Does “caching” the length of an array improve performance in a for loop?

If the body of the loop is not empty, then any possible performance increase will be measured at the most in units of percent.

And if in the body of the loop there is work with DOM, then you will not notice an increase in the length of the array.

Do not save on matches. Write a code that is easier to read.

Funny results

Test code:

 let arr = []; for (var e = 0; e < 1000; e++) arr.push(e); function run(count) { let f = [], s = []; count = parseInt(count) || 0; for (let e = 0; e < count; e++) { let start = Date.now(); for (let i = 0; i < arr.length; i++) null; f.push(Date.now() - start); } for (let e = 0; e < count; e++) { let start = Date.now(), length = arr.length; for (var i = 0; i < length; i++) null; s.push(Date.now() - start); } return { first: f, second: s }; } [1000, 10000, 100000, 1000000].forEach(item => { let tmp = run(item); console.info(`Размер массива: ${arr.length}\nКоличество итераций: ${item}\n\nБез кэширования (среднее): ${tmp.first.reduce((a, e) => a += e) / tmp.first.length}\nС кэшированием (среднее): ${tmp.second.reduce((a, e) => a += e) / tmp.second.length}`); }); 

Chrome 51 :

Firefox 47 :

As you can see, it is more profitable for the fox to cache only small structures, the larger the array, the less benefit (even in the minus, in terms of benefit, is gone).
Chrome is the same, but caches more efficiently, apparently.

Conclusion :
Optimization is a complicated thing. It seems to be there, and it seems like only a minus :)
First of all, write the human-readable code !
Saving man-hours on parsing - How does it work? - override -4 millisecond optimization.
Then you can already think how to optimize access to the length and so on.
Such big things need to be divided into small pieces.

There used to be a test where all iteration options were compared, it was called foreach-vs-loop , now it is not available but you can search by pictures, in short, there is no difference, with caching it even turned out a little slower, which surprisingly was the opposite for faster than direct ps : you can find a saved copy, for example on Yandex , this is certainly not a complete test, but there is a question that interests you. ps: the speed of assign in loop declaration is surprising, I would even say it doesn’t surprise to amaze.