local variables performance

I have the following code:

Code:

---

for (i = 0; i < 1000000; i++) {
  int Side1 = Result[i].X1 - 1;
  int Side2 = Result[i].X2 - 1;

  double Dev1 = Func(Idx, Side1);
  double Dev2 = Func(Idx + 1, Side2);

  .....
}

---

Will it perform better if I take the declarations of the doubles and ints out of the loop? In the general sense.

Probably not. Compilers are very adept at making these kinds of optimizations.

Quote:

---

Originally Posted by KIBO

Will it perform better if I take the declarations of the doubles and ints out of the loop? In the general sense.

---

In a general sense, it would be more efficient to declare the variables outside the loop -- if, instead of being basic data types, they were class objects, then the class constructor/destructor would be called for each pass through the loop.

In this specific case, it doesn't make any difference.

You could try it both ways and see for yourself.

Quote:

---

Originally Posted by cpjust

You could try it both ways and see for yourself.

---

I could but then I would know for one compiler, 1 setting!

Quote:

---

Originally Posted by KIBO

I could but then I would know for one compiler, 1 setting!

---

Yep .... this sort of thing is dependent upon compiler, compiler setting/optimisation, operating system, hardware ...... and the interaction between them.

If the answer really matters in your application (which is unlikely, BTW: it's more likely you're indulging in premature optimisation) then you need to test for every relevant use case.

Quote:

---

Originally Posted by grumpy

If the answer really matters in your application (which is unlikely, BTW: it's more likely you're indulging in premature optimisation) then you need to test for every relevant use case.

---

Or at least for a couple of the most important cases. If your code is very portable, it can be hard to find the right solution for LOTS of platforms without resorting to LOTS of #ifdef.

--
Mats

I don't see how in this case it could ever be a performance hit to lift the declaration out of the loop.

Quote:

---

Originally Posted by KIBO

I could but then I would know for one compiler, 1 setting!

---

That's a given for any non-algorithmic performance optimization.

Quote:

---

Originally Posted by brewbuck

I don't see how in this case it could ever be a performance hit to lift the declaration out of the loop.

---

I'd agree with that. But there is no end to what (stupidity) some compilers will perform.

Quote:

---

Originally Posted by CornedBee

That's a given for any non-algorithmic performance optimization.

---

And that is also something I'd agree with.

--
Mats

Quote:

---

Originally Posted by brewbuck

I don't see how in this case it could ever be a performance hit to lift the declaration out of the loop.

---

I don't recall anyone saying there was. However, the original question was the reverse: whether doing so would give a performance gain - for which there is no guarantee, particularly with optimising compilers.