Complex.h 9.04 KB
Newer Older
LM's avatar
LM committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2010 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// Eigen is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 3 of the License, or (at your option) any later version.
//
// Alternatively, you can redistribute it and/or
// modify it under the terms of the GNU General Public License as
// published by the Free Software Foundation; either version 2 of
// the License, or (at your option) any later version.
//
// Eigen is distributed in the hope that it will be useful, but WITHOUT ANY
// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
// FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License or the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License and a copy of the GNU General Public License along with
// Eigen. If not, see <http://www.gnu.org/licenses/>.

#ifndef EIGEN_COMPLEX_NEON_H
#define EIGEN_COMPLEX_NEON_H

namespace internal {

static uint32x4_t p4ui_CONJ_XOR = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 };
static uint32x2_t p2ui_CONJ_XOR = { 0x00000000, 0x80000000 };

//---------- float ----------
struct Packet2cf
{
  EIGEN_STRONG_INLINE Packet2cf() {}
  EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {}
  Packet4f  v;
};

template<> struct packet_traits<std::complex<float> >  : default_packet_traits
{
  typedef Packet2cf type;
  enum {
    Vectorizable = 1,
    AlignedOnScalar = 1,
    size = 2,

    HasAdd    = 1,
    HasSub    = 1,
    HasMul    = 1,
    HasDiv    = 1,
    HasNegate = 1,
    HasAbs    = 0,
    HasAbs2   = 0,
    HasMin    = 0,
    HasMax    = 0,
    HasSetLinear = 0
  };
};

template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; };

template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>&  from)
{
  float32x2_t r64;
  r64 = vld1_f32((float *)&from);

  return Packet2cf(vcombine_f32(r64, r64));
}

template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(padd<Packet4f>(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(psub<Packet4f>(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate<Packet4f>(a.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a)
{
  Packet4ui b = vreinterpretq_u32_f32(a.v);
  return Packet2cf(vreinterpretq_f32_u32(veorq_u32(b, p4ui_CONJ_XOR)));
}

template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  Packet4f v1, v2;
  float32x2_t a_lo, a_hi;

  // Get the real values of a | a1_re | a1_re | a2_re | a2_re |
  v1 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 0), vdup_lane_f32(vget_high_f32(a.v), 0));
  // Get the real values of a | a1_im | a1_im | a2_im | a2_im |
  v2 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 1), vdup_lane_f32(vget_high_f32(a.v), 1));
  // Multiply the real a with b
  v1 = vmulq_f32(v1, b.v);
  // Multiply the imag a with b
  v2 = vmulq_f32(v2, b.v);
  // Conjugate v2 
  v2 = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(v2), p4ui_CONJ_XOR));
  // Swap real/imag elements in v2.
  a_lo = vrev64_f32(vget_low_f32(v2));
  a_hi = vrev64_f32(vget_high_f32(v2));
  v2 = vcombine_f32(a_lo, a_hi);
  // Add and return the result
  return Packet2cf(vaddq_f32(v1, v2));
}

template<> EIGEN_STRONG_INLINE Packet2cf pand   <Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  return Packet2cf(vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v))));
}
template<> EIGEN_STRONG_INLINE Packet2cf por    <Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  return Packet2cf(vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v))));
}
template<> EIGEN_STRONG_INLINE Packet2cf pxor   <Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  return Packet2cf(vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v))));
}
template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  return Packet2cf(vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v))));
}

template<> EIGEN_STRONG_INLINE Packet2cf pload<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>((const float*)from)); }
template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>((const float*)from)); }

template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); }

template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); }
template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); }

template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> *   addr) { __pld((float *)addr); }

template<> EIGEN_STRONG_INLINE std::complex<float>  pfirst<Packet2cf>(const Packet2cf& a)
{
  std::complex<float> EIGEN_ALIGN16 x[2];
  vst1q_f32((float *)x, a.v);
  return x[0];
}

template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a)
{
  float32x2_t a_lo, a_hi;
  Packet4f a_r128;

  a_lo = vget_low_f32(a.v);
  a_hi = vget_high_f32(a.v);
  a_r128 = vcombine_f32(a_hi, a_lo);

  return Packet2cf(a_r128);
}

template<> EIGEN_STRONG_INLINE Packet2cf pcplxflip<Packet2cf>(const Packet2cf& a)
{
  return Packet2cf(vrev64q_f32(a.v));
}

template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a)
{
  float32x2_t a1, a2;
  std::complex<float> s;

  a1 = vget_low_f32(a.v);
  a2 = vget_high_f32(a.v);
  a2 = vadd_f32(a1, a2);
  vst1_f32((float *)&s, a2);

  return s;
}

template<> EIGEN_STRONG_INLINE Packet2cf preduxp<Packet2cf>(const Packet2cf* vecs)
{
  Packet4f sum1, sum2, sum;

  // Add the first two 64-bit float32x2_t of vecs[0]
  sum1 = vcombine_f32(vget_low_f32(vecs[0].v), vget_low_f32(vecs[1].v));
  sum2 = vcombine_f32(vget_high_f32(vecs[0].v), vget_high_f32(vecs[1].v));
  sum = vaddq_f32(sum1, sum2);

  return Packet2cf(sum);
}

template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a)
{
  float32x2_t a1, a2, v1, v2, prod;
  std::complex<float> s;

  a1 = vget_low_f32(a.v);
  a2 = vget_high_f32(a.v);
   // Get the real values of a | a1_re | a1_re | a2_re | a2_re |
  v1 = vdup_lane_f32(a1, 0);
  // Get the real values of a | a1_im | a1_im | a2_im | a2_im |
  v2 = vdup_lane_f32(a1, 1);
  // Multiply the real a with b
  v1 = vmul_f32(v1, a2);
  // Multiply the imag a with b
  v2 = vmul_f32(v2, a2);
  // Conjugate v2 
  v2 = vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(v2), p2ui_CONJ_XOR));
  // Swap real/imag elements in v2.
  v2 = vrev64_f32(v2);
  // Add v1, v2
  prod = vadd_f32(v1, v2);

  vst1_f32((float *)&s, prod);

  return s;
}

template<int Offset>
struct palign_impl<Offset,Packet2cf>
{
  EIGEN_STRONG_INLINE static void run(Packet2cf& first, const Packet2cf& second)
  {
    if (Offset==1)
    {
      first.v = vextq_f32(first.v, second.v, 2);
    }
  }
};

template<> struct conj_helper<Packet2cf, Packet2cf, false,true>
{
  EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const
  { return padd(pmul(x,y),c); }

  EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const
  {
    return internal::pmul(a, pconj(b));
  }
};

template<> struct conj_helper<Packet2cf, Packet2cf, true,false>
{
  EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const
  { return padd(pmul(x,y),c); }

  EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const
  {
    return internal::pmul(pconj(a), b);
  }
};

template<> struct conj_helper<Packet2cf, Packet2cf, true,true>
{
  EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const
  { return padd(pmul(x,y),c); }

  EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const
  {
    return pconj(internal::pmul(a, b));
  }
};

template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
  // TODO optimize it for AltiVec
  Packet2cf res = conj_helper<Packet2cf,Packet2cf,false,true>().pmul(a,b);
  Packet4f s, rev_s;
  float32x2_t a_lo, a_hi;

  // this computes the norm
  s = vmulq_f32(b.v, b.v);
  a_lo = vrev64_f32(vget_low_f32(s));
  a_hi = vrev64_f32(vget_high_f32(s));
  rev_s = vcombine_f32(a_lo, a_hi);

  return Packet2cf(pdiv(res.v, vaddq_f32(s,rev_s)));
}

} // end namespace internal

#endif // EIGEN_COMPLEX_NEON_H