gcc generates invalid i386 fmulp

[Robert Lipe]

> I fiddled around with the glibc source that generated the invalid
> `fmulp %st(1),%st', and came up with the following test case.  It's
> not all that small, but reducing the complexity seems to make the
> problem go away.

Would anyone like to commit this test case to the testsuite?

It isn't just the Unixware assembler that hurls on the snapshot output
of this.  The OpenServer assemblers (the ELF one is Unixware-derived
since it was licensed from USL, the owner of Unixware at the time, the
COFF one was, I think, developed internally at SCO in the late 80's.)
choke in both COFF and ELF modes.

for m in "-melf " "-mcoff" 
do 
  for o in + - '*' / 
  do 
      ./xgcc $m -B./ -c -O2 -DOP=$o /tmp/am.c 
  done 
done 


/usr/tmp/cca000_7.s:90:invalid operand combination: faddp
/usr/tmp/cca000_B.s:90:invalid operand combination: fsubrp
/usr/tmp/cca000_F.s:90:invalid operand combination: fmulp
/usr/tmp/cca000_J.s:90:invalid operand combination: fdivrp
Assembler: am.c
        aline 89        : Syntax error
Assembler: am.c
        aline 89        : Syntax error
Assembler: am.c
        aline 89        : Syntax error
Assembler: am.c
        aline 89        : Syntax error



> ------------------ fop_pop_bug.c ------------------
> /*
>  * inspired by glibc-2.0.6/sysdeps/libm-ieee754/s_nextafterf.c
>  *
>  * gcc -O2 -S -DOP=+ gives faddp %st(1),%st
>  * gcc -O2 -S -DOP=* gives fmulp %st(1),%st
>  * gcc -O2 -S -DOP=- gives fsubrp %st(1),%st
>  * gcc -O2 -S -DOP=/ gives fdivrp %st(1),%st
>  */
> 
> #ifndef OP
> #define OP *
> #endif
> 
> typedef int int32_t __attribute__ ((__mode__ (  __SI__ ))) ;
> typedef unsigned int u_int32_t __attribute__ ((__mode__ (  __SI__ ))) ;
> 
> typedef union
> {
>   float value;
>   u_int32_t word;
> } ieee_float_shape_type;
> 
> float __nextafterf(float x, float y)
> {
>  int32_t hx,hy,ix,iy;
> 
>  {
>   ieee_float_shape_type gf_u;
>   gf_u.value = x;
>   hx = gf_u.word;
>  }
>  {
>   ieee_float_shape_type gf_u;
>   gf_u.value = y;
>   hy = gf_u.word;
>  }
>  ix = hx&0x7fffffff;
>  iy = hy&0x7fffffff;
> 
>  if ( ix > 0x7f800000 || iy > 0x7f800000 )
>     return x+y;
>  if (x == y) return x;
>  if (ix == 0)
>    {
>     {
>      ieee_float_shape_type sf_u;
>      sf_u.word = (hy&0x80000000) | 1;
>      x = sf_u.value;
>     }
>     y = x*x;
>     if (y == x) return y; else return x;
>    }
>  if (hx >= 0)
>    {
>     if (hx > hy)
>        hx -= 1;
>     else
>        hx += 1;
>    }
>  else
>    {
>     if (hy >= 0 || hx > hy)
>        hx -= 1;
>     else
>        hx += 1;
>    }
>  hy = hx & 0x7f800000;
>  if (hy >= 0x7f800000)
>     return x+x;
>  if (hy < 0x00800000)
>    {
>     y = x OP x;
>     if (y != x)
>       {
>        ieee_float_shape_type sf_u;
>        sf_u.word = hx;
>        y = sf_u.value;
>        return y;
>       }
>    }
>  {
>   ieee_float_shape_type sf_u;
>   sf_u.word = hx;
>   x = sf_u.value;
>  }
>  return x;
> }
> 
> 

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com
              (WEB ADDRESS MAY BE TEMPORARILY UNAVAILABLE)