------- Comment #3 from jakub at gcc dot gnu dot org 2010-08-24 19:57 -------
Reduced testcase:
typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__));
typedef int __v2si __attribute__ ((__vector_size__ (8)));
typedef float __v4sf __attribute__ ((__vector_size__ (16)));
static __inline__ unsigned short bar (unsigned short x)
{
return ((x << 8) | (x >> 8));
}
unsigned int
foo (float *x, short *y)
{
__v4sf b = (__v4sf){ 32767.5f, 32767.5f, 32767.5f, 32767.5f };
__v4sf c = __builtin_ia32_mulps (*(__v4sf *) x, b);
__v4sf d = __builtin_ia32_movhlps (c, c);
__v2si e = __builtin_ia32_cvtps2pi (c);
__v2si f = __builtin_ia32_cvtps2pi (d);
__m64 g = (__m64) __builtin_ia32_packssdw (e, f);
__builtin_memcpy (y, &g, sizeof (short) * 4);
y[0] = bar (y[0]);
}
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45400
