1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2010 Gael Guennebaud <gael.guennebaud@inria.fr> 5 // 6 // This Source Code Form is subject to the terms of the Mozilla 7 // Public License v. 2.0. If a copy of the MPL was not distributed 8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. 9 10 #ifndef EIGEN_COMPLEX_SSE_H 11 #define EIGEN_COMPLEX_SSE_H 12 13 namespace Eigen { 14 15 namespace internal { 16 17 //---------- float ---------- 18 struct Packet2cf 19 { Packet2cfPacket2cf20 EIGEN_STRONG_INLINE Packet2cf() {} Packet2cfPacket2cf21 EIGEN_STRONG_INLINE explicit Packet2cf(const __m128& a) : v(a) {} 22 __m128 v; 23 }; 24 25 template<> struct packet_traits<std::complex<float> > : default_packet_traits 26 { 27 typedef Packet2cf type; 28 enum { 29 Vectorizable = 1, 30 AlignedOnScalar = 1, 31 size = 2, 32 33 HasAdd = 1, 34 HasSub = 1, 35 HasMul = 1, 36 HasDiv = 1, 37 HasNegate = 1, 38 HasAbs = 0, 39 HasAbs2 = 0, 40 HasMin = 0, 41 HasMax = 0, 42 HasSetLinear = 0 43 }; 44 }; 45 46 template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; }; 47 48 template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); } 49 template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); } 50 template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) 51 { 52 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); 53 return Packet2cf(_mm_xor_ps(a.v,mask)); 54 } 55 template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) 56 { 57 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 58 return Packet2cf(_mm_xor_ps(a.v,mask)); 59 } 60 61 template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 62 { 63 // TODO optimize it for SSE3 and 4 64 #ifdef EIGEN_VECTORIZE_SSE3 65 return Packet2cf(_mm_addsub_ps(_mm_mul_ps(_mm_moveldup_ps(a.v), b.v), 66 _mm_mul_ps(_mm_movehdup_ps(a.v), 67 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 68 // return Packet2cf(_mm_addsub_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 69 // _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 70 // vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 71 #else 72 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x00000000,0x80000000,0x00000000)); 73 return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 74 _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 75 vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); 76 #endif 77 } 78 79 template<> EIGEN_STRONG_INLINE Packet2cf pand <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_and_ps(a.v,b.v)); } 80 template<> EIGEN_STRONG_INLINE Packet2cf por <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_or_ps(a.v,b.v)); } 81 template<> EIGEN_STRONG_INLINE Packet2cf pxor <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); } 82 template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(a.v,b.v)); } 83 84 template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(&numext::real_ref(*from))); } 85 template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(&numext::real_ref(*from))); } 86 87 template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) 88 { 89 Packet2cf res; 90 #if EIGEN_GNUC_AT_MOST(4,2) 91 // Workaround annoying "may be used uninitialized in this function" warning with gcc 4.2 92 res.v = _mm_loadl_pi(_mm_set1_ps(0.0f), reinterpret_cast<const __m64*>(&from)); 93 #elif EIGEN_GNUC_AT_LEAST(4,6) 94 // Suppress annoying "may be used uninitialized in this function" warning with gcc >= 4.6 95 #pragma GCC diagnostic push 96 #pragma GCC diagnostic ignored "-Wuninitialized" 97 res.v = _mm_loadl_pi(res.v, (const __m64*)&from); 98 #pragma GCC diagnostic pop 99 #else 100 res.v = _mm_loadl_pi(res.v, (const __m64*)&from); 101 #endif 102 return Packet2cf(_mm_movelh_ps(res.v,res.v)); 103 } 104 105 template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); } 106 107 template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), from.v); } 108 template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), from.v); } 109 110 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> * addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } 111 112 template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a) 113 { 114 #if EIGEN_GNUC_AT_MOST(4,3) 115 // Workaround gcc 4.2 ICE - this is not performance wise ideal, but who cares... 116 // This workaround also fix invalid code generation with gcc 4.3 117 EIGEN_ALIGN16 std::complex<float> res[2]; 118 _mm_store_ps((float*)res, a.v); 119 return res[0]; 120 #else 121 std::complex<float> res; 122 _mm_storel_pi((__m64*)&res, a.v); 123 return res; 124 #endif 125 } 126 127 template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) { return Packet2cf(_mm_castpd_ps(preverse(_mm_castps_pd(a.v)))); } 128 129 template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a) 130 { 131 return pfirst(Packet2cf(_mm_add_ps(a.v, _mm_movehl_ps(a.v,a.v)))); 132 } 133 134 template<> EIGEN_STRONG_INLINE Packet2cf preduxp<Packet2cf>(const Packet2cf* vecs) 135 { 136 return Packet2cf(_mm_add_ps(_mm_movelh_ps(vecs[0].v,vecs[1].v), _mm_movehl_ps(vecs[1].v,vecs[0].v))); 137 } 138 139 template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a) 140 { 141 return pfirst(pmul(a, Packet2cf(_mm_movehl_ps(a.v,a.v)))); 142 } 143 144 template<int Offset> 145 struct palign_impl<Offset,Packet2cf> 146 { 147 static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) 148 { 149 if (Offset==1) 150 { 151 first.v = _mm_movehl_ps(first.v, first.v); 152 first.v = _mm_movelh_ps(first.v, second.v); 153 } 154 } 155 }; 156 157 template<> struct conj_helper<Packet2cf, Packet2cf, false,true> 158 { 159 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 160 { return padd(pmul(x,y),c); } 161 162 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 163 { 164 #ifdef EIGEN_VECTORIZE_SSE3 165 return internal::pmul(a, pconj(b)); 166 #else 167 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 168 return Packet2cf(_mm_add_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), 169 _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 170 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 171 #endif 172 } 173 }; 174 175 template<> struct conj_helper<Packet2cf, Packet2cf, true,false> 176 { 177 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 178 { return padd(pmul(x,y),c); } 179 180 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 181 { 182 #ifdef EIGEN_VECTORIZE_SSE3 183 return internal::pmul(pconj(a), b); 184 #else 185 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 186 return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 187 _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 188 vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); 189 #endif 190 } 191 }; 192 193 template<> struct conj_helper<Packet2cf, Packet2cf, true,true> 194 { 195 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 196 { return padd(pmul(x,y),c); } 197 198 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 199 { 200 #ifdef EIGEN_VECTORIZE_SSE3 201 return pconj(internal::pmul(a, b)); 202 #else 203 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 204 return Packet2cf(_mm_sub_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), 205 _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 206 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 207 #endif 208 } 209 }; 210 211 template<> struct conj_helper<Packet4f, Packet2cf, false,false> 212 { 213 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet4f& x, const Packet2cf& y, const Packet2cf& c) const 214 { return padd(c, pmul(x,y)); } 215 216 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet4f& x, const Packet2cf& y) const 217 { return Packet2cf(Eigen::internal::pmul(x, y.v)); } 218 }; 219 220 template<> struct conj_helper<Packet2cf, Packet4f, false,false> 221 { 222 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet4f& y, const Packet2cf& c) const 223 { return padd(c, pmul(x,y)); } 224 225 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& x, const Packet4f& y) const 226 { return Packet2cf(Eigen::internal::pmul(x.v, y)); } 227 }; 228 229 template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 230 { 231 // TODO optimize it for SSE3 and 4 232 Packet2cf res = conj_helper<Packet2cf,Packet2cf,false,true>().pmul(a,b); 233 __m128 s = _mm_mul_ps(b.v,b.v); 234 return Packet2cf(_mm_div_ps(res.v,_mm_add_ps(s,_mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(s), 0xb1))))); 235 } 236 237 EIGEN_STRONG_INLINE Packet2cf pcplxflip/*<Packet2cf>*/(const Packet2cf& x) 238 { 239 return Packet2cf(vec4f_swizzle1(x.v, 1, 0, 3, 2)); 240 } 241 242 243 //---------- double ---------- 244 struct Packet1cd 245 { 246 EIGEN_STRONG_INLINE Packet1cd() {} 247 EIGEN_STRONG_INLINE explicit Packet1cd(const __m128d& a) : v(a) {} 248 __m128d v; 249 }; 250 251 template<> struct packet_traits<std::complex<double> > : default_packet_traits 252 { 253 typedef Packet1cd type; 254 enum { 255 Vectorizable = 1, 256 AlignedOnScalar = 0, 257 size = 1, 258 259 HasAdd = 1, 260 HasSub = 1, 261 HasMul = 1, 262 HasDiv = 1, 263 HasNegate = 1, 264 HasAbs = 0, 265 HasAbs2 = 0, 266 HasMin = 0, 267 HasMax = 0, 268 HasSetLinear = 0 269 }; 270 }; 271 272 template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1}; }; 273 274 template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } 275 template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } 276 template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(a.v)); } 277 template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) 278 { 279 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 280 return Packet1cd(_mm_xor_pd(a.v,mask)); 281 } 282 283 template<> EIGEN_STRONG_INLINE Packet1cd pmul<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 284 { 285 // TODO optimize it for SSE3 and 4 286 #ifdef EIGEN_VECTORIZE_SSE3 287 return Packet1cd(_mm_addsub_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 288 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 289 vec2d_swizzle1(b.v, 1, 0)))); 290 #else 291 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); 292 return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 293 _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 294 vec2d_swizzle1(b.v, 1, 0)), mask))); 295 #endif 296 } 297 298 template<> EIGEN_STRONG_INLINE Packet1cd pand <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_and_pd(a.v,b.v)); } 299 template<> EIGEN_STRONG_INLINE Packet1cd por <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_or_pd(a.v,b.v)); } 300 template<> EIGEN_STRONG_INLINE Packet1cd pxor <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_xor_pd(a.v,b.v)); } 301 template<> EIGEN_STRONG_INLINE Packet1cd pandnot<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_andnot_pd(a.v,b.v)); } 302 303 // FIXME force unaligned load, this is a temporary fix 304 template<> EIGEN_STRONG_INLINE Packet1cd pload <Packet1cd>(const std::complex<double>* from) 305 { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } 306 template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) 307 { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } 308 template<> EIGEN_STRONG_INLINE Packet1cd pset1<Packet1cd>(const std::complex<double>& from) 309 { /* here we really have to use unaligned loads :( */ return ploadu<Packet1cd>(&from); } 310 311 template<> EIGEN_STRONG_INLINE Packet1cd ploaddup<Packet1cd>(const std::complex<double>* from) { return pset1<Packet1cd>(*from); } 312 313 // FIXME force unaligned store, this is a temporary fix 314 template<> EIGEN_STRONG_INLINE void pstore <std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, from.v); } 315 template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, from.v); } 316 317 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<double> >(const std::complex<double> * addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } 318 319 template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet1cd>(const Packet1cd& a) 320 { 321 EIGEN_ALIGN16 double res[2]; 322 _mm_store_pd(res, a.v); 323 return std::complex<double>(res[0],res[1]); 324 } 325 326 template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } 327 328 template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet1cd>(const Packet1cd& a) 329 { 330 return pfirst(a); 331 } 332 333 template<> EIGEN_STRONG_INLINE Packet1cd preduxp<Packet1cd>(const Packet1cd* vecs) 334 { 335 return vecs[0]; 336 } 337 338 template<> EIGEN_STRONG_INLINE std::complex<double> predux_mul<Packet1cd>(const Packet1cd& a) 339 { 340 return pfirst(a); 341 } 342 343 template<int Offset> 344 struct palign_impl<Offset,Packet1cd> 345 { 346 static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) 347 { 348 // FIXME is it sure we never have to align a Packet1cd? 349 // Even though a std::complex<double> has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... 350 } 351 }; 352 353 template<> struct conj_helper<Packet1cd, Packet1cd, false,true> 354 { 355 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 356 { return padd(pmul(x,y),c); } 357 358 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 359 { 360 #ifdef EIGEN_VECTORIZE_SSE3 361 return internal::pmul(a, pconj(b)); 362 #else 363 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 364 return Packet1cd(_mm_add_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), 365 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 366 vec2d_swizzle1(b.v, 1, 0)))); 367 #endif 368 } 369 }; 370 371 template<> struct conj_helper<Packet1cd, Packet1cd, true,false> 372 { 373 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 374 { return padd(pmul(x,y),c); } 375 376 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 377 { 378 #ifdef EIGEN_VECTORIZE_SSE3 379 return internal::pmul(pconj(a), b); 380 #else 381 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 382 return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 383 _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 384 vec2d_swizzle1(b.v, 1, 0)), mask))); 385 #endif 386 } 387 }; 388 389 template<> struct conj_helper<Packet1cd, Packet1cd, true,true> 390 { 391 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 392 { return padd(pmul(x,y),c); } 393 394 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 395 { 396 #ifdef EIGEN_VECTORIZE_SSE3 397 return pconj(internal::pmul(a, b)); 398 #else 399 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 400 return Packet1cd(_mm_sub_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), 401 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 402 vec2d_swizzle1(b.v, 1, 0)))); 403 #endif 404 } 405 }; 406 407 template<> struct conj_helper<Packet2d, Packet1cd, false,false> 408 { 409 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet2d& x, const Packet1cd& y, const Packet1cd& c) const 410 { return padd(c, pmul(x,y)); } 411 412 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet2d& x, const Packet1cd& y) const 413 { return Packet1cd(Eigen::internal::pmul(x, y.v)); } 414 }; 415 416 template<> struct conj_helper<Packet1cd, Packet2d, false,false> 417 { 418 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet2d& y, const Packet1cd& c) const 419 { return padd(c, pmul(x,y)); } 420 421 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& x, const Packet2d& y) const 422 { return Packet1cd(Eigen::internal::pmul(x.v, y)); } 423 }; 424 425 template<> EIGEN_STRONG_INLINE Packet1cd pdiv<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 426 { 427 // TODO optimize it for SSE3 and 4 428 Packet1cd res = conj_helper<Packet1cd,Packet1cd,false,true>().pmul(a,b); 429 __m128d s = _mm_mul_pd(b.v,b.v); 430 return Packet1cd(_mm_div_pd(res.v, _mm_add_pd(s,_mm_shuffle_pd(s, s, 0x1)))); 431 } 432 433 EIGEN_STRONG_INLINE Packet1cd pcplxflip/*<Packet1cd>*/(const Packet1cd& x) 434 { 435 return Packet1cd(preverse(x.v)); 436 } 437 438 } // end namespace internal 439 440 } // end namespace Eigen 441 442 #endif // EIGEN_COMPLEX_SSE_H 443