Computer Graphics, Virtual Reality & Marketing
struct vector
{
float x,y,z;
};
vector add(const vector &a, const vector &b)
{
vector c;
c.x = a.x + b.x;
c.y = a.y + b.y;
c.z = a.z + b.z;
return c;
}
# include <xmmintrin.h>
struct vector
{
__m128 m;
// Pour acceder aux composantes x, y, z
// vous pouvez simplement faire un cast ((float*)&m)[ 0 1 2 ou 3]
};
vector add(const vector &a, const vector &b)
{
vector c;
c.m = _mm_add_ps(a.m, b.m);
return c;
}
__m128 fastrsqrt( const __m128 v )
{
static const __m128 _half4 = _mm_set_ps1( 0.5f );
static const __m128 _three = _mm_set_ps1( 3.0f );
const __m128 nr = _mm_rsqrt_ps( v ); // Newton-Raphson iteration
const __m128 muls = _mm_mul_ps( _mm_mul_ps( v, nr ), nr );
return _mm_mul_ps( _mm_mul_ps( _half4, nr ), _mm_sub_ps( _three, muls ) );
}
void Normalize4(vector p[4])
{
// Store x, y, and z separately in 3 __m128
__m128 x4 = _mm_set_ps(p[3][0], p[2][0], p[1][0], p[0][0]);
__m128 y4 = _mm_set_ps(p[3][1], p[2][1], p[1][1], p[0][1]);
__m128 z4 = _mm_set_ps(p[3][2], p[2][2], p[1][2], p[0][2]);
x4 = _mm_mul_ps(x4, x4); // x * x
y4 = _mm_mul_ps(y4, y4); // y * y
z4 = _mm_mul_ps(z4, z4); // z * z
const vector l = fastrsqrt(_mm_add_ps(z4, _mm_add_ps(y4, x4))); // sqrt(x*x+y*y+z*z)
p[0].m = _mm_mul_ps(p[0].m, _mm_set_ps1(l[0])); // load results in packet
p[1].m = _mm_mul_ps(p[1].m, _mm_set_ps1(l[1]));
p[2].m = _mm_mul_ps(p[2].m, _mm_set_ps1(l[2]));
p[3].m = _mm_mul_ps(p[3].m, _mm_set_ps1(l[3]));
}
