Andrew Haley writes:
> Richard Bonomo writes:
> >
> > I am hoping someone can give me a little
> > advice in how to proceed. I have
> > some C-code which I compile under
> > gcc-3.4.5 which employs a definition
> > such as this:
> >
> > #define disable() \
> > ({ register int __Status __asm__ ("d"); \
> > __asm__ ("clra ; save CPU status"); \
> > __asm__ volatile ("tfr cc,b" : "=d" (__Status) : : "d", "b"); \
> > __asm__ ("orcc #0x50 ; disable interrupts"); \
> > __Status; })
> >
> > When an attempt is made to compile, upon hitting the
> > spot in the program where this code winds up being
> > inserted, the compiler produces this error message
> >
> > error: asm-specifier for variable `__Status' conflicts with asm clobber list
> >
> > and quits.
> >
> > After doing quite a bit if digging, I do know that for some
> > reason the compiler is probably complaining about the
> > assignment of register "d" (which includes registers
> > "a" and "b" in the Motorola 6809 as its high and
> > low components).
> >
> > However, I am still at quite a loss as to how
> > I am to deal with this: how do I get around this,
> > short of rewriting this and other functions to
> > use ordinary variables as much as possible?
> >
> > Can you shed any light on what, precisely is being
> > objected to, and how to meet that objection?
>
> Sure. You declare a register variable "__Status" that lives in
> register d. You then use that same register as an asm output. gcc is
> complaining about that conflict.
>
> It's easy to fix. Something like this:
>
> #define disable() \
> ({ int __Status; \
> __asm__ volatile ("clra ; save CPU status"); \
> __asm__ volatile ("tfr cc,b" : "=d" (__Status)); \
> __asm__ volatile ("orcc #0x50 ; disable interrupts"); \
> __Status; })
MM, wrong. gcc might put something between the insns. Ok, try
#define disable() \
({ int __Status; \
__asm__ volatile ("clra ; save CPU status\n" \
"tfr cc,b\n" \
"orcc #0x50 ; disable interrupts" \
: "=d" (__Status)); \
__Status; })
