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// Copyright (C) 2010 Konstantinos Margaritis <markos@freevec.org> 6// 7// This Source Code Form is subject to the terms of the Mozilla 8// Public License v. 2.0. If a copy of the MPL was not distributed 9// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. 10 11#ifndef EIGEN_COMPLEX_NEON_H 12#define EIGEN_COMPLEX_NEON_H 13 14namespace Eigen { 15 16namespace internal { 17 18inline uint32x4_t p4ui_CONJ_XOR() { 19// See bug 1325, clang fails to call vld1q_u64. 20#if EIGEN_COMP_CLANG 21 uint32x4_t ret = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 }; 22 return ret; 23#else 24 static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 }; 25 return vld1q_u32( conj_XOR_DATA ); 26#endif 27} 28 29inline uint32x2_t p2ui_CONJ_XOR() { 30 static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000 }; 31 return vld1_u32( conj_XOR_DATA ); 32} 33 34//---------- float ---------- 35struct Packet2cf 36{ 37 EIGEN_STRONG_INLINE Packet2cf() {} 38 EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {} 39 Packet4f v; 40}; 41 42template<> struct packet_traits<std::complex<float> > : default_packet_traits 43{ 44 typedef Packet2cf type; 45 typedef Packet2cf half; 46 enum { 47 Vectorizable = 1, 48 AlignedOnScalar = 1, 49 size = 2, 50 HasHalfPacket = 0, 51 52 HasAdd = 1, 53 HasSub = 1, 54 HasMul = 1, 55 HasDiv = 1, 56 HasNegate = 1, 57 HasAbs = 0, 58 HasAbs2 = 0, 59 HasMin = 0, 60 HasMax = 0, 61 HasSetLinear = 0 62 }; 63}; 64 65template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; 66 67template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) 68{ 69 float32x2_t r64; 70 r64 = vld1_f32((float *)&from); 71 72 return Packet2cf(vcombine_f32(r64, r64)); 73} 74 75template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(padd<Packet4f>(a.v,b.v)); } 76template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(psub<Packet4f>(a.v,b.v)); } 77template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate<Packet4f>(a.v)); } 78template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) 79{ 80 Packet4ui b = vreinterpretq_u32_f32(a.v); 81 return Packet2cf(vreinterpretq_f32_u32(veorq_u32(b, p4ui_CONJ_XOR()))); 82} 83 84template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 85{ 86 Packet4f v1, v2; 87 88 // Get the real values of a | a1_re | a1_re | a2_re | a2_re | 89 v1 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 0), vdup_lane_f32(vget_high_f32(a.v), 0)); 90 // Get the imag values of a | a1_im | a1_im | a2_im | a2_im | 91 v2 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 1), vdup_lane_f32(vget_high_f32(a.v), 1)); 92 // Multiply the real a with b 93 v1 = vmulq_f32(v1, b.v); 94 // Multiply the imag a with b 95 v2 = vmulq_f32(v2, b.v); 96 // Conjugate v2 97 v2 = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(v2), p4ui_CONJ_XOR())); 98 // Swap real/imag elements in v2. 99 v2 = vrev64q_f32(v2); 100 // Add and return the result 101 return Packet2cf(vaddq_f32(v1, v2)); 102} 103 104template<> EIGEN_STRONG_INLINE Packet2cf pand <Packet2cf>(const Packet2cf& a, const Packet2cf& b) 105{ 106 return Packet2cf(vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); 107} 108template<> EIGEN_STRONG_INLINE Packet2cf por <Packet2cf>(const Packet2cf& a, const Packet2cf& b) 109{ 110 return Packet2cf(vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); 111} 112template<> EIGEN_STRONG_INLINE Packet2cf pxor <Packet2cf>(const Packet2cf& a, const Packet2cf& b) 113{ 114 return Packet2cf(vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); 115} 116template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 117{ 118 return Packet2cf(vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); 119} 120 121template<> EIGEN_STRONG_INLINE Packet2cf pload<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>((const float*)from)); } 122template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>((const float*)from)); } 123 124template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); } 125 126template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); } 127template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); } 128 129template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather<std::complex<float>, Packet2cf>(const std::complex<float>* from, Index stride) 130{ 131 Packet4f res = pset1<Packet4f>(0.f); 132 res = vsetq_lane_f32(std::real(from[0*stride]), res, 0); 133 res = vsetq_lane_f32(std::imag(from[0*stride]), res, 1); 134 res = vsetq_lane_f32(std::real(from[1*stride]), res, 2); 135 res = vsetq_lane_f32(std::imag(from[1*stride]), res, 3); 136 return Packet2cf(res); 137} 138 139template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet2cf>(std::complex<float>* to, const Packet2cf& from, Index stride) 140{ 141 to[stride*0] = std::complex<float>(vgetq_lane_f32(from.v, 0), vgetq_lane_f32(from.v, 1)); 142 to[stride*1] = std::complex<float>(vgetq_lane_f32(from.v, 2), vgetq_lane_f32(from.v, 3)); 143} 144 145template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> * addr) { EIGEN_ARM_PREFETCH((float *)addr); } 146 147template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a) 148{ 149 std::complex<float> EIGEN_ALIGN16 x[2]; 150 vst1q_f32((float *)x, a.v); 151 return x[0]; 152} 153 154template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) 155{ 156 float32x2_t a_lo, a_hi; 157 Packet4f a_r128; 158 159 a_lo = vget_low_f32(a.v); 160 a_hi = vget_high_f32(a.v); 161 a_r128 = vcombine_f32(a_hi, a_lo); 162 163 return Packet2cf(a_r128); 164} 165 166template<> EIGEN_STRONG_INLINE Packet2cf pcplxflip<Packet2cf>(const Packet2cf& a) 167{ 168 return Packet2cf(vrev64q_f32(a.v)); 169} 170 171template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a) 172{ 173 float32x2_t a1, a2; 174 std::complex<float> s; 175 176 a1 = vget_low_f32(a.v); 177 a2 = vget_high_f32(a.v); 178 a2 = vadd_f32(a1, a2); 179 vst1_f32((float *)&s, a2); 180 181 return s; 182} 183 184template<> EIGEN_STRONG_INLINE Packet2cf preduxp<Packet2cf>(const Packet2cf* vecs) 185{ 186 Packet4f sum1, sum2, sum; 187 188 // Add the first two 64-bit float32x2_t of vecs[0] 189 sum1 = vcombine_f32(vget_low_f32(vecs[0].v), vget_low_f32(vecs[1].v)); 190 sum2 = vcombine_f32(vget_high_f32(vecs[0].v), vget_high_f32(vecs[1].v)); 191 sum = vaddq_f32(sum1, sum2); 192 193 return Packet2cf(sum); 194} 195 196template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a) 197{ 198 float32x2_t a1, a2, v1, v2, prod; 199 std::complex<float> s; 200 201 a1 = vget_low_f32(a.v); 202 a2 = vget_high_f32(a.v); 203 // Get the real values of a | a1_re | a1_re | a2_re | a2_re | 204 v1 = vdup_lane_f32(a1, 0); 205 // Get the real values of a | a1_im | a1_im | a2_im | a2_im | 206 v2 = vdup_lane_f32(a1, 1); 207 // Multiply the real a with b 208 v1 = vmul_f32(v1, a2); 209 // Multiply the imag a with b 210 v2 = vmul_f32(v2, a2); 211 // Conjugate v2 212 v2 = vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(v2), p2ui_CONJ_XOR())); 213 // Swap real/imag elements in v2. 214 v2 = vrev64_f32(v2); 215 // Add v1, v2 216 prod = vadd_f32(v1, v2); 217 218 vst1_f32((float *)&s, prod); 219 220 return s; 221} 222 223template<int Offset> 224struct palign_impl<Offset,Packet2cf> 225{ 226 EIGEN_STRONG_INLINE static void run(Packet2cf& first, const Packet2cf& second) 227 { 228 if (Offset==1) 229 { 230 first.v = vextq_f32(first.v, second.v, 2); 231 } 232 } 233}; 234 235template<> struct conj_helper<Packet2cf, Packet2cf, false,true> 236{ 237 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 238 { return padd(pmul(x,y),c); } 239 240 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 241 { 242 return internal::pmul(a, pconj(b)); 243 } 244}; 245 246template<> struct conj_helper<Packet2cf, Packet2cf, true,false> 247{ 248 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 249 { return padd(pmul(x,y),c); } 250 251 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 252 { 253 return internal::pmul(pconj(a), b); 254 } 255}; 256 257template<> struct conj_helper<Packet2cf, Packet2cf, true,true> 258{ 259 EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const 260 { return padd(pmul(x,y),c); } 261 262 EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const 263 { 264 return pconj(internal::pmul(a, b)); 265 } 266}; 267 268template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b) 269{ 270 // TODO optimize it for NEON 271 Packet2cf res = conj_helper<Packet2cf,Packet2cf,false,true>().pmul(a,b); 272 Packet4f s, rev_s; 273 274 // this computes the norm 275 s = vmulq_f32(b.v, b.v); 276 rev_s = vrev64q_f32(s); 277 278 return Packet2cf(pdiv(res.v, vaddq_f32(s,rev_s))); 279} 280 281EIGEN_DEVICE_FUNC inline void 282ptranspose(PacketBlock<Packet2cf,2>& kernel) { 283 Packet4f tmp = vcombine_f32(vget_high_f32(kernel.packet[0].v), vget_high_f32(kernel.packet[1].v)); 284 kernel.packet[0].v = vcombine_f32(vget_low_f32(kernel.packet[0].v), vget_low_f32(kernel.packet[1].v)); 285 kernel.packet[1].v = tmp; 286} 287 288//---------- double ---------- 289#if EIGEN_ARCH_ARM64 && !EIGEN_APPLE_DOUBLE_NEON_BUG 290 291// See bug 1325, clang fails to call vld1q_u64. 292#if EIGEN_COMP_CLANG 293 static uint64x2_t p2ul_CONJ_XOR = {0x0, 0x8000000000000000}; 294#else 295 const uint64_t p2ul_conj_XOR_DATA[] = { 0x0, 0x8000000000000000 }; 296 static uint64x2_t p2ul_CONJ_XOR = vld1q_u64( p2ul_conj_XOR_DATA ); 297#endif 298 299struct Packet1cd 300{ 301 EIGEN_STRONG_INLINE Packet1cd() {} 302 EIGEN_STRONG_INLINE explicit Packet1cd(const Packet2d& a) : v(a) {} 303 Packet2d v; 304}; 305 306template<> struct packet_traits<std::complex<double> > : default_packet_traits 307{ 308 typedef Packet1cd type; 309 typedef Packet1cd half; 310 enum { 311 Vectorizable = 1, 312 AlignedOnScalar = 0, 313 size = 1, 314 HasHalfPacket = 0, 315 316 HasAdd = 1, 317 HasSub = 1, 318 HasMul = 1, 319 HasDiv = 1, 320 HasNegate = 1, 321 HasAbs = 0, 322 HasAbs2 = 0, 323 HasMin = 0, 324 HasMax = 0, 325 HasSetLinear = 0 326 }; 327}; 328 329template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; 330 331template<> EIGEN_STRONG_INLINE Packet1cd pload<Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } 332template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } 333 334template<> 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 337template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(padd<Packet2d>(a.v,b.v)); } 338template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(psub<Packet2d>(a.v,b.v)); } 339template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate<Packet2d>(a.v)); } 340template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) { return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v), p2ul_CONJ_XOR))); } 341 342template<> EIGEN_STRONG_INLINE Packet1cd pmul<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 343{ 344 Packet2d v1, v2; 345 346 // Get the real values of a 347 v1 = vdupq_lane_f64(vget_low_f64(a.v), 0); 348 // Get the imag values of a 349 v2 = vdupq_lane_f64(vget_high_f64(a.v), 0); 350 // Multiply the real a with b 351 v1 = vmulq_f64(v1, b.v); 352 // Multiply the imag a with b 353 v2 = vmulq_f64(v2, b.v); 354 // Conjugate v2 355 v2 = vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(v2), p2ul_CONJ_XOR)); 356 // Swap real/imag elements in v2. 357 v2 = preverse<Packet2d>(v2); 358 // Add and return the result 359 return Packet1cd(vaddq_f64(v1, v2)); 360} 361 362template<> EIGEN_STRONG_INLINE Packet1cd pand <Packet1cd>(const Packet1cd& a, const Packet1cd& b) 363{ 364 return Packet1cd(vreinterpretq_f64_u64(vandq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); 365} 366template<> EIGEN_STRONG_INLINE Packet1cd por <Packet1cd>(const Packet1cd& a, const Packet1cd& b) 367{ 368 return Packet1cd(vreinterpretq_f64_u64(vorrq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); 369} 370template<> EIGEN_STRONG_INLINE Packet1cd pxor <Packet1cd>(const Packet1cd& a, const Packet1cd& b) 371{ 372 return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); 373} 374template<> EIGEN_STRONG_INLINE Packet1cd pandnot<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 375{ 376 return Packet1cd(vreinterpretq_f64_u64(vbicq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); 377} 378 379template<> EIGEN_STRONG_INLINE Packet1cd ploaddup<Packet1cd>(const std::complex<double>* from) { return pset1<Packet1cd>(*from); } 380 381template<> EIGEN_STRONG_INLINE void pstore <std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, from.v); } 382template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<double> >(std::complex<double> * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, from.v); } 383 384template<> EIGEN_STRONG_INLINE void prefetch<std::complex<double> >(const std::complex<double> * addr) { EIGEN_ARM_PREFETCH((double *)addr); } 385 386template<> EIGEN_DEVICE_FUNC inline Packet1cd pgather<std::complex<double>, Packet1cd>(const std::complex<double>* from, Index stride) 387{ 388 Packet2d res = pset1<Packet2d>(0.0); 389 res = vsetq_lane_f64(std::real(from[0*stride]), res, 0); 390 res = vsetq_lane_f64(std::imag(from[0*stride]), res, 1); 391 return Packet1cd(res); 392} 393 394template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<double>, Packet1cd>(std::complex<double>* to, const Packet1cd& from, Index stride) 395{ 396 to[stride*0] = std::complex<double>(vgetq_lane_f64(from.v, 0), vgetq_lane_f64(from.v, 1)); 397} 398 399 400template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet1cd>(const Packet1cd& a) 401{ 402 std::complex<double> EIGEN_ALIGN16 res; 403 pstore<std::complex<double> >(&res, a); 404 405 return res; 406} 407 408template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } 409 410template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet1cd>(const Packet1cd& a) { return pfirst(a); } 411 412template<> EIGEN_STRONG_INLINE Packet1cd preduxp<Packet1cd>(const Packet1cd* vecs) { return vecs[0]; } 413 414template<> EIGEN_STRONG_INLINE std::complex<double> predux_mul<Packet1cd>(const Packet1cd& a) { return pfirst(a); } 415 416template<int Offset> 417struct palign_impl<Offset,Packet1cd> 418{ 419 static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) 420 { 421 // FIXME is it sure we never have to align a Packet1cd? 422 // Even though a std::complex<double> has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... 423 } 424}; 425 426template<> struct conj_helper<Packet1cd, Packet1cd, false,true> 427{ 428 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 429 { return padd(pmul(x,y),c); } 430 431 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 432 { 433 return internal::pmul(a, pconj(b)); 434 } 435}; 436 437template<> struct conj_helper<Packet1cd, Packet1cd, true,false> 438{ 439 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 440 { return padd(pmul(x,y),c); } 441 442 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 443 { 444 return internal::pmul(pconj(a), b); 445 } 446}; 447 448template<> struct conj_helper<Packet1cd, Packet1cd, true,true> 449{ 450 EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const 451 { return padd(pmul(x,y),c); } 452 453 EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const 454 { 455 return pconj(internal::pmul(a, b)); 456 } 457}; 458 459template<> EIGEN_STRONG_INLINE Packet1cd pdiv<Packet1cd>(const Packet1cd& a, const Packet1cd& b) 460{ 461 // TODO optimize it for NEON 462 Packet1cd res = conj_helper<Packet1cd,Packet1cd,false,true>().pmul(a,b); 463 Packet2d s = pmul<Packet2d>(b.v, b.v); 464 Packet2d rev_s = preverse<Packet2d>(s); 465 466 return Packet1cd(pdiv(res.v, padd<Packet2d>(s,rev_s))); 467} 468 469EIGEN_STRONG_INLINE Packet1cd pcplxflip/*<Packet1cd>*/(const Packet1cd& x) 470{ 471 return Packet1cd(preverse(Packet2d(x.v))); 472} 473 474EIGEN_STRONG_INLINE void ptranspose(PacketBlock<Packet1cd,2>& kernel) 475{ 476 Packet2d tmp = vcombine_f64(vget_high_f64(kernel.packet[0].v), vget_high_f64(kernel.packet[1].v)); 477 kernel.packet[0].v = vcombine_f64(vget_low_f64(kernel.packet[0].v), vget_low_f64(kernel.packet[1].v)); 478 kernel.packet[1].v = tmp; 479} 480#endif // EIGEN_ARCH_ARM64 481 482} // end namespace internal 483 484} // end namespace Eigen 485 486#endif // EIGEN_COMPLEX_NEON_H 487