Re: [avr-gcc-list] inline assembler problem

[Theodore A. Roth]

On Tue, 20 Jan 2004, Ewout Boks wrote:

>
>
> That is the inline asm problem I asked Joerg about yesterday. It is an
> unecessary warning since your code looks perfectly allright.
>
> I believe that it has been solved in post 3.3.1 versions of avr-gcc, so
> update the compiler is your solution.

The warning are correct and the posted code will not work as he expects.
The reason it won't work is discussed in this FAQ entry:

  http://jubal.westnet.com/AVR/doc/avr-libc-user-manual/FAQ.html#faq_port_pass

> > Since some of you wanted to view the whole testprogram in order to help
> > me track the error I've struggeling with, I post it here. The compiler
> > I'm using is still the avr-gcc-3.3.1.
> > The functions in the code below I did implement in order to have some
> > simple functions to learn from.
> >
> >
> > #include <inttypes.h>
> >
> > #define PORT_A 0x1B
> > #define DDR_A 0x1A
> > #define PIN_A 0x19

Change these to:

#define PORT_A PORTA
#define DDR_A DDRA
#define PIN_A PINA

> >
> > uint8_t read(const uint8_t adr) {

In order to pass the io reg, it must be done via a pointer, change this
to:

uint8_t read (volatile uint8_t *adr) {

> >   uint8_t p;
> >   asm volatile ("in %0, %1" : "=r" (p) : "I" (adr));
> >   return p;
> > }

This is impossible. Read the FAQ entry linked above.

> >
> > uint8_t fastSBI(const uint8_t port, const uint8_t bit) {
> >   asm volatile("sbi %0, %1" : : "I" (port), "I" (bit));
> > }
> >
> > int main(void) {
> >   uint8_t port_a = PORT_A;
> >   uint8_t ddr_a = DDR_A;
> >   uint8_t pin = 3;
> >   uint8_t data;
> >
> >   data = read(port_a);
> >   fastSBI(ddr_a, pin);
> >
> >   return 0;
> > }
> >
> > When trying to compile this I get the following errors:
> > inlasm.c: In function `read':
> > inlasm.c:11: warning: asm operand 1 probably doesn't match constraints
> > inlasm.c:11: error: impossible constraint in `asm'
> > inlasm.c: In function `fastSBI':
> > inlasm.c:17: warning: asm operand 0 probably doesn't match constraints
> > inlasm.c:17: warning: asm operand 1 probably doesn't match constraints

I don't think that gcc is lying to you about the "impossible
constraint".

Ted Roth