1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2010 Gael Guennebaud <[email protected]> 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 Packet4f 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 HasSqrt = 1, 44 HasAbs = 0, 45 HasAbs2 = 0, 46 HasMin = 0, 47 HasMax = 0, 48 HasSetLinear = 0, 49 HasBlend = 1 50 }; 51 }; 52 #endif 53 54 template<> struct unpacket_traits<Packet2cf> { 55 typedef std::complex<float> type; 56 typedef Packet2cf half; 57 typedef Packet4f as_real; 58 enum { 59 size=2, 60 alignment=Aligned16, 61 vectorizable=true, 62 masked_load_available=false, 63 masked_store_available=false 64 }; 65 }; 66 67 template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); } 68 template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); } 69 70 template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) 71 { 72 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); 73 return Packet2cf(_mm_xor_ps(a.v,mask)); 74 } 75 template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) 76 { 77 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); 78 return Packet2cf(_mm_xor_ps(a.v,mask)); 79 } 80 81 template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 82 { 83 #ifdef EIGEN_VECTORIZE_SSE3 84 return Packet2cf(_mm_addsub_ps(_mm_mul_ps(_mm_moveldup_ps(a.v), b.v), 85 _mm_mul_ps(_mm_movehdup_ps(a.v), 86 vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 87 // return Packet2cf(_mm_addsub_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 88 // _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 89 // vec4f_swizzle1(b.v, 1, 0, 3, 2)))); 90 #else 91 const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x00000000,0x80000000,0x00000000)); 92 return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), 93 _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), 94 vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); 95 #endif 96 } 97 98 template<> EIGEN_STRONG_INLINE Packet2cf ptrue <Packet2cf>(const Packet2cf& a) { return Packet2cf(ptrue(Packet4f(a.v))); } 99 template<> EIGEN_STRONG_INLINE Packet2cf pand <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_and_ps(a.v,b.v)); } 100 template<> EIGEN_STRONG_INLINE Packet2cf por <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_or_ps(a.v,b.v)); } 101 template<> EIGEN_STRONG_INLINE Packet2cf pxor <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); } 102 template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(b.v,a.v)); } 103 104 template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(&numext::real_ref(*from))); } 105 template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(&numext::real_ref(*from))); } 106 107 template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) 108 { 109 Packet2cf res; 110 #ifdef EIGEN_VECTORIZE_SSE3 111 res.v = _mm_castpd_ps(_mm_loaddup_pd(reinterpret_cast<double const*>(&from))); 112 #else 113 res.v = _mm_castpd_ps(_mm_load_sd(reinterpret_cast<double const*>(&from))); 114 res.v = _mm_movelh_ps(res.v, res.v); 115 #endif 116 return res; 117 } 118 119 template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); } 120 121 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)); } 122 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)); } 123 124 125 template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather<std::complex<float>, Packet2cf>(const std::complex<float>* from, Index stride) 126 { 127 return Packet2cf(_mm_set_ps(std::imag(from[1*stride]), std::real(from[1*stride]), 128 std::imag(from[0*stride]), std::real(from[0*stride]))); 129 } 130 131 template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet2cf>(std::complex<float>* to, const Packet2cf& from, Index stride) 132 { 133 to[stride*0] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 0)), 134 _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 1))); 135 to[stride*1] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 2)), 136 _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 3))); 137 } 138 139 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } 140 141 template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a) 142 { 143 #if EIGEN_GNUC_AT_MOST(4,3) 144 // Workaround gcc 4.2 ICE - this is not performance wise ideal, but who cares... 145 // This workaround also fix invalid code generation with gcc 4.3 146 EIGEN_ALIGN16 std::complex<float> res[2]; 147 _mm_store_ps((float*)res, a.v); 148 return res[0]; 149 #else 150 std::complex<float> res; 151 _mm_storel_pi((__m64*)&res, a.v); 152 return res; 153 #endif 154 } 155 156 template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) { return Packet2cf(_mm_castpd_ps(preverse(Packet2d(_mm_castps_pd(a.v))))); } 157 158 template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a) 159 { 160 return pfirst(Packet2cf(_mm_add_ps(a.v, _mm_movehl_ps(a.v,a.v)))); 161 } 162 163 template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a) 164 { 165 return pfirst(pmul(a, Packet2cf(_mm_movehl_ps(a.v,a.v)))); 166 } 167 168 EIGEN_STRONG_INLINE Packet2cf pcplxflip/* <Packet2cf> */(const Packet2cf& x) 169 { 170 return Packet2cf(vec4f_swizzle1(x.v, 1, 0, 3, 2)); 171 } 172 173 EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) 174 175 template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 176 { 177 // TODO optimize it for SSE3 and 4 178 Packet2cf res = pmul(a, pconj(b)); 179 __m128 s = _mm_mul_ps(b.v,b.v); 180 return Packet2cf(_mm_div_ps(res.v,_mm_add_ps(s,vec4f_swizzle1(s, 1, 0, 3, 2)))); 181 } 182 183 184 185 //---------- double ---------- 186 struct Packet1cd 187 { 188 EIGEN_STRONG_INLINE Packet1cd() {} 189 EIGEN_STRONG_INLINE explicit Packet1cd(const __m128d& a) : v(a) {} 190 Packet2d v; 191 }; 192 193 // Use the packet_traits defined in AVX/PacketMath.h instead if we're going 194 // to leverage AVX instructions. 195 #ifndef EIGEN_VECTORIZE_AVX 196 template<> struct packet_traits<std::complex<double> > : default_packet_traits 197 { 198 typedef Packet1cd type; 199 typedef Packet1cd half; 200 enum { 201 Vectorizable = 1, 202 AlignedOnScalar = 0, 203 size = 1, 204 HasHalfPacket = 0, 205 206 HasAdd = 1, 207 HasSub = 1, 208 HasMul = 1, 209 HasDiv = 1, 210 HasNegate = 1, 211 HasSqrt = 1, 212 HasAbs = 0, 213 HasAbs2 = 0, 214 HasMin = 0, 215 HasMax = 0, 216 HasSetLinear = 0 217 }; 218 }; 219 #endif 220 221 template<> struct unpacket_traits<Packet1cd> { 222 typedef std::complex<double> type; 223 typedef Packet1cd half; 224 typedef Packet2d as_real; 225 enum { 226 size=1, 227 alignment=Aligned16, 228 vectorizable=true, 229 masked_load_available=false, 230 masked_store_available=false 231 }; 232 }; 233 234 template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } 235 template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } 236 template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } 237 template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) 238 { 239 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); 240 return Packet1cd(_mm_xor_pd(a.v,mask)); 241 } 242 243 template<> EIGEN_STRONG_INLINE Packet1cd pmul<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 244 { 245 #ifdef EIGEN_VECTORIZE_SSE3 246 return Packet1cd(_mm_addsub_pd(_mm_mul_pd(_mm_movedup_pd(a.v), b.v), 247 _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 248 vec2d_swizzle1(b.v, 1, 0)))); 249 #else 250 const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); 251 return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), 252 _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), 253 vec2d_swizzle1(b.v, 1, 0)), mask))); 254 #endif 255 } 256 257 template<> EIGEN_STRONG_INLINE Packet1cd ptrue <Packet1cd>(const Packet1cd& a) { return Packet1cd(ptrue(Packet2d(a.v))); } 258 template<> EIGEN_STRONG_INLINE Packet1cd pand <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_and_pd(a.v,b.v)); } 259 template<> EIGEN_STRONG_INLINE Packet1cd por <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_or_pd(a.v,b.v)); } 260 template<> EIGEN_STRONG_INLINE Packet1cd pxor <Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_xor_pd(a.v,b.v)); } 261 template<> EIGEN_STRONG_INLINE Packet1cd pandnot<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_andnot_pd(b.v,a.v)); } 262 263 // FIXME force unaligned load, this is a temporary fix 264 template<> EIGEN_STRONG_INLINE Packet1cd pload <Packet1cd>(const std::complex<double>* from) 265 { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } 266 template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) 267 { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } 268 template<> EIGEN_STRONG_INLINE Packet1cd pset1<Packet1cd>(const std::complex<double>& from) 269 { /* here we really have to use unaligned loads :( */ return ploadu<Packet1cd>(&from); } 270 271 template<> EIGEN_STRONG_INLINE Packet1cd ploaddup<Packet1cd>(const std::complex<double>* from) { return pset1<Packet1cd>(*from); } 272 273 // FIXME force unaligned store, this is a temporary fix 274 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)); } 275 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)); } 276 277 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<double> >(const std::complex<double> * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } 278 279 template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet1cd>(const Packet1cd& a) 280 { 281 EIGEN_ALIGN16 double res[2]; 282 _mm_store_pd(res, a.v); 283 return std::complex<double>(res[0],res[1]); 284 } 285 286 template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } 287 288 template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet1cd>(const Packet1cd& a) 289 { 290 return pfirst(a); 291 } 292 293 template<> EIGEN_STRONG_INLINE std::complex<double> predux_mul<Packet1cd>(const Packet1cd& a) 294 { 295 return pfirst(a); 296 } 297 298 EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) 299 300 template<> EIGEN_STRONG_INLINE Packet1cd pdiv<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 301 { 302 // TODO optimize it for SSE3 and 4 303 Packet1cd res = pmul(a,pconj(b)); 304 __m128d s = _mm_mul_pd(b.v,b.v); 305 return Packet1cd(_mm_div_pd(res.v, _mm_add_pd(s,_mm_shuffle_pd(s, s, 0x1)))); 306 } 307 308 EIGEN_STRONG_INLINE Packet1cd pcplxflip/* <Packet1cd> */(const Packet1cd& x) 309 { 310 return Packet1cd(preverse(Packet2d(x.v))); 311 } 312 313 EIGEN_DEVICE_FUNC inline void 314 ptranspose(PacketBlock<Packet2cf,2>& kernel) { 315 __m128d w1 = _mm_castps_pd(kernel.packet[0].v); 316 __m128d w2 = _mm_castps_pd(kernel.packet[1].v); 317 318 __m128 tmp = _mm_castpd_ps(_mm_unpackhi_pd(w1, w2)); 319 kernel.packet[0].v = _mm_castpd_ps(_mm_unpacklo_pd(w1, w2)); 320 kernel.packet[1].v = tmp; 321 } 322 323 template<> EIGEN_STRONG_INLINE Packet2cf pcmp_eq(const Packet2cf& a, const Packet2cf& b) 324 { 325 __m128 eq = _mm_cmpeq_ps(a.v, b.v); 326 return Packet2cf(pand<Packet4f>(eq, vec4f_swizzle1(eq, 1, 0, 3, 2))); 327 } 328 329 template<> EIGEN_STRONG_INLINE Packet1cd pcmp_eq(const Packet1cd& a, const Packet1cd& b) 330 { 331 __m128d eq = _mm_cmpeq_pd(a.v, b.v); 332 return Packet1cd(pand<Packet2d>(eq, vec2d_swizzle1(eq, 1, 0))); 333 } 334 335 template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { 336 __m128d result = pblend<Packet2d>(ifPacket, _mm_castps_pd(thenPacket.v), _mm_castps_pd(elsePacket.v)); 337 return Packet2cf(_mm_castpd_ps(result)); 338 } 339 340 template<> EIGEN_STRONG_INLINE Packet1cd psqrt<Packet1cd>(const Packet1cd& a) { 341 return psqrt_complex<Packet1cd>(a); 342 } 343 344 template<> EIGEN_STRONG_INLINE Packet2cf psqrt<Packet2cf>(const Packet2cf& a) { 345 return psqrt_complex<Packet2cf>(a); 346 } 347 348 } // end namespace internal 349 } // end namespace Eigen 350 351 #endif // EIGEN_COMPLEX_SSE_H 352