Question

In C:

Lets say function “Myfuny()” has 50 line of codes in which other smaller functions also get called. Which one of the following code would be more efficient?

void myfunction(long *a, long *b);
int i;
for(i=0;i<8;i++)
   myfunction(&a, &b);

or

myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);
myfunction(&a, &b);  

any help would be appreciated.

Answer 1

That’s premature optimization, you just shouldn’t care…

Now, from a code maintenance point of view the first form (with the loop) is definitely better.

From a run-time point of view and if the function is inline and defined in the same compilation unit, and with a compiler that does not unroll the loop itself, and if code is already in instruction cache (I don’t know for moon phases, I still believe it shouldn’t have any noticable effect) the second one may be marginally fastest.

As you can see, there is many conditions for it to be fastest, so you shouldn’t do that. There is probably many other parameters to optimize in your program that would have a much greater effect for code speed than this one. Any change that would affect algorithmic complexity of the program will have a much greater effect. More generally speaking any code change that does not affect algorithmic complexity is probably premature optimization.

If you really want to be sure, measure. On x86 you can use the kind of trick I used in this question to get a fairly accurate measure. The trick is to read a processor register that count the number of cycles spent. The question also illustrate how code optimization questions can become tricky, even for very simple problems.