11 #ifndef EIGEN_GEOMETRY_SSE_H 12 #define EIGEN_GEOMETRY_SSE_H 18 template<
class Derived,
class OtherDerived>
19 struct quat_product<Architecture::SSE, Derived, OtherDerived, float,
Aligned16>
21 static inline Quaternion<float> run(
const QuaternionBase<Derived>& _a,
const QuaternionBase<OtherDerived>& _b)
23 Quaternion<float> res;
24 const __m128 mask = _mm_setr_ps(0.f,0.f,0.f,-0.f);
25 __m128 a = _a.coeffs().template packet<Aligned16>(0);
26 __m128 b = _b.coeffs().template packet<Aligned16>(0);
27 __m128 s1 = _mm_mul_ps(vec4f_swizzle1(a,1,2,0,2),vec4f_swizzle1(b,2,0,1,2));
28 __m128 s2 = _mm_mul_ps(vec4f_swizzle1(a,3,3,3,1),vec4f_swizzle1(b,0,1,2,1));
30 _mm_add_ps(_mm_sub_ps(_mm_mul_ps(a,vec4f_swizzle1(b,3,3,3,3)),
31 _mm_mul_ps(vec4f_swizzle1(a,2,0,1,0),
32 vec4f_swizzle1(b,1,2,0,0))),
33 _mm_xor_ps(mask,_mm_add_ps(s1,s2))));
39 template<
class Derived,
int Alignment>
40 struct quat_conj<Architecture::SSE, Derived, float, Alignment>
42 static inline Quaternion<float> run(
const QuaternionBase<Derived>& q)
44 Quaternion<float> res;
45 const __m128 mask = _mm_setr_ps(-0.f,-0.f,-0.f,0.f);
46 pstore(&res.x(), _mm_xor_ps(mask, q.coeffs().template packet<Alignment>(0)));
52 template<
typename VectorLhs,
typename VectorRhs>
53 struct cross3_impl<Architecture::SSE,VectorLhs,VectorRhs,float,true>
55 static inline typename plain_matrix_type<VectorLhs>::type
56 run(
const VectorLhs& lhs,
const VectorRhs& rhs)
58 __m128 a = lhs.template packet<traits<VectorLhs>::Alignment>(0);
59 __m128 b = rhs.template packet<traits<VectorRhs>::Alignment>(0);
60 __m128 mul1=_mm_mul_ps(vec4f_swizzle1(a,1,2,0,3),vec4f_swizzle1(b,2,0,1,3));
61 __m128 mul2=_mm_mul_ps(vec4f_swizzle1(a,2,0,1,3),vec4f_swizzle1(b,1,2,0,3));
62 typename plain_matrix_type<VectorLhs>::type res;
63 pstore(&res.x(),_mm_sub_ps(mul1,mul2));
71 template<
class Derived,
class OtherDerived,
int Alignment>
72 struct quat_product<Architecture::SSE, Derived, OtherDerived, double, Alignment>
74 static inline Quaternion<double> run(
const QuaternionBase<Derived>& _a,
const QuaternionBase<OtherDerived>& _b)
76 const Packet2d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0));
78 Quaternion<double> res;
80 const double* a = _a.coeffs().data();
81 Packet2d b_xy = _b.coeffs().template packet<Alignment>(0);
82 Packet2d b_zw = _b.coeffs().template packet<Alignment>(2);
83 Packet2d a_xx = pset1<Packet2d>(a[0]);
84 Packet2d a_yy = pset1<Packet2d>(a[1]);
85 Packet2d a_zz = pset1<Packet2d>(a[2]);
86 Packet2d a_ww = pset1<Packet2d>(a[3]);
96 t1 = padd(pmul(a_ww, b_xy), pmul(a_yy, b_zw));
97 t2 = psub(pmul(a_zz, b_xy), pmul(a_xx, b_zw));
98 #ifdef EIGEN_VECTORIZE_SSE3 99 EIGEN_UNUSED_VARIABLE(mask)
100 pstore(&res.x(), _mm_addsub_pd(t1, preverse(t2)));
102 pstore(&res.x(), padd(t1, pxor(mask,preverse(t2))));
110 t1 = psub(pmul(a_ww, b_zw), pmul(a_yy, b_xy));
111 t2 = padd(pmul(a_zz, b_zw), pmul(a_xx, b_xy));
112 #ifdef EIGEN_VECTORIZE_SSE3 113 EIGEN_UNUSED_VARIABLE(mask)
114 pstore(&res.z(), preverse(_mm_addsub_pd(preverse(t1), t2)));
116 pstore(&res.z(), psub(t1, pxor(mask,preverse(t2))));
123 template<
class Derived,
int Alignment>
124 struct quat_conj<Architecture::SSE, Derived, double, Alignment>
126 static inline Quaternion<double> run(
const QuaternionBase<Derived>& q)
128 Quaternion<double> res;
129 const __m128d mask0 = _mm_setr_pd(-0.,-0.);
130 const __m128d mask2 = _mm_setr_pd(-0.,0.);
131 pstore(&res.x(), _mm_xor_pd(mask0, q.coeffs().template packet<Alignment>(0)));
132 pstore(&res.z(), _mm_xor_pd(mask2, q.coeffs().template packet<Alignment>(2)));
141 #endif // EIGEN_GEOMETRY_SSE_H Definition: Constants.h:230
Namespace containing all symbols from the Eigen library.
Definition: Core:271
Definition: Eigen_Colamd.h:50