RE: Inline functiion in GCC

[jeroen dobbelaere <jeroen dot dobbelaere at acunia dot com>]

Hi Venkatachala Upadhya,

Your inline problem would probably work better if you would add optimization when compiling.
(-O2 for example).

I've changed your example into the following :
The assembly function was changed in such way, that gcc can reorder
instructions at will.



#include <stdio.h>
#include <stdlib.h>

__inline__ static int add(int x, int y)
{
   __asm__
   (
        "qadd    %0, %1, %2 \n;"
        : "=r" (x)
        : "0" (x << 16), "r" (y << 16)

   ) ;
   return (x >> 16);
}

int main (void)
{
 int x=10, y=20, z=30;
 int sum1, sum2, sum3;

 sum1 = add (x, y);
 sum2 = add (y, z);
 sum3 = add (sum1, sum2);

 printf("sum1 = %d sum2 = %d sum3 = %d\n", sum1, sum2, sum3);
}



The resulting code with '-O2' is :
(my arm-linux-gcc compiler has an 'xscale' target (armv5) as default)

> arm-linux-gcc -O2 -finline-functions -Winline -I. test.c -c -o test.o
> arm-linux-objdump -d test.o

test.o:     file format elf32-littlearm

Disassembly of section .text:

00000000 <main>:
   0:   e3a00705        mov     r0, #1310720    ; 0x140000
   4:   e1a02000        mov     r2, r0
   8:   e3a0381e        mov     r3, #1966080    ; 0x1e0000
   c:   e3a0180a        mov     r1, #655360     ; 0xa0000
  10:   e1032052        qadd    r2, r2, r3
  14:   e1001051        qadd    r1, r1, r0
  18:   e1a01841        mov     r1, r1, asr #16
  1c:   e1a02842        mov     r2, r2, asr #16
  20:   e1a00802        mov     r0, r2, lsl #16
  24:   e1a03801        mov     r3, r1, lsl #16
  28:   e1003053        qadd    r3, r3, r0
  2c:   e59f0004        ldr     r0, [pc, #4]    ; 38 <main+0x38>
  30:   e1a03843        mov     r3, r3, asr #16
  34:   eafffffe        b       0 <main>
  38:   00000000        andeq   r0, r0, r0


Greetings,
--
Jeroen Dobbelaere
Embedded Software Engineer

Acunia International
http://www.acunia.com