前言
在上一篇文章 《Android图形渲染原理(上)》中,详细的讲解了图像消费者,我们已经了解了Android中的图像元数据是如何被SurfaceFlinger,HWComposer或者OpenGL ES消费的,那么,图像元数据又是怎么生成的呢?这一篇文章就来详细介绍Android中的图像生产者——SKIA,OPenGL ES,Vulkan,他们是Android中最重要的三支画笔。
图像生产者
OpenGL ES
什么是OpenGL呢?OpenGL是一套图像编程接口,对于开发者来说,其实就是一套C语言编写的API接口,通过调用这些函数,便可以调用显卡来进行计算机的图形开发。虽然OpenGL是一套API接口,但它并没有具体的实现这些接口,接口的实现是由显卡的驱动程序来完成的。在前一篇文章中介绍过,显卡驱动是其他模块和显卡沟通的入口,开发者通过调用OpenGL的图像编程接口发出渲染命令,这些渲染命令被称为DrawCall,显卡驱动会将渲染命令翻译能GPU能理解的数据,然后通知GPU读取数据进行操作。OpenGL ES又是什么呢?它是为了更好的适应嵌入式等硬件较差的设备,推出的OpenGL的剪裁版,基本和OpenGL是一致的。Android从4.0开始默认开启硬件加速,也就是默认使用OpenGL ES来进行图形的生成和渲染工作。
我们接着来看看如何使用OpenGL ES。
如何使用OpenGL ES?
想要在Android上使用OpenGL ES,我们要先了解EGL。OpenGL虽然是跨平台的,但是在各个平台上也不能直接使用,因为每个平台的窗口都是不一样的,而EGL就是适配Android本地窗口系统和OpenGL ES桥接层。
OpenGL ES 定义了平台无关的 GL 绘图指令,EGL则定义了控制 displays,contexts 以及 surfaces 的统一的平台接口
那么如何使用EGL和OpenGL ES生成图形呢?其实比较简单,主要有这三步
- EGL初始化Display,Context和Surface
- OpenGL ES调用绘制指令
- EGL提交绘制后的buffer
我们详细来看一下每一步的流程
1,EGL进行初始化:主要初始化Display,Context 和Surface三个元素就可以了。
- Display(EGLDisplay) 是对实际显示设备的抽象
//创建于本地窗口系统的连接
EGLDisplay display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
//初始化display
eglInitialize(display, NULL, NULL);
- Context (EGLContext) 存储 OpenGL ES绘图的一些状态信息
/* create an EGL rendering context */
context = eglCreateContext(display, config, EGL_NO_CONTEXT, NULL);
- Surface(EGLSurface)是对用来存储图像的内存区域
//设置Surface配置
eglChooseConfig(display, attribute_list, &config, 1, &num_config);
//创建本地窗口
native_window = createNativeWindow();
//创建surface
surface = eglCreateWindowSurface(display, config, native_window, NULL);
- 初始化完成后,需要绑定上下文
//绑定上下文
eglMakeCurrent(display, surface, surface, context);
2,OpenGL ES调用绘制指令:主要通过使用 OpenGL ES API ——gl_*(),接口进行绘制图形
//绘制点
glBegin(GL_POINTS);
glVertex3f(0.7f,-0.5f,0.0f); //入参为三维坐标
glVertex3f(0.6f,-0.7f,0.0f);
glVertex3f(0.6f,-0.8f,0.0f);
glEnd();
//绘制线
glBegin(GL_LINE_STRIP);
glVertex3f(-1.0f,1.0f,0.0f);
glVertex3f(-0.5f,0.5f,0.0f);
glVertex3f(-0.7f,0.5f,0.0f);
glEnd();
//……
3,EGL提交绘制后的buffer:通过eglSwapBuffer()进行双缓冲buffer的切换
EGLBoolean res = eglSwapBuffers(mDisplay, mSurface);
swapBuffer切换缓冲区buffer后,显卡就会对Buffer中的图像进行渲染处理。此时,我们的图像就能显示出来了。
我们看一个完整的使用流程Demo
#include <stdlib.h>
#include <unistd.h>
#include <EGL/egl.h>
#include <GLES/gl.h>
typedef ... NativeWindowType;
extern NativeWindowType createNativeWindow(void);
static EGLint const attribute_list[] = {
EGL_RED_SIZE, 1,
EGL_GREEN_SIZE, 1,
EGL_BLUE_SIZE, 1,
EGL_NONE
};
int main(int argc, char ** argv)
{
EGLDisplay display;
EGLConfig config;
EGLContext context;
EGLSurface surface;
NativeWindowType native_window;
EGLint num_config;
/* get an EGL display connection */
display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
/* initialize the EGL display connection */
eglInitialize(display, NULL, NULL);
/* get an appropriate EGL frame buffer configuration */
eglChooseConfig(display, attribute_list, &config, 1, &num_config);
/* create an EGL rendering context */
context = eglCreateContext(display, config, EGL_NO_CONTEXT, NULL);
/* create a native window */
native_window = createNativeWindow();
/* create an EGL window surface */
surface = eglCreateWindowSurface(display, config, native_window, NULL);
/* connect the context to the surface */
eglMakeCurrent(display, surface, surface, context);
/* clear the color buffer */
glClearColor(1.0, 1.0, 0.0, 1.0);
glClear(GL_COLOR_BUFFER_BIT);
glFlush();
eglSwapBuffers(display, surface);
sleep(10);
return EXIT_SUCCESS;
}
介绍完EGL和OpenGL的使用方式了,我们可以开始看Android是如何通过它进行界面的绘制的,这里会列举两个场景:开机动画,硬件加速来详细的讲解OpenGL ES作为图像生产者,是如何生产,即如何绘制图像的。
OpenGL ES播放开机动画
当Android系统启动时,会启动Init进程,Init进程会启动Zygote,ServerManager,SurfaceFlinger等服务。随着SurfaceFlinger的启动,我们的开机动画也会开始启动。先看看SurfaceFlinger的初始化函数。
//文件-->/frameworks/native/services/surfaceflinger/SurfaceFlinger.cpp
void SurfaceFlinger::init() {
...
mStartBootAnimThread = new StartBootAnimThread();
if (mStartBootAnimThread->Start() != NO_ERROR) {
ALOGE("Run StartBootAnimThread failed!");
}
}
//文件-->/frameworks/native/services/surfaceflinger/StartBootAnimThread.cpp
status_t StartBootAnimThread::Start() {
return run("SurfaceFlinger::StartBootAnimThread", PRIORITY_NORMAL);
}
bool StartBootAnimThread::threadLoop() {
property_set("service.bootanim.exit", "0");
property_set("ctl.start", "bootanim");
// Exit immediately
return false;
}
从上面的代码可以看到,SurfaceFlinger的init函数中会启动BootAnimThread线程,BootAnimThread线程会通过property_set来发送通知,它是一种Socket方式的IPC通信机制,对Android IPC通信感兴趣的可以看看我的这篇文章《掌握Android进程间通信机制》,这里就不过多讲解了。init进程会接收到bootanim的通知,然后启动我们的动画线程BootAnimation。
了解了前面的流程,我们开始看BootAnimation这个类,Android的开机动画的逻辑都在这个类中。我们先看看构造函数和onFirsetRef函数,这是这个类创建时最先执行的两个函数:
//文件-->/frameworks/base/cmds/bootanimation/BootAnimation.cpp
BootAnimation::BootAnimation() : Thread(false), mClockEnabled(true), mTimeIsAccurate(false),
mTimeFormat12Hour(false), mTimeCheckThread(NULL) {
//创建SurfaceComposerClient
mSession = new SurfaceComposerClient();
//……
}
void BootAnimation::onFirstRef() {
status_t err = mSession->linkToComposerDeath(this);
if (err == NO_ERROR) {
run("BootAnimation", PRIORITY_DISPLAY);
}
}
构造函数中创建了SurfaceComposerClient,SurfaceComposerClient是SurfaceFlinger的客户端代理,我们可以通过它来和SurfaceFlinger建立通信。构造函数执行完后就会执行onFirsetRef()函数,这个函数会启动BootAnimation线程
接着看BootAnimation线程的初始化函数readyToRun。
//文件-->/frameworks/base/cmds/bootanimation/BootAnimation.cpp
status_t BootAnimation::readyToRun() {
mAssets.addDefaultAssets();
sp<IBinder> dtoken(SurfaceComposerClient::getBuiltInDisplay(
ISurfaceComposer::eDisplayIdMain));
DisplayInfo dinfo;
//获取屏幕信息
status_t status = SurfaceComposerClient::getDisplayInfo(dtoken, &dinfo);
if (status)
return -1;
// 通知SurfaceFlinger创建Surface,创建成功会返回一个SurfaceControl代理
sp<SurfaceControl> control = session()->createSurface(String8("BootAnimation"),
dinfo.w, dinfo.h, PIXEL_FORMAT_RGB_565);
SurfaceComposerClient::openGlobalTransaction();
//设置这个layer在SurfaceFlinger中的层级顺序
control->setLayer(0x40000000);
//获取surface
sp<Surface> s = control->getSurface();
// 以下是EGL的初始化流程
const EGLint attribs[] = {
EGL_RED_SIZE, 8,
EGL_GREEN_SIZE, 8,
EGL_BLUE_SIZE, 8,
EGL_DEPTH_SIZE, 0,
EGL_NONE
};
EGLint w, h;
EGLint numConfigs;
EGLConfig config;
EGLSurface surface;
EGLContext context;
//步骤1:获取Display
EGLDisplay display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
//步骤2:初始化EGL
eglInitialize(display, 0, 0);
//步骤3:选择参数
eglChooseConfig(display, attribs, &config, 1, &numConfigs);
//步骤4:传入SurfaceFlinger生成的surface,并以此构造EGLSurface
surface = eglCreateWindowSurface(display, config, s.get(), NULL);
//步骤5:构造egl上下文
context = eglCreateContext(display, config, NULL, NULL);
//步骤6:绑定EGL上下文
if (eglMakeCurrent(display, surface, surface, context) == EGL_FALSE)
return NO_INIT;
//……
}
通过readyToRun函数可以看到,里面主要做了两件事情:初始化Surface,初始化EGL,EGL的初始化流程和上面OpenGL ES使用中讲的流程是一样的,这里就不详细讲了,主要简单介绍一下Surface初始化的流程,详细的流程会在下一篇文章图像缓冲区中讲,它的步骤如下:
- 创建SurfaceComponentClient
- 通过SurfaceComponentClient通知SurfaceFlinger创建Surface,并返回SurfaceControl
- 有了SurfaceControl之后,我们就可以设置这块Surface的层级等属性,并能获取到这块Surface。
- 获取到Surface后,将Surface绑定到EGL中去
Surface也创建好了,EGL也创建好了,此时我们就可以通过OpenGL来生成图像——也就是开机动画了,我们接着看看线程的执行方法threadLoop函数中是如何播放的动画的。
//文件-->/frameworks/base/cmds/bootanimation/BootAnimation.cpp
bool BootAnimation::threadLoop()
{
bool r;
if (mZipFileName.isEmpty()) {
r = android(); //Android默认动画
} else {
r = movie(); //自定义动画
}
//动画播放完后的释放工作
eglMakeCurrent(mDisplay, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
eglDestroyContext(mDisplay, mContext);
eglDestroySurface(mDisplay, mSurface);
mFlingerSurface.clear();
mFlingerSurfaceControl.clear();
eglTerminate(mDisplay);
eglReleaseThread();
IPCThreadState::self()->stopProcess();
return r;
}
函数中会判断是否有自定义的开机动画文件,如果没有就播放默认的动画,有就播放自定义的动画,播放完成后就是释放和清除的操作。默认动画和自定义动画的播放方式其实差不多,我们以自定义动画为例,看看具体的实现流程。
//文件-->/frameworks/base/cmds/bootanimation/BootAnimation.cpp
bool BootAnimation::movie()
{
//根据文件路径加载动画文件
Animation* animation = loadAnimation(mZipFileName);
if (animation == NULL)
return false;
//……
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
// 调用OpenGL清理屏幕
glShadeModel(GL_FLAT);
glDisable(GL_DITHER);
glDisable(GL_SCISSOR_TEST);
glDisable(GL_BLEND);
glBindTexture(GL_TEXTURE_2D, 0);
glEnable(GL_TEXTURE_2D);
glTexEnvx(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
//……
//播放动画
playAnimation(*animation);
//……
//释放动画
releaseAnimation(animation);
return false;
}
movie函数主要做的事情如下
- 通过文件路径加载动画
- 调用OpenGL做清屏操作
- 调用playAnimation函数播放动画。
- 停止播放动画后通过releaseAnimation释放资源
我们接着看playAnimation函数
//文件-->/frameworks/base/cmds/bootanimation/BootAnimation.cpp
bool BootAnimation::playAnimation(const Animation& animation)
{
const size_t pcount = animation.parts.size();
nsecs_t frameDuration = s2ns(1) / animation.fps;
const int animationX = (mWidth - animation.width) / 2;
const int animationY = (mHeight - animation.height) / 2;
//遍历动画片段
for (size_t i=0 ; i<pcount ; i++) {
const Animation::Part& part(animation.parts[i]);
const size_t fcount = part.frames.size();
glBindTexture(GL_TEXTURE_2D, 0);
// Handle animation package
if (part.animation != NULL) {
playAnimation(*part.animation);
if (exitPending())
break;
continue; //to next part
}
//循环动画片段
for (int r=0 ; !part.count || r<part.count ; r++) {
// Exit any non playuntil complete parts immediately
if(exitPending() && !part.playUntilComplete)
break;
//启动音频线程,播放音频文件
if (r == 0 && part.audioData && playSoundsAllowed()) {
if (mInitAudioThread != nullptr) {
mInitAudioThread->join();
}
audioplay::playClip(part.audioData, part.audioLength);
}
glClearColor(
part.backgroundColor[0],
part.backgroundColor[1],
part.backgroundColor[2],
1.0f);
//按照frameDuration频率,循环绘制开机动画图片纹理
for (size_t j=0 ; j<fcount && (!exitPending() || part.playUntilComplete) ; j++) {
const Animation::Frame& frame(part.frames[j]);
nsecs_t lastFrame = systemTime();
if (r > 0) {
glBindTexture(GL_TEXTURE_2D, frame.tid);
} else {
if (part.count != 1) {
//生成纹理
glGenTextures(1, &frame.tid);
//绑定纹理
glBindTexture(GL_TEXTURE_2D, frame.tid);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
}
int w, h;
initTexture(frame.map, &w, &h);
}
const int xc = animationX + frame.trimX;
const int yc = animationY + frame.trimY;
Region clearReg(Rect(mWidth, mHeight));
clearReg.subtractSelf(Rect(xc, yc, xc+frame.trimWidth, yc+frame.trimHeight));
if (!clearReg.isEmpty()) {
Region::const_iterator head(clearReg.begin());
Region::const_iterator tail(clearReg.end());
glEnable(GL_SCISSOR_TEST);
while (head != tail) {
const Rect& r2(*head++);
glScissor(r2.left, mHeight - r2.bottom, r2.width(), r2.height());
glClear(GL_COLOR_BUFFER_BIT);
}
glDisable(GL_SCISSOR_TEST);
}
// 绘制纹理
glDrawTexiOES(xc, mHeight - (yc + frame.trimHeight),
0, frame.trimWidth, frame.trimHeight);
if (mClockEnabled && mTimeIsAccurate && validClock(part)) {
drawClock(animation.clockFont, part.clockPosX, part.clockPosY);
}
eglSwapBuffers(mDisplay, mSurface);
nsecs_t now = systemTime();
nsecs_t delay = frameDuration - (now - lastFrame);
//ALOGD("%lld, %lld", ns2ms(now - lastFrame), ns2ms(delay));
lastFrame = now;
if (delay > 0) {
struct timespec spec;
spec.tv_sec = (now + delay) / 1000000000;
spec.tv_nsec = (now + delay) % 1000000000;
int err;
do {
err = clock_nanosleep(CLOCK_MONOTONIC, TIMER_ABSTIME, &spec, NULL);
} while (err<0 && errno == EINTR);
}
checkExit();
}
//休眠
usleep(part.pause * ns2us(frameDuration));
// 动画退出条件判断
if(exitPending() && !part.count)
break;
}
}
// 释放纹理
for (const Animation::Part& part : animation.parts) {
if (part.count != 1) {
const size_t fcount = part.frames.size();
for (size_t j = 0; j < fcount; j++) {
const Animation::Frame& frame(part.frames[j]);
glDeleteTextures(1, &frame.tid);
}
}
}
// 关闭和视频音频
audioplay::setPlaying(false);
audioplay::destroy();
return true;
}
从上面的源码可以看到,playAnimation函数播放动画的原理,其实就是按照一定的频率,循环调用glDrawTexiOES函数,绘制图片纹理,同时调用音频播放模块播放音频。
通过OpenGL ES播放动画的案例就讲完了,我们也了解了通过OpenGL来播放视频的一种方式,我们接着看第二个案例,Activity界面如何通过OpenGL来进行硬件加速,也就是硬件绘制绘制的。
OpenGL ES进行硬件加速
我们知道,Activity界面的显示需要经历Measure测量,Layout布局,和Draw绘制三个过程,而Draw绘制流程又分为软件绘制和硬件绘制,硬件绘制便是通过OpenGL ES进行的。我们直接看看硬件绘制流程里,OpenGL ES是如何来进行绘制的,它的入口在ViewRootImpl的performDraw函数中。
//文件-->/frameworks/base/core/java/android/view/ViewRootImpl.java
private void performDraw() {
//……
draw(fullRedrawNeeded);
//……
}
private void draw(boolean fullRedrawNeeded) {
Surface surface = mSurface;
if (!surface.isValid()) {
return;
}
//……
if (!dirty.isEmpty() || mIsAnimating || accessibilityFocusDirty) {
if (!dirty.isEmpty() || mIsAnimating || accessibilityFocusDirty) {
if (mAttachInfo.mThreadedRenderer != null && mAttachInfo.mThreadedRenderer.isEnabled()) {
//……
//硬件渲染
mAttachInfo.mThreadedRenderer.draw(mView, mAttachInfo, this);
} else {
//……
//软件渲染
if (!drawSoftware(surface, mAttachInfo, xOffset, yOffset, scalingRequired, dirty)) {
return;
}
}
}
//……
}
//……
}
从上面的代码可以看到,硬件渲染是通过mThreadedRenderer.draw方法进行的,在分析mThreadedRenderer.draw函数之前,我们需要先了解ThreadedRenderer是什么,它的创建要在Measure,Layout和Draw的流程之前,当我们在Activity的onCreate回调中执行setContentView函数时,最终会执行ViewRootImpl的setView方法,ThreadedRenderer就是在这个此时被创建的。
//文件-->/frameworks/base/core/java/android/view/ViewRootImpl.java
public void setView(View view, WindowManager.LayoutParams attrs, View panelParentView) {
synchronized (this) {
if (mView == null) {
mView = view;
//……
if (mSurfaceHolder == null) {
enableHardwareAcceleration(attrs);
}
//……
}
}
}
private void enableHardwareAcceleration(WindowManager.LayoutParams attrs) {
mAttachInfo.mHardwareAccelerated = false;
mAttachInfo.mHardwareAccelerationRequested = false;
// 兼容模式下不开启硬件加速
if (mTranslator != null) return;
final boolean hardwareAccelerated =
(attrs.flags & WindowManager.LayoutParams.FLAG_HARDWARE_ACCELERATED) != 0;
if (hardwareAccelerated) {
if (!ThreadedRenderer.isAvailable()) {
return;
}
//……
if (fakeHwAccelerated) {
//……
} else if (!ThreadedRenderer.sRendererDisabled
|| (ThreadedRenderer.sSystemRendererDisabled && forceHwAccelerated)) {
//……
//创建ThreadedRenderer
mAttachInfo.mThreadedRenderer = ThreadedRenderer.create(mContext, translucent,
attrs.getTitle().toString());
if (mAttachInfo.mThreadedRenderer != null) {
mAttachInfo.mHardwareAccelerated =
mAttachInfo.mHardwareAccelerationRequested = true;
}
}
}
}
可以看到,当RootViewImpl在调用setView的时候,会开启硬件加速,并通过ThreadedRenderer.create函数来创建ThreadedRenderer。
我们继续看看ThreadedRenderer这个类的实现。
//文件-->/frameworks/base/core/java/android/view/ThreadedRenderer.java
public static ThreadedRenderer create(Context context, boolean translucent, String name) {
ThreadedRenderer renderer = null;
if (isAvailable()) {
renderer = new ThreadedRenderer(context, translucent, name);
}
return renderer;
}
ThreadedRenderer(Context context, boolean translucent, String name) {
//……
//创建RootRenderNode
long rootNodePtr = nCreateRootRenderNode();
mRootNode = RenderNode.adopt(rootNodePtr);
mRootNode.setClipToBounds(false);
mIsOpaque = !translucent;
//创建RenderProxy
mNativeProxy = nCreateProxy(translucent, rootNodePtr);
nSetName(mNativeProxy, name);
//启动GraphicsStatsService,统计渲染信息
ProcessInitializer.sInstance.init(context, mNativeProxy);
loadSystemProperties();
}
ThreadedRenderer的构造函数中主要做了这两件事情:
- 通过JNI方法nCreateRootRenderNode在Native创建RootRenderNode,每一个View都对应了一个RenderNode,它包含了这个View及其子view的DisplayList,DisplayList包含了是可以让openGL识别的渲染指令,这些渲染指令被封装成了一条条OP。
//文件-->/frameworks/base/core/jni/android_view_ThreadedRenderer.cpp
static jlong android_view_ThreadedRenderer_createRootRenderNode(JNIEnv* env, jobject clazz) {
RootRenderNode* node = new RootRenderNode(env);
node->incStrong(0);
node->setName("RootRenderNode");
return reinterpret_cast<jlong>(node);
}
- 通过Jni方法nCreateProxy在Native层的RenderProxy,它就是用来跟渲染线程进行通信的句柄,我们看下nCreateProxy的Native实现
//文件-->/frameworks/base/core/jni/android_view_ThreadedRenderer.cpp
static jlong android_view_ThreadedRenderer_createProxy(JNIEnv* env, jobject clazz,
jboolean translucent, jlong rootRenderNodePtr) {
RootRenderNode* rootRenderNode = reinterpret_cast<RootRenderNode*>(rootRenderNodePtr);
ContextFactoryImpl factory(rootRenderNode);
return (jlong) new RenderProxy(translucent, rootRenderNode, &factory);
}
//文件-->/frameworks/base/libs/hwui/renderthread/RenderProxy.cpp
RenderProxy::RenderProxy(bool translucent, RenderNode* rootRenderNode, IContextFactory* contextFactory)
: mRenderThread(RenderThread::getInstance())
, mContext(nullptr) {
SETUP_TASK(createContext);
args->translucent = translucent;
args->rootRenderNode = rootRenderNode;
args->thread = &mRenderThread;
args->contextFactory = contextFactory;
mContext = (CanvasContext*) postAndWait(task);
mDrawFrameTask.setContext(&mRenderThread, mContext, rootRenderNode);
}
从RenderProxy构造函数可以看到,通过RenderThread::getInstance()创建了RenderThread,也就是硬件绘制的渲染线程。相比于在主线程进行的软件绘制,硬件加速会新建一个线程,这样能减轻主线程的工作量。
了解了ThreadedRenderer的创建和初始化流程,我们继续回到渲染的流程mThreadedRenderer.draw这个函数中来,先看看这个函数的源码。
//文件-->/frameworks/base/core/java/android/view/ThreadedRenderer.java
void draw(View view, AttachInfo attachInfo, DrawCallbacks callbacks) {
attachInfo.mIgnoreDirtyState = true;
final Choreographer choreographer = attachInfo.mViewRootImpl.mChoreographer;
choreographer.mFrameInfo.markDrawStart();
//1,构建RootView的DisplayList
updateRootDisplayList(view, callbacks);
attachInfo.mIgnoreDirtyState = false;
//…… 窗口动画处理
final long[] frameInfo = choreographer.mFrameInfo.mFrameInfo;
//2,通知渲染
int syncResult = nSyncAndDrawFrame(mNativeProxy, frameInfo, frameInfo.length);
//…… 渲染失败的处理
}
这个流程我们只需要关心这两件事情:
- 构建DisplayList
- **绘制DisplayList****
经过这两步,界面就显示出来。我们详细看一下这这两步的流程:
构建DisplayList
1,通过updateRootDisplayList函数构建根view的DisplayList,DisplayList在前面提到过,它包含了可以让openGL识别的渲染指令,先看看函数的实现
//文件-->/frameworks/base/core/java/android/view/ThreadedRenderer.java
private void updateRootDisplayList(View view, DrawCallbacks callbacks) {
Trace.traceBegin(Trace.TRACE_TAG_VIEW, "Record View#draw()");
//构建View的DisplayList
updateViewTreeDisplayList(view);
if (mRootNodeNeedsUpdate || !mRootNode.isValid()) {
//获取DisplayListCanvas
DisplayListCanvas canvas = mRootNode.start(mSurfaceWidth, mSurfaceHeight);
try {
final int saveCount = canvas.save();
canvas.translate(mInsetLeft, mInsetTop);
callbacks.onPreDraw(canvas);
canvas.insertReorderBarrier();
//合并和优化DisplayList
canvas.drawRenderNode(view.updateDisplayListIfDirty());
canvas.insertInorderBarrier();
callbacks.onPostDraw(canvas);
canvas.restoreToCount(saveCount);
mRootNodeNeedsUpdate = false;
} finally {
//更新RootRenderNode
mRootNode.end(canvas);
}
}
Trace.traceEnd(Trace.TRACE_TAG_VIEW);
}
updateRootDisplayList函数的主要流程有这几步:
- 构建根View的DisplayList
- 合并和优化DisplayList
构建根View的DisplayList
我们先看第一步构建根View的DisplayList的源码。
//文件-->/frameworks/base/core/java/android/view/ThreadedRenderer.java
private void updateViewTreeDisplayList(View view) {
view.mPrivateFlags |= View.PFLAG_DRAWN;
view.mRecreateDisplayList = (view.mPrivateFlags & View.PFLAG_INVALIDATED)
== View.PFLAG_INVALIDATED;
view.mPrivateFlags &= ~View.PFLAG_INVALIDATED;
view.updateDisplayListIfDirty();
view.mRecreateDisplayList = false;
}
//文件-->/frameworks/base/core/java/android/view/View.java
public RenderNode updateDisplayListIfDirty() {
final RenderNode renderNode = mRenderNode;
if (!canHaveDisplayList()) {
return renderNode;
}
//判断硬件加速是否可用
if ((mPrivateFlags & PFLAG_DRAWING_CACHE_VALID) == 0
|| !renderNode.isValid()
|| (mRecreateDisplayList)) {
//…… 不需要更新displaylist时,则直接返回renderNode
//获取DisplayListCanvas
final DisplayListCanvas canvas = renderNode.start(width, height);
try {
if (layerType == LAYER_TYPE_SOFTWARE) {
//如果强制开启了软件绘制,比如一些不支持硬件加速的组件,或者静止了硬件加速的组件,会转换成bitmap后,交给硬件渲染
buildDrawingCache(true);
Bitmap cache = getDrawingCache(true);
if (cache != null) {
canvas.drawBitmap(cache, 0, 0, mLayerPaint);
}
} else {
if ((mPrivateFlags & PFLAG_SKIP_DRAW) == PFLAG_SKIP_DRAW) {
//递归子View构建或更新displaylist
dispatchDraw(canvas);
} else {
//调用自身的draw方法
draw(canvas);
}
}
} finally {
//讲DisplayListCanvas内容绑定到renderNode上
renderNode.end(canvas);
setDisplayListProperties(renderNode);
}
} else {
mPrivateFlags |= PFLAG_DRAWN | PFLAG_DRAWING_CACHE_VALID;
mPrivateFlags &= ~PFLAG_DIRTY_MASK;
}
return renderNode;
}
可以看到updateDisplayListIfDirty主要做的事情有这几件
- 获取DisplayListCanvas
- 判断组件是否支持硬件加速,不支持则转换成bitmap后交给DisplayListCanvas
- 递归子View执行DisplayList的构建
- 调用自身的draw方法,交给DisplayListCanvas进行绘制
- 返回RenderNode
看到这里可能会有人疑问,为什么构建更新DisplayList函数updateDisplayListIfDirty中并没有看到DisplayList,返回对象也不是DisplayList,而是RenderNode?这个DisplayList其实是在Native层创建的,在前面提到过RenderNode其实包含了DisplayList,renderNode.end(canvas)函数会将DisplayList绑定到renderNode中。而DisplayListCanvas的作用,就是在Native层创建DisplayList。那么我们接着看DisplayListCanvas这个类。
//文件-->/frameworks/base/core/java/android/view/RenderNode.java
public DisplayListCanvas start(int width, int height) {
return DisplayListCanvas.obtain(this, width, height);
}
//文件-->/frameworks/base/core/java/android/view/DisplayListCanvas.java
static DisplayListCanvas obtain(@NonNull RenderNode node, int width, int height) {
if (node == null) throw new IllegalArgumentException("node cannot be null");
DisplayListCanvas canvas = sPool.acquire();
if (canvas == null) {
canvas = new DisplayListCanvas(node, width, height);
} else {
nResetDisplayListCanvas(canvas.mNativeCanvasWrapper, node.mNativeRenderNode,
width, height);
}
canvas.mNode = node;
canvas.mWidth = width;
canvas.mHeight = height;
return canvas;
}
private DisplayListCanvas(@NonNull RenderNode node, int width, int height) {
super(nCreateDisplayListCanvas(node.mNativeRenderNode, width, height));
mDensity = 0; // disable bitmap density scaling
}
我们通过RenderNode.start方法获取一个DisplayListCanvas,RenderNode会通过obtain来创建或从缓存中获取DisplayListCanvas,这是一种享元模式。DisplayListCanvas的构造函数里,会通过JNI方法nCreateDisplayListCanvas创建native的Canvas,我们接着看一下Native的流程
//文件-->/frameworks/base/core/jni/android_view_DisplayListCanvas.cpp
static jlong android_view_DisplayListCanvas_createDisplayListCanvas(jlong renderNodePtr,
jint width, jint height) {
RenderNode* renderNode = reinterpret_cast<RenderNode*>(renderNodePtr);
return reinterpret_cast<jlong>(Canvas::create_recording_canvas(width, height, renderNode));
}
//文件-->/frameworks/base/libs/hwui/hwui/Canvas.cpp
Canvas* Canvas::create_recording_canvas(int width, int height, uirenderer::RenderNode* renderNode) {
if (uirenderer::Properties::isSkiaEnabled()) {
return new uirenderer::skiapipeline::SkiaRecordingCanvas(renderNode, width, height);
}
return new uirenderer::RecordingCanvas(width, height);
}
可以看到,java层的DisplayListCanvas对应了native层RecordingCanvas或者SkiaRecordingCanvas
这里简单介绍一下这两个Canvas的区别,在Android8之前,HWUI中通过OpenGL对绘制操作进行封装后,直接送GPU进行渲染。Android 8.0开始,HWUI进行了重构,增加了RenderPipeline的概念,RenderPipeline有三种类型,分别为Skia,OpenGL和Vulkan,分别对应不同的渲染。并且Android8.0开始强化和重视Skia的地位,Android10版本后,所有通过硬件加速的渲染,都是通过SKIA进行封装,然后再经过OpenGL或Vulkan,最后交给GPU渲染。我讲解的源码是8.0的源码,可以看到,其实已经可以通过配置,来开启skiapipeline了。
为了更容易的讲解如何通过OpenGL进行硬件渲染,这里我还是以RecordingCanvas来讲解,这里列举几个RecordingCanvas中的常规操作
//文件-->/frameworks/base/libs/hwui/RecordingCanvas.cpp
//绘制点
void RecordingCanvas::drawPoints(const float* points, int floatCount, const SkPaint& paint) {
if (CC_UNLIKELY(floatCount < 2 || paint.nothingToDraw())) return;
floatCount &= ~0x1; // round down to nearest two
addOp(alloc().create_trivial<PointsOp>(
calcBoundsOfPoints(points, floatCount),
*mState.currentSnapshot()->transform,
getRecordedClip(),
refPaint(&paint), refBuffer<float>(points, floatCount), floatCount));
}
struct PointsOp : RecordedOp {
PointsOp(BASE_PARAMS, const float* points, const int floatCount)
: SUPER(PointsOp)
, points(points)
, floatCount(floatCount) {}
const float* points;
const int floatCount;
};
//绘制线
void RecordingCanvas::drawLines(const float* points, int floatCount, const SkPaint& paint) {
if (CC_UNLIKELY(floatCount < 4 || paint.nothingToDraw())) return;
floatCount &= ~0x3; // round down to nearest four
addOp(alloc().create_trivial<LinesOp>(
calcBoundsOfPoints(points, floatCount),
*mState.currentSnapshot()->transform,
getRecordedClip(),
refPaint(&paint), refBuffer<float>(points, floatCount), floatCount));
}
struct LinesOp : RecordedOp {
LinesOp(BASE_PARAMS, const float* points, const int floatCount)
: SUPER(LinesOp)
, points(points)
, floatCount(floatCount) {}
const float* points;
const int floatCount;
};
//绘制矩阵
void RecordingCanvas::drawRect(float left, float top, float right, float bottom, const SkPaint& paint) {
if (CC_UNLIKELY(paint.nothingToDraw())) return;
addOp(alloc().create_trivial<RectOp>(
Rect(left, top, right, bottom),
*(mState.currentSnapshot()->transform),
getRecordedClip(),
refPaint(&paint)));
}
struct RectOp : RecordedOp {
RectOp(BASE_PARAMS)
: SUPER(RectOp) {}
};
struct RoundRectOp : RecordedOp {
RoundRectOp(BASE_PARAMS, float rx, float ry)
: SUPER(RoundRectOp)
, rx(rx)
, ry(ry) {}
const float rx;
const float ry;
};
int RecordingCanvas::addOp(RecordedOp* op) {
// skip op with empty clip
if (op->localClip && op->localClip->rect.isEmpty()) {
// NOTE: this rejection happens after op construction/content ref-ing, so content ref'd
// and held by renderthread isn't affected by clip rejection.
// Could rewind alloc here if desired, but callers would have to not touch op afterwards.
return -1;
}
int insertIndex = mDisplayList->ops.size();
mDisplayList->ops.push_back(op);
if (mDeferredBarrierType != DeferredBarrierType::None) {
// op is first in new chunk
mDisplayList->chunks.emplace_back();
DisplayList::Chunk& newChunk = mDisplayList->chunks.back();
newChunk.beginOpIndex = insertIndex;
newChunk.endOpIndex = insertIndex + 1;
newChunk.reorderChildren = (mDeferredBarrierType == DeferredBarrierType::OutOfOrder);
newChunk.reorderClip = mDeferredBarrierClip;
int nextChildIndex = mDisplayList->children.size();
newChunk.beginChildIndex = newChunk.endChildIndex = nextChildIndex;
mDeferredBarrierType = DeferredBarrierType::None;
} else {
// standard case - append to existing chunk
mDisplayList->chunks.back().endOpIndex = insertIndex + 1;
}
return insertIndex;
}
可以看到,我们通过RecordingCanvas绘制的图元,都被封装成了一个个能够让GPU能够识别的OP,这些OP都存储在了mDisplayList中。这就回答了前面的疑问,为什么updateDisplayListIfDirty没有看到DisplayList,因为DisplayListCanvas通过调用Natice层的RecordingCanvas,更新了Natice层的mDisplayList。
我们在接着看renderNode.end(canvas)函数,如何将Natice层的DisplayList绑定到renderNode中。
//文件-->/frameworks/base/core/java/android/view/RenderNode.java
public void end(DisplayListCanvas canvas) {
long displayList = canvas.finishRecording();
nSetDisplayList(mNativeRenderNode, displayList);
canvas.recycle();
}
这里通过JNI方法nSetDisplayList进行了DisplayList和RenderNode的绑定,此时,我们就能理解我在前面说的:RenderNode包含了这个View及其子view的DisplayList,DisplayList包含了一条条可以让openGL识别的渲染指令——OP操作,它是一个基本的能让GPU识别的绘制元素。
合并和优化DisplayList
updateViewTreeDisplayList花了比较大精力,将所有的View的DisplayList已经创建好了,DisplayList里的DrawOP树也创建好了,为什么还要在调用canvas.drawRenderNode(view.updateDisplayListIfDirty())这个函数呢?这个函数的主要功能是对前面构建的DisplayList做优化和合并处理,我们看看具体的实现细节。
//文件-->/frameworks/base/core/java/android/view/DisplayListCanvas.java
public void drawRenderNode(RenderNode renderNode) {
nDrawRenderNode(mNativeCanvasWrapper, renderNode.getNativeDisplayList());
}
//文件-->/frameworks/base/core/jni/android_view_DisplayListCanvas.cpp
static void android_view_DisplayListCanvas_drawRenderNode(jlong canvasPtr, jlong renderNodePtr) {
Canvas* canvas = reinterpret_cast<Canvas*>(canvasPtr);
RenderNode* renderNode = reinterpret_cast<RenderNode*>(renderNodePtr);
canvas->drawRenderNode(renderNode);
}
//文件-->/frameworks/base/libs/hwui/RecordingCanvas.cpp
void RecordingCanvas::drawRenderNode(RenderNode* renderNode) {
auto&& stagingProps = renderNode->stagingProperties();
RenderNodeOp* op = alloc().create_trivial<RenderNodeOp>(
Rect(stagingProps.getWidth(), stagingProps.getHeight()),
*(mState.currentSnapshot()->transform),
getRecordedClip(),
renderNode);
int opIndex = addOp(op);
if (CC_LIKELY(opIndex >= 0)) {
int childIndex = mDisplayList->addChild(op);
// update the chunk's child indices
DisplayList::Chunk& chunk = mDisplayList->chunks.back();
chunk.endChildIndex = childIndex + 1;
if (renderNode->stagingProperties().isProjectionReceiver()) {
// use staging property, since recording on UI thread
mDisplayList->projectionReceiveIndex = opIndex;
}
}
}
可以看到,最终执行到了RecordingCanvas中的drawRenderNode函数,这个函数会对DisplayList做合并和优化。
绘制DisplayList
经过比较长的篇幅,我们把mThreadedRenderer.draw函数中的第一个流程,构建DisplayList说完,现在开始说第二个流程,nSyncAndDrawFrame进行帧绘制,这个流程结束,我们的界面就能在屏幕上显示出来了。nSyncAndDrawFrame是一个native方法,我们看看它的实现
static int android_view_ThreadedRenderer_syncAndDrawFrame(JNIEnv* env, jobject clazz,
jlong proxyPtr, jlongArray frameInfo, jint frameInfoSize) {
LOG_ALWAYS_FATAL_IF(frameInfoSize != UI_THREAD_FRAME_INFO_SIZE,
"Mismatched size expectations, given %d expected %d",
frameInfoSize, UI_THREAD_FRAME_INFO_SIZE);
RenderProxy* proxy = reinterpret_cast<RenderProxy*>(proxyPtr);
env->GetLongArrayRegion(frameInfo, 0, frameInfoSize, proxy->frameInfo());
return proxy->syncAndDrawFrame();
}
int RenderProxy::syncAndDrawFrame() {
return mDrawFrameTask.drawFrame();
}
nSyncAndDrawFrame函数调用了RenderProxy的syncAndDrawFrame,syncAndDrawFrame调用了DrawFrameTask.drawFrame()方法
//文件-->/frameworks/base/libs/hwui/renderthread/DrawFrameTask.cpp
int DrawFrameTask::drawFrame() {
LOG_ALWAYS_FATAL_IF(!mContext, "Cannot drawFrame with no CanvasContext!");
mSyncResult = SyncResult::OK;
mSyncQueued = systemTime(CLOCK_MONOTONIC);
postAndWait();
return mSyncResult;
}
void DrawFrameTask::postAndWait() {
AutoMutex _lock(mLock);
mRenderThread->queue(this);
mSignal.wait(mLock);
}
void DrawFrameTask::run() {
ATRACE_NAME("DrawFrame");
bool canUnblockUiThread;
bool canDrawThisFrame;
{
TreeInfo info(TreeInfo::MODE_FULL, *mContext);
canUnblockUiThread = syncFrameState(info);
canDrawThisFrame = info.out.canDrawThisFrame;
}
// Grab a copy of everything we need
CanvasContext* context = mContext;
// From this point on anything in "this" is *UNSAFE TO ACCESS*
if (canUnblockUiThread) {
unblockUiThread();
}
if (CC_LIKELY(canDrawThisFrame)) {
context->draw();
} else {
// wait on fences so tasks don't overlap next frame
context->waitOnFences();
}
if (!canUnblockUiThread) {
unblockUiThread();
}
}
DrawFrameTask做了两件事情
- 调用syncFrameState函数同步frame信息
- 调用CanvasContext.draw()函数进行绘制
同步Frame信息
我们先看看第一件事情,同步Frame信息,它主要的工作是将主线程的RenderNode同步到RenderNode来,在前面讲mAttachInfo.mThreadedRenderer.draw函数中,第一步会将DisplayList构建完毕,然后绑定到RenderNode中,这个RenderNode是在主线程创建的。而我们的DrawFrameTask,是在native层的RenderThread中执行的,所以需要讲数据同步过来。
//文件-->/frameworks/base/libs/hwui/renderthread/DrawFrameTask.cpp
bool DrawFrameTask::syncFrameState(TreeInfo& info) {
ATRACE_CALL();
int64_t vsync = mFrameInfo[static_cast<int>(FrameInfoIndex::Vsync)];
mRenderThread->timeLord().vsyncReceived(vsync);
bool canDraw = mContext->makeCurrent();
mContext->unpinImages();
for (size_t i = 0; i < mLayers.size(); i++) {
mLayers[i]->apply();
}
mLayers.clear();
mContext->prepareTree(info, mFrameInfo, mSyncQueued, mTargetNode);
//……
// If prepareTextures is false, we ran out of texture cache space
return info.prepareTextures;
}
这里调用了mContext->prepareTree函数,mContext在下面会详细讲,我们这里先看看这个方法的实现。
//文件-->/frameworks/base/libs/hwui/renderthread/CanvasContext.cpp
void CanvasContext::prepareTree(TreeInfo& info, int64_t* uiFrameInfo,
int64_t syncQueued, RenderNode* target) {
//……
for (const sp<RenderNode>& node : mRenderNodes) {
// Only the primary target node will be drawn full - all other nodes would get drawn in
// real time mode. In case of a window, the primary node is the window content and the other
// node(s) are non client / filler nodes.
info.mode = (node.get() == target ? TreeInfo::MODE_FULL : TreeInfo::MODE_RT_ONLY);
node->prepareTree(info);
GL_CHECKPOINT(MODERATE);
}
//……
}
void RenderNode::prepareTree(TreeInfo& info) {
bool functorsNeedLayer = Properties::debugOverdraw;
prepareTreeImpl(info, functorsNeedLayer);
}
void RenderNode::prepareTreeImpl(TreeInfo& info, bool functorsNeedLayer) {
info.damageAccumulator->pushTransform(this);
if (info.mode == TreeInfo::MODE_FULL) {
// 同步属性
pushStagingPropertiesChanges(info);
}
// layer
prepareLayer(info, animatorDirtyMask);
//同步DrawOpTree
if (info.mode == TreeInfo::MODE_FULL) {
pushStagingDisplayListChanges(info);
}
//递归处理子View
prepareSubTree(info, childFunctorsNeedLayer, mDisplayListData);
// push
pushLayerUpdate(info);
info.damageAccumulator->popTransform();
}
同步Frame的操作完成了,我们接着看最后绘制的流程。
进行绘制
图形的硬件渲染,是通过调用CanvasContext的draw方法来进行绘制的,CanvasContext是什么呢?
它是渲染的上下文,CanvasContext可以选择不同的渲染模式进行渲染,这是策略模式的设计。我们看一下CanvasContext的create方法,可以看到,方法中会根据渲染类型,创建不同的渲染管道,总共有三种渲染管道——OpenGL,SKiaGL和SkiaVulkan。
CanvasContext* CanvasContext::create(RenderThread& thread,
bool translucent, RenderNode* rootRenderNode, IContextFactory* contextFactory) {
auto renderType = Properties::getRenderPipelineType();
switch (renderType) {
case RenderPipelineType::OpenGL:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<OpenGLPipeline>(thread));
case RenderPipelineType::SkiaGL:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<skiapipeline::SkiaOpenGLPipeline>(thread));
case RenderPipelineType::SkiaVulkan:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<skiapipeline::SkiaVulkanPipeline>(thread));
default:
LOG_ALWAYS_FATAL("canvas context type %d not supported", (int32_t) renderType);
break;
}
return nullptr;
}
我们这里这里只看通过OpenGL进行渲染的OpenGLPipeline
OpenGLPipeline::OpenGLPipeline(RenderThread& thread)
: mEglManager(thread.eglManager())
, mRenderThread(thread) {
}
在OpenGLPipeline的构造函数里面,创建了EglManager,EglManager将我们对EGL的操作全部封装好了,我们看看EglManager的初始化方法
//文件-->/frameworks/base/libs/hwui/renderthread/EglManager.cpp
void EglManager::initialize() {
if (hasEglContext()) return;
ATRACE_NAME("Creating EGLContext");
//获取 EGL Display 对象
mEglDisplay = eglGetDisplay(EGL_DEFAULT_DISPLAY);
LOG_ALWAYS_FATAL_IF(mEglDisplay == EGL_NO_DISPLAY,
"Failed to get EGL_DEFAULT_DISPLAY! err=%s", eglErrorString());
EGLint major, minor;
//初始化与 EGLDisplay 之间的连接
LOG_ALWAYS_FATAL_IF(eglInitialize(mEglDisplay, &major, &minor) == EGL_FALSE,
"Failed to initialize display %p! err=%s", mEglDisplay, eglErrorString());
//……
//EGL配置设置
loadConfig();
//创建EGL上下文
createContext();
//创建离屏渲染Buffer
createPBufferSurface();
//绑定上下文
makeCurrent(mPBufferSurface);
DeviceInfo::initialize();
mRenderThread.renderState().onGLContextCreated();
}
在这里我们看到了熟悉的身影,EglManager中的初始化流程和前面所有EGL初始化的流程都是一样的。但在初始化的流程中,我们没看到WindowSurface的设置,只看到了PBufferSurface的创建,它是一个离屏渲染的Buffer,这里简单介绍一下WindowSurface和PbufferSurface
- WindowSurface:是和窗口相关的,也就是在屏幕上的一块显示区的封装,渲染后即显示在界面上。
- PbufferSurface:在显存中开辟一个空间,将渲染后的数据(帧)存放在这里。
可以看到没有WindowSurface,OpenGL ES渲染的图形是没法显示在界面上的。其实EglManager已经封装了初始化WindowSurface的方法。
//文件-->/frameworks/base/libs/hwui/renderthread/EglManager.cpp
EGLSurface EglManager::createSurface(EGLNativeWindowType window) {
initialize();
EGLint attribs[] = {
#ifdef ANDROID_ENABLE_LINEAR_BLENDING
EGL_GL_COLORSPACE_KHR, EGL_GL_COLORSPACE_SRGB_KHR,
EGL_COLORSPACE, EGL_COLORSPACE_sRGB,
#endif
EGL_NONE
};
EGLSurface surface = eglCreateWindowSurface(mEglDisplay, mEglConfig, window, attribs);
LOG_ALWAYS_FATAL_IF(surface == EGL_NO_SURFACE,
"Failed to create EGLSurface for window %p, eglErr = %s",
(void*) window, eglErrorString());
if (mSwapBehavior != SwapBehavior::Preserved) {
LOG_ALWAYS_FATAL_IF(eglSurfaceAttrib(mEglDisplay, surface, EGL_SWAP_BEHAVIOR, EGL_BUFFER_DESTROYED) == EGL_FALSE,
"Failed to set swap behavior to destroyed for window %p, eglErr = %s",
(void*) window, eglErrorString());
}
return surface;
}
这个surface又是什么时候设置的呢?在activity的界面显示流程中,当我们setView后,ViewRootImpl会执行performTraveserl函数,然后执行Measure测量,Layout布局,和Draw绘制的流程,setView函数在前面讲过,会开启硬件加速,创建ThreadedRenderer,draw函数也讲过,measure,layout的流程就不在这儿说了,它和OpgenGL没关系,其实performTraveserl函数里,同时也设置了EGL的Surface,可见这个函数是多么重要的一个函数,我们看一下。
private void performTraversals() {
//……
if (mAttachInfo.mThreadedRenderer != null) {
try {
//调用ThreadedRenderer initialize函数
hwInitialized = mAttachInfo.mThreadedRenderer.initialize(
mSurface);
if (hwInitialized && (host.mPrivateFlags
& View.PFLAG_REQUEST_TRANSPARENT_REGIONS) == 0) {
// Don't pre-allocate if transparent regions
// are requested as they may not be needed
mSurface.allocateBuffers();
}
} catch (OutOfResourcesException e) {
handleOutOfResourcesException(e);
return;
}
}
//……
}
boolean initialize(Surface surface) throws OutOfResourcesException {
boolean status = !mInitialized;
mInitialized = true;
updateEnabledState(surface);
nInitialize(mNativeProxy, surface);
return status;
}
ThreadedRenderer的initialize函数调用了native层的initialize方法。
static void android_view_ThreadedRenderer_initialize(JNIEnv* env, jobject clazz,
jlong proxyPtr, jobject jsurface) {
RenderProxy* proxy = reinterpret_cast<RenderProxy*>(proxyPtr);
sp<Surface> surface = android_view_Surface_getSurface(env, jsurface);
proxy->initialize(surface);
}
void RenderProxy::initialize(const sp<Surface>& surface) {
SETUP_TASK(initialize);
args->context = mContext;
args->surface = surface.get();
post(task);
}
void CanvasContext::initialize(Surface* surface) {
setSurface(surface);
}
void CanvasContext::setSurface(Surface* surface) {
ATRACE_CALL();
mNativeSurface = surface;
bool hasSurface = mRenderPipeline->setSurface(surface, mSwapBehavior);
mFrameNumber = -1;
if (hasSurface) {
mHaveNewSurface = true;
mSwapHistory.clear();
} else {
mRenderThread.removeFrameCallback(this);
}
}
从这里可以看到,EGL的Surface在很早之前就已经设置好了。
此时我们的流程中,EGL的初始化工作都已经完成了,现在可以开始绘制了,我们回到DrawFrameTask::run的draw流程上来
void CanvasContext::draw() {
SkRect dirty;
mDamageAccumulator.finish(&dirty);
mCurrentFrameInfo->markIssueDrawCommandsStart();
Frame frame = mRenderPipeline->getFrame();
SkRect windowDirty = computeDirtyRect(frame, &dirty);
//调用OpenGL的draw函数
bool drew = mRenderPipeline->draw(frame, windowDirty, dirty, mLightGeometry, &mLayerUpdateQueue,
mContentDrawBounds, mOpaque, mLightInfo, mRenderNodes, &(profiler()));
waitOnFences();
bool requireSwap = false;
//交换缓冲区
bool didSwap = mRenderPipeline->swapBuffers(frame, drew, windowDirty, mCurrentFrameInfo,
&requireSwap);
mIsDirty = false;
//……
}
这里调用mRenderPipeline的draw方法,其实就是调用了OpenGL的draw方法,然后调用mRenderPipeline->swapBuffers进行缓存区交换
//文件-->/frameworks/base/libs/hwui/renderthread/OpenGLPipeline.cpp
bool OpenGLPipeline::draw(const Frame& frame, const SkRect& screenDirty, const SkRect& dirty,
const FrameBuilder::LightGeometry& lightGeometry,
LayerUpdateQueue* layerUpdateQueue,
const Rect& contentDrawBounds, bool opaque,
const BakedOpRenderer::LightInfo& lightInfo,
const std::vector< sp<RenderNode> >& renderNodes,
FrameInfoVisualizer* profiler) {
//……
//BakedOpRenderer用于替代之前的OpenGLRenderer
BakedOpRenderer renderer(caches, mRenderThread.renderState(),
opaque, lightInfo);
frameBuilder.replayBakedOps<BakedOpDispatcher>(renderer);
//调用GPU进行渲染
drew = renderer.didDraw();
//……
return drew;
}
bool OpenGLPipeline::swapBuffers(const Frame& frame, bool drew, const SkRect& screenDirty,
FrameInfo* currentFrameInfo, bool* requireSwap) {
GL_CHECKPOINT(LOW);
// Even if we decided to cancel the frame, from the perspective of jank
// metrics the frame was swapped at this point
currentFrameInfo->markSwapBuffers();
*requireSwap = drew || mEglManager.damageRequiresSwap();
if (*requireSwap && (CC_UNLIKELY(!mEglManager.swapBuffers(frame, screenDirty)))) {
return false;
}
return *requireSwap;
}
至此,通过OpenGL ES进行硬件渲染的主要流程结束了。看完了两个例子,是不是对OpenGL ES作为图像生产者是如何生产图像已经了解了呢?我们接着看下一个图像生产者Skia。
Skia
Skia是谷歌开源的一款跨平台的2D图形引擎,目前谷歌的Chrome浏览器、Android、Flutter、以及火狐浏览器、火狐操作系统和其它许多产品都使用它作为图形引擎,它作为Android系统第三方软件,放在external/skia/ 目录下。虽然Android从4.0开始默认开启了硬件加速,但不代表Skia的作用就不大了,其实Skia在Android中的地位是越来越重要了,从Android 8开始,我们可以选择使用Skia进行硬件加速,Android 9开始就默认使用Skia来进行硬件加速。Skia的硬件加速主要是通过 copybit 模块调用OpenGL或者SKia来实现分。
由于Skia的硬件加速也是通过Copybit模块调用的OpenGL或者Vulkan接口,所以我们这儿只说说Skia通过cpu绘制的,也就是软绘的方式。还是老规则,先看看Skia要如何使用
如何使用Skia?
OpenGL ES的使用要配合EGL,需要初始化Display,surface,context等,用法还是比较繁琐的,Skia在使用上就方便很多了。掌握Skia绘制三要素:画板SKCanvas 、画纸SiBitmap、画笔Skpaint,我们就能很轻松的用Skia来绘制图形。
下面详细的解释Skia的绘图三要素
- SKBitmap用来存储图形数据,它封装了与位图相关的一系列操作
SkBitmap bitmap = new SkBitmap();
//设置位图格式及宽高
bitmap->setConfig(SkBitmap::kRGB_565_Config,800,480);
//分配位图所占空间
bitmap->allocPixels();
- SKCanvas 封装了所有画图操作的函数,通过调用这些函数,我们就能实现绘制操作。
//使用前传入bitmap
SkCanvas canvas(bitmap);
//移位,缩放,旋转,变形操作
translate(SkiaScalar dx, SkiaScalar dy);
scale(SkScalar sx, SkScalar sy);
rotate(SkScalar degrees);
skew(SkScalar sx, SkScalar sy);
//绘制操作
drawARGB(u8 a, u8 r, u8 g, u8 b....) //给定透明度以及红,绿,兰3色,填充整个可绘制区域。
drawColor(SkColor color...) //给定颜色color, 填充整个绘制区域。
drawPaint(SkPaint& paint) //用指定的画笔填充整个区域。
drawPoint(...)//根据各种不同参数绘制不同的点。
drawLine(x0, y0, x1, y1, paint) //画线,起点(x0, y0), 终点(x1, y1), 使用paint作为画笔。
drawRect(rect, paint) //画矩形,矩形大小由rect指定,画笔由paint指定。
drawRectCoords(left, top, right, bottom, paint),//给定4个边界画矩阵。
drawOval(SkRect& oval, SkPaint& paint) //画椭圆,椭圆大小由oval矩形指定。
//……其他操作
- Skpaint用来设置绘制内容的风格,样式,颜色等信息
setAntiAlias: 设置画笔的锯齿效果。
setColor: 设置画笔颜色
setARGB: 设置画笔的a,r,p,g值。
setAlpha: 设置Alpha值
setTextSize: 设置字体尺寸。
setStyle: 设置画笔风格,空心或者实心。
setStrokeWidth: 设置空心的边框宽度。
getColor: 得到画笔的颜色
getAlpha: 得到画笔的Alpha值。
我们看一个完整的使用Demo
void draw() {
SkBitmap bitmap = new SkBitmap();
//设置位图格式及宽高
bitmap->setConfig(SkBitmap::kRGB_565_Config,800,480);
//分配位图所占空间
bitmap->allocPixels();
//使用前传入bitmap
SkCanvas canvas(bitmap);
//定义画笔
SkPaint paint1, paint2, paint3;
paint1.setAntiAlias(true);
paint1.setColor(SkColorSetRGB(255, 0, 0));
paint1.setStyle(SkPaint::kFill_Style);
paint2.setAntiAlias(true);
paint2.setColor(SkColorSetRGB(0, 136, 0));
paint2.setStyle(SkPaint::kStroke_Style);
paint2.setStrokeWidth(SkIntToScalar(3));
paint3.setAntiAlias(true);
paint3.setColor(SkColorSetRGB(136, 136, 136));
sk_sp<SkTextBlob> blob1 =
SkTextBlob::MakeFromString("Skia!", SkFont(nullptr, 64.0f, 1.0f, 0.0f));
sk_sp<SkTextBlob> blob2 =
SkTextBlob::MakeFromString("Skia!", SkFont(nullptr, 64.0f, 1.5f, 0.0f));
canvas->clear(SK_ColorWHITE);
canvas->drawTextBlob(blob1.get(), 20.0f, 64.0f, paint1);
canvas->drawTextBlob(blob1.get(), 20.0f, 144.0f, paint2);
canvas->drawTextBlob(blob2.get(), 20.0f, 224.0f, paint3);
}
这个Demo的效果如下
了解了Skia如何使用,我们接着看两个场景:Skia进行软件绘制,Flutter界面绘制
Skia进行软件绘制
在上面我讲了通过使用OpenGL渲染的硬件绘制方式,这里会接着讲使用Skia渲染的软件绘制方式,虽然Android默认开启了硬件加速,但是由于硬件加速会有耗电和内存的问题,一些系统应用和常驻应用依然是使用的软件绘制的方式,软绘入口还是在draw方法中。
//文件-->/frameworks/base/core/java/android/view/ViewRootImpl.java
private void performDraw() {
//……
draw(fullRedrawNeeded);
//……
}
private void draw(boolean fullRedrawNeeded) {
Surface surface = mSurface;
if (!surface.isValid()) {
return;
}
//……
if (!dirty.isEmpty() || mIsAnimating || accessibilityFocusDirty) {
if (!dirty.isEmpty() || mIsAnimating || accessibilityFocusDirty) {
if (mAttachInfo.mThreadedRenderer != null && mAttachInfo.mThreadedRenderer.isEnabled()) {
//……
//硬件渲染
mAttachInfo.mThreadedRenderer.draw(mView, mAttachInfo, this);
} else {
//……
//软件渲染
if (!drawSoftware(surface, mAttachInfo, xOffset, yOffset, scalingRequired, dirty)) {
return;
}
}
}
//……
}
//……
}
我们来看看drawSoftware函数的实现
private boolean drawSoftware(Surface surface, AttachInfo attachInfo, int xoff, int yoff,
boolean scalingRequired, Rect dirty) {
// Draw with software renderer.
final Canvas canvas;
//……
canvas = mSurface.lockCanvas(dirty);
//……
mView.draw(canvas);
//……
surface.unlockCanvasAndPost(canvas);
//……
return true;
}
drawSoftware函数的流程主要为三步
- 通过mSurface.lockCanvas获取Canvas
- 通过draw方法,将根View及其子View遍历绘制到Canvas上
- 通过surface.unlockCanvasAndPost将绘制内容提交给surfaceFlinger进行合成
Lock Surface
我们先来看第一步,这个Canvas对应着Native层的SKCanvas。
//文件-->/frameworks/base/core/java/android/view/Surface.java
public Canvas lockCanvas(Rect inOutDirty)
throws Surface.OutOfResourcesException, IllegalArgumentException {
synchronized (mLock) {
checkNotReleasedLocked();
if (mLockedObject != 0) {
throw new IllegalArgumentException("Surface was already locked");
}
mLockedObject = nativeLockCanvas(mNativeObject, mCanvas, inOutDirty);
return mCanvas;
}
}
lockCanvas函数中通过JNI函数nativeLockCanvas,创建Nativce层的Canvas,nativeLockCanvas的入参mNativeObject对应着Native层的Surface,关于Surface和Buffer的知识,在下一篇图形缓冲区中会详细简介,这里不做太多介绍。我们直接着看nativeLockCanvas的实现。
static jlong nativeLockCanvas(JNIEnv* env, jclass clazz,
jlong nativeObject, jobject canvasObj, jobject dirtyRectObj) {
sp<Surface> surface(reinterpret_cast<Surface *>(nativeObject));
if (!isSurfaceValid(surface)) {
doThrowIAE(env);
return 0;
}
Rect dirtyRect(Rect::EMPTY_RECT);
Rect* dirtyRectPtr = NULL;
if (dirtyRectObj) {
dirtyRect.left = env->GetIntField(dirtyRectObj, gRectClassInfo.left);
dirtyRect.top = env->GetIntField(dirtyRectObj, gRectClassInfo.top);
dirtyRect.right = env->GetIntField(dirtyRectObj, gRectClassInfo.right);
dirtyRect.bottom = env->GetIntField(dirtyRectObj, gRectClassInfo.bottom);
dirtyRectPtr = &dirtyRect;
}
ANativeWindow_Buffer outBuffer;
//1,获取用来存储图形绘制的buffer
status_t err = surface->lock(&outBuffer, dirtyRectPtr);
if (err < 0) {
const char* const exception = (err == NO_MEMORY) ?
OutOfResourcesException :
"java/lang/IllegalArgumentException";
jniThrowException(env, exception, NULL);
return 0;
}
SkImageInfo info = SkImageInfo::Make(outBuffer.width, outBuffer.height,
convertPixelFormat(outBuffer.format),
outBuffer.format == PIXEL_FORMAT_RGBX_8888
? kOpaque_SkAlphaType : kPremul_SkAlphaType,
GraphicsJNI::defaultColorSpace());
SkBitmap bitmap;
ssize_t bpr = outBuffer.stride * bytesPerPixel(outBuffer.format);
bitmap.setInfo(info, bpr);
if (outBuffer.width > 0 && outBuffer.height > 0) {
//将上一个buffer里的图形数据复制到当前bitmap中
bitmap.setPixels(outBuffer.bits);
} else {
// be safe with an empty bitmap.
bitmap.setPixels(NULL);
}
//2,创建一个SKCanvas
Canvas* nativeCanvas = GraphicsJNI::getNativeCanvas(env, canvasObj);
//3,给SKCanvas设置Bitmap
nativeCanvas->setBitmap(bitmap);
//如果指定了脏区,则设定脏区的区域
if (dirtyRectPtr) {
nativeCanvas->clipRect(dirtyRect.left, dirtyRect.top,
dirtyRect.right, dirtyRect.bottom, SkClipOp::kIntersect);
}
if (dirtyRectObj) {
env->SetIntField(dirtyRectObj, gRectClassInfo.left, dirtyRect.left);
env->SetIntField(dirtyRectObj, gRectClassInfo.top, dirtyRect.top);
env->SetIntField(dirtyRectObj, gRectClassInfo.right, dirtyRect.right);
env->SetIntField(dirtyRectObj, gRectClassInfo.bottom, dirtyRect.bottom);
}
sp<Surface> lockedSurface(surface);
lockedSurface->incStrong(&sRefBaseOwner);
return (jlong) lockedSurface.get();
}
nativeLockCanvas主要做了这几件事情
- 通过surface->lock函数获取绘制用的Buffer
- 根据Buffer信息创建SKBitmap
- 根据SKBitmap,创建并初始化SKCanvas
通过nativeLockCanvas,我们就创建好了SKCanvas了,并且设置了可以绘制图形的bitmap,此时我们就可以通过SKCanvas往bitmap里面绘制图形,mView.draw()函数,就做了这件事情。
绘制
我们接着看看View中的draw()函数
//文件-->/frameworks/base/core/java/android/view/View.java
public void draw(Canvas canvas) {
final int privateFlags = mPrivateFlags;
final boolean dirtyOpaque = (privateFlags & PFLAG_DIRTY_MASK) == PFLAG_DIRTY_OPAQUE &&
(mAttachInfo == null || !mAttachInfo.mIgnoreDirtyState);
mPrivateFlags = (privateFlags & ~PFLAG_DIRTY_MASK) | PFLAG_DRAWN;
int saveCount;
//1,绘制背景
if (!dirtyOpaque) {
drawBackground(canvas);
}
final int viewFlags = mViewFlags;
boolean horizontalEdges = (viewFlags & FADING_EDGE_HORIZONTAL) != 0;
boolean verticalEdges = (viewFlags & FADING_EDGE_VERTICAL) != 0;
if (!verticalEdges && !horizontalEdges) {
// 2,绘制当前view的图形
if (!dirtyOpaque) onDraw(canvas);
// 3,绘制子view的图形
dispatchDraw(canvas);
drawAutofilledHighlight(canvas);
// Overlay is part of the content and draws beneath Foreground
if (mOverlay != null && !mOverlay.isEmpty()) {
mOverlay.getOverlayView().dispatchDraw(canvas);
}
//4,绘制decorations,如滚动条,前景等 Step 6, draw decorations (foreground, scrollbars)
onDrawForeground(canvas);
// 5,绘制焦点的高亮
drawDefaultFocusHighlight(canvas);
if (debugDraw()) {
debugDrawFocus(canvas);
}
// we're done...
return;
}
//……
}
draw函数中做了这几件事情
- 绘制背景
- 绘制当前view
- 遍历绘制子view
- 绘制前景
我们可以看看Canvas里的绘制方法,这些绘制方法都是JNI方法,并且一一对应着SKCanvas中的绘制方法
//文件-->/frameworks/base/graphics/java/android/graphics/Canvas.java
//……
private static native void nDrawBitmap(long nativeCanvas, int[] colors, int offset, int stride,
float x, float y, int width, int height, boolean hasAlpha, long nativePaintOrZero);
private static native void nDrawColor(long nativeCanvas, int color, int mode);
private static native void nDrawPaint(long nativeCanvas, long nativePaint);
private static native void nDrawPoint(long canvasHandle, float x, float y, long paintHandle);
private static native void nDrawPoints(long canvasHandle, float[] pts, int offset, int count,
long paintHandle);
private static native void nDrawLine(long nativeCanvas, float startX, float startY, float stopX,
float stopY, long nativePaint);
private static native void nDrawLines(long canvasHandle, float[] pts, int offset, int count,
long paintHandle);
private static native void nDrawRect(long nativeCanvas, float left, float top, float right,
float bottom, long nativePaint);
private static native void nDrawOval(long nativeCanvas, float left, float top, float right,
float bottom, long nativePaint);
private static native void nDrawCircle(long nativeCanvas, float cx, float cy, float radius,
long nativePaint);
private static native void nDrawArc(long nativeCanvas, float left, float top, float right,
float bottom, float startAngle, float sweep, boolean useCenter, long nativePaint);
private static native void nDrawRoundRect(long nativeCanvas, float left, float top, float right,
float bottom, float rx, float ry, long nativePaint);
//……
Post Surface
软件绘制的最后一步,通过surface.unlockCanvasAndPost将绘制内容提交给surfaceFlinger绘制,将绘制出来的图形提交给SurfaceFlinger,然后SurfaceFlinger作为消费者处理图形后,我们的界面就显示出来了。
public void unlockCanvasAndPost(Canvas canvas) {
synchronized (mLock) {
checkNotReleasedLocked();
if (mHwuiContext != null) {
mHwuiContext.unlockAndPost(canvas);
} else {
unlockSwCanvasAndPost(canvas);
}
}
}
private void unlockSwCanvasAndPost(Canvas canvas) {
if (canvas != mCanvas) {
throw new IllegalArgumentException("canvas object must be the same instance that "
+ "was previously returned by lockCanvas");
}
if (mNativeObject != mLockedObject) {
Log.w(TAG, "WARNING: Surface's mNativeObject (0x" +
Long.toHexString(mNativeObject) + ") != mLockedObject (0x" +
Long.toHexString(mLockedObject) +")");
}
if (mLockedObject == 0) {
throw new IllegalStateException("Surface was not locked");
}
try {
nativeUnlockCanvasAndPost(mLockedObject, canvas);
} finally {
nativeRelease(mLockedObject);
mLockedObject = 0;
}
}
这里调用了Native函数nativeUnlockCanvasAndPost,我们接着往下看。
static void nativeUnlockCanvasAndPost(JNIEnv* env, jclass clazz,
jlong nativeObject, jobject canvasObj) {
sp<Surface> surface(reinterpret_cast<Surface *>(nativeObject));
if (!isSurfaceValid(surface)) {
return;
}
// detach the canvas from the surface
Canvas* nativeCanvas = GraphicsJNI::getNativeCanvas(env, canvasObj);
nativeCanvas->setBitmap(SkBitmap());
// unlock surface
status_t err = surface->unlockAndPost();
if (err < 0) {
doThrowIAE(env);
}
}
在这里,surface->unlockAndPost()函数就会将Skia绘制出来的图像传递给SurfaceFlinger进行合成。通过skia进行软件绘制的流程已经讲完了,至于如何通过Surface获取缓冲区,在缓冲区绘制完数据后,surface->unlockAndPost()又如何通知SurfaceFlinger,这一点在下一篇文章的图形缓冲区中会详细的讲解。
可以看到,Skia软件绘制的流程比硬件绘制要简单很多,我们接着看看Skia进行Flutter绘制的案例。
Skia进行Flutter的界面绘制
在讲解Flutter如何通过Skia生产图像之前,先简单介绍一下Flutter,Flutter的架构分为Framework层,Engine层和Embedder三层。
Framework层使用dart语言实现,包括UI,文本,图片,按钮等Widgets,渲染,动画,手势等。
Engine使用C++实现,主要包括渲染引擎Skia, Dart虚拟机和文字排版Tex等模块。
Embedder是一个嵌入层,通过该层把Flutter嵌入到各个平台上去,Embedder的主要工作包括渲染Surface设置, 线程设置,以及插件等
了解了Flutter的架构,我们在接着了解Flutter显示一个界面的流程。我们知道在Android中,显示一个界面需要将XML界面布局解析成ViewGroup,然后再经过测量Measure,布局Layout和绘制Draw的流程。Flutter和Android的显示不太一样,它会将通过Dart语言编写的Widget界面布局转换成ElementTree和Render ObjectTree。ElementTree相当于是ViewGroup,Render ObjectTree相当于是经过Measure和Layout流程之后的ViewGroup。这种模式在很多场景上都有使用,比如Webview,在渲染界面时,也会创建一颗Dom树,render树和RenderObject,这样的好处是可以通过Diff比较改变过的组件,然后渲染时,只对改变过的组件做渲染,同时对跨平台友好,可以通过这种树的形式来抽象出不同平台的公共部分。
讲完了上面两个背景,我们直接来看Flutter是如何使用Skia来绘制界面的。
下面是一个Flutter页面的Demo
import 'package:flutter/material.dart';
void main() => runApp(MyApp());
class MyApp extends StatelessWidget {
@override
Widget build(BuildContext context) {
return MaterialApp(
title: 'Flutter Demo',
theme: ThemeData(
primarySwatch: Colors.blue,
),
home: const MyHomePage(title: 'Flutter Demo Home Page'),
);
}
}
这个页面是一个WidgetTree,相当于我们Activity的xml,widget树会转换成ElementTree和RenderObjectTree,我们看看入口函数runApp时如何进行树的转换的。
//文件-->/packages/flutter/lib/src/widgets
void runApp(Widget app) {
WidgetsFlutterBinding.ensureInitialized()
..scheduleAttachRootWidget(app)
..scheduleWarmUpFrame();
}
void scheduleAttachRootWidget(Widget rootWidget) {
Timer.run(() {
attachRootWidget(rootWidget);
});
}
void attachRootWidget(Widget rootWidget) {
_readyToProduceFrames = true;
_renderViewElement = RenderObjectToWidgetAdapter<RenderBox>(
container: renderView,
debugShortDescription: '[root]',
child: rootWidget,
).attachToRenderTree(buildOwner, renderViewElement as RenderObjectToWidgetElement<RenderBox>);
}
接着看attachToRenderTree函数
RenderObjectToWidgetElement<T> attachToRenderTree(BuildOwner owner, [RenderObjectToWidgetElement<T> element]) {
if (element == null) {
owner.lockState(() {
element = createElement(); //创建rootElement
element.assignOwner(owner); //绑定BuildOwner
});
owner.buildScope(element, () { //子widget的初始化从这里开始
element.mount(null, null); // 初始化子Widget前,先执行rootElement的mount方法
});
} else {
...
}
return element;
}
void mount(Element parent, dynamic newSlot) {
super.mount(parent, newSlot);
_renderObject = widget.createRenderObject(this);
attachRenderObject(newSlot);
_dirty = false;
}
从代码中可以看到,Widget都被转换成了Element,Element接着调用了mount方法,在mount方法中,可以看到Widget又被转换成了RenderObject,此时Widget Tree的ElementTree和RenderObject便都生成完了。
前面提到了RenderObject类似于经过了Measure和Layout流程的ViewGroup,RenderObject的Measure和Layout就不在这儿说了,那么还剩一个流程Draw流程,同样是在RenderObject中进行的,它的入口在RenderObject的paint函数中。
// 绘制入口,从 view 根节点开始,逐个绘制所有子节点
@override
void paint(PaintingContext context, Offset offset) {
if (child != null)
context.paintChild(child, offset);
}
可以看到,RenderObject通过PaintingContext来进行了图形的绘制,我们接着来了解一下PaintingContext是什么。
//文件-->/packages/flutter/lib/src/rendering/object.dart
import 'dart:ui' as ui show PictureRecorder;
class PaintingContext extends ClipContext {
@protected
PaintingContext(this._containerLayer, this.estimatedBounds)
final ContainerLayer _containerLayer;
final Rect estimatedBounds;
PictureLayer _currentLayer;
ui.PictureRecorder _recorder;
Canvas _canvas;
@override
Canvas get canvas {
if (_canvas == null)
_startRecording();
return _canvas;
}
void _startRecording() {
_currentLayer = PictureLayer(estimatedBounds);
_recorder = ui.PictureRecorder();
_canvas = Canvas(_recorder);
_containerLayer.append(_currentLayer);
}
void stopRecordingIfNeeded() {
if (!_isRecording)
return;
_currentLayer.picture = _recorder.endRecording();
_currentLayer = null;
_recorder = null;
_canvas = null;
}
可以看到,PaintingContext是绘制的上下文,前面讲OpenGL进行硬件加速时提到的CanvasContext,它也是绘制的上下文,里面封装了Skia,Opengl或者Vulkan的渲染管线。这里的PaintingContext则封装了Skia。
我们可以通过CanvasContext的get canvas函数获取Canvas,它调用了_startRecording函数中,函数中创建了PictureRecorder和Canvas,这两个类都是位于dart:ui库中,dart:ui位于engine层,在前面架构中提到,Flutter分为Framewrok,Engine和embened三层,Engine中包含了Skia,dart虚拟机和Text。dart:ui就是位于Engine层的。
我们接着去Engine层的代码看看Canvas的实现。
//文件-->engine-master\lib\ui\canvas.dart
Canvas(PictureRecorder recorder, [ Rect? cullRect ]) : assert(recorder != null) { // ignore: unnecessary_null_comparison
if (recorder.isRecording)
throw ArgumentError('"recorder" must not already be associated with another Canvas.');
_recorder = recorder;
_recorder!._canvas = this;
cullRect ??= Rect.largest;
_constructor(recorder, cullRect.left, cullRect.top, cullRect.right, cullRect.bottom);
}
void _constructor(PictureRecorder recorder,
double left,
double top,
double right,
double bottom) native 'Canvas_constructor';
这里Canvas调用了Canvas_constructor这一个native方法,我们接着看这个native方法的实现。
//文件-->engine-master\lib\ui\painting\engine.cc
static void Canvas_constructor(Dart_NativeArguments args) {
UIDartState::ThrowIfUIOperationsProhibited();
DartCallConstructor(&Canvas::Create, args);
}
fml::RefPtr<Canvas> Canvas::Create(PictureRecorder* recorder,
double left,
double top,
double right,
double bottom) {
if (!recorder) {
Dart_ThrowException(
ToDart("Canvas constructor called with non-genuine PictureRecorder."));
return nullptr;
}
fml::RefPtr<Canvas> canvas = fml::MakeRefCounted<Canvas>(
recorder->BeginRecording(SkRect::MakeLTRB(left, top, right, bottom)));
recorder->set_canvas(canvas);
return canvas;
}
Canvas::Canvas(SkCanvas* canvas) : canvas_(canvas) {}
可以看到,这里通过PictureRecorder->BeginRecording创建了SKCanvas,这其实是SKCanvas的另外一种使用方式,这里我简单的介绍一个使用demo。
Picture createSolidRectanglePicture(
Color color, double width, double height)
{
PictureRecorder recorder = PictureRecorder();
Canvas canvas = Canvas(recorder);
Paint paint = Paint();
paint.color = color;
canvas.drawRect(Rect.fromLTWH(0, 0, width, height), paint);
return recorder.endRecording();
}
这个demo的效果如下图,它创建Skia的方式就和Flutter创建Skia的方式是一样的。
此时,我们的SKCanvas创建好了,并且直接通过PaintingContext的get canvas函数就能获取到,那么获取到SKCanvas后直接调用Canvas的绘制api,就可以将图像绘制出来了。
Flutter界面显示的全流程是比较复杂的,Flutter是完全是自建的一套图像显示流程,无法通过Android的SurfaceFlinger进行图像合成,也无法使用Android的Gralloc模块分配图像缓冲区,所以它需要有自己的图像生产者,有自己的图形消费者,也有自己的图形缓冲区,这里面就有非常多的流程,比如如何接收VSync,如何处理及合成Layer,如何创建图像缓冲区,这里只是对Flutter的图像生产者的部分做了一个初步的介绍,关于Flutter更深入一步的细节,就不在这里继续讲解了。后面我会专门写一系列文章来详细讲解Flutter。
Vulkan
与OpenGL相比,Vulkan可以更详细的向显卡描述你的应用程序打算做什么,从而可以获得更好的性能和更小的驱动开销,作为OpenGL的替代者,它设计之初就是为了跨平台实现的,可以同时在Windows、Linux和Android开发。甚至在Mac OS系统上运行。Android在7.0开始,便增加了对Vulkan的支持,Vulkan一定是未来的趋势,因为它比OpenGL的性能更好更强大。下面我们就了解一下,如何使用Vulkan来生产图像。
如何使用Vulkan?
Vulkan的使用和OpenGL类似,同样是三步:初始化,绘制,提交buffer下面来看一下具体的流程
1,初始化Vulkan实例,物理设备和任务队列以及Surface
- 创建Instances实例
VkInstanceCreateInfo instance_create_info = {
VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
nullptr,
0,
&application_info,
0,
nullptr,
static_cast<uint32_t>(desired_extensions.size()),
desired_extensions.size() > 0 ? &desired_extensions[0] : nullptr
};
VkInstance inst;
VkResult result = vkCreateInstance( &instance_create_info, nullptr, &inst );
- 初始化物理设备,也就是我们的显卡设备,Vulkna的设计是支持多GPU的,这里选择第一个设备就行了。
uint32_t extensions_count = 0;
VkResult result = VK_SUCCESS;
//获取所有可用物理设备,并选择第一个
result = vkEnumerateDeviceExtensionProperties( physical_device, nullptr, &extensions_count, &available_extensions[0]);
if( (result != VK_SUCCESS) ||
(extensions_count == 0) ) {
std::cout << "Could not get the number of device extensions." << std::endl;
return false;
}
- 获取queue,Vulkan的所有操作,从绘图到上传纹理,都需要将命令提交到队列中
uint32_t queue_families_count = 0;
//获取队列簇,并选择第一个
queue_families.resize( queue_families_count );
vkGetPhysicalDeviceQueueFamilyProperties( physical_device, &queue_families_count, &queue_families[0] );
if( queue_families_count == 0 ) {
std::cout << "Could not acquire properties of queue families." << std::endl;
return false;
}
- 初始化逻辑设备,在选择要使用的物理设备之后,我们需要设置一个逻辑设备用于交互。
VkResult result = vkCreateDevice( physical_device, &device_create_info, nullptr, &logical_device );
if( (result != VK_SUCCESS) ||
(logical_device == VK_NULL_HANDLE) ) {
std::cout << "Could not create logical device." << std::endl;
return false;
}
return true;
- 上述初始完毕后,接着初始化Surface,然后我们就可以使用Vulkan进行绘制了
#ifdef VK_USE_PLATFORM_WIN32_KHR
//创建WIN32的surface,如果是Android,需要使用VkAndroidSurfaceCreateInfoKHR
VkWin32SurfaceCreateInfoKHR surface_create_info = {
VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR,
nullptr,
0,
window_parameters.HInstance,
window_parameters.HWnd
};
VkResult result = vkCreateWin32SurfaceKHR( instance, &surface_create_info, nullptr, &presentation_surface );
2,通过vkCmdDraw函数进行图像绘制
void vkCmdDraw(
//在Vulkan中,像绘画命令、内存转换等操作并不是直接通过方法调用去完成的,而是需要把所有的操作放在Command Buffer里
VkCommandBuffer commandBuffer,
uint32_t vertexCount, //顶点数量
uint32_t instanceCount, // 要画的instance数量,没有:置1
uint32_t firstVertex,// vertex buffer中第一个位置 和 vertex Shader 里gl_vertexIndex 相关。
uint32_t firstInstance);// 同firstVertex 类似。
3,提交buffer
if (vkQueueSubmit(graphicsQueue, 1, &submitInfo, VK_NULL_HANDLE) != VK_SUCCESS) {
throw std::runtime_error("failed to submit draw command buffer!");
}
我在这里比较浅显的介绍了Vulkan的用法,但上面介绍的只是Vulkan的一点皮毛,Vulkan的使用比OpenGL要复杂的很多,机制也复杂很多,如果想进一步了解Vulkan还是得专门去深入研究。虽然只介绍了一点皮毛,但已经可以让我们去了解Vulkan这一图像生产者,是如何在Android系统中生产图像的,下面就来看看吧。
Vulkan进行硬件加速
在前面讲OpenGL 进行硬件加速时,提到了CanvasContext,它会根据渲染的类型选择不同的渲染管线,Android是通过Vulkan或者还是通过OpenGL渲染,主要是CanvasContext里选择的渲染管线的不同。
CanvasContext* CanvasContext::create(RenderThread& thread,
bool translucent, RenderNode* rootRenderNode, IContextFactory* contextFactory) {
auto renderType = Properties::getRenderPipelineType();
switch (renderType) {
case RenderPipelineType::OpenGL:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<OpenGLPipeline>(thread));
case RenderPipelineType::SkiaGL:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<skiapipeline::SkiaOpenGLPipeline>(thread));
case RenderPipelineType::SkiaVulkan:
return new CanvasContext(thread, translucent, rootRenderNode, contextFactory,
std::make_unique<skiapipeline::SkiaVulkanPipeline>(thread));
default:
LOG_ALWAYS_FATAL("canvas context type %d not supported", (int32_t) renderType);
break;
}
return nullptr;
}
我们这里直接看SkiaVulkanPipeline。
//文件->/frameworks/base/libs/hwui/pipeline/skia/SkiaVulkanPipeline.cpp
SkiaVulkanPipeline::SkiaVulkanPipeline(renderthread::RenderThread& thread)
: SkiaPipeline(thread), mVkManager(thread.vulkanManager()) {}
SkiaVulkanPipeline的构造函数中初始化了VulkanManager,VulkanManager是对Vulkan使用的封装,和前面讲到的OpenGLPipeline中的EglManager类似。我们看一下VulkanManager的初始化函数。
//文件-->/frameworks/base/libs/hwui/renderthread/VulkanManager.cpp
void VulkanManager::initialize() {
if (hasVkContext()) {
return;
}
auto canPresent = [](VkInstance, VkPhysicalDevice, uint32_t) { return true; };
mBackendContext.reset(GrVkBackendContext::Create(vkGetInstanceProcAddr, vkGetDeviceProcAddr,
&mPresentQueueIndex, canPresent));
//……
}
初始化函数中我们主要关注GrVkBackendContext::Create方法。
// Create the base Vulkan objects needed by the GrVkGpu object
const GrVkBackendContext* GrVkBackendContext::Create(uint32_t* presentQueueIndexPtr,
CanPresentFn canPresent,
GrVkInterface::GetProc getProc) {
//……
const VkInstanceCreateInfo instance_create = {
VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO, // sType
nullptr, // pNext
0, // flags
&app_info, // pApplicationInfo
(uint32_t) instanceLayerNames.count(), // enabledLayerNameCount
instanceLayerNames.begin(), // ppEnabledLayerNames
(uint32_t) instanceExtensionNames.count(), // enabledExtensionNameCount
instanceExtensionNames.begin(), // ppEnabledExtensionNames
};
ACQUIRE_VK_PROC(CreateInstance, VK_NULL_HANDLE, VK_NULL_HANDLE);
//1,创建Vulkan实例
err = grVkCreateInstance(&instance_create, nullptr, &inst);
if (err < 0) {
SkDebugf("vkCreateInstance failed: %d\n", err);
return nullptr;
}
uint32_t gpuCount;
//2,查询可用物理设备
err = grVkEnumeratePhysicalDevices(inst, &gpuCount, nullptr);
if (err) {
//……
}
//……
gpuCount = 1;
//3,选择物理设备
err = grVkEnumeratePhysicalDevices(inst, &gpuCount, &physDev);
if (err) {
//……
}
//4,查询队列簇
uint32_t queueCount;
grVkGetPhysicalDeviceQueueFamilyProperties(physDev, &queueCount, nullptr);
if (!queueCount) {
//……
return nullptr;
}
SkAutoMalloc queuePropsAlloc(queueCount * sizeof(VkQueueFamilyProperties));
// now get the actual queue props
VkQueueFamilyProperties* queueProps = (VkQueueFamilyProperties*)queuePropsAlloc.get();
//5,选择队列簇
grVkGetPhysicalDeviceQueueFamilyProperties(physDev, &queueCount, queueProps);
//……
// iterate to find the graphics queue
const VkDeviceCreateInfo deviceInfo = {
VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO, // sType
nullptr, // pNext
0, // VkDeviceCreateFlags
queueInfoCount, // queueCreateInfoCount
queueInfo, // pQueueCreateInfos
(uint32_t) deviceLayerNames.count(), // layerCount
deviceLayerNames.begin(), // ppEnabledLayerNames
(uint32_t) deviceExtensionNames.count(), // extensionCount
deviceExtensionNames.begin(), // ppEnabledExtensionNames
&deviceFeatures // ppEnabledFeatures
};
//6,创建逻辑设备
err = grVkCreateDevice(physDev, &deviceInfo, nullptr, &device);
if (err) {
SkDebugf("CreateDevice failed: %d\n", err);
grVkDestroyInstance(inst, nullptr);
return nullptr;
}
auto interface =
sk_make_sp<GrVkInterface>(getProc, inst, device, extensionFlags);
if (!interface->validate(extensionFlags)) {
SkDebugf("Vulkan interface validation failed\n");
grVkDeviceWaitIdle(device);
grVkDestroyDevice(device, nullptr);
grVkDestroyInstance(inst, nullptr);
return nullptr;
}
VkQueue queue;
grVkGetDeviceQueue(device, graphicsQueueIndex, 0, &queue);
GrVkBackendContext* ctx = new GrVkBackendContext();
ctx->fInstance = inst;
ctx->fPhysicalDevice = physDev;
ctx->fDevice = device;
ctx->fQueue = queue;
ctx->fGraphicsQueueIndex = graphicsQueueIndex;
ctx->fMinAPIVersion = kGrVkMinimumVersion;
ctx->fExtensions = extensionFlags;
ctx->fFeatures = featureFlags;
ctx->fInterface.reset(interface.release());
ctx->fOwnsInstanceAndDevice = true;
return ctx;
}
可以看到,GrVkBackendContext::Create中所作的事情就是初始化Vulkan,初始化的流程和前面介绍如何使用Vulkan中初始化流程都是一样的,这些都是通用的流程。
初始化完成,我们接着看看Vulkan如何绑定Surface,只有绑定了Surface,我们才能使用Vulkan进行图像绘制。
//文件-->/frameworks/base/libs/hwui/renderthread/VulkanManager.cpp
VulkanSurface* VulkanManager::createSurface(ANativeWindow* window) {
initialize();
if (!window) {
return nullptr;
}
VulkanSurface* surface = new VulkanSurface();
VkAndroidSurfaceCreateInfoKHR surfaceCreateInfo;
memset(&surfaceCreateInfo, 0, sizeof(VkAndroidSurfaceCreateInfoKHR));
surfaceCreateInfo.sType = VK_STRUCTURE_TYPE_ANDROID_SURFACE_CREATE_INFO_KHR;
surfaceCreateInfo.pNext = nullptr;
surfaceCreateInfo.flags = 0;
surfaceCreateInfo.window = window;
VkResult res = mCreateAndroidSurfaceKHR(mBackendContext->fInstance, &surfaceCreateInfo, nullptr,
&surface->mVkSurface);
if (VK_SUCCESS != res) {
delete surface;
return nullptr;
}
SkDEBUGCODE(VkBool32 supported; res = mGetPhysicalDeviceSurfaceSupportKHR(
mBackendContext->fPhysicalDevice, mPresentQueueIndex,
surface->mVkSurface, &supported);
// All physical devices and queue families on Android must be capable of
// presentation with any
// native window.
SkASSERT(VK_SUCCESS == res && supported););
if (!createSwapchain(surface)) {
destroySurface(surface);
return nullptr;
}
return surface;
}
可以看到,这个创建了VulkanSurface,并绑定了ANativeWindow,ANativeWindow是Android的原生窗口,在前面介绍OpenGL进行硬件渲染时,也提到过createSurface这个函数,它是在performDraw被执行的,在这里就不重复说了。
接下来就是调用Vulkan的api进行绘制的图像的流程
bool SkiaVulkanPipeline::draw(const Frame& frame, const SkRect& screenDirty, const SkRect& dirty,
const FrameBuilder::LightGeometry& lightGeometry,
LayerUpdateQueue* layerUpdateQueue, const Rect& contentDrawBounds,
bool opaque, bool wideColorGamut,
const BakedOpRenderer::LightInfo& lightInfo,
const std::vector<sp<RenderNode>>& renderNodes,
FrameInfoVisualizer* profiler) {
sk_sp<SkSurface> backBuffer = mVkSurface->getBackBufferSurface();
if (backBuffer.get() == nullptr) {
return false;
}
SkiaPipeline::updateLighting(lightGeometry, lightInfo);
renderFrame(*layerUpdateQueue, dirty, renderNodes, opaque, wideColorGamut, contentDrawBounds,
backBuffer);
layerUpdateQueue->clear();
// Draw visual debugging features
if (CC_UNLIKELY(Properties::showDirtyRegions ||
ProfileType::None != Properties::getProfileType())) {
SkCanvas* profileCanvas = backBuffer->getCanvas();
SkiaProfileRenderer profileRenderer(profileCanvas);
profiler->draw(profileRenderer);
profileCanvas->flush();
}
// Log memory statistics
if (CC_UNLIKELY(Properties::debugLevel != kDebugDisabled)) {
dumpResourceCacheUsage();
}
return true;
}
最后通过swapBuffers提交绘制内容
void VulkanManager::swapBuffers(VulkanSurface* surface) {
if (CC_UNLIKELY(Properties::waitForGpuCompletion)) {
ATRACE_NAME("Finishing GPU work");
mDeviceWaitIdle(mBackendContext->fDevice);
}
SkASSERT(surface->mBackbuffers);
VulkanSurface::BackbufferInfo* backbuffer =
surface->mBackbuffers + surface->mCurrentBackbufferIndex;
GrVkImageInfo* imageInfo;
SkSurface* skSurface = surface->mImageInfos[backbuffer->mImageIndex].mSurface.get();
skSurface->getRenderTargetHandle((GrBackendObject*)&imageInfo,
SkSurface::kFlushRead_BackendHandleAccess);
// Check to make sure we never change the actually wrapped image
SkASSERT(imageInfo->fImage == surface->mImages[backbuffer->mImageIndex]);
// We need to transition the image to VK_IMAGE_LAYOUT_PRESENT_SRC_KHR and make sure that all
// previous work is complete for before presenting. So we first add the necessary barrier here.
VkImageLayout layout = imageInfo->fImageLayout;
VkPipelineStageFlags srcStageMask = layoutToPipelineStageFlags(layout);
VkPipelineStageFlags dstStageMask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
VkAccessFlags srcAccessMask = layoutToSrcAccessMask(layout);
VkAccessFlags dstAccessMask = VK_ACCESS_MEMORY_READ_BIT;
VkImageMemoryBarrier imageMemoryBarrier = {
VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, // sType
NULL, // pNext
srcAccessMask, // outputMask
dstAccessMask, // inputMask
layout, // oldLayout
VK_IMAGE_LAYOUT_PRESENT_SRC_KHR, // newLayout
mBackendContext->fGraphicsQueueIndex, // srcQueueFamilyIndex
mPresentQueueIndex, // dstQueueFamilyIndex
surface->mImages[backbuffer->mImageIndex], // image
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1} // subresourceRange
};
mResetCommandBuffer(backbuffer->mTransitionCmdBuffers[1], 0);
VkCommandBufferBeginInfo info;
memset(&info, 0, sizeof(VkCommandBufferBeginInfo));
info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
info.flags = 0;
mBeginCommandBuffer(backbuffer->mTransitionCmdBuffers[1], &info);
mCmdPipelineBarrier(backbuffer->mTransitionCmdBuffers[1], srcStageMask, dstStageMask, 0, 0,
nullptr, 0, nullptr, 1, &imageMemoryBarrier);
mEndCommandBuffer(backbuffer->mTransitionCmdBuffers[1]);
surface->mImageInfos[backbuffer->mImageIndex].mImageLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR;
// insert the layout transfer into the queue and wait on the acquire
VkSubmitInfo submitInfo;
memset(&submitInfo, 0, sizeof(VkSubmitInfo));
submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submitInfo.waitSemaphoreCount = 0;
submitInfo.pWaitDstStageMask = 0;
submitInfo.commandBufferCount = 1;
submitInfo.pCommandBuffers = &backbuffer->mTransitionCmdBuffers[1];
submitInfo.signalSemaphoreCount = 1;
// When this command buffer finishes we will signal this semaphore so that we know it is now
// safe to present the image to the screen.
submitInfo.pSignalSemaphores = &backbuffer->mRenderSemaphore;
// Attach second fence to submission here so we can track when the command buffer finishes.
mQueueSubmit(mBackendContext->fQueue, 1, &submitInfo, backbuffer->mUsageFences[1]);
// Submit present operation to present queue. We use a semaphore here to make sure all rendering
// to the image is complete and that the layout has been change to present on the graphics
// queue.
const VkPresentInfoKHR presentInfo = {
VK_STRUCTURE_TYPE_PRESENT_INFO_KHR, // sType
NULL, // pNext
1, // waitSemaphoreCount
&backbuffer->mRenderSemaphore, // pWaitSemaphores
1, // swapchainCount
&surface->mSwapchain, // pSwapchains
&backbuffer->mImageIndex, // pImageIndices
NULL // pResults
};
mQueuePresentKHR(mPresentQueue, &presentInfo);
surface->mBackbuffer.reset();
surface->mImageInfos[backbuffer->mImageIndex].mLastUsed = surface->mCurrentTime;
surface->mImageInfos[backbuffer->mImageIndex].mInvalid = false;
surface->mCurrentTime++;
}
这些流程都和OpenGL是一样的,初始化,绑定Surface,绘制,提交,所以就不细说了,对Vulkan有兴趣的,可以深入的去研究。至此Android中的另一个图像生产者Vulkan生产图像的流程也讲完了。
结尾
OpenGL,Skia,Vulkan都是跨平台的图形生产者,我们不仅仅可以在Android设备上使用,我们也可以在IOS设备上使用,也可以在Windows设备上使用,使用的流程基本和上面一致,但是需要适配设备的原生窗口和缓冲,所以掌握了Android是如何绘制图像的,我们也具备了掌握其他任何设备上是如何绘制图像的能力。
在下一篇文章中,我会介绍Android图像渲染原理的最后一部分:图像缓冲区。这三部分如果都能掌握,我们基本就能掌握Android中图像绘制的原理了。