catapult/tracing/tracing/extras/rail/rail_interaction_record.html - platform/external/chromium-trace - Git at Google

 <!DOCTYPE html>
 <!--
 Copyright (c) 2015 The Chromium Authors. All rights reserved.
 Use of this source code is governed by a BSD-style license that can be
 found in the LICENSE file.
 -->
 <link rel="import" href="/tracing/base/base.html">
 <link rel="import" href="/tracing/base/statistics.html">
 <link rel="import" href="/tracing/core/auditor.html">
 <link rel="import" href="/tracing/model/model.html">
 <link rel="import" href="/tracing/base/range_utils.html">
 <link rel="import" href="/tracing/extras/chrome/chrome_model_helper.html">

 <script>
 'use strict';

 /**
  * @fileoverview Base class for trace data Auditors.
  */
 tr.exportTo('tr.e.rail', function() {
   var ColorScheme = tr.b.ColorScheme;

   // When computing an IR's RAIL score, the IR's comfort and efficiency are
   // averaged together such that the lower score has a higher weight.
   // Without knowing which sub-score is lower, comfort is
   // theoretically twice as important as efficiency. If the entire web were to
   // eventually achieve relatively high comfort scores such that comfort was
   // less of a concern than efficiency, then this number could be lowered. If
   // further thought suggests that comfort is even more than twice as important
   // as efficiency, then this number could be raised.
   // Must be greater than 0.
   var COMFORT_IMPORTANCE = 2;

   // We need an up-front list of all IR types in order to keep various groupings
   // stable, in presence of only a portion of interactions in a given trace.
   var ALL_RAIL_TYPE_NAMES = [
     'rail_response',
     'rail_animate',
     'rail_idle',
     'rail_load'
   ];

   var DOES_RAIL_TYPE_NAME_EXIST = {};
   ALL_RAIL_TYPE_NAMES.forEach(function(railTypeName) {
     DOES_RAIL_TYPE_NAME_EXIST[railTypeName] = true;
   });

   var RAIL_ORDER = [];
   ALL_RAIL_TYPE_NAMES.forEach(function(railTypeName) {
     RAIL_ORDER.push(railTypeName.toUpperCase());
     RAIL_ORDER.push(userFriendlyRailTypeName(railTypeName).toUpperCase());
   });


   function RAILInteractionRecord(
       parentModel, title, railTypeName, start, duration) {
     if (!DOES_RAIL_TYPE_NAME_EXIST[railTypeName])
       throw new Error(railTypeName + ' is not listed in ALL_RAIL_TYPE_NAMES');

     var colorId = ColorScheme.getColorIdForReservedName(railTypeName);
     this.railTypeName_ = railTypeName;
     this.name = '';
     tr.model.InteractionRecord.call(
         this, parentModel, title, colorId, start, duration);
   }

   RAILInteractionRecord.prototype = {
     __proto__: tr.model.InteractionRecord.prototype,

     updateArgs: function() {
       var args = {};

       var layoutSlices = this.associatedEvents.filter(function(event) {
         return event.title === 'FrameView::layout';
       });
       var timeInLayout = tr.b.Statistics.sum(layoutSlices, function(event) {
         return event.duration;
       });

       args['layoutInfo'] = {
         'timeInLayout': timeInLayout
       };

       this.args = args;
     },

     get railTypeName() {
       return this.railTypeName_;
     },

     /**
      * Returns the overall rail score, from 0 to 1.
      *
      * RAILScore for an interaction merges the user's comfort with the
      * efficiency, in order to create a perception-oriented measure
      * of how users percieve speed during this interaction.
      *
      *  0 means a bad user experience.
      *  1 means a perfect user experience.
      */
     get railScore() {
       var comfort = this.normalizedUserComfort;
       var efficiency = this.normalizedEfficiency;
       return weightedAverage2(comfort, efficiency, COMFORT_IMPORTANCE);
     },

     /**
      * Measures the comfort the user experienced, from 0 to 1.
      *
      * A user performs an interaction with an expectation in mind.
      * When we meet their expectations, we get perfect comfort.
      * When we don't live up to their expectations, comfort goes down.
      */
     get normalizedUserComfort() {
       throw new Error('Not implemented');
     },

     /**
      * Returns the sum of the number of CPU ms spent by this IR.
      */
     get rawCpuMs() {
       var cpuMs = 0;
       this.associatedEvents.forEach(function(event) {
         if (event.cpuSelfTime)
           cpuMs += event.cpuSelfTime;
       });
       return cpuMs;
     },

     /**
      * Returns a number between 0 and 1 representing how efficiently this IR
      * used CPU resources. 0 is maximally in-efficient, 1 is maximally
      * efficient.
      */
     get normalizedCpuEfficiency() {
       var minCpuMs = this.duration * this.minCpuFraction;
       var maxCpuMs = this.duration * this.maxCpuFraction;
       var normalizedCpu = tr.b.normalize(this.rawCpuMs, minCpuMs, maxCpuMs);
       return 1 - tr.b.clamp(normalizedCpu, 0, 1);
     },

     /**
      * The minimum fraction of a CPU that can be spent on this IR before the
      * efficiency score will be impacted.
      * If less CPU ms than this is spent on this IR, then
      * normalizedCpuEfficiency will be 1.
      */
     get minCpuFraction() {
       return 0.5;
     },

     /**
      * The maximum fraction of a CPU that can be spent on this IR.
      * If more CPU ms than this is spent on this IR, then
      * normalizedCpuEfficiency will be 0.
      */
     get maxCpuFraction() {
       return 1.5;
     },

     /**
      * Measures the efficiency of the interaction from 0 to 1.
      *
      * Efficiency is a notion of how well we used the machine's limited
      * resources in service of this interaction. If we used it perfectly,
      * we would get a 1.0. If we used everything that there was to use ---
      * power, memory, cpu, then we'd get a zero.
      */
     get normalizedEfficiency() {
       return this.normalizedCpuEfficiency;
     }
   };

   // Returns a weighted average of numbers between 0 and 1.
   // The lower input has a higher weight.
   // If the first input should have a higher weight a priori its relationship to
   // the other input, then set opt_apriori > 1.
   // This function is graphed at http://goo.gl/XMWUKA
   function weightedAverage2(x, y, opt_apriori) {
     var numerator = 0;
     var denominator = 0;

     var xWeight = (opt_apriori || 1) * Math.exp(1 - x);
     numerator += xWeight * x;
     denominator += xWeight;

     var yWeight = Math.exp(1 - y);
     numerator += yWeight * y;
     denominator += yWeight;

     return numerator / denominator;
   }

   // A user friendly name is currently formed by dropping the rail_ prefix and
   // capitalizing.
   function userFriendlyRailTypeName(railTypeName) {
     if (railTypeName.length < 6 || railTypeName.indexOf('rail_') != 0)
       return railTypeName;
     return railTypeName[5].toUpperCase() + railTypeName.slice(6);
   }

   // Compare two rail type names or rail user-friendly names so they are sorted
   // in R,A,I,L order. Capitalization is ignore. Non rail names are sorted
   // lexicographically after rail names.
   function railCompare(name1, name2) {
     var i1 = RAIL_ORDER.indexOf(name1.toUpperCase());
     var i2 = RAIL_ORDER.indexOf(name2.toUpperCase());
     if (i1 == -1 && i2 == -1)
       return name1.localeCompare(name2);
     if (i1 == -1)
       return 1;   // i2 is a RAIL name but not i1.
     if (i2 == -1)
       return -1;  // i1 is a RAIL name but not i2.
     // Two rail names.
     return i1 - i2;
   }

   return {
     RAILInteractionRecord: RAILInteractionRecord,
     weightedAverage2: weightedAverage2,
     userFriendlyRailTypeName: userFriendlyRailTypeName,
     railCompare: railCompare,
     ALL_RAIL_TYPE_NAMES: ALL_RAIL_TYPE_NAMES
   };
 });
 </script>
	<!DOCTYPE html>
	<!--
	Copyright (c) 2015 The Chromium Authors. All rights reserved.
	Use of this source code is governed by a BSD-style license that can be
	found in the LICENSE file.
	-->
	<link rel="import" href="/tracing/base/base.html">
	<link rel="import" href="/tracing/base/statistics.html">
	<link rel="import" href="/tracing/core/auditor.html">
	<link rel="import" href="/tracing/model/model.html">
	<link rel="import" href="/tracing/base/range_utils.html">
	<link rel="import" href="/tracing/extras/chrome/chrome_model_helper.html">

	<script>
	'use strict';

	/**
	* @fileoverview Base class for trace data Auditors.
	*/
	tr.exportTo('tr.e.rail', function() {
	var ColorScheme = tr.b.ColorScheme;

	// When computing an IR's RAIL score, the IR's comfort and efficiency are
	// averaged together such that the lower score has a higher weight.
	// Without knowing which sub-score is lower, comfort is
	// theoretically twice as important as efficiency. If the entire web were to
	// eventually achieve relatively high comfort scores such that comfort was
	// less of a concern than efficiency, then this number could be lowered. If
	// further thought suggests that comfort is even more than twice as important
	// as efficiency, then this number could be raised.
	// Must be greater than 0.
	var COMFORT_IMPORTANCE = 2;

	// We need an up-front list of all IR types in order to keep various groupings
	// stable, in presence of only a portion of interactions in a given trace.
	var ALL_RAIL_TYPE_NAMES = [
	'rail_response',
	'rail_animate',
	'rail_idle',
	'rail_load'
	];

	var DOES_RAIL_TYPE_NAME_EXIST = {};
	ALL_RAIL_TYPE_NAMES.forEach(function(railTypeName) {
	DOES_RAIL_TYPE_NAME_EXIST[railTypeName] = true;
	});

	var RAIL_ORDER = [];
	ALL_RAIL_TYPE_NAMES.forEach(function(railTypeName) {
	RAIL_ORDER.push(railTypeName.toUpperCase());
	RAIL_ORDER.push(userFriendlyRailTypeName(railTypeName).toUpperCase());
	});



	function RAILInteractionRecord(
	parentModel, title, railTypeName, start, duration) {
	if (!DOES_RAIL_TYPE_NAME_EXIST[railTypeName])
	throw new Error(railTypeName + ' is not listed in ALL_RAIL_TYPE_NAMES');

	var colorId = ColorScheme.getColorIdForReservedName(railTypeName);
	this.railTypeName_ = railTypeName;
	this.name = '';
	tr.model.InteractionRecord.call(
	this, parentModel, title, colorId, start, duration);
	}

	RAILInteractionRecord.prototype = {
	__proto__: tr.model.InteractionRecord.prototype,

	updateArgs: function() {
	var args = {};

	var layoutSlices = this.associatedEvents.filter(function(event) {
	return event.title === 'FrameView::layout';
	});
	var timeInLayout = tr.b.Statistics.sum(layoutSlices, function(event) {
	return event.duration;
	});

	args['layoutInfo'] = {
	'timeInLayout': timeInLayout
	};

	this.args = args;
	},

	get railTypeName() {
	return this.railTypeName_;
	},

	/**
	* Returns the overall rail score, from 0 to 1.
	*
	* RAILScore for an interaction merges the user's comfort with the
	* efficiency, in order to create a perception-oriented measure
	* of how users percieve speed during this interaction.
	*
	* 0 means a bad user experience.
	* 1 means a perfect user experience.
	*/
	get railScore() {
	var comfort = this.normalizedUserComfort;
	var efficiency = this.normalizedEfficiency;
	return weightedAverage2(comfort, efficiency, COMFORT_IMPORTANCE);
	},

	/**
	* Measures the comfort the user experienced, from 0 to 1.
	*
	* A user performs an interaction with an expectation in mind.
	* When we meet their expectations, we get perfect comfort.
	* When we don't live up to their expectations, comfort goes down.
	*/
	get normalizedUserComfort() {
	throw new Error('Not implemented');
	},

	/**
	* Returns the sum of the number of CPU ms spent by this IR.
	*/
	get rawCpuMs() {
	var cpuMs = 0;
	this.associatedEvents.forEach(function(event) {
	if (event.cpuSelfTime)
	cpuMs += event.cpuSelfTime;
	});
	return cpuMs;
	},

	/**
	* Returns a number between 0 and 1 representing how efficiently this IR
	* used CPU resources. 0 is maximally in-efficient, 1 is maximally
	* efficient.
	*/
	get normalizedCpuEfficiency() {
	var minCpuMs = this.duration * this.minCpuFraction;
	var maxCpuMs = this.duration * this.maxCpuFraction;
	var normalizedCpu = tr.b.normalize(this.rawCpuMs, minCpuMs, maxCpuMs);
	return 1 - tr.b.clamp(normalizedCpu, 0, 1);
	},

	/**
	* The minimum fraction of a CPU that can be spent on this IR before the
	* efficiency score will be impacted.
	* If less CPU ms than this is spent on this IR, then
	* normalizedCpuEfficiency will be 1.
	*/
	get minCpuFraction() {
	return 0.5;
	},

	/**
	* The maximum fraction of a CPU that can be spent on this IR.
	* If more CPU ms than this is spent on this IR, then
	* normalizedCpuEfficiency will be 0.
	*/
	get maxCpuFraction() {
	return 1.5;
	},

	/**
	* Measures the efficiency of the interaction from 0 to 1.
	*
	* Efficiency is a notion of how well we used the machine's limited
	* resources in service of this interaction. If we used it perfectly,
	* we would get a 1.0. If we used everything that there was to use ---
	* power, memory, cpu, then we'd get a zero.
	*/
	get normalizedEfficiency() {
	return this.normalizedCpuEfficiency;
	}
	};

	// Returns a weighted average of numbers between 0 and 1.
	// The lower input has a higher weight.
	// If the first input should have a higher weight a priori its relationship to
	// the other input, then set opt_apriori > 1.
	// This function is graphed at http://goo.gl/XMWUKA
	function weightedAverage2(x, y, opt_apriori) {
	var numerator = 0;
	var denominator = 0;

	var xWeight = (opt_apriori \|\| 1) * Math.exp(1 - x);
	numerator += xWeight * x;
	denominator += xWeight;

	var yWeight = Math.exp(1 - y);
	numerator += yWeight * y;
	denominator += yWeight;

	return numerator / denominator;
	}

	// A user friendly name is currently formed by dropping the rail_ prefix and
	// capitalizing.
	function userFriendlyRailTypeName(railTypeName) {
	if (railTypeName.length < 6 \|\| railTypeName.indexOf('rail_') != 0)
	return railTypeName;
	return railTypeName[5].toUpperCase() + railTypeName.slice(6);
	}

	// Compare two rail type names or rail user-friendly names so they are sorted
	// in R,A,I,L order. Capitalization is ignore. Non rail names are sorted
	// lexicographically after rail names.
	function railCompare(name1, name2) {
	var i1 = RAIL_ORDER.indexOf(name1.toUpperCase());
	var i2 = RAIL_ORDER.indexOf(name2.toUpperCase());
	if (i1 == -1 && i2 == -1)
	return name1.localeCompare(name2);
	if (i1 == -1)
	return 1; // i2 is a RAIL name but not i1.
	if (i2 == -1)
	return -1; // i1 is a RAIL name but not i2.
	// Two rail names.
	return i1 - i2;
	}

	return {
	RAILInteractionRecord: RAILInteractionRecord,
	weightedAverage2: weightedAverage2,
	userFriendlyRailTypeName: userFriendlyRailTypeName,
	railCompare: railCompare,
	ALL_RAIL_TYPE_NAMES: ALL_RAIL_TYPE_NAMES
	};
	});
	</script>