In my quest to optimize my game engine I have discovered optimization i have been doing affecting each browser differently, in a lot of cases making one browser worse and the other better!
Currently I'm trying to optimize looping as i do alot of it and depending on the way this is done can have a big effect on the performance of my engine.
Based on the results here http://jsperf.com/for-vs-while-loop-iterating/3
It seems a reverse for loop in chrome is the fastest
var l = foo.length;
for (var i = l; i--;) {
}
And in firefox a forward for loop is fastest
var l = foo.length;
for (var i = 0; i < l; i++) {
}
Now in order to use the correct one per browser I'm doing something like this
function foreach(func, iterations){
var browser = $.browser;
var i;
if (browser.webkit)
{
for(i=iterations;i--;)
{
func(i);
}
}
else
{
for (i = 0; i < iterations; i++)
{
func(i);
}
}
}
but it seems there may be alot of overhead here that may hurt performance.
If you were to provide different ways of looping for different browsers what would you do?
EDIT: seems there was a bug in my testing where i was doing one too many loops on the forward loop and now chrome seems to be the fastest there also, I may not need to optimize the loops but it may still be worth while as mention in another comment incase browser versions change performance again
Unfortunately, if your goal is the best performance loop on each browser, the very last thing you want to do is introduce function calls into it. There's no way you can define your foreach function such that it will be anything like as fast as the straight for loop. Calling the iteration function will wash out any gains you might get.
A quick jsperf can confirm this easily enough. For instance, run this one on Chrome or a recent version of Firefox or Opera. It compares looping forward with for, backward with for, or using the browser's built-in Array#forEach function (part of ECMAScript5). (I think we can assume any foreach function you build will be slower than the built-in one.) As you can see, the forEach version is dramatically slower than either of the for loops, which makes sense: Calling functions isn't very expensive, but it isn't free.
Mind you, what you're doing in the loop probably washes out the difference between counting up and count down. What I'd do is figure out what's fastest on the slower browsers, and use that. You said, for instance, that a reverse loop is faster in Chrome but a forward loop is faster in Firefox. As Chrome's V8 is dramatically faster than Firefox's SpiderMonkey (at the moment, these things are constantly in flux), pick the forward loop, as it's faster on the slower engine.
Otherwise, you're into needing to do preprocessing on your code and churning out a different version tailored to each browser.
I don't think your overhad is as big as you feel, but what you can do is do your test only once:
var foreach;
if (forwardIsFaster) {
foreach = function (func, iterations) {
// loop forwards...
};
} else {
foreach = function (func, iterations) {
// loop backwards...
};
}
That said, I'm not sure using browser sniffing is the best solution here; maybe instead do a test loop on startup, measure which solution is faster, and choose the one that turns out to be faster.
When you start doing something significant inside the loop (e.g. just printing the variable), the order of iteration doesn't matter, see:
http://jsperf.com/for-vs-while-loop-iterating/4
So stop caring about this and if (and only if) your code is slow in any place, just profile it and optimize that part.
You could make foreach return a function, which would be an iterator. So you would have, once in your script, var iterator = foreach($.browser.webkit);. From then on, you would just use iterator(iterations, callback), which would no longer be executing any conditionals.
The key, basically, is that the user's browser won't change, so the result of the execution of that conditional needs to be evaluated only once per script.
Related
I have a computationally heavy function which is called many times in a loop:
function func() { // Some fluff
for(let i = 0; i < 1000; i++) {
i *= 10
i /= 10
}
}
function run() {
for(let i = 0; i < 100000; i++) {
func()
}
}
run()
When I profile this script using Chrome's DevTools, I get this:
run has a self time of 887ms out of the total time of 1015ms even though the only thing it does is repeatedly call func.
I would expect func to have the majority of self time since it's a leaf function.
Why is this?
(V8 developer here.)
the function was automatically inlined after some time when it became "hot".
Correct. As soon as run is optimized, the optimizer decides to inline func into it. After that, as far as the profiler is concerned, all the time is spent in run.
(To verify this, run the snippet in d8 or node with --trace-turbo-inlining.)
Side note: getting to optimized code for run takes a bit longer than usual in this case, because the function never returns to get called again (which is the best time to switch to optimized code). The system waits a bit for that to happen, and when it doesn't happen, run is eventually "on-stack replaced". This is a typical pattern that occurs frequently in small tests and benchmarks, and rarely in real-world code.
Doesn't this just show that performing 100000 function calls are more expensive than a 1000 iterations of two simple arithmetic operations -- which makes sense?
No, it doesn't show that; that's just one particular way how one could be misled by this microbenchmark.
Personally, I'm mildly disappointed to see (with --print-opt-code) that the compiler didn't realize that i *= 10; i /= 10; is a no-op and could be dropped entirely. That would have been another great way to be misled here. Oh well, there's probably some reason why it's harder than it looks for the compiler to figure out that that transformation would be both applicable and safe...
I've seen plenty of questions that suggest using:
for (var i = 0; i < myArray.length; i++){ /* ... */ }
instead of:
for (var i in myArray){ /* ... */ }
for arrays, due to inconsistent iteration (see here).
However, I can't seem to find anything that seems to prefer the object oriented loop:
myArray.forEach(function(item, index){ /* ... */ });
Which seems way more intuitive to me.
For my current project, IE8 compatibilty is important, and I'm considering using Mozilla's polyfill, however I'm not 100% sure how this will work.
Are there any differences between the standard for loop (the first example above) and the Array.prototype.forEach implementation by modern browsers?
Are there any difference between modern browser implementations and Mozilla's implementation linked to above (with special regard to IE8)?
Performance is not as much of an issue, just consistency with which properties are iterated over.
The most substantive difference between the for loop and the forEach method is that, with the former, you may break out of the loop. You can simulate continue by simply returning from the function passed to forEach, but there is no way to stop looping altogether.
Aside from that, the two accomplish effectively the same functionality. Another minor difference involves the scope of the index (and all containing variables) in the for loop, due to variable hoisting.
// 'i' is scoped to the containing function
for (var i = 0; i < arr.length; i++) { ... }
// 'i' is scoped to the internal function
arr.forEach(function (el, i) { ... });
However, I find that forEach is much more expressive—it represents your intent to iterate through each element of an array, and it provides you with a reference to the element, not just the index. Overall, it mostly comes down to personal taste, but if you can use forEach, I would recommend using it.
There are a few more substantial differences between the two versions, specifically regarding performance. In fact, the simple for loop performs considerably better than the forEach method, as demonstrated by this jsperf test.
Whether or not such performance is necessary for you is up to you to decide, and in most cases, I would favor expressiveness over speed. This speed difference is likely due to the minor semantic differences between the basic loop and the method when operating on sparse arrays, as noted in this answer.
If you don't need the behavior of forEach and/or you need to break out of the loop early, you can use Lo-Dash's _.each as an alternative, which will also work cross-browser. If you're using jQuery, it also provides a similar $.each, just note the differences in arguments passed to the callback function in each variation.
(As for the forEach polyfill, it should work in older browsers without problems, if you choose to go that route.)
You can use your custom foreach function which will perform much better then Array.forEach
You should add this once to your code. This will add new function to the Array.
function foreach(fn) {
var arr = this;
var len = arr.length;
for(var i=0; i<len; ++i) {
fn(arr[i], i);
}
}
Object.defineProperty(Array.prototype, 'customForEach', {
enumerable: false,
value: foreach
});
Then you can use it anywhere like the Array.forEach
[1,2,3].customForEach(function(val, i){
});
The only difference it is 3 times faster. https://jsperf.com/native-arr-foreach-vs-custom-foreach
UPDATE: In new Chrome version the performance of .forEach() was improved. However, the solution can give the additional performance in other browsers.
It is suggested by many developers (e.g. Kyle Simpson) to use .forEach to indicate that the array will have a side effect and .map for pure functions. for loops fit well as a general-purpose solution for known number of loops or any other case that doesn't fit as it is easier to communicate because of its broad support across the majority of programming languages.
e.g.
/* For Loop known number of iterations */
const numberOfSeasons = 4;
for (let i = 0; i < numberOfSeasons; i++) {
//Do Something
}
/* Pure transformation */
const arrayToBeUppercased = ['www', 'html', 'js', 'us'];
const acronyms = arrayToBeUppercased.map((el) => el.toUpperCase));
/* Impure, side-effects with .forEach */
const acronymsHolder = [];
['www', 'html', 'js', 'us'].forEach((el) => acronymsHolder.push(el.toUpperCase()));
Convention wise, this seems best, however the community hasn't really settled on a convention on the newer iteration protocol for in loops. Generally, I think it's a good idea to follow the FP concepts that the JS community seems to be open to adopting.
I've put together a jsperf test that compares for loops that iterate over an array with caching the array length condition vs not caching. I thought that caching the variable before to avoid recalculating the array length each iteration would be faster, but the jsperf test says otherwise. Can someone explain why this is? I also thought that including the array length variable definition (when cached) within the for loop's initialization would reduce the time since the parse doesn't need to look up the "var" keyword twice, but that also doesn't appear to be the case.
example without caching:
for(var i = 0; i < testArray.length; i++){
//
}
example with caching
var len = testArray.length;
for(var i = 0; i < len; i++){
//
}
example with caching variable defined in for loop initialization
for(var i = 0, len=testArray.length; i < len; i++){
//
}
http://jsperf.com/for-loop-condition-caching
Can someone explain why this is?
This optimization and case is extremely common so modern JavaScript engines will automatically perform this optimization for you.
Some notes:
This is not the case when iterating a NodeList (such as the result of querySelectorAll
This is an overkill micro optimization for most code paths anyway, usually the body of the loop takes more time than this comparison anyway.
The performance of the scenarios that you posted depends on how smart the JS engine's optimizer is. An engine with a very dump optimizer (or no optimizer at all) will likely be faster when you are using the variable, but you can't even rely on that. After all, length's type is well-known, while a variable can be anything and may require additional checks.
Given a perfect optimizer, all three examples should have the same performance. And as your examples are pretty simple, modern engines should reach that point soon.
I've seen plenty of questions that suggest using:
for (var i = 0; i < myArray.length; i++){ /* ... */ }
instead of:
for (var i in myArray){ /* ... */ }
for arrays, due to inconsistent iteration (see here).
However, I can't seem to find anything that seems to prefer the object oriented loop:
myArray.forEach(function(item, index){ /* ... */ });
Which seems way more intuitive to me.
For my current project, IE8 compatibilty is important, and I'm considering using Mozilla's polyfill, however I'm not 100% sure how this will work.
Are there any differences between the standard for loop (the first example above) and the Array.prototype.forEach implementation by modern browsers?
Are there any difference between modern browser implementations and Mozilla's implementation linked to above (with special regard to IE8)?
Performance is not as much of an issue, just consistency with which properties are iterated over.
The most substantive difference between the for loop and the forEach method is that, with the former, you may break out of the loop. You can simulate continue by simply returning from the function passed to forEach, but there is no way to stop looping altogether.
Aside from that, the two accomplish effectively the same functionality. Another minor difference involves the scope of the index (and all containing variables) in the for loop, due to variable hoisting.
// 'i' is scoped to the containing function
for (var i = 0; i < arr.length; i++) { ... }
// 'i' is scoped to the internal function
arr.forEach(function (el, i) { ... });
However, I find that forEach is much more expressive—it represents your intent to iterate through each element of an array, and it provides you with a reference to the element, not just the index. Overall, it mostly comes down to personal taste, but if you can use forEach, I would recommend using it.
There are a few more substantial differences between the two versions, specifically regarding performance. In fact, the simple for loop performs considerably better than the forEach method, as demonstrated by this jsperf test.
Whether or not such performance is necessary for you is up to you to decide, and in most cases, I would favor expressiveness over speed. This speed difference is likely due to the minor semantic differences between the basic loop and the method when operating on sparse arrays, as noted in this answer.
If you don't need the behavior of forEach and/or you need to break out of the loop early, you can use Lo-Dash's _.each as an alternative, which will also work cross-browser. If you're using jQuery, it also provides a similar $.each, just note the differences in arguments passed to the callback function in each variation.
(As for the forEach polyfill, it should work in older browsers without problems, if you choose to go that route.)
You can use your custom foreach function which will perform much better then Array.forEach
You should add this once to your code. This will add new function to the Array.
function foreach(fn) {
var arr = this;
var len = arr.length;
for(var i=0; i<len; ++i) {
fn(arr[i], i);
}
}
Object.defineProperty(Array.prototype, 'customForEach', {
enumerable: false,
value: foreach
});
Then you can use it anywhere like the Array.forEach
[1,2,3].customForEach(function(val, i){
});
The only difference it is 3 times faster. https://jsperf.com/native-arr-foreach-vs-custom-foreach
UPDATE: In new Chrome version the performance of .forEach() was improved. However, the solution can give the additional performance in other browsers.
It is suggested by many developers (e.g. Kyle Simpson) to use .forEach to indicate that the array will have a side effect and .map for pure functions. for loops fit well as a general-purpose solution for known number of loops or any other case that doesn't fit as it is easier to communicate because of its broad support across the majority of programming languages.
e.g.
/* For Loop known number of iterations */
const numberOfSeasons = 4;
for (let i = 0; i < numberOfSeasons; i++) {
//Do Something
}
/* Pure transformation */
const arrayToBeUppercased = ['www', 'html', 'js', 'us'];
const acronyms = arrayToBeUppercased.map((el) => el.toUpperCase));
/* Impure, side-effects with .forEach */
const acronymsHolder = [];
['www', 'html', 'js', 'us'].forEach((el) => acronymsHolder.push(el.toUpperCase()));
Convention wise, this seems best, however the community hasn't really settled on a convention on the newer iteration protocol for in loops. Generally, I think it's a good idea to follow the FP concepts that the JS community seems to be open to adopting.
1)As we know, there's no side-effect with map() and reduce(). Nowadays, we also have muti-core on cell phone. So is it more efficient to use them?
2)On the other hand, there's only 1 thread for js to execute on most of the browsers. Therefor map() and reduce() are prepared for server-side scripting?
I just tested this today, using map and reduce over floating point numbers, with the latest node.js version, and the answer is that map and reduce was two orders of magnitude slower than an regular for loop.
var r = array.map(x => x*x).reduce( (total,num) => total+num,0);
~11,000ms
var r = 0.0;
array.forEach( (x,i,a) => r += x*x )
~300ms
var r = 0.0;
for (var j = 0; j < array.length;j++){
var x = array[j];
r += x*x;
}
~35ms
EDIT: One should note that this difference is much less in Firefox, and may be much less in future version of Node / Chrome as well.
1)
there's no side-effect with map() and reduce()
Well. You very well can implement map and reduce callbacks having side effects. Nothing prevents it and in the current state of JavaScript it's not even considered as bad practice.
2)
there's only 1 thread for js to execute on most of the browsers
There's only one thread in all today's JS engines, even when they run server-side (in fact there can be more but in isolation, not accessing the same array).
So the fact there is no side effect wouldn't make array modifications parallelisable at all. No JS engine can do otherwise than call the callback sequentially on standard arrays.
Note : as pointed by zirak, there's this not standard Mozilla ParallelArray thing which could help making parallel execution. I don't know if there's something similar on V8.
As long as the dimensions of the array are very low ( in order of 10's) then there is not much a difference in the performance, but when the size of the array increases to a very large value, then using conventional for loop is a great and better method, because we just need to loop through the elements and get a value at a specific index as we point, but in other methods, not only we are getting the value at an index, but also additional information such as the index, (in map, reduce, forEach) and accumulator value (in reduce). And these methods as need an a callback function, they both take up a great deal of stock on the memory for callbacks which further reduces the performance speed.
You can check the justification of this with the following script.
Just see the values that are console logged.
var scripts=[];
// GLOBAL variales declaration
var a=[];
function preload() {
for(var i=0;i<100000;i++) a[i]=i;
}
preload();
// TEST function 0
scripts.push(function() {
var sum=0;
a.forEach(function(v) {
sum+=v;
});
//console.log(sum);
});
// TEST function 1
scripts.push(function() {
a.reduce(function(acc,v) {
return acc+v;
});
});
// TEST function 2
scripts.push(function() {
var sum=0;
for(var i=0;i<a.length;i++) {
sum+=a[i];
}
});
// EVALUATION
scripts.forEach(function(f,index) {
var date=new Date();
for(var i=0;i<10000;i++) {
f();
}
console.log("call "+index+" "+(new Date()-date));
});
It is easily overlooked, but the key to getting the benefits of MapReduce is to
A) Exploit the optimized shuffle. Often, your map and reduce functions can be implemented in a slow language, as long as the shuffle - the most expensive operation - is well optimized, it will still be fast and scalable.
B) Exploit the checkpointing functionality to recover from node failures (but hopefully, your CPU cores won't fail).
So in the end, map-reduce is actually neither about the map, nor the reduce functions. It's about the framework around it; which will give you good performance even with bad "map" and "reduce" functions (unless you lose control over your data set size in the shuffle step!).
The gains to be obtained from doing a multi-threaded map-reduce on a single node are fairly low, and most likely there are much better ways of parallelizing your shop for shared memory architectures than map-reduce...
Unfortunately, there is a lot of hype (and too little understanding) surrounding mapreduce these days. If you look up the original paper, it goes into detail about "Backup Tasks", "Machine Failures" and "locality optimization" (neither of which makes sense for an in-memory single-host use case).
Just because it has a "map" and a "reduce" doesn't make it a "mapreduce" yet.
It's only a MapReduce if it has an optimized shuffle, node crash and straggler recovery.