Re: [avr-gcc-list] generic queue library for AVR GCC?

[E. Weddington]

Bruce D. Lightner wrote:

> E. Weddington wrote:
>
> > I agree with Bruce's method; it's better in that it preserves the 
> > original state of the interrupt flag.
>
>
> I believe that the "push/pop" method did as well...

You're right. I misspoke. (and mentioned this in another response).

> > The push/pop is not required, and probably wasteful.
> >
> > Though you don't have to do it in inline assembly. It can be done all 
> > in C:
> >
> > #include <avr/io.h>
> > #include <avr/interrupt.h>
> > #include <inttypes.h>
> > ....
> > {
> >    uint8_t sreg = SREG;
> >    cli();
> >    sm_qin = sm_qout = 0;  // reset queue pointers.
> >    SREG = sreg;  }
>
>
> Yeah.  I like yours better.  The resultant AVR code is identical.
>
> I've got to "get onboard" with the "new" gcc 3.x AVR I/O syntax.
> I still have a number of active mixed AVR assembly/avr-gcc
> projects, which only work with the gcc 2.95 I/O definitions, so
> I guess I gravitate to inline assembly more often than I probably
> should. :-)
Oh, well that makes sense....

> How about this...
>
>     #include <avr/io.h>
>     #include <avr/interrupt.h>
>     #include <inttypes.h>
>     #define begin_critical_section()  SREG; cli()
>     #define end_critical_section(val) SREG = val
>     ....
>     {
>         unsigned char sreg = begin_critical_section();
>         sm_qin = sm_qout = 0;   // reset queue pointers
>         end_critical_section(sreg);
>     }
>
> This gets you the same, optimally efficient AVR code.
Absolutely. :-)

Eric