Re: [avr-gcc-list] Can this be optimised?

[Theodore A. Roth]

On Wed, 28 Jul 2004, Royce & Sharal Pereira wrote:

> Hi, all,
>
> Can this be optimised somehow?
> The following code:-
> //------------------------------
> unsigned int tmp;
> //......etc
>  tmp= ADCL |(ADCH<<8);
> //------------------------------
> gets compiled(correctly) as:
>
>  1376 04d0 84B1        in r24,36-0x20
>  1377 04d2 282F        mov r18,r24
>  1378 04d4 3327        clr r19
>  1379 04d6 85B1        in r24,37-0x20
>  1380 04d8 9927        clr r25
>  1381 04da 982F        mov r25,r24
>  1382 04dc 8827        clr r24
>  1383 04de 282B        or r18,r24
>  1384 04e0 392B        or r19,r25
>  1385 04e2 3093 0000   sts (tmp)+1,r19
>  1386 04e6 2093 0000   sts tmp,r18
>
> Is there a way (without using assembly) to shorten this to:
>
>               in     rxx,     36-0x20
>               sts   tmp,    rxx
>               in     ryy,     37-0x20
>               sts   tmp+1, ryy
>
> Or is in-line assembly the best way?

Sure. Do it like this:

  int main (void)
  {
     volatile uint16_t foo;

     foo = ADC;

     return (int)foo;
  }

Compiling with -Os yields this asm code:

int main (void)
{
  da:   cd ef           ldi     r28, 0xFD       ; 253
  dc:   d0 e1           ldi     r29, 0x10       ; 16
  de:   de bf           out     0x3e, r29       ; 62
  e0:   cd bf           out     0x3d, r28       ; 61
    volatile uint16_t foo;

    foo = ADC;
  e2:   84 b1           in      r24, 0x04       ; 4
  e4:   95 b1           in      r25, 0x05       ; 5
  e6:   89 83           std     Y+1, r24        ; 0x01
  e8:   9a 83           std     Y+2, r25        ; 0x02

    return (int)foo;
  ea:   89 81           ldd     r24, Y+1        ; 0x01
  ec:   9a 81           ldd     r25, Y+2        ; 0x02
  ee:   0c 94 79 00     jmp     0xf2



---
Ted Roth
PGP Key ID: 0x18F846E9
Jabber ID: address@hidden