Dnia 13 lipca 2009 16:35 Andrew Haley <aph@xxxxxxxxxx> napisał(a):
> ami_stuff wrote:
> >>> Ok, so it is a GCC's bug?
> >> It's perhaps an optimization that gcc doesn't do. But then
> >> there is an infinite number of optimizations that gcc doesn't do.
> >
> > How should I modify GCC's inline to repleace MULH function (to make it compatible)?
>
> What for? I posted sample code that should work.
Could you tell me what I do wrong?
#include <stdio.h>
#include <stdint.h>
#define MULH_ORG(a,b) (((int64_t)(a) * (int64_t)(b))>>32)
#define MULH_ASM(xh, xl, a, b)
__asm__ ("| Inlined umul_ppmm \n\t"
" move%.l %2, %/d0 \n\t"
" move%.l %3, %/d1 \n\t"
" move%.l %/d0, %/d2 \n\t"
" swap %/d0 \n\t"
" move%.l %/d1, %/d3 \n\t"
" swap %/d1 \n\t"
" move%.w %/d2, %/d4 \n\t"
" mulu %/d3, %/d4 \n\t"
" mulu %/d1, %/d2 \n\t"
" mulu %/d1, %/d2 \n\t"
" mulu %/d0, %/d1 \n\t"
" move%.l %/d4, %/d0 \n\t"
" eor%.w %/d0, %/d0 \n\t"
" swap %/d0 \n\t"
" add%.l %/d0, %/d2 \n\t"
" add%.l %/d3, %/d2 \n\t"
" jcc 1f \n\t"
" add%.l %#65536,%/d1 \n\t"
"1: swap %/d2 \n\t"
" moveq %#0, %/d0 \n\t"
" move%.w %/d2, %/d0 \n\t"
" move%.w %/d4, %/d2 \n\t"
" move%.l %/d2, %1 \n\t"
" add%.l %/d1, %/d0 \n\t"
" move%.l %/d0, %0 \n\t"
: "=g" ((USItype) (xh)),
"=g" ((USItype) (xl))
: "g" ((USItype) (a)),
"g" ((USItype) (b))
: "d0", "d1", "d2", "d3", "d4")
main ()
{
printf("result org: %d\n",m(5555555,5555555));
printf("result asm: %d\n",MULH_ASM(5555555,5555555));
}
Error:
mulhtest.c: In function `main':
mulhtest.c:64: warning: assignment makes pointer from integer without a cast
mulhtest.c:71: error: syntax error before "asm"
