Re: [avr-libc-dev] RFC: avr/bits.h

[Bob Paddock]

On Tue, 01 Mar 2005 10:00:13 -0700, E. Weddington <address@hidden> wrote:

> Bob Paddock wrote:
>
> > I wanted to do a test case

Doing the test case right in the first place always help.
You can not us a BIT() inside of a bit_set(). :-(

Lets try a correct test case with your foo() added:

#include <inttypes.h>
#include <avr/io.h>

#include "bits.h"

uint16_t shorttest_u16 = (uint16_t) 0;
uint32_t longtest_u32  = (uint32_t) 0;

void foo(void);
void foo(void)
{
   bit_set(PORTD, 3);
}

int main( void )
{
    bit_set( shorttest_u16, 0 );
    bit_set( shorttest_u16, 1 );
    bit_set( shorttest_u16, 2 );
    bit_set( shorttest_u16, 3 );
    bit_set( shorttest_u16, 4 );

    bit_set( shorttest_u16, 5 );
    bit_set( shorttest_u16, 6 );
    bit_set( shorttest_u16, 7 );
    bit_set( shorttest_u16, 8 );
    bit_set( shorttest_u16, 9 );
    bit_set( shorttest_u16, 10 );
    bit_set( shorttest_u16, 11 );
    bit_set( shorttest_u16, 12 );
    bit_set( shorttest_u16, 13 );
    bit_set( shorttest_u16, 14 );
    bit_set( shorttest_u16, 15 );

    bit_set( longtest_u32, 0 );

    bit_set( longtest_u32, 7 );
    bit_set( longtest_u32, 8 );

    bit_set( longtest_u32, 15 );
    bit_set( longtest_u32, 16 );

    bit_set( longtest_u32, 31 );

    foo();
}


> #define bit_32(x)  ((uint32_t)1 << (x))

I used these:

#define BIT(bitpos) ((uint32_t)1<<(bitpos))
#define bit_set(var, bit) ((var) |= BIT(bit))
#define bit_clear(var, bit) ((var) &= ~BIT(bit))
#define bit_toggle(var, bit) ((var) ^= BIT(bit))
#define bit_read(var, bit) ((var) & BIT(bit))

All of the code lookes reasonable now that I don't
have recursive macros, even got the proper sbi for PORTD.


  34                    foo:
  35                    .LFB2:
  36                    .LM1:
  37                    /* prologue: frame size=0 */
  38                    /* prologue end (size=0) */
  39                    .LM2:
  40 0000 939A                  sbi 50-0x20,3
  41                    /* epilogue: frame size=0 */

Is there some case where the construct ((uint32_t)1<<(bitpos))
is really going to promote to 32 bit code?