Android GUI 的更新过程

Android GUI 的更新过程

2011-05-25 17:41

 

From: http://www.linuxgraphics.cn/android/gui_update_flow.html

组件图

相关组件如下图所示:

• ViewRoot

在private void draw(boolean fullRedrawNeeded)中，会调用lockCanvas，从而获取一个Canvas对象，然后调用递归调用子窗口(View)的draw函数去绘制自己，最后调用unlockCanvasAndPost让Surface把自己更新到屏幕上。

    canvas = surface.lockCanvas(dirty);    mView.draw(canvas);    surface.unlockCanvasAndPost(canvas);

• android/view/Surface

它主要对一些native函数的包装。

    private native Canvas lockCanvasNative(Rect dirty);    public native   void unlockCanvasAndPost(Canvas canvas);

• android_view_Surface.cpp

这是native函数的实现。

Surface_lockCanvas它先锁住Surface，从而获取Surface相关的信息，如果格式、宽度、高度和像素缓冲区。然后创建一个bitmap，这个bitmap和View共享一个像素缓冲区，这样View就能直接把自己绘制到Surface上了。

    status_t err = surface->lock(&info, &dirtyRegion);    bitmap.setPixels(info.bits);    nativeCanvas->setBitmapDevice(bitmap);Surface_unlockCanvasAndPost它断开SkCanvas与Surface之间的关系，然后调用Surface的unlockAndPost。

    nativeCanvas->setBitmapDevice(SkBitmap());    status_t err = surface->unlockAndPost();

• surfaceflinger_client/Surface.cpp

Surface::lock调用dequeueBuffer获取一个可用的Buffer，dequeueBuffer会调用SharedBufferClient::dequeue等到backbuffer可用，然后锁住这个Buffer并填充相关信息。

    status_t err = dequeueBuffer(&backBuffer);    err = lockBuffer(backBuffer.get());Surface::unlockAndPost先unlock Buffer，然后调queueBuffer显示Buffer。

    status_t err = mLockedBuffer->unlock();    err = queueBuffer(mLockedBuffer.get());Surface::queueBuffer是比较有意思的，它设置更新的区域，然后通知服务端(SurfaceFlinger)。    mSharedBufferClient->setDirtyRegion(bufIdx, mDirtyRegion);    err = mSharedBufferClient->queue(bufIdx);    client->signalServer();通过binder发送请求给服务：

    virtual void signal() const    {        Parcel data, reply;        data.writeInterfaceToken(ISurfaceComposer::getInterfaceDescriptor());        remote()->transact(BnSurfaceComposer::SIGNAL, data, &reply, IBinder::FLAG_ONEWAY);    }

• SurfaceFlinger

    BnSurfaceComposer::onTransact    case SIGNAL: {        CHECK_INTERFACE(ISurfaceComposer, data, reply);        signal();    } break;这是SurfaceFlinger中处理SIGNAL的代码，它调用SurfaceFlinger::signal，再调用 SurfaceFlinger::signalEvent，最后调用MessageQueue::invalidate唤醒是 SurfaceFlinger的主循环。

在主循环中waitForEvent返回，再handleRepaint去合成个Layer/Surface。

    bool SurfaceFlinger::threadLoop()    {        waitForEvent();        handleRepaint();        postFramebuffer();    }SurfaceFlinger::handleRepaint会调用composeSurfaces把需要绘制各个Surface合并起来，绘制到FrameBuffer的BackBuffer上。

    for (size_t i=0 ; i<count ; ++i) {        const sp<LayerBase>& layer = layers[i];        const Region& visibleRegion(layer->visibleRegionScreen);        if (!visibleRegion.isEmpty())  {            /* broncho cy add */            #ifdef USEOVERLAY            mToppest[1] = mToppest[0];            mToppest[0] = layer->getIdentity();            #endif            /* broncho end */            const Region clip(dirty.intersect(visibleRegion));            if (!clip.isEmpty()) {                layer->draw(clip);            }        }    }SurfaceFlinger::postFramebuffer最后把FrameBuffer的BackBuffer显示到屏幕上：

hw.flip(mInvalidRegion); -->eglSwapBuffers(dpy, surface);