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 // Use the packet_traits defined in AVX/PacketMath.h instead if we're going 26 // to leverage AVX instructions. 27 #ifndef EIGEN_VECTORIZE_AVX 28 template<> struct packet_traits<std::complex<float> > : default_packet_traits 29 { 30 typedef Packet2cf type; 31 typedef Packet2cf half; 32 enum { 33 Vectorizable = 1, 34 AlignedOnScalar = 1, 35 size = 2, 36 HasHalfPacket = 0, 37 38 HasAdd = 1, 39 HasSub = 1, 40 HasMul = 1, 41 HasDiv = 1, 42 HasNegate = 1, 43 HasAbs = 0, 44 HasAbs2 = 0, 45 HasMin = 0, 46 HasMax = 0, 47 HasSetLinear = 0, 48 HasBlend = 1 49 }; 50 }; 51 #endif 52 53 template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; 54 55 template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); } 56 template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); } 57 template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) 58 { 59 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); 60 return Packet2cf(_mm_xor_ps(a.v,mask)); 61 } 62 template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) 63 { 64 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 65 return Packet2cf(_mm_xor_ps(a.v,mask)); 66 } 67 68 template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 69 { 70 #ifdef EIGEN_VECTORIZE_SSE3 71 return Packet2cf(_mm_addsub_ps(_mm_mul_ps(_mm_moveldup_ps(a.v), b.v), 72 _mm_mul_ps(_mm_movehdup_ps(a.v), 73 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 74 // return Packet2cf(_mm_addsub_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 75 // _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 76 // vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 77 #else 78 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x00000000,0x80000000,0x00000000)); 79 return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 80 _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 81 vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); 82 #endif 83 } 84 85 template<> EIGEN_STRONG_INLINE Packet2cf pand <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_and_ps(a.v,b.v)); } 86 template<> EIGEN_STRONG_INLINE Packet2cf por <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_or_ps(a.v,b.v)); } 87 template<> EIGEN_STRONG_INLINE Packet2cf pxor <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); } 88 template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(a.v,b.v)); } 89 90 template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(&numext::real_ref(*from))); } 91 template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(&numext::real_ref(*from))); } 92 93 template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) 94 { 95 Packet2cf res; 96 #if EIGEN_GNUC_AT_MOST(4,2) 97 // Workaround annoying "may be used uninitialized in this function" warning with gcc 4.2 98 res.v = _mm_loadl_pi(_mm_set1_ps(0.0f), reinterpret_cast<const __m64*>(&from)); 99 #elif EIGEN_GNUC_AT_LEAST(4,6) 100 // Suppress annoying "may be used uninitialized in this function" warning with gcc >= 4.6 101 #pragma GCC diagnostic push 102 #pragma GCC diagnostic ignored "-Wuninitialized" 103 res.v = _mm_loadl_pi(res.v, (const __m64*)&from); 104 #pragma GCC diagnostic pop 105 #else 106 res.v = _mm_loadl_pi(res.v, (const __m64*)&from); 107 #endif 108 return Packet2cf(_mm_movelh_ps(res.v,res.v)); 109 } 110 111 template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); } 112 113 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), Packet4f(from.v)); } 114 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), Packet4f(from.v)); } 115 116 117 template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather<std::complex<float>, Packet2cf>(const std::complex<float>* from, Index stride) 118 { 119 return Packet2cf(_mm_set_ps(std::imag(from[1*stride]), std::real(from[1*stride]), 120 std::imag(from[0*stride]), std::real(from[0*stride]))); 121 } 122 123 template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet2cf>(std::complex<float>* to, const Packet2cf& from, Index stride) 124 { 125 to[stride*0] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 0)), 126 _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 1))); 127 to[stride*1] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 2)), 128 _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 3))); 129 } 130 131 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> * addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } 132 133 template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a) 134 { 135 #if EIGEN_GNUC_AT_MOST(4,3) 136 // Workaround gcc 4.2 ICE - this is not performance wise ideal, but who cares... 137 // This workaround also fix invalid code generation with gcc 4.3 138 EIGEN_ALIGN16 std::complex<float> res[2]; 139 _mm_store_ps((float*)res, a.v); 140 return res[0]; 141 #else 142 std::complex<float> res; 143 _mm_storel_pi((__m64*)&res, a.v); 144 return res; 145 #endif 146 } 147 148 template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) { return Packet2cf(_mm_castpd_ps(preverse(Packet2d(_mm_castps_pd(a.v))))); } 149 150 template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a) 151 { 152 return pfirst(Packet2cf(_mm_add_ps(a.v, _mm_movehl_ps(a.v,a.v)))); 153 } 154 155 template<> EIGEN_STRONG_INLINE Packet2cf preduxp<Packet2cf>(const Packet2cf* vecs) 156 { 157 return Packet2cf(_mm_add_ps(_mm_movelh_ps(vecs[0].v,vecs[1].v), _mm_movehl_ps(vecs[1].v,vecs[0].v))); 158 } 159 160 template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a) 161 { 162 return pfirst(pmul(a, Packet2cf(_mm_movehl_ps(a.v,a.v)))); 163 } 164 165 template<int Offset> 166 struct palign_impl<Offset,Packet2cf> 167 { 168 static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) 169 { 170 if (Offset==1) 171 { 172 first.v = _mm_movehl_ps(first.v, first.v); 173 first.v = _mm_movelh_ps(first.v, second.v); 174 } 175 } 176 }; 177 178 template<> struct conj_helper<Packet2cf, Packet2cf, false,true> 179 { 180 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 181 { return padd(pmul(x,y),c); } 182 183 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 184 { 185 #ifdef EIGEN_VECTORIZE_SSE3 186 return internal::pmul(a, pconj(b)); 187 #else 188 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 189 return Packet2cf(_mm_add_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), 190 _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 191 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 192 #endif 193 } 194 }; 195 196 template<> struct conj_helper<Packet2cf, Packet2cf, true,false> 197 { 198 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 199 { return padd(pmul(x,y),c); } 200 201 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 202 { 203 #ifdef EIGEN_VECTORIZE_SSE3 204 return internal::pmul(pconj(a), b); 205 #else 206 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 207 return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 208 _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 209 vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); 210 #endif 211 } 212 }; 213 214 template<> struct conj_helper<Packet2cf, Packet2cf, true,true> 215 { 216 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 217 { return padd(pmul(x,y),c); } 218 219 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 220 { 221 #ifdef EIGEN_VECTORIZE_SSE3 222 return pconj(internal::pmul(a, b)); 223 #else 224 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 225 return Packet2cf(_mm_sub_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), 226 _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 227 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 228 #endif 229 } 230 }; 231 232 template<> struct conj_helper<Packet4f, Packet2cf, false,false> 233 { 234 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet4f& x, const Packet2cf& y, const Packet2cf& c) const 235 { return padd(c, pmul(x,y)); } 236 237 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet4f& x, const Packet2cf& y) const 238 { return Packet2cf(Eigen::internal::pmul<Packet4f>(x, y.v)); } 239 }; 240 241 template<> struct conj_helper<Packet2cf, Packet4f, false,false> 242 { 243 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet4f& y, const Packet2cf& c) const 244 { return padd(c, pmul(x,y)); } 245 246 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& x, const Packet4f& y) const 247 { return Packet2cf(Eigen::internal::pmul<Packet4f>(x.v, y)); } 248 }; 249 250 template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 251 { 252 // TODO optimize it for SSE3 and 4 253 Packet2cf res = conj_helper<Packet2cf,Packet2cf,false,true>().pmul(a,b); 254 __m128 s = _mm_mul_ps(b.v,b.v); 255 return Packet2cf(_mm_div_ps(res.v,_mm_add_ps(s,_mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(s), 0xb1))))); 256 } 257 258 EIGEN_STRONG_INLINE Packet2cf pcplxflip/* <Packet2cf> */(const Packet2cf& x) 259 { 260 return Packet2cf(vec4f_swizzle1(x.v, 1, 0, 3, 2)); 261 } 262 263 264 //---------- double ---------- 265 struct Packet1cd 266 { 267 EIGEN_STRONG_INLINE Packet1cd() {} 268 EIGEN_STRONG_INLINE explicit Packet1cd(const __m128d& a) : v(a) {} 269 __m128d v; 270 }; 271 272 // Use the packet_traits defined in AVX/PacketMath.h instead if we're going 273 // to leverage AVX instructions. 274 #ifndef EIGEN_VECTORIZE_AVX 275 template<> struct packet_traits<std::complex<double> > : default_packet_traits 276 { 277 typedef Packet1cd type; 278 typedef Packet1cd half; 279 enum { 280 Vectorizable = 1, 281 AlignedOnScalar = 0, 282 size = 1, 283 HasHalfPacket = 0, 284 285 HasAdd = 1, 286 HasSub = 1, 287 HasMul = 1, 288 HasDiv = 1, 289 HasNegate = 1, 290 HasAbs = 0, 291 HasAbs2 = 0, 292 HasMin = 0, 293 HasMax = 0, 294 HasSetLinear = 0 295 }; 296 }; 297 #endif 298 299 template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; 300 301 template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } 302 template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } 303 template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } 304 template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) 305 { 306 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 307 return Packet1cd(_mm_xor_pd(a.v,mask)); 308 } 309 310 template<> EIGEN_STRONG_INLINE Packet1cd pmul<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 311 { 312 #ifdef EIGEN_VECTORIZE_SSE3 313 return Packet1cd(_mm_addsub_pd(_mm_mul_pd(_mm_movedup_pd(a.v), b.v), 314 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 315 vec2d_swizzle1(b.v, 1, 0)))); 316 #else 317 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); 318 return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 319 _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 320 vec2d_swizzle1(b.v, 1, 0)), mask))); 321 #endif 322 } 323 324 template<> EIGEN_STRONG_INLINE Packet1cd pand <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_and_pd(a.v,b.v)); } 325 template<> EIGEN_STRONG_INLINE Packet1cd por <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_or_pd(a.v,b.v)); } 326 template<> EIGEN_STRONG_INLINE Packet1cd pxor <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_xor_pd(a.v,b.v)); } 327 template<> EIGEN_STRONG_INLINE Packet1cd pandnot<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_andnot_pd(a.v,b.v)); } 328 329 // FIXME force unaligned load, this is a temporary fix 330 template<> EIGEN_STRONG_INLINE Packet1cd pload <Packet1cd>(const std::complex<double>* from) 331 { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } 332 template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) 333 { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } 334 template<> EIGEN_STRONG_INLINE Packet1cd pset1<Packet1cd>(const std::complex<double>& from) 335 { /* here we really have to use unaligned loads :( */ return ploadu<Packet1cd>(&from); } 336 337 template<> EIGEN_STRONG_INLINE Packet1cd ploaddup<Packet1cd>(const std::complex<double>* from) { return pset1<Packet1cd>(*from); } 338 339 // FIXME force unaligned store, this is a temporary fix 340 template<> EIGEN_STRONG_INLINE void pstore <std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, Packet2d(from.v)); } 341 template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, Packet2d(from.v)); } 342 343 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<double> >(const std::complex<double> * addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } 344 345 template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet1cd>(const Packet1cd& a) 346 { 347 EIGEN_ALIGN16 double res[2]; 348 _mm_store_pd(res, a.v); 349 return std::complex<double>(res[0],res[1]); 350 } 351 352 template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } 353 354 template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet1cd>(const Packet1cd& a) 355 { 356 return pfirst(a); 357 } 358 359 template<> EIGEN_STRONG_INLINE Packet1cd preduxp<Packet1cd>(const Packet1cd* vecs) 360 { 361 return vecs[0]; 362 } 363 364 template<> EIGEN_STRONG_INLINE std::complex<double> predux_mul<Packet1cd>(const Packet1cd& a) 365 { 366 return pfirst(a); 367 } 368 369 template<int Offset> 370 struct palign_impl<Offset,Packet1cd> 371 { 372 static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) 373 { 374 // FIXME is it sure we never have to align a Packet1cd? 375 // Even though a std::complex<double> has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... 376 } 377 }; 378 379 template<> struct conj_helper<Packet1cd, Packet1cd, false,true> 380 { 381 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 382 { return padd(pmul(x,y),c); } 383 384 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 385 { 386 #ifdef EIGEN_VECTORIZE_SSE3 387 return internal::pmul(a, pconj(b)); 388 #else 389 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 390 return Packet1cd(_mm_add_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), 391 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 392 vec2d_swizzle1(b.v, 1, 0)))); 393 #endif 394 } 395 }; 396 397 template<> struct conj_helper<Packet1cd, Packet1cd, true,false> 398 { 399 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 400 { return padd(pmul(x,y),c); } 401 402 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 403 { 404 #ifdef EIGEN_VECTORIZE_SSE3 405 return internal::pmul(pconj(a), b); 406 #else 407 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 408 return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 409 _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 410 vec2d_swizzle1(b.v, 1, 0)), mask))); 411 #endif 412 } 413 }; 414 415 template<> struct conj_helper<Packet1cd, Packet1cd, true,true> 416 { 417 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 418 { return padd(pmul(x,y),c); } 419 420 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 421 { 422 #ifdef EIGEN_VECTORIZE_SSE3 423 return pconj(internal::pmul(a, b)); 424 #else 425 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 426 return Packet1cd(_mm_sub_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), 427 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 428 vec2d_swizzle1(b.v, 1, 0)))); 429 #endif 430 } 431 }; 432 433 template<> struct conj_helper<Packet2d, Packet1cd, false,false> 434 { 435 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet2d& x, const Packet1cd& y, const Packet1cd& c) const 436 { return padd(c, pmul(x,y)); } 437 438 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet2d& x, const Packet1cd& y) const 439 { return Packet1cd(Eigen::internal::pmul<Packet2d>(x, y.v)); } 440 }; 441 442 template<> struct conj_helper<Packet1cd, Packet2d, false,false> 443 { 444 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet2d& y, const Packet1cd& c) const 445 { return padd(c, pmul(x,y)); } 446 447 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& x, const Packet2d& y) const 448 { return Packet1cd(Eigen::internal::pmul<Packet2d>(x.v, y)); } 449 }; 450 451 template<> EIGEN_STRONG_INLINE Packet1cd pdiv<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 452 { 453 // TODO optimize it for SSE3 and 4 454 Packet1cd res = conj_helper<Packet1cd,Packet1cd,false,true>().pmul(a,b); 455 __m128d s = _mm_mul_pd(b.v,b.v); 456 return Packet1cd(_mm_div_pd(res.v, _mm_add_pd(s,_mm_shuffle_pd(s, s, 0x1)))); 457 } 458 459 EIGEN_STRONG_INLINE Packet1cd pcplxflip/* <Packet1cd> */(const Packet1cd& x) 460 { 461 return Packet1cd(preverse(Packet2d(x.v))); 462 } 463 464 EIGEN_DEVICE_FUNC inline void 465 ptranspose(PacketBlock<Packet2cf,2>& kernel) { 466 __m128d w1 = _mm_castps_pd(kernel.packet[0].v); 467 __m128d w2 = _mm_castps_pd(kernel.packet[1].v); 468 469 __m128 tmp = _mm_castpd_ps(_mm_unpackhi_pd(w1, w2)); 470 kernel.packet[0].v = _mm_castpd_ps(_mm_unpacklo_pd(w1, w2)); 471 kernel.packet[1].v = tmp; 472 } 473 474 template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { 475 __m128d result = pblend<Packet2d>(ifPacket, _mm_castps_pd(thenPacket.v), _mm_castps_pd(elsePacket.v)); 476 return Packet2cf(_mm_castpd_ps(result)); 477 } 478 479 template<> EIGEN_STRONG_INLINE Packet2cf pinsertfirst(const Packet2cf& a, std::complex<float> b) 480 { 481 return Packet2cf(_mm_loadl_pi(a.v, reinterpret_cast<const __m64*>(&b))); 482 } 483 484 template<> EIGEN_STRONG_INLINE Packet1cd pinsertfirst(const Packet1cd&, std::complex<double> b) 485 { 486 return pset1<Packet1cd>(b); 487 } 488 489 template<> EIGEN_STRONG_INLINE Packet2cf pinsertlast(const Packet2cf& a, std::complex<float> b) 490 { 491 return Packet2cf(_mm_loadh_pi(a.v, reinterpret_cast<const __m64*>(&b))); 492 } 493 494 template<> EIGEN_STRONG_INLINE Packet1cd pinsertlast(const Packet1cd&, std::complex<double> b) 495 { 496 return pset1<Packet1cd>(b); 497 } 498 499 } // end namespace internal 500 501 } // end namespace Eigen 502 503 #endif // EIGEN_COMPLEX_SSE_H 504