Android 丢帧原理以及办法

接近年底,想分享点儿东西给大家。

Android UI绘制过程

开发中的卡顿我想没跟人都遇到过,之前也是搜博客看看怎么个解决办法,没有认真研究过,今天我打算跟大家聊一聊。

先从View 说吧。相信大家应该都知道View的绘制过程,measure,layout,draw。丢帧一定是在16ms内没有把这些事儿干完就对了,这里我们简单的分一下,主要是计算时间,以及绘图时间。

计算时间:这里的measure,layout的过程,都是会向下递归计算的,学过数据结构的话,应该知道,深搜的代价是很大的。所以尽量让树的高度降低,这里就引出扁平化布局。

绘图时间:这里需要着重讲一下,因为有时候这才是我们UI卡顿的主要原因。在这里我们要把android的试图看成是三维的,就像photoshop的图层一样。android在绘制的时候就会一层一层的“粉刷”,好了,那么造成卡顿,也就是丢帧,说白了最后没有在16ms内做完。好了,让我们剖析一下:

invalidate():

我们知道invalidate 是用来请求View 重绘的

1
2
3
4
5
6
7
8
// Propagate the damage rectangle to the parent view.
final AttachInfo ai = mAttachInfo;
final ViewParent p = mParent;
if (p != null && ai != null && l < r && t < b) {
final Rect damage = ai.mTmpInvalRect;
damage.set(l, t, r, b);
p.invalidateChild(this, damage);
}

invalidateInternal
这里可以看出来draw的过程其实就是拿到AttachInfo 里面包含着绘制信息,以及将绘制区域拿到,通过parent去绘制。让我们跟进去。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
public ViewParent invalidateChildInParent(final int[] location, final Rect dirty) {
if ((mPrivateFlags & PFLAG_DRAWN) == PFLAG_DRAWN ||
(mPrivateFlags & PFLAG_DRAWING_CACHE_VALID) == PFLAG_DRAWING_CACHE_VALID) {
if ((mGroupFlags & (FLAG_OPTIMIZE_INVALIDATE | FLAG_ANIMATION_DONE)) !=
FLAG_OPTIMIZE_INVALIDATE) {
dirty.offset(location[CHILD_LEFT_INDEX] - mScrollX,
location[CHILD_TOP_INDEX] - mScrollY);
if ((mGroupFlags & FLAG_CLIP_CHILDREN) == 0) {
dirty.union(0, 0, mRight - mLeft, mBottom - mTop);
}
final int left = mLeft;
final int top = mTop;
if ((mGroupFlags & FLAG_CLIP_CHILDREN) == FLAG_CLIP_CHILDREN) {
if (!dirty.intersect(0, 0, mRight - left, mBottom - top)) {
dirty.setEmpty();
}
}
mPrivateFlags &= ~PFLAG_DRAWING_CACHE_VALID;
location[CHILD_LEFT_INDEX] = left;
location[CHILD_TOP_INDEX] = top;
if (mLayerType != LAYER_TYPE_NONE) {
mPrivateFlags |= PFLAG_INVALIDATED;
}
return mParent;
} else {
mPrivateFlags &= ~PFLAG_DRAWN & ~PFLAG_DRAWING_CACHE_VALID;
location[CHILD_LEFT_INDEX] = mLeft;
location[CHILD_TOP_INDEX] = mTop;
if ((mGroupFlags & FLAG_CLIP_CHILDREN) == FLAG_CLIP_CHILDREN) {
dirty.set(0, 0, mRight - mLeft, mBottom - mTop);
} else {
// in case the dirty rect extends outside the bounds of this container
dirty.union(0, 0, mRight - mLeft, mBottom - mTop);
}
if (mLayerType != LAYER_TYPE_NONE) {
mPrivateFlags |= PFLAG_INVALIDATED;
}
return mParent;
}
}
return null;
}

invalidateChildInParent
这里的dirty代表你绘制的这块区域是否透明。

1
2
3
4
5
6
void invalidate() {
mDirty.set(0, 0, mWidth, mHeight);
if (!mWillDrawSoon) {
scheduleTraversals();
}
}

invalidate
这里我们看到了个关键函数 scheduleTraversals ,为什么说神奇。我们看一下

1
2
3
4
5
6
7
8
9
10
11
12
13
void scheduleTraversals() {
if (!mTraversalScheduled) {
mTraversalScheduled = true;
mTraversalBarrier = mHandler.getLooper().getQueue().postSyncBarrier();
mChoreographer.postCallback(
Choreographer.CALLBACK_TRAVERSAL, mTraversalRunnable, null);
if (!mUnbufferedInputDispatch) {
scheduleConsumeBatchedInput();
}
notifyRendererOfFramePending();
pokeDrawLockIfNeeded();
}
}

scheduleTraversals
这里最重要的是Choreographer 这个,我们最终算出来的绘制信息都要通过它回调,开始他会注册一个广播用来接收时钟信息,然后他会在内部建立一个UI绘制队列:CallbackQueue,我们在外部CallBack的时候,会将我们的绘制信息作为CallbackRecord 然后会在接收到一个时钟信号的时候进行doFrame操作,并打印Traces信息,从而来绘制一帧。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
private static final class CallbackRecord {
public CallbackRecord next;
public long dueTime;
public Object action; // Runnable or FrameCallback
public Object token;
public void run(long frameTimeNanos) {
if (token == FRAME_CALLBACK_TOKEN) {
((FrameCallback)action).doFrame(frameTimeNanos);
} else {
((Runnable)action).run();
}
}
}
private final class CallbackQueue {
private CallbackRecord mHead;
public boolean hasDueCallbacksLocked(long now) {
return mHead != null && mHead.dueTime <= now;
}
public CallbackRecord extractDueCallbacksLocked(long now) {
CallbackRecord callbacks = mHead;
if (callbacks == null || callbacks.dueTime > now) {
return null;
}
CallbackRecord last = callbacks;
CallbackRecord next = last.next;
while (next != null) {
if (next.dueTime > now) {
last.next = null;
break;
}
last = next;
next = next.next;
}
mHead = next;
return callbacks;
}
public void addCallbackLocked(long dueTime, Object action, Object token) {
CallbackRecord callback = obtainCallbackLocked(dueTime, action, token);
CallbackRecord entry = mHead;
if (entry == null) {
mHead = callback;
return;
}
if (dueTime < entry.dueTime) {
callback.next = entry;
mHead = callback;
return;
}
while (entry.next != null) {
if (dueTime < entry.next.dueTime) {
callback.next = entry.next;
break;
}
entry = entry.next;
}
entry.next = callback;
}
public void removeCallbacksLocked(Object action, Object token) {
CallbackRecord predecessor = null;
for (CallbackRecord callback = mHead; callback != null;) {
final CallbackRecord next = callback.next;
if ((action == null || callback.action == action)
&& (token == null || callback.token == token)) {
if (predecessor != null) {
predecessor.next = next;
} else {
mHead = next;
}
recycleCallbackLocked(callback);
} else {
predecessor = callback;
}
callback = next;
}
}
}

CallbackQueue and CallbackRecord

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
private void postCallbackDelayedInternal(int callbackType,
Object action, Object token, long delayMillis) {
if (DEBUG_FRAMES) {
Log.d(TAG, "PostCallback: type=" + callbackType
+ ", action=" + action + ", token=" + token
+ ", delayMillis=" + delayMillis);
}
synchronized (mLock) {
final long now = SystemClock.uptimeMillis();
final long dueTime = now + delayMillis;
mCallbackQueues[callbackType].addCallbackLocked(dueTime, action, token);
if (dueTime <= now) {
scheduleFrameLocked(now);
} else {
Message msg = mHandler.obtainMessage(MSG_DO_SCHEDULE_CALLBACK, action);
msg.arg1 = callbackType;
msg.setAsynchronous(true);
mHandler.sendMessageAtTime(msg, dueTime);
}
}
}

postCallbackDelayedInternal
可以看到这里我们把我们的绘制内容扔到队列里,等待轮训。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
private final class FrameDisplayEventReceiver extends DisplayEventReceiver
implements Runnable {
private boolean mHavePendingVsync;
private long mTimestampNanos;
private int mFrame;
public FrameDisplayEventReceiver(Looper looper) {
super(looper);
}
@Override
public void onVsync(long timestampNanos, int builtInDisplayId, int frame) {
// Ignore vsync from secondary display.
// This can be problematic because the call to scheduleVsync() is a one-shot.
// We need to ensure that we will still receive the vsync from the primary
// display which is the one we really care about. Ideally we should schedule
// vsync for a particular display.
// At this time Surface Flinger won't send us vsyncs for secondary displays
// but that could change in the future so let's log a message to help us remember
// that we need to fix this.
if (builtInDisplayId != SurfaceControl.BUILT_IN_DISPLAY_ID_MAIN) {
Log.d(TAG, "Received vsync from secondary display, but we don't support "
+ "this case yet. Choreographer needs a way to explicitly request "
+ "vsync for a specific display to ensure it doesn't lose track "
+ "of its scheduled vsync.");
scheduleVsync();
return;
}
// Post the vsync event to the Handler.
// The idea is to prevent incoming vsync events from completely starving
// the message queue. If there are no messages in the queue with timestamps
// earlier than the frame time, then the vsync event will be processed immediately.
// Otherwise, messages that predate the vsync event will be handled first.
long now = System.nanoTime();
if (timestampNanos > now) {
Log.w(TAG, "Frame time is " + ((timestampNanos - now) * 0.000001f)
+ " ms in the future! Check that graphics HAL is generating vsync "
+ "timestamps using the correct timebase.");
timestampNanos = now;
}
if (mHavePendingVsync) {
Log.w(TAG, "Already have a pending vsync event. There should only be "
+ "one at a time.");
} else {
mHavePendingVsync = true;
}
mTimestampNanos = timestampNanos;
mFrame = frame;
Message msg = Message.obtain(mHandler, this);
msg.setAsynchronous(true);
mHandler.sendMessageAtTime(msg, timestampNanos / TimeUtils.NANOS_PER_MS);
}
@Override
public void run() {
mHavePendingVsync = false;
doFrame(mTimestampNanos, mFrame);
}
}

FrameDisplayEventReceiver
接收时钟脉冲信号的广播,16ms一次,我们的目的就是在这个时钟脉冲里搞定整个 view

Android 动画

Animator,ScrollTo,offsetLeftAndRight,这里面我们先单列这几项,都是同一个原理。这里我们可以大胆的猜想,一定是频繁执行我们的 Choreographer.CallBack 来绘制,因为只要在16ms内绘制成功,那就是流畅的动画。下面我们验证一下

ScrollTo:

我们先看一下 View 中这个方法

1
2
3
4
5
6
7
8
9
10
11
12
13
public void scrollTo(int x, int y) {
if (mScrollX != x || mScrollY != y) {
int oldX = mScrollX;
int oldY = mScrollY;
mScrollX = x;
mScrollY = y;
invalidateParentCaches();
onScrollChanged(mScrollX, mScrollY, oldX, oldY);
if (!awakenScrollBars()) {
postInvalidateOnAnimation();
}
}
}

scrollTo
很简单,我们都可以看懂,开始位置,结束位置,这里我们重点关注 postInvalidateOnAnimation() 这个方法

1
2
3
4
5
6
7
8
public void postInvalidateOnAnimation() {
// We try only with the AttachInfo because there's no point in invalidating
// if we are not attached to our window
final AttachInfo attachInfo = mAttachInfo;
if (attachInfo != null) {
attachInfo.mViewRootImpl.dispatchInvalidateOnAnimation(this);
}
}

postInvalidateOnAnimation
我们可以看到,这里的动画过程绘制他还是扔到了ViewRootImpl 代理做这件事。

1
2
3
public void dispatchInvalidateRectOnAnimation(AttachInfo.InvalidateInfo info) {
mInvalidateOnAnimationRunnable.addViewRect(info);
}

dispatchInvalidateOnAnimation
这里我们看到他开了个线程 mInvalidateOnAnimationRunnable 去添加我们这个将要绘制的 view,接下来我们继续庖丁解牛

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
final class InvalidateOnAnimationRunnable implements Runnable {
private boolean mPosted;
private final ArrayList<View> mViews = new ArrayList<View>();
private final ArrayList<AttachInfo.InvalidateInfo> mViewRects =
new ArrayList<AttachInfo.InvalidateInfo>();
private View[] mTempViews;
private AttachInfo.InvalidateInfo[] mTempViewRects;
public void addView(View view) {
synchronized (this) {
mViews.add(view);
postIfNeededLocked();
}
}
public void addViewRect(AttachInfo.InvalidateInfo info) {
synchronized (this) {
mViewRects.add(info);
postIfNeededLocked();
}
}
public void removeView(View view) {
synchronized (this) {
mViews.remove(view);
for (int i = mViewRects.size(); i-- > 0; ) {
AttachInfo.InvalidateInfo info = mViewRects.get(i);
if (info.target == view) {
mViewRects.remove(i);
info.recycle();
}
}
if (mPosted && mViews.isEmpty() && mViewRects.isEmpty()) {
mChoreographer.removeCallbacks(Choreographer.CALLBACK_ANIMATION, this, null);
mPosted = false;
}
}
}
@Override
public void run() {
final int viewCount;
final int viewRectCount;
synchronized (this) {
mPosted = false;
viewCount = mViews.size();
if (viewCount != 0) {
mTempViews = mViews.toArray(mTempViews != null
? mTempViews : new View[viewCount]);
mViews.clear();
}
viewRectCount = mViewRects.size();
if (viewRectCount != 0) {
mTempViewRects = mViewRects.toArray(mTempViewRects != null
? mTempViewRects : new AttachInfo.InvalidateInfo[viewRectCount]);
mViewRects.clear();
}
}
for (int i = 0; i < viewCount; i++) {
mTempViews[i].invalidate();
mTempViews[i] = null;
}
for (int i = 0; i < viewRectCount; i++) {
final View.AttachInfo.InvalidateInfo info = mTempViewRects[i];
info.target.invalidate(info.left, info.top, info.right, info.bottom);
info.recycle();
}
}
private void postIfNeededLocked() {
if (!mPosted) {
mChoreographer.postCallback(Choreographer.CALLBACK_ANIMATION, this, null);
mPosted = true;
}
}
}

InvalidateOnAnimationRunnable
终于,应了我们的猜想,ViewRootImpl 有一个专门执行动画绘制操作的线程,我们可以看到 run() 里面不断地CallBack,然后回收,当然里面有些线程锁啥的不涉及本文就不细说了。

ValueAnimator:

这里我们有个 AnimationHandler 来执行动画操作,这其中我们可以看到

1
2
3
4
5
6
7
for (int i = 0; i < numAnims; ++i) {
ValueAnimator anim = mTmpAnimations.get(i);
if (mAnimations.contains(anim) && anim.doAnimationFrame(frameTime)) {
mEndingAnims.add(anim);
}
}
mTmpAnimations.clear();

doAnimationFrame
这里在不断循环我们所有的anim,并在不断执行 scheduleAnimation 方法

1
2
3
4
5
6
private void scheduleAnimation() {
if (!mAnimationScheduled) {
mChoreographer.postCallback(Choreographer.CALLBACK_ANIMATION, mAnimate, null);
mAnimationScheduled = true;
}
}

scheduleAnimation
剩下的大家自己翻阅源码把。

这里总结一下。我们所有界面上视图的变化都是都是 ViewRootImpl 把需要重绘的东西填充 Choreographer 中的 mCallbackQueues 队列,然后在时钟脉冲的广播下进行轮训执行。

既然提到队列,假如我们在16ms内大量的填充 AttachInfo 之类的绘制OBJ,就会导致无法再一次时钟脉冲内绘制完毕,就会在造成丢帧,UI阻塞。

避免 Android UI 卡顿解决办法

解决办法:分析了好多,这里说两个方法。

1.避免重绘,这里避免图层(View)迭代。这里我们可以去开发者模式中对“显示GPU视图更新”打钩

Alt text

过度绘制

Alt text

优化以后
这里引用 http://hukai.me/android-performance-render 这篇博客的作者,盗个图。😂

这里可以进行,选择制定画布绘制,而不是整个view去绘制。可以在onDraw中进行限制,去限制绘制区域,例如

canvas.clipRect(100,100,350,600, Region.Op.INTERSECT);

2.扁平化布局,归根结底也是减少 mCallbackQueues 队列大小。保证尽量在16ms内绘制完毕,再有就是可以减少视图 ViewTree 的高度,减少时间复杂度,从而优化计算过程

Alt text

xml代码

Alt text

优化后的xml代码
*附:

Alt text

绘制层级
通过打开刚才说的开发者选项,来根据颜色来判断页面绘制情况。

距离回家还有8 个小时,17年希望可以发觉更多的东西给大家,并且希望大家可以积极指出文章中的错误。祝大家新年快乐!😄