Code comparison

Which one of the following two pieces of code should be 开发者_运维技巧preferred over other? and on what basis the decision should be taken generally?

MAX_LIMIT might vary from 1000 to 5000 in various calls of this function.

for (i=0;i<MAX_LIMIT;++i)
{
    for (j=0;j<MAX_LIMIT;++j)
    {
         anObj.setMatrix(i,j,0);
    }
}
for (i=0;i<MAX_LIMIT;++i)
{
    anObj.setMatrix(i,i,1);
}

vs

for (i=0;i<MAX_LIMIT;++i)
{
    for (j=0;j<MAX_LIMIT;++j)
    {
         if(i==j)
         {
             anObj.setMatrix(i,j,1);
         }
         else
         {
             anObj.setMatrix(i,j,0);
         }
    }
}

Thanks.

The perfomance of the two should be asymptoticaly equal, as both run in O(n^2). You should probably prefer the one with most readability.

The second, which does what it says. However, I'd strongly prefer

for (i=0;i<MAX_LIMIT;++i)
{
    for (j=0;j<MAX_LIMIT;++j)
    {
         anObj.setMatrix(i, j, i==j ? 1 : 0);
    }
}

In my opinion I would go with the second of the two. It is best to limit the number of times you have to loop through something. Since you're looking at at least O(n^2) runtime I would definately limity yourself to only 2 loops.

Better for what? I think the second version is easiest to understand at a glance, although other people may have other opinions.

If you're interested in performance, it should be easy to time the two methods. (It may even be relevant.)

Taking a look at this for performance, I'd suspect the second method would be faster, because it will work through the array one cache line at a time. The first method will work through the array one cache line at a time, and then go on to execute N cache faults. My guess is that this is more significant than putting an extra conditional in the loop.

So, if array initialization takes a significant amount of time compared to other operations, and this matters, time each one and see. If not, then go with the more readable version.