Learnin' C

fib.c

#include <stdlib.h>

int main(void) {
	int edi = 5; // nth fib num
	int ebx = 0;
	int eax = 1;
	int edx; // tmp

	start_loop:
	--edi;
	
	if (edi==0) { goto loop_exit; }

	edx = eax;
	eax += ebx;
	ebx = edx;

	goto start_loop;

	loop_exit:

	return ebx;
}

fib_prototype.s

.section .data
.section .text
.globl _start
_start:
movl $10, %edi # number of fibnum we hunt
movl $0, %ebx # 1st fibnum
movl $1, %eax # 2nd fibnum

arb_label:
## Check if %edi is 0
## (thus meaning %ebx is nth fibnum)
## and if so, exit loop
decl %edi
cmpl $0, %edi
je loop_exit

## core Fibonnaci algorithm
## (hold %eax in tmp register %edx,
## eax = eax + ebx,
## ebx = edx)
movl %eax, %edx
addl %ebx, %eax
movl %edx, %ebx

jmp arb_label

## Once loop completes...
loop_exit:
movl $1, %eax
int $0x80

@informatimago

My handwritten assembly code written without optimization in mind outperforms GCC's assembly compilation of your code in terms of speed.I am using the Linux time command. How strange. I don't know why.

edit: Well, I have an idea of why (your program makes use of a pricey function call that gives my function-less version a slight edge). But what I mean to say is, I don't know why GCC at -O3 can't optimize away the function call entirely.