compbio
Complex.h
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 {
20  EIGEN_STRONG_INLINE Packet2cf() {}
21  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
Eigen::half
Definition: Half.h:76
Eigen::Aligned16
Data pointer is aligned on a 16 bytes boundary.
Definition: Constants.h:230
Eigen::internal::conj_helper
Definition: BlasUtil.h:61
Eigen::internal::unpacket_traits
Definition: XprHelper.h:158
Eigen
Namespace containing all symbols from the Eigen library.
Definition: bench_norm.cpp:85
std
Definition: Half.h:529
Eigen::internal::packet_traits
Definition: GenericPacketMath.h:96
Eigen::internal::Packet1cd
Definition: Complex.h:265
Eigen::internal::default_packet_traits
Definition: GenericPacketMath.h:42
Eigen::internal::conj_helper< Packet1cd, Packet1cd, false, true >
Definition: Complex.h:379
Eigen::Index
EIGEN_DEFAULT_DENSE_INDEX_TYPE Index
The Index type as used for the API.
Definition: Meta.h:33
Eigen::internal::PacketBlock
Definition: GenericPacketMath.h:539
internal
Definition: BandTriangularSolver.h:13
complex
Definition: datatypes.h:12
Eigen::internal::palign_impl
Definition: GenericPacketMath.h:492
Eigen::internal::Packet2cf
Definition: Complex.h:30
Eigen::internal::Selector
Definition: GenericPacketMath.h:552
Generated by   doxygen 1.8.12