首页
学习
活动
专区
工具
TVP
发布
社区首页 >问答首页 >Chrome requestAnimationFrame问题

Chrome requestAnimationFrame问题
EN

Stack Overflow用户
提问于 2013-10-16 12:59:24
回答 2查看 11.5K关注 0票数 21

相关主题:requestAnimationFrame garbage collection

我一直致力于在我为触摸设备构建的小工具中实现流畅的动画效果,我找到的工具之一就是Chrome Memory Timeline screen。

这对我在rAF循环中评估内存消耗有一点帮助,但我对在Chrome30中观察到的行为的一些方面感到困扰。

在最初进入运行rAF循环的页面时,我看到了以下内容。

看起来没问题。如果我已经完成了我的工作,并消除了内部循环中的对象分配,那么就不应该有锯齿。这是与链接主题一致的行为,也就是说,无论您何时使用rAF,Chrome都有一个内置的泄漏。(呀!)

当我开始在页面中做各种事情时,它会变得更加有趣。

我并没有做什么不同的事情,只是临时添加了两个元素,使CSS3 3D变换样式应用于几个帧,然后我停止与它们交互。

我们在这里看到的是Chrome报告说,突然之间,每个rAF触发(16ms)都会导致Animation Frame Fired x 3

这种重复以及它这样做的速率单调增加,直到页面刷新。

您已经可以在第二个截图中看到,在从Animation Frame FiredAnimation Frame Fired x 3的初始跳跃之后,锯齿斜率急剧增加。

过了一会儿,它跳到了x 21

看起来我的代码运行了一大堆额外的时间,但所有这些额外的多次运行只是浪费了热量,丢弃了计算。

当我在拍摄第三个录像带时,我的Macbook预热得很厉害。不久之后,在我能够将时间线移动到末尾(大约8分钟)以查看x数量增加到什么程度之前,检查员窗口变得完全没有响应,并且提示我我的页面变得没有响应,必须终止。

下面是在页面中运行的完整代码:

// ============================================================================
// Copyright (c) 2013 Steven Lu

// 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.
// ============================================================================

// This is meant to be a true velocity verlet integrator, which means sending
// in for the force and torque a function (not a value). If the forces provided
// are evaluated at the current time step then I think we are left with plain
// old Euler integration.  This is a 3 DOF integrator that is meant for use
// with 2D rigid bodies, but it should be equally useful for modeling 3d point
// dynamics.

// this attempts to minimize memory waste by operating on state in-place.

function vel_verlet_3(state, acc, dt) {
  var x = state[0],
      y = state[1],
      z = state[2],
      vx = state[3],
      vy = state[4],
      vz = state[5],
      ax = state[6],
      ay = state[7],
      az = state[8],
      x1 = x + vx * dt + 0.5 * ax * dt * dt,
      y1 = y + vy * dt + 0.5 * ay * dt * dt,
      z1 = z + vz * dt + 0.5 * az * dt * dt,  // eqn 1
      a1 = acc(x1, y1, z1),
      ax1 = a1[0],
      ay1 = a1[1],
      az1 = a1[2];
  state[0] = x1;
  state[1] = y1;
  state[2] = z1;
  state[3] = vx + 0.5 * (ax + ax1) * dt,
  state[4] = vy + 0.5 * (ay + ay1) * dt,
  state[5] = vz + 0.5 * (az + az1) * dt; // eqn 2
  state[6] = ax1;
  state[7] = ay1;
  state[8] = az1;
}

// velocity indepedent acc --- shit this is gonna need to change soon
var acc = function(x, y, z) {
  return [0,0,0];
};
$("#lock").click(function() {
  var values = [Number($('#ax').val()), Number($('#ay').val()), Number($('#az').val())];
  acc = function() {
    return values;
  };
});

// Obtain the sin and cos from an angle.
// Allocate nothing.
function getRotation(angle, cs) {
  cs[0] = Math.cos(angle);
  cs[1] = Math.sin(angle);
}

// Provide the localpoint as [x,y].
// Allocate nothing.
function global(bodystate, localpoint, returnpoint) {
  getRotation(bodystate[2], returnpoint);
  // now returnpoint contains cosine+sine of angle.
  var px = bodystate[0], py = bodystate[1];
  var x = localpoint[0], y = localpoint[1];
  // console.log('global():', cs, [px, py], localpoint, 'with', [x,y]);
  // [ c -s px ]   [x]
  // [ s  c py ] * [y]
  //               [1]
  var c = returnpoint[0];
  var s = returnpoint[1];
  returnpoint[0] = c * x - s * y + px;
  returnpoint[1] = s * x + c * y + py;
}

function local(bodystate, globalpoint, returnpoint) {
  getRotation(bodystate[2], returnpoint);
  // now returnpoint contains cosine+sine of angle
  var px = bodystate[0], py = bodystate[1];
  var x = globalpoint[0], y = globalpoint[1];
  // console.log('local():', cs, [px, py], globalpoint, 'with', [x,y]);
  // [  c s ]   [x - px]
  // [ -s c ] * [y - py]
  var xx = x - px, yy = y - py;
  var c = returnpoint[0], s = returnpoint[1];
  returnpoint[0] = c * xx + s * yy;
  returnpoint[1] = -s * xx + c * yy;
}

var cumulativeOffset = function(element) {
  var top = 0, left = 0;
  do {
    top += element.offsetTop || 0;
    left += element.offsetLeft || 0;
    element = element.offsetParent;
  } while (element);
  return {
    top: top,
    left: left
  };
};

// helper to create/assign position debugger (handles a single point)
// offset here is a boundingclientrect offset and needs window.scrollXY correction
var hasDPOffsetRun = false;
var dpoff = false;
function debugPoint(position, id, color, offset) {
  if (offset) {
    position[0] += offset.left;
    position[1] += offset.top;
  }
  // if (position[0] >= 0) { console.log('debugPoint:', id, color, position); }
  var element = $('#point' + id);
  if (!element.length) {
    element = $('<div></div>')
    .attr('id', 'point' + id)
    .css({
          pointerEvents: 'none',
          position: 'absolute',
          backgroundColor: color,
          border: '#fff 1px solid',
          top: -2,
          left: -2,
          width: 2,
          height: 2,
          borderRadius: 300,
          boxShadow: '0 0 6px 0 ' + color
        });
    $('body').append(
        $('<div></div>')
        .addClass('debugpointcontainer')
        .css({
          position: 'absolute',
          top: 0,
          left: 0
        })
      .append(element)
    );
    if (!hasDPOffsetRun) {
      // determine the offset of the body-appended absolute element. body's margin
      // is the primary offender that tends to throw a wrench into our shit.
      var dpoffset = $('.debugpointcontainer')[0].getBoundingClientRect();
      dpoff = [dpoffset.left + window.scrollX, dpoffset.top + window.scrollY];
      hasDPOffsetRun = true;
    }
  }
  if (dpoff) {
    position[0] -= dpoff[0];
    position[1] -= dpoff[1];
  }
  // set position
  element[0].style.webkitTransform = 'translate3d(' + position[0] + 'px,' + position[1] + 'px,0)';
}

var elements_tracked = [];

/*
var globaleventhandler = function(event) {
  var t = event.target;
  if (false) { // t is a child of a tracked element...

  }
};

// when the library is loaded the global event handler for GRAB is not
// installed. It is lazily installed when GRAB_global is first called, and so
// if you only ever call GRAB then the document does not get any handlers
// attached to it.  This will remain unimplemented as it's not clear what the
// semantics for defining behavior are. It's much more straightforward to use
// the direct API
function GRAB_global(element, custom_behavior) {
  // this is the entry point that will initialize a grabbable element all state
  // for the element will be accessible through its __GRAB__ element through
  // the DOM, and the DOM is never accessed (other than through initial
  // assignment) by the code.

  // event handlers are attached to the document, so use GRAB_direct if your
  // webpage relies on preventing event bubbling.
  if (elements_tracked.indexOf(element) !== -1) {
    console.log('You tried to call GRAB() on an element more than once.',
                element, 'existing elements:', elements_tracked);
  }
  elements_tracked.push(element);
  if (elements_tracked.length === 1) { // this is the initial call
    document.addEventListener('touchstart', globaleventhandler, true);
    document.addEventListener('mousedown', globaleventhandler, true);
  }
}

// cleanup function cleans everything up, returning behavior to normal.
// may provide a boolean true argument to indicate that you want the CSS 3D
// transform value to be cleared
function GRAB_global_remove(cleartransform) {
  document.removeEventListener('touchstart', globaleventhandler, true);
  document.removeEventListener('mousedown', globaleventhandler, true);
}

*/

var mousedownelement = false;
var stop = false;
// there is only one mouse, and the only time when we need to handle release
// of pointer is when the one mouse is let go somewhere far away.
function GRAB(element, onfinish, center_of_mass) {
  // This version directly assigns the event handlers to the element
  // it is less efficient but more "portable" and self-contained, and also
  // potentially more friendly by using a regular event handler rather than
  // a capture event handler, so that you can customize the grabbing behavior
  // better and also more easily define it per element
  var offset = center_of_mass;
  var pageOffset = cumulativeOffset(element);
  var bcrOffset = element.getBoundingClientRect();
  bcrOffset = {
    left: bcrOffset.left + window.scrollX,
    right: bcrOffset.right + window.scrollX,
    top: bcrOffset.top + window.scrollY,
    bottom: bcrOffset.bottom + window.scrollY
  };
  if (!offset) {
    offset = [element.offsetWidth / 2, element.offsetHeight / 2];
  }
  var model = {
    state: [0, 0, 0, 0, 0, 0, 0, 0, 0],
    offset: offset,
    pageoffset: bcrOffset // remember, these values are pre-window.scroll[XY]-corrected
  };
  element.__GRAB__ = model;
  var eventhandlertouchstart = function(event) {
    // set
    var et0 = event.touches[0];
    model.anchor = [0,0];
    local(model.state, [et0.pageX - bcrOffset.left - offset[0], et0.pageY - bcrOffset.top - offset[1]], model.anchor);
    debugPoint([et0.pageX, et0.pageY], 1, 'red');
    event.preventDefault();
    requestAnimationFrame(step);
  };
  var eventhandlermousedown = function(event) {
    console.log('todo: reject right clicks');
    // console.log('a', document.body.scrollLeft);
    // set
    // model.anchor = [event.offsetX - offset[0], event.offsetY - offset[1]];
    model.anchor = [0,0];
    var globalwithoffset = [event.pageX - bcrOffset.left - offset[0], event.pageY - bcrOffset.top - offset[1]];
    local(model.state, globalwithoffset, model.anchor);
    debugPoint([event.pageX, event.pageY], 1, 'red');
    mousedownelement = element;
    requestAnimationFrame(step);
  };
  var eventhandlertouchend = function(event) {
    // clear
    model.anchor = false;
    requestAnimationFrame(step);
  };
  element.addEventListener('touchstart', eventhandlertouchstart, false);
  element.addEventListener('mousedown', eventhandlermousedown, false);
  element.addEventListener('touchend', eventhandlertouchend, false);
  elements_tracked.push(element);
  // assign some favorable properties to grabbable element.
  element.style.webkitTouchCallout = 'none';
  element.style.webkitUserSelect = 'none';
  // TODO: figure out the proper values for these
  element.style.MozUserSelect = 'none';
  element.style.msUserSelect = 'none';
  element.style.MsUserSelect = 'none';
}
document.addEventListener('mouseup', function() {
  if (mousedownelement) {
    mousedownelement.__GRAB__.anchor = false;
    mousedownelement = false;
    requestAnimationFrame(step);
  }
}, false);

function GRAB_remove(element, cleartransform) {}
// unimpld
function GRAB_remove_all(cleartransform) {}

GRAB($('#content2')[0]);

(function() {
  var requestAnimationFrame = window.mozRequestAnimationFrame ||
      window.webkitRequestAnimationFrame ||
      window.msRequestAnimationFrame ||
      window.requestAnimationFrame;
  window.requestAnimationFrame = requestAnimationFrame;
})();

var now = function() { return window.performance ? performance.now() : Date.now(); };
var lasttime = 0;
var abs = Math.abs;
var dt = 0;
var scratch0 = [0,0];
var scratch1 = [0,0]; // memory pool
var step = function(time) {
  dt = (time - lasttime) * 0.001;
  if (time < 1e12) {
    // highres timer
  } else {
    // ms since unix epoch
    if (dt > 1e9) {
      dt = 0;
    }
  }
  // console.log('dt: ' + dt);
  lasttime = time;
  var foundnotstopped = false;
  for (var i = 0; i < elements_tracked.length; ++i) {
    var e = elements_tracked[i];
    var data = e.__GRAB__;
    if (data.anchor) {
      global(data.state, data.anchor, scratch0);
      scratch1[0] = scratch0[0] + data.offset[0];
      scratch1[1] = scratch0[1] + data.offset[1];
      //console.log("output of global", point);
      debugPoint(scratch1,
                 0, 'blue', data.pageoffset);
    } else {
      scratch1[0] = -1000;
      scratch1[1] = -1000;
      debugPoint(scratch1, 0, 'blue');
    }
    // timestep is dynamic and based on reported time. clamped to 100ms.
    if (dt > 0.3) {
      //console.log('clamped from ' + dt + ' @' + now());
      dt = 0.3;
    }
    vel_verlet_3(data.state, acc, dt);
    e.style.webkitTransform = 'translate3d(' + data.state[0] + 'px,' + data.state[1] + 'px,0)' +
        'rotateZ(' + data.state[2] + 'rad)';
  }
  requestAnimationFrame(step);
};

requestAnimationFrame(step);

为了完整,下面是测试页面HTML:

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="utf-8" />
    <meta http-equiv="cache-control" content="max-age=0" />
    <meta http-equiv="cache-control" content="no-cache" />
    <meta http-equiv="expires" content="0" />
    <meta http-equiv="expires" content="Tue, 01 Jan 1980 1:00:00 GMT" />
    <meta http-equiv="pragma" content="no-cache" />
    <title>symplectic integrator test page</title>
    <script src="zepto.js"></script>
    <script src="d3.v3.js"></script>
    <style type='text/css'>
        body {
            position: relative;
            margin: 80px;
        }
        #content {
            width: 800px;
            height: 40px;
            display: inline-block;
            background: lightgreen;
            padding: 20px;
            margin: 30px;
            border: green dashed 1px;
        }
        #content2 {
            top: 200px;
            width: 600px;
            height: 200px;
            display: inline-block;
            background: lightblue;
            padding: 20px;
            margin: 30px;
            border: blue dashed 1px;
        }
    </style>
</head>
<body>
    <div id='scrolling-placeholder' style='background-color: #eee; height: 1000px;'></div>
    <label>dt:<input id='dt' type='number' step='0.001' value='0.016666666' /></label>
    <label>ax:<input id='ax' type='number' step='0.25' value='0' /></label>
    <label>ay:<input id='ay' type='number' step='0.25' value='0' /></label>
    <label>t:<input id='az' type='number' step='0.01' value='0' /></label>
    <button id='lock'>Set</button>
    <button id='zerof' onclick='$("#ax,#ay,#az").val(0);'>Zero forces</button>
    <button id='zerov'>Zero velocities</button>
    <div>
        <span id='content'>content</span>
        <span id='content2'>content2</span>
    </div>
    <div id='debuglog'></div>
    <script src="rb2.js"></script>
</body>
</html>

这应该能满足任何“给我们看代码”的要求。

现在我不敢拿我的性命去赌,但我很确定我至少在以正确的方式使用rAF方面做了一件不错的工作。我没有滥用任何东西,到目前为止,我已经将代码改进为对Javascript内存分配非常轻量级。

所以,真的,绝对没有理由让Chrome拿着这个,试图像火箭一样把我的笔记本电脑送入轨道。没什么原因。

总的来说,Safari似乎能更好地处理它(它不会最终死亡),我还会注意到,iOS通常能够保持200x600px的div以60fps的速度进行平移和旋转。

然而,我承认我从来没有见过Chrome真的像这样死掉,除非我把它记录在内存时间线上。

在这一点上,我有点挠头。这可能只是一些意想不到的,不可预见的与这个特定的开发工具功能的交互(据我所知,这是唯一的一种)。

因此,我尝试了一些新的方法,至少可以通过内存时间线额外回调触发来帮助研究这个问题:

添加了这些行。

window.rafbuf = [];
var step = function(time) {
  window.rafbuf.push(time);

这基本上记录了我的rAF例程( step()函数)被调用的所有次数。

当它正常运行时,它大约每16.7毫秒写一次时间。

我得到了这个:

这清楚地表明它正在使用相同的时间输入参数重新运行step()至少22次,就像时间线试图告诉我的那样。

所以我敢说你,互联网,告诉我这是有意为之的行为。:)

EN

回答 2

Stack Overflow用户

发布于 2014-12-10 05:18:08

我认为您有一个问题,因为您在每个mousedownmouseup事件上都调用requestAnimationFrame(step);。由于您的step()函数也(应该如此)调用requestAnimationFrame(step);,因此您实际上为每个mousedownmouseup事件启动了新的“动画循环”,并且由于您从未停止它们,因此它们会累积起来。

我可以看到您还在代码的末尾启动了“动画循环”。如果你想在发生鼠标事件时立即重绘,你应该将绘图移出step()函数,并直接从鼠标事件处理程序调用它。

类似这样的东西:

function redraw() { 
  // drawing logic
}
function onmousedown() {
  // ...
  redraw()
}
function onmouseup() {
  // ...
  redraw()
}

function step() {
  redraw();
  requestAnimationFrame(step);
}
requestAnimationFrame(step);
票数 6
EN

Stack Overflow用户

发布于 2013-11-13 10:01:31

我为http://www.testufo.com创建了动画,还在http://www.testufo.com/animation-time-graph创建了一个requestAnimationFrame()一致性检查器

http://www.testufo.com/browser.html上列出了支持requestAnimationFrame()与计算机显示器刷新率(即使不是60 at )自动同步的web浏览器列表。这意味着在75 it的监视器上,requestAnimationFrame()现在在支持的浏览器上每秒被调用75次,前提是网页当前处于前台,并且CPU/图形性能允许。

Chrome30的某个版本有一个令人讨厌的requestAnimationFrame()错误:http://www.blurbusters.com/blurbusters-helping-google-debug-chrome-30-animation-bugs/

Chrome 29和Chrome 31运行良好,新版本的Chrome 30也是如此。幸运的是,据我所知,chrome 33 Canary似乎更完全地解决了我所看到的问题。它可以更流畅地运行动画,而不需要调用requestAnimationFrame()。

此外,我还注意到电源管理(CPU减速/节流太省电)会对requestAnimationFrame()的回调速率造成严重影响……它在帧渲染时间( http://www.testufo.com/#test=animation-time-graph&measure=rendering )中表现为奇怪的向上/向下尖峰

票数 2
EN
页面原文内容由Stack Overflow提供。腾讯云小微IT领域专用引擎提供翻译支持
原文链接:

https://stackoverflow.com/questions/19395565

复制
相关文章

相似问题

领券
问题归档专栏文章快讯文章归档关键词归档开发者手册归档开发者手册 Section 归档