose)
 {
     assert(scene);
     assert(compose);
@@ -245,7 +282,7 @@ void WgCompositor::renderScene(WgContext& context, WgRenderStorage* scene, WgCom
 }
 
 
-void WgCompositor::composeScene(WgContext& context, WgRenderStorage* src, WgRenderStorage* mask, WgCompose* cmp)
+void WgCompositor::composeScene(WgContext& context, WgRenderTarget* src, WgRenderTarget* mask, WgCompose* cmp)
 {
     assert(cmp);
     assert(src);
@@ -257,15 +294,19 @@ void WgCompositor::composeScene(WgContext& context, WgRenderStorage* src, WgRend
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, src->bindGroupTexure, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, mask->bindGroupTexure, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.scene_compose[(uint32_t)cmp->method]);
-    meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &meshDataBlit);
 }
 
 
-void WgCompositor::blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderStorage* src, WGPUTextureView dstView) {
+void WgCompositor::blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderTarget* src, WGPUTextureView dstView)
+{
+    assert(!renderPassEncoder);
     const WGPURenderPassDepthStencilAttachment depthStencilAttachment{ 
         .view = texViewDepthStencil,
-        .depthLoadOp = WGPULoadOp_Load, .depthStoreOp = WGPUStoreOp_Discard,
+        .depthLoadOp = WGPULoadOp_Load,
-        .stencilLoadOp = WGPULoadOp_Load, .stencilStoreOp = WGPUStoreOp_Discard
+        .depthStoreOp = WGPUStoreOp_Discard,
+        .stencilLoadOp = WGPULoadOp_Load,
+        .stencilStoreOp = WGPUStoreOp_Discard
     };
     const WGPURenderPassColorAttachment colorAttachment { 
         .view = dstView,
@@ -274,15 +315,56 @@ void WgCompositor::blit(WgContext& context, WGPUCommandEncoder encoder, WgRender
         .storeOp = WGPUStoreOp_Store,
     };
     const WGPURenderPassDescriptor renderPassDesc{ .colorAttachmentCount = 1, .colorAttachments = &colorAttachment, .depthStencilAttachment = &depthStencilAttachment };
-    WGPURenderPassEncoder renderPass = wgpuCommandEncoderBeginRenderPass(encoder, &renderPassDesc);
+    renderPassEncoder = wgpuCommandEncoderBeginRenderPass(encoder, &renderPassDesc);
-    wgpuRenderPassEncoderSetBindGroup(renderPass, 0, src->bindGroupTexure, 0, nullptr);
+    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, src->bindGroupTexure, 0, nullptr);
-    wgpuRenderPassEncoderSetPipeline(renderPass, pipelines.blit);
+    wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.blit);
-    meshData.drawImage(context, renderPass);
+    drawMeshImage(context, &meshDataBlit);
-    wgpuRenderPassEncoderEnd(renderPass);
+    wgpuRenderPassEncoderEnd(renderPassEncoder);
-    wgpuRenderPassEncoderRelease(renderPass);
+    wgpuRenderPassEncoderRelease(renderPassEncoder);
+    renderPassEncoder = nullptr;
 }
 
 
+void WgCompositor::drawMesh(WgContext& context, WgMeshData* meshData)
+{
+    assert(meshData);
+    assert(renderPassEncoder);
+    uint64_t icount = meshData->ibuffer.count;
+    uint64_t vsize = meshData->vbuffer.count * sizeof(Point);
+    uint64_t isize = icount * sizeof(uint32_t);
+    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, stageBuffer.vbuffer_gpu, meshData->voffset, vsize);
+    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, stageBuffer.ibuffer_gpu, WGPUIndexFormat_Uint32, meshData->ioffset, isize);
+    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, icount, 1, 0, 0, 0);
+};
+
+
+void WgCompositor::drawMeshFan(WgContext& context, WgMeshData* meshData)
+{
+    assert(meshData);
+    assert(renderPassEncoder);
+    uint64_t icount = (meshData->vbuffer.count - 2) * 3;
+    uint64_t vsize = meshData->vbuffer.count * sizeof(Point);
+    uint64_t isize = icount * sizeof(uint32_t);
+    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, stageBuffer.vbuffer_gpu, meshData->voffset, vsize);
+    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, context.bufferIndexFan, WGPUIndexFormat_Uint32, 0, isize);
+    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, icount, 1, 0, 0, 0);
+};
+
+
+void WgCompositor::drawMeshImage(WgContext& context, WgMeshData* meshData)
+{
+    assert(meshData);
+    assert(renderPassEncoder);
+    uint64_t icount = meshData->ibuffer.count;
+    uint64_t vsize = meshData->vbuffer.count * sizeof(Point);
+    uint64_t isize = icount * sizeof(uint32_t);
+    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 0, stageBuffer.vbuffer_gpu, meshData->voffset, vsize);
+    wgpuRenderPassEncoderSetVertexBuffer(renderPassEncoder, 1, stageBuffer.vbuffer_gpu, meshData->toffset, vsize);
+    wgpuRenderPassEncoderSetIndexBuffer(renderPassEncoder, stageBuffer.ibuffer_gpu, WGPUIndexFormat_Uint32, meshData->ioffset, isize);
+    wgpuRenderPassEncoderDrawIndexed(renderPassEncoder, icount, 1, 0, 0, 0);
+};
+
+
 void WgCompositor::drawShape(WgContext& context, WgRenderDataShape* renderData)
 {
     assert(renderData);
@@ -298,7 +380,7 @@ void WgCompositor::drawShape(WgContext& context, WgRenderDataShape* renderData)
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, stencilPipeline);
     // draw to stencil (first pass)
     ARRAY_FOREACH(p, renderData->meshGroupShapes.meshes)
-        (*p)->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, (*p));
     // setup fill rules
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
@@ -315,7 +397,7 @@ void WgCompositor::drawShape(WgContext& context, WgRenderDataShape* renderData)
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial);
     }
     // draw to color (second pass)
-    renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+    drawMeshFan(context, &renderData->meshDataBBox);
 }
 
 
@@ -325,10 +407,10 @@ void WgCompositor::blendShape(WgContext& context, WgRenderDataShape* renderData,
     assert(renderPassEncoder);
     assert(renderData->meshGroupShapes.meshes.count == renderData->meshGroupShapesBBox.meshes.count);
     if (renderData->renderSettingsShape.skip || renderData->meshGroupShapes.meshes.count == 0 || renderData->viewport.invalid()) return;
-    // copy current render target data to dst storage
+    // copy current render target data to dst target
-    WgRenderStorage *target = currentTarget;
+    WgRenderTarget *target = currentTarget;
     endRenderPass();
-    copyTexture(&storageTemp0, target);
+    copyTexture(&targetTemp0, target);
     beginRenderPass(commandEncoder, target, false);
     // render shape with blend settings
     wgpuRenderPassEncoderSetScissorRect(renderPassEncoder, renderData->viewport.x(), renderData->viewport.y(), renderData->viewport.w(), renderData->viewport.h());
@@ -340,12 +422,12 @@ void WgCompositor::blendShape(WgContext& context, WgRenderDataShape* renderData,
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, stencilPipeline);
     // draw to stencil (first pass)
     ARRAY_FOREACH(p, renderData->meshGroupShapes.meshes)
-        (*p)->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, (*p));
     // setup fill rules
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
-    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, storageTemp0.bindGroupTexure, 0, nullptr);
+    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, targetTemp0.bindGroupTexure, 0, nullptr);
     uint32_t blendMethodInd = (uint32_t)blendMethod;
     WgRenderSettings& settings = renderData->renderSettingsShape;
     if (settings.fillType == WgRenderSettingsType::Solid) {
@@ -359,7 +441,7 @@ void WgCompositor::blendShape(WgContext& context, WgRenderDataShape* renderData,
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial_blend[blendMethodInd]);
     }
     // draw to color (second pass)
-    renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+    drawMeshFan(context, &renderData->meshDataBBox);
 }
 
 
@@ -379,12 +461,12 @@ void WgCompositor::clipShape(WgContext& context, WgRenderDataShape* renderData)
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, stencilPipeline);
     // draw to stencil (first pass)
     ARRAY_FOREACH(p, renderData->meshGroupShapes.meshes)
-        (*p)->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, (*p));
     // merge depth and stencil buffer
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[128], 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.merge_depth_stencil);
-    renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+    drawMeshFan(context, &renderData->meshDataBBox);
     // setup fill rules
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
@@ -401,7 +483,7 @@ void WgCompositor::clipShape(WgContext& context, WgRenderDataShape* renderData)
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial);
     }
     // draw to color (second pass)
-    renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+    drawMeshFan(context, &renderData->meshDataBBox);
 }
 
 
@@ -421,7 +503,7 @@ void WgCompositor::drawStrokes(WgContext& context, WgRenderDataShape* renderData
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
         // draw to stencil (first pass)
-        renderData->meshGroupStrokes.meshes[i]->draw(context, renderPassEncoder);
+        drawMesh(context, renderData->meshGroupStrokes.meshes[i]);
         // setup fill rules
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
@@ -438,7 +520,7 @@ void WgCompositor::drawStrokes(WgContext& context, WgRenderDataShape* renderData
             wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial);
         }
         // draw to color (second pass)
-        renderData->meshGroupStrokesBBox.meshes[i]->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, renderData->meshGroupStrokesBBox.meshes[i]);
     }
 }
 
@@ -450,10 +532,10 @@ void WgCompositor::blendStrokes(WgContext& context, WgRenderDataShape* renderDat
     assert(renderData->meshGroupStrokes.meshes.count == renderData->meshGroupStrokesBBox.meshes.count);
     if (renderData->renderSettingsStroke.skip || renderData->meshGroupStrokes.meshes.count == 0 || renderData->viewport.invalid()) return;
 
-    // copy current render target data to dst storage
+    // copy current render target data to dst target
-    WgRenderStorage *target = currentTarget;
+    WgRenderTarget *target = currentTarget;
     endRenderPass();
-    copyTexture(&storageTemp0, target);
+    copyTexture(&targetTemp0, target);
     beginRenderPass(commandEncoder, target, false);
     wgpuRenderPassEncoderSetScissorRect(renderPassEncoder, renderData->viewport.x(), renderData->viewport.y(), renderData->viewport.w(), renderData->viewport.h());
     // draw strokes to stencil (first pass)
@@ -464,12 +546,12 @@ void WgCompositor::blendStrokes(WgContext& context, WgRenderDataShape* renderDat
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
         // draw to stencil (first pass)
-        renderData->meshGroupStrokes.meshes[i]->draw(context, renderPassEncoder);
+        drawMesh(context, renderData->meshGroupStrokes.meshes[i]);
         // setup fill rules
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
-        wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, storageTemp0.bindGroupTexure, 0, nullptr);
+        wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, targetTemp0.bindGroupTexure, 0, nullptr);
         uint32_t blendMethodInd = (uint32_t)blendMethod;
         WgRenderSettings& settings = renderData->renderSettingsStroke;
         if (settings.fillType == WgRenderSettingsType::Solid) {
@@ -483,7 +565,7 @@ void WgCompositor::blendStrokes(WgContext& context, WgRenderDataShape* renderDat
             wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial_blend[blendMethodInd]);
         }
         // draw to color (second pass)
-        renderData->meshGroupStrokesBBox.meshes[i]->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, renderData->meshGroupStrokesBBox.meshes[i]);
     }
 };
 
@@ -505,12 +587,12 @@ void WgCompositor::clipStrokes(WgContext& context, WgRenderDataShape* renderData
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
         // draw to stencil (first pass)
-        renderData->meshGroupStrokes.meshes[i]->draw(context, renderPassEncoder);
+        drawMesh(context, renderData->meshGroupStrokes.meshes[i]);
         // merge depth and stencil buffer
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[128], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.merge_depth_stencil);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
         // setup fill rules
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
@@ -527,7 +609,7 @@ void WgCompositor::clipStrokes(WgContext& context, WgRenderDataShape* renderData
             wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.radial);
         }
         // draw to color (second pass)
-        renderData->meshGroupStrokesBBox.meshes[i]->drawFan(context, renderPassEncoder);
+        drawMeshFan(context, renderData->meshGroupStrokesBBox.meshes[i]);
     }
 }
 
@@ -543,14 +625,14 @@ void WgCompositor::drawImage(WgContext& context, WgRenderDataPicture* renderData
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
     // draw image
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, renderData->bindGroupPicture, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.image);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
 }
 
 
@@ -560,26 +642,26 @@ void WgCompositor::blendImage(WgContext& context, WgRenderDataPicture* renderDat
     assert(renderPassEncoder);
     if (renderData->viewport.invalid()) return;
     wgpuRenderPassEncoderSetScissorRect(renderPassEncoder, renderData->viewport.x(), renderData->viewport.y(), renderData->viewport.w(), renderData->viewport.h());
-    // copy current render target data to dst storage
+    // copy current render target data to dst target
-    WgRenderStorage *target = currentTarget;
+    WgRenderTarget *target = currentTarget;
     endRenderPass();
-    copyTexture(&storageTemp0, target);
+    copyTexture(&targetTemp0, target);
     beginRenderPass(commandEncoder, target, false);
     // setup stencil rules
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 255);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
     // blend image
     uint32_t blendMethodInd = (uint32_t)blendMethod;
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, renderData->bindGroupPicture, 0, nullptr);
-    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, storageTemp0.bindGroupTexure, 0, nullptr);
+    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 3, targetTemp0.bindGroupTexure, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.image_blend[blendMethodInd]);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
 };
 
 
@@ -594,23 +676,23 @@ void WgCompositor::clipImage(WgContext& context, WgRenderDataPicture* renderData
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
     // merge depth and stencil buffer
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[128], 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.merge_depth_stencil);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
     // draw image
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, bindGroupViewMat, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, renderData->bindGroupPicture, 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.image);
-    renderData->meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &renderData->meshData);
 }
 
 
-void WgCompositor::drawScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose)
+void WgCompositor::drawScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose)
 {
     assert(scene);
     assert(compose);
@@ -622,19 +704,19 @@ void WgCompositor::drawScene(WgContext& context, WgRenderStorage* scene, WgCompo
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, scene->bindGroupTexure, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, bindGroupOpacities[compose->opacity], 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.scene);
-    meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &meshDataBlit);
 }
 
 
-void WgCompositor::blendScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose)
+void WgCompositor::blendScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose)
 {
     assert(scene);
     assert(compose);
     assert(currentTarget);
-    // copy current render target data to dst storage
+    // copy current render target data to dst target
-    WgRenderStorage *target = currentTarget;
+    WgRenderTarget *target = currentTarget;
     endRenderPass();
-    copyTexture(&storageTemp0, target);
+    copyTexture(&targetTemp0, target);
     beginRenderPass(commandEncoder, target, false);
     // blend scene
     uint32_t blendMethodInd = (uint32_t)compose->blend;
@@ -642,10 +724,10 @@ void WgCompositor::blendScene(WgContext& context, WgRenderStorage* scene, WgComp
     wgpuRenderPassEncoderSetScissorRect(renderPassEncoder, rect.x(), rect.y(), rect.w(), rect.h());
     wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 0, scene->bindGroupTexure, 0, nullptr);
-    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, storageTemp0.bindGroupTexure, 0, nullptr);
+    wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, targetTemp0.bindGroupTexure, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[compose->opacity], 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.scene_blend[blendMethodInd]);
-    meshData.drawImage(context, renderPassEncoder);
+    drawMeshImage(context, &meshDataBlit);
 }
 
 
@@ -658,13 +740,13 @@ void WgCompositor::markupClipPath(WgContext& context, WgRenderDataShape* renderD
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 255);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.direct);
         ARRAY_FOREACH(p, renderData->meshGroupStrokes.meshes)
-            (*p)->draw(context, renderPassEncoder);
+            drawMesh(context, (*p));
     } else {
         WGPURenderPipeline stencilPipeline = (renderData->fillRule == FillRule::NonZero) ? pipelines.nonzero : pipelines.evenodd;
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, stencilPipeline);
         ARRAY_FOREACH(p, renderData->meshGroupShapes.meshes)
-            (*p)->drawFan(context, renderPassEncoder);
+            drawMeshFan(context, (*p));
     }
 }
 
@@ -686,7 +768,7 @@ void WgCompositor::renderClipPath(WgContext& context, WgRenderDataPaint* paint)
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData0->bindGroupPaint, 0, nullptr);
     wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[128], 0, nullptr);
     wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.copy_stencil_to_depth);
-    renderData0->meshDataBBox.drawFan(context, renderPassEncoder);
+    drawMeshFan(context, &renderData0->meshDataBBox);
     // merge clip pathes with AND logic
     for (auto p = paint->clips.begin() + 1; p < paint->clips.end(); ++p) {
         // get render data
@@ -698,31 +780,31 @@ void WgCompositor::renderClipPath(WgContext& context, WgRenderDataPaint* paint)
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[190], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.copy_stencil_to_depth_interm);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
         // copy depth to stencil
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 1);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[190], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.copy_depth_to_stencil);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
         // clear depth current (keep stencil)
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[255], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.clear_depth);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
         // clear depth original (keep stencil)
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData0->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[255], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.clear_depth);
-        renderData0->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData0->meshDataBBox);
         // copy stencil to depth (clear stencil)
         wgpuRenderPassEncoderSetStencilReference(renderPassEncoder, 0);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[128], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.copy_stencil_to_depth);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
     }
 }
 
@@ -743,12 +825,12 @@ void WgCompositor::clearClipPath(WgContext& context, WgRenderDataPaint* paint)
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 1, renderData->bindGroupPaint, 0, nullptr);
         wgpuRenderPassEncoderSetBindGroup(renderPassEncoder, 2, bindGroupOpacities[255], 0, nullptr);
         wgpuRenderPassEncoderSetPipeline(renderPassEncoder, pipelines.clear_depth);
-        renderData->meshDataBBox.drawFan(context, renderPassEncoder);
+        drawMeshFan(context, &renderData->meshDataBBox);
     }
 }
 
 
-bool WgCompositor::gaussianBlur(WgContext& context, WgRenderStorage* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose)
+bool WgCompositor::gaussianBlur(WgContext& context, WgRenderTarget* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose)
 {
     assert(dst);
     assert(params);
@@ -758,8 +840,8 @@ bool WgCompositor::gaussianBlur(WgContext& context, WgRenderStorage* dst, const
     auto renderData = (WgRenderDataEffectParams*)params->rd;
     auto aabb = compose->aabb;
     auto viewport = compose->rdViewport;
-    WgRenderStorage* sbuff = dst;
+    WgRenderTarget* sbuff = dst;
-    WgRenderStorage* dbuff = &storageTemp0;
+    WgRenderTarget* dbuff = &targetTemp0;
 
     // begin compute pass
     WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass gaussian blur" };
@@ -791,14 +873,14 @@ bool WgCompositor::gaussianBlur(WgContext& context, WgRenderStorage* dst, const
     wgpuComputePassEncoderRelease(computePassEncoder);
 
     // if final result stored in intermidiate buffer we must copy result to destination buffer
-    if (sbuff == &storageTemp0) 
+    if (sbuff == &targetTemp0) 
         copyTexture(sbuff, dbuff, aabb);
 
     return true;
 }
 
 
-bool WgCompositor::dropShadow(WgContext& context, WgRenderStorage* dst, const RenderEffectDropShadow* params, const WgCompose* compose)
+bool WgCompositor::dropShadow(WgContext& context, WgRenderTarget* dst, const RenderEffectDropShadow* params, const WgCompose* compose)
 {
     assert(dst);
     assert(params);
@@ -811,9 +893,9 @@ bool WgCompositor::dropShadow(WgContext& context, WgRenderStorage* dst, const Re
     auto viewport = compose->rdViewport;
 
     { // apply blur
-        copyTexture(&storageTemp1, dst, aabb);
+        copyTexture(&targetTemp1, dst, aabb);
-        WgRenderStorage* sbuff = &storageTemp1;
+        WgRenderTarget* sbuff = &targetTemp1;
-        WgRenderStorage* dbuff = &storageTemp0;
+        WgRenderTarget* dbuff = &targetTemp0;
         WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass drop shadow blur" };
         WGPUComputePassEncoder computePassEncoder = wgpuCommandEncoderBeginComputePass(commandEncoder, &computePassDesc);
         // horizontal blur
@@ -837,7 +919,7 @@ bool WgCompositor::dropShadow(WgContext& context, WgRenderStorage* dst, const Re
     }
     
     { // blend origin (temp0), shadow (temp1) to destination
-        copyTexture(&storageTemp0, dst, aabb);
+        copyTexture(&targetTemp0, dst, aabb);
         WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass drop shadow blend" };
         WGPUComputePassEncoder computePassEncoder = wgpuCommandEncoderBeginComputePass(commandEncoder, &computePassDesc);
         wgpuComputePassEncoderSetBindGroup(computePassEncoder, 0, bindGroupStorageTemp, 0, nullptr);
@@ -854,7 +936,7 @@ bool WgCompositor::dropShadow(WgContext& context, WgRenderStorage* dst, const Re
 }
 
 
-bool WgCompositor::fillEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectFill* params, const WgCompose* compose)
+bool WgCompositor::fillEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectFill* params, const WgCompose* compose)
 {
     assert(dst);
     assert(params);
@@ -862,7 +944,7 @@ bool WgCompositor::fillEffect(WgContext& context, WgRenderStorage* dst, const Re
     assert(compose->rdViewport);
     assert(!renderPassEncoder);
 
-    copyTexture(&storageTemp0, dst, compose->aabb);
+    copyTexture(&targetTemp0, dst, compose->aabb);
     WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass fill" };
     WGPUComputePassEncoder computePassEncoder = wgpuCommandEncoderBeginComputePass(commandEncoder, &computePassDesc);
     wgpuComputePassEncoderSetBindGroup(computePassEncoder, 0, bindGroupStorageTemp, 0, nullptr);
@@ -878,7 +960,7 @@ bool WgCompositor::fillEffect(WgContext& context, WgRenderStorage* dst, const Re
 }
 
 
-bool WgCompositor::tintEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTint* params, const WgCompose* compose)
+bool WgCompositor::tintEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectTint* params, const WgCompose* compose)
 {
     assert(dst);
     assert(params);
@@ -886,7 +968,7 @@ bool WgCompositor::tintEffect(WgContext& context, WgRenderStorage* dst, const Re
     assert(compose->rdViewport);
     assert(!renderPassEncoder);
 
-    copyTexture(&storageTemp0, dst, compose->aabb);
+    copyTexture(&targetTemp0, dst, compose->aabb);
     WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass tint" };
     WGPUComputePassEncoder computePassEncoder = wgpuCommandEncoderBeginComputePass(commandEncoder, &computePassDesc);
     wgpuComputePassEncoderSetBindGroup(computePassEncoder, 0, bindGroupStorageTemp, 0, nullptr);
@@ -901,7 +983,7 @@ bool WgCompositor::tintEffect(WgContext& context, WgRenderStorage* dst, const Re
     return true;
 }
 
-bool WgCompositor::tritoneEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTritone* params, const WgCompose* compose)
+bool WgCompositor::tritoneEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectTritone* params, const WgCompose* compose)
 {
     assert(dst);
     assert(params);
@@ -909,7 +991,7 @@ bool WgCompositor::tritoneEffect(WgContext& context, WgRenderStorage* dst, const
     assert(compose->rdViewport);
     assert(!renderPassEncoder);
 
-    copyTexture(&storageTemp0, dst, compose->aabb);
+    copyTexture(&targetTemp0, dst, compose->aabb);
     WGPUComputePassDescriptor computePassDesc{ .label = "Compute pass tritone" };
     WGPUComputePassEncoder computePassEncoder = wgpuCommandEncoderBeginComputePass(commandEncoder, &computePassDesc);
     wgpuComputePassEncoderSetBindGroup(computePassEncoder, 0, bindGroupStorageTemp, 0, nullptr);

src/renderer/wg_engine/tvgWgCompositor.h

@@ -38,6 +38,8 @@ class WgCompositor
 private:
     // pipelines
     WgPipelines pipelines{};
+    // stage buffers
+    WgRenderDataStageBuffer stageBuffer{};
     // global stencil/depth buffer handles
     WGPUTexture texDepthStencil{};
     WGPUTextureView texViewDepthStencil{};
@@ -52,21 +54,26 @@ private:
     // current render pass handles
     WGPURenderPassEncoder renderPassEncoder{};
     WGPUCommandEncoder commandEncoder{};
-    WgRenderStorage* currentTarget{};
+    WgRenderTarget* currentTarget{};
-    // intermediate render storages
+    // intermediate render targets
-    WgRenderStorage storageTemp0;
+    WgRenderTarget targetTemp0;
-    WgRenderStorage storageTemp1;
+    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 WgRenderTarget* dst, const WgRenderTarget* src);
-    void copyTexture(const WgRenderStorage* dst, const WgRenderStorage* src, const RenderRegion& region);
+    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 +91,8 @@ private:
     void clipImage(WgContext& context, WgRenderDataPicture* renderData);
 
     // scenes
-    void drawScene(WgContext& context, WgRenderStorage* scene, WgCompose* compose);
+    void drawScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose);
-    void blendScene(WgContext& context, WgRenderStorage* src, 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 +107,32 @@ 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();
 
+    // request shapes for drawing (staging)
+        // stage data
+    void reset(WgContext& context);
+    void flush(WgContext& context);
+
+    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 renderScene(WgContext& context, WgRenderTarget* scene, WgCompose* compose);
-    void composeScene(WgContext& context, WgRenderStorage* src, WgRenderStorage* mask, WgCompose* compose);
+    void composeScene(WgContext& context, WgRenderTarget* src, WgRenderTarget* mask, WgCompose* compose);
 
-    // blit render storage to texture view (f.e. screen buffer)
+    // blit render target to texture view (f.e. screen buffer)
-    void blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderStorage* src, WGPUTextureView dstView);
+    void blit(WgContext& context, WGPUCommandEncoder encoder, WgRenderTarget* src, WGPUTextureView dstView);
 
     // effects
-    bool gaussianBlur(WgContext& context, WgRenderStorage* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose);
+    bool gaussianBlur(WgContext& context, WgRenderTarget* dst, const RenderEffectGaussianBlur* params, const WgCompose* compose);
-    bool dropShadow(WgContext& context, WgRenderStorage* dst, const RenderEffectDropShadow* params, const WgCompose* compose);
+    bool dropShadow(WgContext& context, WgRenderTarget* dst, const RenderEffectDropShadow* params, const WgCompose* compose);
-    bool fillEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectFill* params, const WgCompose* compose);
+    bool fillEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectFill* params, const WgCompose* compose);
-    bool tintEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTint* params, const WgCompose* compose);
+    bool tintEffect(WgContext& context, WgRenderTarget* dst, const RenderEffectTint* params, const WgCompose* compose);
-    bool tritoneEffect(WgContext& context, WgRenderStorage* dst, const RenderEffectTritone* 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/tvgWgRenderData.cpp

@@ -30,95 +30,88 @@
 // 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)
 {
     assert(vertexBuffer.count > 2);
-    vertexCount = vertexBuffer.count;
+    // setup vertex data
-    indexCount = (vertexBuffer.count - 2) * 3;
+    vbuffer.reserve(vertexBuffer.count);
-    context.allocateBufferVertex(bufferPosition, (float*)vertexBuffer.data, vertexCount * sizeof(float) * 2);
+    vbuffer.count = vertexBuffer.count;
-    context.allocateBufferIndexFan(vertexCount);
+    memcpy(vbuffer.data, vertexBuffer.data, sizeof(vertexBuffer.data[0])*vertexBuffer.count);
+    // setup tex coords data
+    tbuffer.clear();
+    context.allocateBufferIndexFan(vbuffer.count);
 }
 
 
 void WgMeshData::update(WgContext& context, const WgIndexedVertexBuffer& vertexBufferInd)
 {
     assert(vertexBufferInd.vcount > 2);
-    vertexCount = vertexBufferInd.vcount;
+    // setup vertex data
-    indexCount = vertexBufferInd.icount;
+    vbuffer.reserve(vertexBufferInd.vcount);
-    if (vertexCount > 0) context.allocateBufferVertex(bufferPosition, (float*)vertexBufferInd.vbuff, vertexCount * sizeof(float) * 2);
+    vbuffer.count = vertexBufferInd.vcount;
-    if (indexCount > 0) context.allocateBufferIndex(bufferIndex, vertexBufferInd.ibuff, indexCount * sizeof(uint32_t));
+    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)
 {
-    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));
+    // setup vertex data
-    context.allocateBufferIndexFan(vertexCount);
+    vbuffer.reserve(4);
+    vbuffer.count = 4;
+    memcpy(vbuffer.data, data, sizeof(data));
+    // setup tex coords data
+    tbuffer.clear();
+    context.allocateBufferIndexFan(vbuffer.count);
 }
 
 
 void WgMeshData::imageBox(WgContext& context, 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));
+    // setup vertex data
-    context.allocateBufferVertex(bufferTexCoord, tdata, sizeof(tdata));
+    vbuffer.reserve(4);
-    context.allocateBufferIndex(bufferIndex, idata, sizeof(idata));
+    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)
 {
-    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));
+    // setup vertex data
-    context.allocateBufferVertex(bufferTexCoord, tdata, sizeof(tdata));
+    vbuffer.reserve(4);
-    context.allocateBufferIndex(bufferIndex, idata, sizeof(idata));
+    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
 //***********************************************************************
@@ -681,3 +674,94 @@ void WgRenderDataEffectParamsPool::release(WgContext& context)
     mPool.clear();
     mList.clear();
 }
+
+//***********************************************************************
+// WgRenderDataStageBuffer
+//***********************************************************************
+
+void WgRenderDataStageBuffer::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]);
+    // 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 WgRenderDataStageBuffer::append(WgMeshDataGroup* meshDataGroup)
+{
+    ARRAY_FOREACH(p, meshDataGroup->meshes) append(*p);
+}
+
+
+void WgRenderDataStageBuffer::append(WgRenderDataShape* renderDataShape)
+{
+    append(&renderDataShape->meshGroupShapes);
+    append(&renderDataShape->meshGroupShapesBBox);
+    append(&renderDataShape->meshGroupStrokes);
+    append(&renderDataShape->meshGroupStrokesBBox);
+    append(&renderDataShape->meshDataBBox);
+    ARRAY_FOREACH(p, renderDataShape->clips)
+        append((WgRenderDataShape* )(*p));
+}
+
+
+void WgRenderDataStageBuffer::append(WgRenderDataPicture* renderDataPicture)
+{
+    append(&renderDataPicture->meshData);
+    ARRAY_FOREACH(p, renderDataPicture->clips)
+        append((WgRenderDataShape* )(*p));
+}
+
+
+void WgRenderDataStageBuffer::release(WgContext& context)
+{
+    context.releaseBuffer(vbuffer_gpu);
+    context.releaseBuffer(ibuffer_gpu);
+}
+
+
+void WgRenderDataStageBuffer::clear()
+{
+    vbuffer.clear();
+    ibuffer.clear();
+}
+
+
+void WgRenderDataStageBuffer::flush(WgContext& context) 
+{
+    context.allocateBufferVertex(vbuffer_gpu, (float *)vbuffer.data, vbuffer.count);
+    context.allocateBufferIndex(ibuffer_gpu, (uint32_t *)ibuffer.data, ibuffer.count);
+}
+
+
+void WgRenderDataStageBuffer::bind(WGPURenderPassEncoder renderPass, size_t voffset, size_t toffset)
+{
+    wgpuRenderPassEncoderSetVertexBuffer(renderPass, 0, vbuffer_gpu, voffset, vbuffer.count - voffset);
+    wgpuRenderPassEncoderSetVertexBuffer(renderPass, 1, vbuffer_gpu, toffset, vbuffer.count - toffset);
+    wgpuRenderPassEncoderSetIndexBuffer(renderPass, ibuffer_gpu, WGPUIndexFormat_Uint32, 0, ibuffer.count);
+}

src/renderer/wg_engine/tvgWgRenderData.h

@@ -28,22 +28,19 @@
 #include "tvgWgShaderTypes.h"
 
 struct WgMeshData {
-    WGPUBuffer bufferPosition{};
+    Array<Point> vbuffer;
-    WGPUBuffer bufferTexCoord{};
+    Array<Point> tbuffer;
-    WGPUBuffer bufferIndex{};
+    Array<uint32_t> ibuffer;
-    size_t vertexCount{};
+    size_t voffset{};
-    size_t indexCount{};
+    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);
+    void release(WgContext& context) {};
 };
 
 class WgMeshDataPool {
@@ -224,4 +221,23 @@ public:
     void release(WgContext& context);
 };
 
+class WgRenderDataStageBuffer {
+private:
+    Array<uint8_t> vbuffer;
+    Array<uint8_t> ibuffer;
+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);
+    void bind(WGPURenderPassEncoder renderPass, size_t voffset, size_t toffset);
+};
+
 #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;
+        renderTarget = new WgRenderTarget;
-        renderStorage->initialize(context, width, height);
+        renderTarget->initialize(context, width, height);
-        list.push(renderStorage);
+        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<WgRenderTarget*> list;
-    Array<WgRenderStorage*> pool;
+    Array<WgRenderTarget*> pool;
     uint32_t width{};
     uint32_t height{};
 public:
-    WgRenderStorage* allocate(WgContext& context);
+    WgRenderTarget* allocate(WgContext& context);
-    void free(WgContext& context, WgRenderStorage* renderTarget);
+    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

@@ -47,13 +47,13 @@ void WgRenderer::release()
     mRenderDataEffectParamsPool.release(mContext);
     WgMeshDataPool::gMeshDataPool->release(mContext);
 
-    // clear render storage pool
+    // clear render  pool
-    mRenderStoragePool.release(mContext);
+    mRenderTargetPool.release(mContext);
 
     // clear rendering tree stacks
-    mCompositorStack.clear();
+    mCompositorList.clear();
-    mRenderStorageStack.clear();
+    mRenderTargetStack.clear();
-    mRenderStorageRoot.release(mContext);
+    mRenderTargetRoot.release(mContext);
 
     // release context handles
     mCompositor.release(mContext);
@@ -195,50 +195,71 @@ RenderData WgRenderer::prepare(RenderSurface* surface, RenderData data, const Ma
 
 bool WgRenderer::preRender()
 {
+    // invalidate context
     if (mContext.invalid()) return false;
-
+    // reset stage data
-    // push rot render storage to the render tree stack
+    mCompositor.reset(mContext);
-    assert(mRenderStorageStack.count == 0);
+    // push root render target to the render tree stack
-    mRenderStorageStack.push(&mRenderStorageRoot);
+    assert(mRenderTargetStack.count == 0);
-    // create command encoder for drawing
+    mRenderTargetStack.push(&mRenderTargetRoot);
-    WGPUCommandEncoderDescriptor commandEncoderDesc{};
+    // create root compose settings
-    mCommandEncoder = wgpuDeviceCreateCommandEncoder(mContext.device, &commandEncoderDesc);
+    WgCompose* compose = new WgCompose();
-    // start root render pass
+    compose->aabb = { { 0, 0 }, { (int32_t)mTargetSurface.w, (int32_t)mTargetSurface.h } };
-    mCompositor.beginRenderPass(mCommandEncoder, mRenderStorageStack.last(), true);
+    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
+    WgPaintTask* paintTask = new WgPaintTask((WgRenderDataPaint*)data, mBlendMethod);
-    mCompositor.renderShape(mContext, (WgRenderDataShape*)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
+    WgPaintTask* paintTask = new WgPaintTask((WgRenderDataPaint*)data, mBlendMethod);
-    mCompositor.renderImage(mContext, (WgRenderDataPicture*)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
+    // flush stage data to gpu
-    mCompositor.endRenderPass();
+    mCompositor.flush(mContext);
-    // release command encoder
+    // create command encoder for drawing
-    const WGPUCommandBufferDescriptor commandBufferDesc{};
+    WGPUCommandEncoder commandEncoder = mContext.createCommandEncoder();
-    WGPUCommandBuffer commandsBuffer = wgpuCommandEncoderFinish(mCommandEncoder, &commandBufferDesc);
+    // run rendering (all the fun is here)
-    wgpuQueueSubmit(mContext.queue, 1, &commandsBuffer);
+    WgSceneTask* sceneTaskRoot = mSceneTaskStack.last();
-    wgpuCommandBufferRelease(commandsBuffer);
+    sceneTaskRoot->run(mContext, mCompositor, commandEncoder);
-    wgpuCommandEncoderRelease(mCommandEncoder);
+    // execute and release command encoder
-    // pop root render storage to the render tree stack
+    mContext.submitCommandEncoder(commandEncoder);
-    mRenderStorageStack.pop();
+    mContext.releaseCommandEncoder(commandEncoder);
-    assert(mRenderStorageStack.count == 0);
+    // clear the render tasks tree
-    // clear viewport list and store allocated handles to pool
+    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();
@@ -325,18 +346,11 @@ bool WgRenderer::sync()
     WGPUTextureView dstTextureView = mContext.createTextureView(dstTexture);
 
     // create command encoder
-    const WGPUCommandEncoderDescriptor commandEncoderDesc{};
+    WGPUCommandEncoder commandEncoder = mContext.createCommandEncoder();
-    WGPUCommandEncoder commandEncoder = wgpuDeviceCreateCommandEncoder(mContext.device, &commandEncoderDesc);
-
     // show root offscreen buffer
-    mCompositor.blit(mContext, commandEncoder, &mRenderStorageRoot, dstTextureView);
+    mCompositor.blit(mContext, commandEncoder, &mRenderTargetRoot, dstTextureView);
-
+    mContext.submitCommandEncoder(commandEncoder);
-    // release command encoder
+    mContext.releaseCommandEncoder(commandEncoder);
-    const WGPUCommandBufferDescriptor commandBufferDesc{};
-    WGPUCommandBuffer commandsBuffer = wgpuCommandEncoderFinish(commandEncoder, &commandBufferDesc);
-    wgpuQueueSubmit(mContext.queue, 1, &commandsBuffer);
-    wgpuCommandBufferRelease(commandsBuffer);
-    wgpuCommandEncoderRelease(commandEncoder);
 
     // release dest buffer view
     mContext.releaseTextureView(dstTextureView);
@@ -367,8 +381,8 @@ bool WgRenderer::target(WGPUDevice device, WGPUInstance instance, void* target,
         mContext.initialize(instance, device);
 
         // initialize render tree instances
-        mRenderStoragePool.initialize(mContext, width, height);
+        mRenderTargetPool.initialize(mContext, width, height);
-        mRenderStorageRoot.initialize(mContext, width, height);
+        mRenderTargetRoot.initialize(mContext, width, height);
         mCompositor.initialize(mContext, width, height);
 
         // store target properties
@@ -387,12 +401,12 @@ 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);
+        mRenderTargetPool.release(mContext);
-        mRenderStorageRoot.release(mContext);
+        mRenderTargetRoot.release(mContext);
         clearTargets();
 
-        mRenderStoragePool.initialize(mContext, width, height);
+        mRenderTargetPool.initialize(mContext, width, height);
-        mRenderStorageRoot.initialize(mContext, width, height);
+        mRenderTargetRoot.initialize(mContext, width, height);
         mCompositor.resize(mContext, width, height);
 
         // store target properties
@@ -447,7 +461,7 @@ 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;
 }
 
@@ -459,57 +473,52 @@ bool WgRenderer::beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_
     compose->method = method;
     compose->opacity = opacity;
     compose->blend = mBlendMethod;
-    // end current render pass
+    WgSceneTask* sceneTaskCurrent = mSceneTaskStack.last();
-    mCompositor.endRenderPass();
+    // allocate new render target and push to the render tree stack
-    // allocate new render storage and push to the render tree stack
+    WgRenderTarget* renderTarget = mRenderTargetPool.allocate(mContext);
-    WgRenderStorage* storage = mRenderStoragePool.allocate(mContext);
+    mRenderTargetStack.push(renderTarget);
-    mRenderStorageStack.push(storage);
+    // create and setup new scene task
-    // begin newly added render pass
+    WgSceneTask* sceneTask = new WgSceneTask(renderTarget, compose, sceneTaskCurrent);
-    WGPUColor color{};
+    sceneTaskCurrent->children.push(sceneTask);
-    if ((compose->method == MaskMethod::None) && (compose->blend != BlendMethod::Normal)) color = { 1.0, 1.0, 1.0, 0.0 };
+    mRenderTaskList.push(sceneTask);
-    mCompositor.beginRenderPass(mCommandEncoder, mRenderStorageStack.last(), true, color);
+    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) {
+    if (cmp->method == MaskMethod::None) {
-        // get source and destination render storages
+        // get source and destination render targets
-        WgRenderStorage* src = mRenderStorageStack.last();
+        WgRenderTarget* src = mRenderTargetStack.last();
-        mRenderStorageStack.pop();
+        mRenderTargetStack.pop();
-        WgRenderStorage* dst = mRenderStorageStack.last();
+        // pop source scene
-        // begin previous render pass
+        WgSceneTask* srcScene = mSceneTaskStack.last();
-        mCompositor.beginRenderPass(mCommandEncoder, dst, false);
+        mSceneTaskStack.pop();
-        // apply composition
+        // setup render target compose destitations
-        mCompositor.renderScene(mContext, src, comp);
+        srcScene->renderTargetDst = mSceneTaskStack.last()->renderTarget;
+        srcScene->renderTargetMsk =  nullptr;
         // back render targets to the pool
-        mRenderStoragePool.free(mContext, src);
+        mRenderTargetPool.free(mContext, src);
-    } else { // finish composition
+    } else { // finish scene composition
-        // get source, mask and destination render storages
+        // get source, mask and destination render targets
-        WgRenderStorage* src = mRenderStorageStack.last();
+        WgRenderTarget* src = mRenderTargetStack.last();
-        mRenderStorageStack.pop();
+        mRenderTargetStack.pop();
-        WgRenderStorage* msk = mRenderStorageStack.last();
+        WgRenderTarget* msk = mRenderTargetStack.last();
-        mRenderStorageStack.pop();
+        mRenderTargetStack.pop();
-        WgRenderStorage* dst = mRenderStorageStack.last();
+        // get source and mask scenes
-        // begin previous render pass
+        WgSceneTask* srcScene = mSceneTaskStack.last();
-        mCompositor.beginRenderPass(mCommandEncoder, dst, false);
+        mSceneTaskStack.pop();
-        // apply composition
+        WgSceneTask* mskScene = mSceneTaskStack.last();
-        mCompositor.composeScene(mContext, src, msk, comp);
+        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);
+        mRenderTargetPool.free(mContext, src);
-        mRenderStoragePool.free(mContext, msk);
+        mRenderTargetPool.free(mContext, msk);
     }
-
-    // delete current compositor settings
-    delete mCompositorStack.last();
-    mCompositorStack.pop();
-
     return true;
 }
 
@@ -606,20 +615,8 @@ 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
+    mSceneTaskStack.last()->effect = effect;
-    mCompositor.endRenderPass();
+    return true;
-    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;
 }
 
 

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;
+    WgRenderTarget mRenderTargetRoot;
-    Array<WgCompose*> mCompositorStack;
+    Array<WgCompose*> mCompositorList;
-    Array<WgRenderStorage*> mRenderStorageStack;
+    Array<WgRenderTarget*> mRenderTargetStack;
     Array<WgRenderDataViewport*> mRenderDataViewportList;
+    Array<WgSceneTask*> mSceneTaskStack;
+    Array<WgRenderTask*> mRenderTaskList;
 
-    // render storage pool
+    // render target pool
-    WgRenderStoragePool mRenderStoragePool;
+    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