Re: 68040 inline asm code problems

• Date: Thu, 16 Jan 97 15:02:33 PST
• From:
  art@acc.com (Art Berggreen)
• Reply-To:
  art@acc.com (Art Berggreen)

>
>
>Hello,
>
>I'm having a problem either in compiling or undestanding the 
>gcc compiler. I'm running a 68k-rtems cross for my RTEMS project
>which works well. However, in adding some bus error handlers that
>I got from the HIDEOS MVME162 project, I ran into problems getting 
>them to compile. Here is an example:
>
>
>IRQHANDLER* SysGetVbr()
>{
>   unsigned long x;
>   asm  volatile ("movec vbr,%0" : "=g" (x)  );  
>   return (IRQHANDLER*)x;
>}
>
>/**************************** access exception ************************/
>	asm (" .text");
>	asm ("_irq_access:");
> 
>	/* Save context */
>	asm ("moveml #0xfffe,sp@-");     /* Save regs */
> 	asm ("movel sp@(62),a2 ");   /* this is pc */
>	asm ("addql #2,sp@(62)");    /* pc+x to skip illegal instruction */
>	asm ("movl #1,_int_occurred");
>	asm ("moveml sp@+,#0x7fff");
>	asm ("rte");
>
>
>compiled with "m68k-rtems-gcc -c"
>
>produces this:
>/tmp/cca14729.s: Assembler messages:
>/tmp/cca14729.s:10: Error: operands mismatch -- statement `movec vbr,-4(%a6)' ignored
>/tmp/cca14729.s:20: Error: parse error -- statement `moveml #0xfffe,sp@-' ignored
>/tmp/cca14729.s:21: Error: parse error -- statement `movel sp@(62),a2' ignored
>/tmp/cca14729.s:22: Error: parse error -- statement `addql #2,sp@(62)' ignored
>/tmp/cca14729.s:24: Error: parse error -- statement `moveml sp@+,#0x7fff' ignored
>
>
>Am I missing something in the command line, or in the file itself?

I think that for the first error, your output constraint "=g" is too general.
The movec instruction requires a register, but the C compiler was trying to
use a stack location (-4(%a6)).  Try using "=r".

For the others, try using %sp instead of sp to reference the stack pointer.

Art

• Re: 68040 inline asm code problems
• From: dean@abraham.nellis.af.mil

• Prev: Re: 68040 inline asm code problems
• Next: Re: 68040 inline asm code problems
• Index: Cross gcc Mailing list
• Thread: Cross gcc Mailing list