For loop speed performance in javascript

Question

I went throught http://www.youtube.com/watch?v=mHtdZgou0qU speed up your javascript.

So i did this personal speed test:

    var count = 50000000;
    var testDummy;

    // test 1
    testDummy = 0;
    var test1Start = new Date().getTime();
    var i;
    for (i=0;i<count;i++) {
        testDummy++;
    }
    var test1End = new Date().getTime();
    var test1Total = (test1End-test1Start);

    // test 2
    testDummy = 0;
    var test2Start = new Date().getTime();
    var i
    for (i=count; i--;) {
        testDummy++;
    }
    var test2End = new Date().getTime();
    var test2Total = (test2End-test2Start);


    debug(
      "test1\n" + 
      "total: " + test1Total + "\n" + 
      "test2\n" + 
      "total: " + test2Total 
    );

I get not significant results, like sometimes they are even and sometimes not.

My question is, if i use for loop like this: "for(i=count;i--;)" is it really faster ? Am i doing something wrong in my tests.

Thanks for your help!

I think you got something wrong here, for(i=count;i--;) doesn't make any sense... — MarioVW
– MarioVW, Commented Feb 4, 2011 at 21:35
Sorry, it does make sense (when it reaches 0 casts to false)... however I don't see how this could be faster — MarioVW
– MarioVW, Commented Feb 4, 2011 at 21:36
I suppose the idea is that instead of executing two statements per loop iteration (test, increment) you can combine the increment step into the test step and execute only one statement per iteration... — maerics
– maerics, Commented Feb 4, 2011 at 21:39
The 3rd arguments is absent so the usual operation time is skipped, and the 2nd argument is faster since there is not operator comparaison. I mean that usually if you do as a 2nd arg, i<length, then it compares i with length, then it compares the result with false to stop the for loop. In this case there is only one comparaison with false, if 'i' is 0 then it is automatically false. — developerGuile
– developerGuile, Commented Feb 4, 2011 at 21:46
As I mentioned in my answer, you may be more likely to get that optimization using predecrement instead of postdecrement. — Dan Breslau
– Dan Breslau, Commented Feb 4, 2011 at 21:55

Dan Breslau · Accepted Answer · 2011-02-05 18:40:35Z

4

(I'd write this as a comment, but it'd be too long.)

First: Worrying about the efficiency of a for loop is almost always a waste of (your own) time. What's inside of the loop usually has much more impact on performance than the details of how the loop is specified.

Second: What browser(s) did you test with? Different browsers will show different performance profiles; even different versions of the same browser will differ.

Third: It's not out of the question that the JavaScript engine optimized your loops out of the picture. A JavaScript compiler could simply look at the loop and decide to set testDummy to 50000000 and be done with it.

Fourth: If you really want to split hairs on performance, I'd try for(i=count; --i != 0;) as well as for(i=count;i--;). The former may save a machine instruction or two, because executing the subtraction (in the predecrement step) may automatically set a hardware flag indicating that the result was 0. That flag is potentially wasted when you're using the postdecrement operator, because it wouldn't be examined until the start of the next iteration. (The chances that you'd be able to notice the difference are slim to none.)

edited Feb 5, 2011 at 18:40

answered Feb 4, 2011 at 21:42

Dan Breslau

11.5k2 gold badges37 silver badges44 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

developerGuile Over a year ago

I test with Firefox, Chrome, Internet explorer 8 and safari. thanks for the comment/answer. I need to proces bunch of data and it as to be lightning fast. So performance gain is important event for 50ms-150ms.

Matt Ball Over a year ago

@developer: how much data are you really dealing with?

developerGuile Over a year ago

well, i'm manipulating around 200 to 500 string element (words) and make combinaison of words in many differents way, and then count frequency of each combinaison (you can compare it as a tagcloud). I build arrays for combinaison of multiples size (2-tuples, 3-tuples, etc.)

Dan Breslau Over a year ago

@developerGuile: Then instead of agonizing over the for loop, you'd be far better off thinking about how to do the string manipulation most efficiently. Any for loop optimization you employ is unlikely to be noticeable when compared to the time used in the string manipulations. (If you find otherwise, I'd really like to hear about it!)

BGerrissen · Accepted Answer · 2011-02-04 22:24:56Z

3

Well...

for( i=0 ; i < len ; i++ )

is practically the same as

for( i = len ; i-- ; )

Lets describe it:

case 1: let i be 0
boolean expression
let i be i + 1

case 2: let i be len
let i be i - 1
cast i to boolean (type coersion) and interpret it.

The difference should be minute and depends entirely on how efficient type coersion is compared to a normal boolean expression.

Incidentally, test this:]

var i = count;
while( i-- ) {}

answered Feb 4, 2011 at 22:24

BGerrissen

21.7k4 gold badges41 silver badges40 bronze badges

Comments

simshaun · Accepted Answer · 2011-02-04 21:43:03Z

0

There's nothing wrong with your tests.

The blocks that you are testing are very-near identical, meaning the difference in execution speeds are going to be trivial. In both examples, a variable (i) is set to a fixed value and looped until it reaches a fixed value (count). The only thing that differs is i++ and i--, which in terms of speed, I think are practically the same.

The thing you have to be careful of (not do) is calculate the "loop until" value inside the loop definition.

answered Feb 4, 2011 at 21:43

simshaun

21.5k1 gold badge59 silver badges73 bronze badges

1 Comment

developerGuile Over a year ago

I'm trying to gain speed by skipping the 3rd argument and taking out a comparison step in the 2nd argument. That is what Nicholas C. Zakas talks about in his youtube video but can't reproduce a test to see the gain by myself.

svlada · Accepted Answer · 2012-06-22 08:57:12Z

I have made some tests too, here are the results.

In many articles, books authors propose that "optimized" loops are faster.

It seems that modern browsers have some optimizations for "normal" loops.

Firefox 13.0.1

Normal Loop: 0.887
Opt1: 1.025
Opt2: 1.098
Opt3: 1.399

Chrome 19.0.1

Normal Loop: 3.349
Opt1: 3.12
Opt2: 3.109
Opt3: 3.095

IE8

Over 12sec...
Repeatedly crashed during tests.

<script type="text/javascript">

function p(p) { console.log(p); }
// function p(p) { document.write(p); }

var testFn = function(num, niz, fn) {
    var start = new Date().getTime();

    fn(num, niz);

    var result = (new Date().getTime() - start) / 1000;

    return result;
}

function normalLoop(num, niz) {
    for (var i = 0; i < niz.length; i++) {
        niz[i] = 'a' + i;
    }
}

function opt1(num, niz) {
    var len = niz.length;
    for (var i = 0; i < len; i++) {
        niz[i] = 'a' + i;
    }
}

function opt2(num, niz) {
    for (var i = niz.length; i--;) {
        niz[i] = 'a' + i;
    }
}

function opt3(num, niz) {
    while(i--) {
        niz[i] = 'a' + i;
    }
}

var niz = [];

var num = 10000000;

for (var i = 0; i < num; i++) { niz.push(i); };

p('Normal Loop: ' + testFn(num, niz, normalLoop));
p('Opt1: ' + testFn(num, niz, opt1));
p('Opt2: ' + testFn(num, niz, opt2));
p('Opt3: ' + testFn(num, niz, opt3));

</script>

Collectives™ on Stack Overflow

For loop speed performance in javascript

4 Answers 4

4 Comments

Comments

1 Comment

Comments

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

4 Comments

Comments

1 Comment

Comments

Related