Thread: Pointers: why not?

I've been thinking about this for some time, and am very curious:

Other than to prevent the function being called from being able to overwrite the variable in question, is there any reason why you would specifically want to pass a variable to a function, instead of a pointer to that variable? It would seem that using pointers as much as possible would be good for performance/footprint, but I don't want to overdo it if there are other considerations.

I hope this isn't a terribly stupid question; I've yet to read up on some of the more advanced aspects of C, and it's possible my question is answered there...

I'm not trying to fix something, I'm trying to determine when pointers are/are not a good idea when trying to make data available to functions.

And yeah, I know that passing, like, pointer to a char might not be the most wildly intelligent of ideas, unless you REALLY have a good reason... but say with floats or the like, I thought I'd like to know if there are any reason I'd particularly want to pass the actual data instead of a pointer to it.

A common programming error occurs when a pointer is passed and the function attempts arithmetic operations with the value. Which doesn't operate properly because it's not using a variable's value, rather it's modifying an address. Value++ produces a completely different result than pointertoValue++.

That's just one reason off the top of my head. However really the predominate reason is just to ensure and enforce the priniciples of least priveledge.

>I thought having a pointer would let you manipulate the original variable in the calling function, thus preventing one from having to create a second variable.

For a 32-bit address space, if a value is passed it is added to the stack (whatever the size), if a pointer is passed, the 32-bit address is added to the stack, so, in the case that you are deciding to pass either an int, or a pointer to an int, they would incur the same overhead it would seem. However, you'd still have to dereference within the function, making it an extra instruction.

Also, as mentioned, access rights help determine it. In general, pointers are not passed if unnecessary, and I would think that the reason I mentioned above (if anything else) would suffice.

Everyone's talking about performance but I haven't seen anyone mention the usage of the pointer inside the function. If you had a function like:

Code:

int sum(int *a, int *b)
{
  return *a + *b;
}

The CPU has to fetch the value inside each pointer variable, and then it has to fetch the value at that memory address. a is one fetch, *a is two fetches. For more complex functions you can see how it would be much more efficient to just have the value instead of a pointer to the value.

The address is a memory address somewhere in your computer's memory. If the function has few enough variables and the CPU is able to hold that kind of variable in a register and the compiler is smart enough to keep it in the register, then yeah...it doesn't require a fetch.

But *a would still need to look at 2 registers instead of 1, which admittedly is much faster than 2 memory fetches, but you better have a really simple function with the right variable types in order to count on it happening.

Originally Posted by Aerie

I'm not trying to fix something, I'm trying to determine when pointers are/are not a good idea when trying to make data available to functions.

And yeah, I know that passing, like, pointer to a char might not be the most wildly intelligent of ideas, unless you REALLY have a good reason... but say with floats or the like, I thought I'd like to know if there are any reason I'd particularly want to pass the actual data instead of a pointer to it.

Pointers are good when:
- There is no alternative (arrays)
- The function have to modify the original value of a variable (write access)
- The data are bigger that an int (typically, structures with at last 2 members)