switch versus if [C++]

[arras]

Are there any difference between switch() and if() in terms of effectivity?

For example:

Code: Select all

if(a == 0) ...do something;
else
if(a == 1) ...do something;
else ...do something;

Code: Select all

switch(a)
{
   case 0: ...do something; break;
   case 1: ...do something; break;
   default: ...do something;
}

[hybrid]

Yes, the if cascade will do the test in each cycle, the switch will only test once and jump directly into the proper case statement.

[arras]

OK, thanks hybrid.

[Kriolyth]

"switch" under most compilers will generate a jump-table, so after evaluation it will take only one CPU instruction to get to the needed branch.
"if" is hardly optimized at all, so there is a huge chance that every if-condition will be tested.
(actually repeating hybrid)

[sp00n]

As i remember a standard asm code by M.Abrash when he was discovering "if vs switch" there was a result: use "if ... else" statement only if you have less than 4 conditions (max example: if ... else if ...), in other cases use a switch ('cause 3 jumps better than jump-table for 3 cases, but j-table for 4 cases better than 4 if-jumps )
But in modern CPU's it itsn't critical problem, i think)

[arras]

Thanks mates...

[MasterGod]

If we're already talking about switches, it's the same syntax for Java right?

[Halifax]

Also, it is best to eliminate as many branching conditions as you can in your code. Even with branch hinting, if the code is in a heavily used loop, it may create for a slowdown or bottleneck. (If too many are present.)

To make what I'm saying a little bit easier, say we have a conditional 'jpz' (jump zero) instruction. If the test case is true it executes in 7 cycles, but if the test case is not true it takes 12 cycles to execute.

[Swarmer]

Switches are also much cleaner and easier to read in my opinion, which is just as important.

[arras]

What about this code?:

Code: Select all

void somefunction(int a, int b, int c)
{
   if(a == 10)
   {
      a = 20;
      return;
   }
   
   if(b == 4)
   {
      b = 0;
      return;
   }
   
   if(c < 1)
   {
      c = 1;
      return;
   }
}

compared to:

Code: Select all

void someotherfunction(int a, int b, int c)
{
   if(a == NULL)
   {
      a = 20;
   }
   else
   if(b == 4)
   {
      b = 0;
   }
   else
   if(c < 1)
   {
      c = 1;
   }
}

[vitek]

That function has no external side effects and no return, so it will be the same. Both implementations will end up being an empty function.

If you are actually curious about this kind of thing, find out what compile flag you need to use to generate assembly, then use it. You really have to test for yourself because the answers will depend on the code, the optimization level, and the compiler.

Travis

[wyrmmage]

Switches are also much cleaner and easier to read in my opinion, which is just as important.

yep; the other big advantage to them is that if you change the name of the variable (or you put it inside of an object), you only have to change one line of code versus quite a few. I find myself doing this quite often when I'm first setting up a game

-wyrmmage

[d3jake]

Would there be a way to adapt the switch statement to your movement code in your free flight demo?
(This code if nobody has any clue what I'm talking about)

Code: Select all

    // direction control
        if(keys[irr::KEY_LEFT])
        {
            turn(shuttle, 0.25);
        }
        if(keys[irr::KEY_RIGHT])
        {
            turn(shuttle, -0.25);
        }
        if(keys[irr::KEY_UP])
        {
            pitch(shuttle, -0.25);
        }
        if(keys[irr::KEY_DOWN])
        {
            pitch(shuttle, 0.25);
        }
        if(keys[irr::KEY_COMMA])
        {
            roll(shuttle, 0.45);
        }
       // And so on...

[Halifax]

No there is no particular, easy, way to do that, and you wouldn't want to do that. That code is engineered that way so as to allow multiple key presses from a boolean array. Thus the user can press up and left, and it will result in an up and left movement.

[arras]

Thanks once again for answers.

That code is engineered that way so as to allow multiple key presses from a boolean array. Thus the user can press up and left, and it will result in an up and left movement.

Exactly. Also user input is not going to happen much too often (maximum once per loop). You do not have to struggle for such subtle optimization. You need that in parts of code which are going to be called say few dozens or hundred times per loop.