Is prefix faster than postfix?

**dwks** · 07-27-2005

Is

Code:

++x;

faster than

Code:

x++;

? I thought it might be, since ++x can do the increment right then and there, while x++ has to do it later.

**itsme86** · 07-27-2005

gcc seems to use the same number of instructions for both. I tested with the following 2 programs:

Code:

itsme@itsme:~/C$ cat prefix.c
int main(void) { int i = 0; int j; j = ++i; }
itsme@itsme:~/C$ cat postfix.c
int main(void) { int i = 0; int j; j = i++; }

Then I compiled each with: gcc -S <filename>

I won't spam the post by actually pasting the entire .s files, but only a few lines changed, and I'll post those:

Code:

// prefix.s
        leal    -4(%ebp), %eax
        incl    (%eax)
        movl    -4(%ebp), %eax

Code:

// postfix.s
        movl    -4(%ebp), %edx
        leal    -4(%ebp), %eax
        incl    (%eax)

Those lines are in the same place as they are in the other file, so there's the same number of instuctions. The rest of the instructions are exactly the same.

**dwks** · 07-27-2005

Hmm, okay, thanks for going to so much trouble.

**Rashakil Fol** · 07-27-2005

You'll find that prefix is faster than postfix for some non-primitive datatypes in C++. (Yes I know this is the C forum.)

**kermit** · 07-27-2005

Originally Posted by dwks

Hmm, okay, thanks for going to so much trouble.

Code:

gcc -S some_prog.c

**dwks** · 07-28-2005

prefix is faster than postfix for some non-primitive datatypes in C++

I know, that's why I was wondering if prefix was faster for ints too.

**grumpy** · 07-28-2005

The only real guarantee is that prefix will be no slower than postfix.

In principle, x++ (for x an int) must be expanded to something like;

Code:

   int temp = x;
   x = x+1;
   /* allow value of temp to be accessible */

while ++x requires no temporary storage of the original value of x. A literal interpretation of that requirement would make postfix slower than prefix operations.

However, compilers (or, more accurately, compiler writers) are smarter than that. Most mainstream compilers are able to recognise cases where the original value of x need not be accessible, and therefore avoid storing it's value temporarily. Which is the reason that the example given by itsme86 will usually exhibit no measurable difference in terms of number of instructions or performance.

Thread: Is prefix faster than postfix?

Thread Tools

Search Thread

Display

Is prefix faster than postfix?

Similar Threads

Little question about postfix and prefix

problem with infix 2 postfix converter & postfix calculator

postfix and prefix

postfix and prefix...???

Data Structures / prefix and postfix