Is one big for loop faster than than two small ones?

Question

In the code, I have made two for loops:

for(int y = 0;i < HEIGHT; y++){
for(int y = 0;i < WIDTH; y++){
//lots of switch cases
}
}

and I wonder, is it faster to use one for loop like this? :

for(int i = 0;i < HEIGHT*WIDTH; i++){
//lots of switch cases
}

You won't be able to measure the difference unless the HEIGHT and WIDTH are really huge and the amount of work done in the switch is really small. Otherwise the time spent "executing the loop statement" is an insignificant fraction. You should write code that is logically clear UNTIL you have actual measurements that support doing otherwise. — user14644949, Nov 24 '20 at 13:23
And of course, your two-loops version won't work. You mixed up x, y, and i. — Ralf Kleberhoff, Nov 24 '20 at 13:26

score 2 · Accepted Answer · answered Nov 24 '20 at 14:24

The usual newbie misconception: java performance doesn't come out of "clever java source code".

The key part is: the optimisations that the just-in-time compiler does at runtime. And that starts with: the JIT deciding whether code is invoked often enough to be worth optimizing.

Thus, the real answer to (most) java performance questions is:

write simple code that is correct and readable
as that enables the JIT to optimize things when necessary AND
it also enables you to improve your code later on, in case you really wrote code that is for this or that reason leading to performance issues

(and that 3rd point, you figure that because your users/tests unveil "bad performance", and then you did a lot of measuring to identify the true root cause of that problem)

Thus, as said: focus on writing correct code that is easy to read and understand, and do not worry about performance until much later.

score 1 · Answer 2 · answered Nov 24 '20 at 14:33

No.

As mentioned, you should determine your end condition before you use in it a loop, meaning that you should have

int total = height*width;
for( int i=0; i < total; ++i ) {
    // ...
}

so that total isn't constantly computed during the loop. That said, everyone does it and I would expect the compiler to optimize your version into this version anyway.

The real issue, as I see it, is that in the loop you may have to reverse the computation from the total, meaning that you might have to have

int total = height*width;
for( int i=0; i < total; ++i ) {
    int x = total / width;
    int y = total % height;
    call(x,y);
}

This would be slower than

for( int i=0; i < width; ++i ) {
    for( int j=0; i < height; ++i ) {
        call(i,j);
    }
}

which should be faster. This does not take into consideration "branching operations" in CPU's. See Branchless Programming: Why "If" is Sloowww... and what we can do about it!.

EDIT: And yes, the most salient comment in this thread is that you should worry about readability much more performance.

Is one big for loop faster than than two small ones?

2 Answers2