Please, help us to better know about our user community by answering the following short survey: https://forms.gle/wpyrxWi18ox9Z5ae9
Eigen  3.4.0
 
Loading...
Searching...
No Matches
NEON/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// 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{
20// See bug 1325, clang fails to call vld1q_u64.
21#if EIGEN_COMP_CLANG || EIGEN_COMP_CASTXML
22 uint32x4_t ret = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 };
23 return ret;
24#else
25 static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 };
26 return vld1q_u32( conj_XOR_DATA );
27#endif
28}
29
30inline uint32x2_t p2ui_CONJ_XOR()
31{
32 static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000 };
33 return vld1_u32( conj_XOR_DATA );
34}
35
36//---------- float ----------
37
38struct Packet1cf
39{
40 EIGEN_STRONG_INLINE Packet1cf() {}
41 EIGEN_STRONG_INLINE explicit Packet1cf(const Packet2f& a) : v(a) {}
42 Packet2f v;
43};
44struct Packet2cf
45{
46 EIGEN_STRONG_INLINE Packet2cf() {}
47 EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {}
48 Packet4f v;
49};
50
51template<> struct packet_traits<std::complex<float> > : default_packet_traits
52{
53 typedef Packet2cf type;
54 typedef Packet1cf half;
55 enum
56 {
57 Vectorizable = 1,
58 AlignedOnScalar = 1,
59 size = 2,
60 HasHalfPacket = 1,
61
62 HasAdd = 1,
63 HasSub = 1,
64 HasMul = 1,
65 HasDiv = 1,
66 HasNegate = 1,
67 HasAbs = 0,
68 HasAbs2 = 0,
69 HasMin = 0,
70 HasMax = 0,
71 HasSetLinear = 0
72 };
73};
74
75template<> struct unpacket_traits<Packet1cf>
76{
77 typedef std::complex<float> type;
78 typedef Packet1cf half;
79 typedef Packet2f as_real;
80 enum
81 {
82 size = 1,
83 alignment = Aligned16,
84 vectorizable = true,
85 masked_load_available = false,
86 masked_store_available = false
87 };
88};
89template<> struct unpacket_traits<Packet2cf>
90{
91 typedef std::complex<float> type;
92 typedef Packet1cf half;
93 typedef Packet4f as_real;
94 enum
95 {
96 size = 2,
97 alignment = Aligned16,
98 vectorizable = true,
99 masked_load_available = false,
100 masked_store_available = false
101 };
102};
103
104template<> EIGEN_STRONG_INLINE Packet1cf pcast<float,Packet1cf>(const float& a)
105{ return Packet1cf(vset_lane_f32(a, vdup_n_f32(0.f), 0)); }
106template<> EIGEN_STRONG_INLINE Packet2cf pcast<Packet2f,Packet2cf>(const Packet2f& a)
107{ return Packet2cf(vreinterpretq_f32_u64(vmovl_u32(vreinterpret_u32_f32(a)))); }
108
109template<> EIGEN_STRONG_INLINE Packet1cf pset1<Packet1cf>(const std::complex<float>& from)
110{ return Packet1cf(vld1_f32(reinterpret_cast<const float*>(&from))); }
111template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from)
112{
113 const float32x2_t r64 = vld1_f32(reinterpret_cast<const float*>(&from));
114 return Packet2cf(vcombine_f32(r64, r64));
115}
116
117template<> EIGEN_STRONG_INLINE Packet1cf padd<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
118{ return Packet1cf(padd<Packet2f>(a.v, b.v)); }
119template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
120{ return Packet2cf(padd<Packet4f>(a.v, b.v)); }
121
122template<> EIGEN_STRONG_INLINE Packet1cf psub<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
123{ return Packet1cf(psub<Packet2f>(a.v, b.v)); }
124template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
125{ return Packet2cf(psub<Packet4f>(a.v, b.v)); }
126
127template<> EIGEN_STRONG_INLINE Packet1cf pnegate(const Packet1cf& a) { return Packet1cf(pnegate<Packet2f>(a.v)); }
128template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate<Packet4f>(a.v)); }
129
130template<> EIGEN_STRONG_INLINE Packet1cf pconj(const Packet1cf& a)
131{
132 const Packet2ui b = vreinterpret_u32_f32(a.v);
133 return Packet1cf(vreinterpret_f32_u32(veor_u32(b, p2ui_CONJ_XOR())));
134}
135template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a)
136{
137 const Packet4ui b = vreinterpretq_u32_f32(a.v);
138 return Packet2cf(vreinterpretq_f32_u32(veorq_u32(b, p4ui_CONJ_XOR())));
139}
140
141template<> EIGEN_STRONG_INLINE Packet1cf pmul<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
142{
143 Packet2f v1, v2;
144
145 // Get the real values of a | a1_re | a1_re |
146 v1 = vdup_lane_f32(a.v, 0);
147 // Get the imag values of a | a1_im | a1_im |
148 v2 = vdup_lane_f32(a.v, 1);
149 // Multiply the real a with b
150 v1 = vmul_f32(v1, b.v);
151 // Multiply the imag a with b
152 v2 = vmul_f32(v2, b.v);
153 // Conjugate v2
154 v2 = vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(v2), p2ui_CONJ_XOR()));
155 // Swap real/imag elements in v2.
156 v2 = vrev64_f32(v2);
157 // Add and return the result
158 return Packet1cf(vadd_f32(v1, v2));
159}
160template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
161{
162 Packet4f v1, v2;
163
164 // Get the real values of a | a1_re | a1_re | a2_re | a2_re |
165 v1 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 0), vdup_lane_f32(vget_high_f32(a.v), 0));
166 // Get the imag values of a | a1_im | a1_im | a2_im | a2_im |
167 v2 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 1), vdup_lane_f32(vget_high_f32(a.v), 1));
168 // Multiply the real a with b
169 v1 = vmulq_f32(v1, b.v);
170 // Multiply the imag a with b
171 v2 = vmulq_f32(v2, b.v);
172 // Conjugate v2
173 v2 = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(v2), p4ui_CONJ_XOR()));
174 // Swap real/imag elements in v2.
175 v2 = vrev64q_f32(v2);
176 // Add and return the result
177 return Packet2cf(vaddq_f32(v1, v2));
178}
179
180template<> EIGEN_STRONG_INLINE Packet1cf pcmp_eq(const Packet1cf& a, const Packet1cf& b)
181{
182 // Compare real and imaginary parts of a and b to get the mask vector:
183 // [re(a[0])==re(b[0]), im(a[0])==im(b[0])]
184 Packet2f eq = pcmp_eq<Packet2f>(a.v, b.v);
185 // Swap real/imag elements in the mask in to get:
186 // [im(a[0])==im(b[0]), re(a[0])==re(b[0])]
187 Packet2f eq_swapped = vrev64_f32(eq);
188 // Return re(a)==re(b) && im(a)==im(b) by computing bitwise AND of eq and eq_swapped
189 return Packet1cf(pand<Packet2f>(eq, eq_swapped));
190}
191template<> EIGEN_STRONG_INLINE Packet2cf pcmp_eq(const Packet2cf& a, const Packet2cf& b)
192{
193 // Compare real and imaginary parts of a and b to get the mask vector:
194 // [re(a[0])==re(b[0]), im(a[0])==im(b[0]), re(a[1])==re(b[1]), im(a[1])==im(b[1])]
195 Packet4f eq = pcmp_eq<Packet4f>(a.v, b.v);
196 // Swap real/imag elements in the mask in to get:
197 // [im(a[0])==im(b[0]), re(a[0])==re(b[0]), im(a[1])==im(b[1]), re(a[1])==re(b[1])]
198 Packet4f eq_swapped = vrev64q_f32(eq);
199 // Return re(a)==re(b) && im(a)==im(b) by computing bitwise AND of eq and eq_swapped
200 return Packet2cf(pand<Packet4f>(eq, eq_swapped));
201}
202
203template<> EIGEN_STRONG_INLINE Packet1cf pand<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
204{ return Packet1cf(vreinterpret_f32_u32(vand_u32(vreinterpret_u32_f32(a.v), vreinterpret_u32_f32(b.v)))); }
205template<> EIGEN_STRONG_INLINE Packet2cf pand<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
206{ return Packet2cf(vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(a.v), vreinterpretq_u32_f32(b.v)))); }
207
208template<> EIGEN_STRONG_INLINE Packet1cf por<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
209{ return Packet1cf(vreinterpret_f32_u32(vorr_u32(vreinterpret_u32_f32(a.v), vreinterpret_u32_f32(b.v)))); }
210template<> EIGEN_STRONG_INLINE Packet2cf por<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
211{ return Packet2cf(vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a.v), vreinterpretq_u32_f32(b.v)))); }
212
213template<> EIGEN_STRONG_INLINE Packet1cf pxor<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
214{ return Packet1cf(vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(a.v), vreinterpret_u32_f32(b.v)))); }
215template<> EIGEN_STRONG_INLINE Packet2cf pxor<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
216{ return Packet2cf(vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a.v), vreinterpretq_u32_f32(b.v)))); }
217
218template<> EIGEN_STRONG_INLINE Packet1cf pandnot<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
219{ return Packet1cf(vreinterpret_f32_u32(vbic_u32(vreinterpret_u32_f32(a.v), vreinterpret_u32_f32(b.v)))); }
220template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
221{ return Packet2cf(vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a.v), vreinterpretq_u32_f32(b.v)))); }
222
223template<> EIGEN_STRONG_INLINE Packet1cf pload<Packet1cf>(const std::complex<float>* from)
224{ EIGEN_DEBUG_ALIGNED_LOAD return Packet1cf(pload<Packet2f>((const float*)from)); }
225template<> EIGEN_STRONG_INLINE Packet2cf pload<Packet2cf>(const std::complex<float>* from)
226{ EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(reinterpret_cast<const float*>(from))); }
227
228template<> EIGEN_STRONG_INLINE Packet1cf ploadu<Packet1cf>(const std::complex<float>* from)
229{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cf(ploadu<Packet2f>((const float*)from)); }
230template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from)
231{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(reinterpret_cast<const float*>(from))); }
232
233template<> EIGEN_STRONG_INLINE Packet1cf ploaddup<Packet1cf>(const std::complex<float>* from)
234{ return pset1<Packet1cf>(*from); }
235template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from)
236{ return pset1<Packet2cf>(*from); }
237
238template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *to, const Packet1cf& from)
239{ EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); }
240template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *to, const Packet2cf& from)
241{ EIGEN_DEBUG_ALIGNED_STORE pstore(reinterpret_cast<float*>(to), from.v); }
242
243template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *to, const Packet1cf& from)
244{ EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); }
245template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *to, const Packet2cf& from)
246{ EIGEN_DEBUG_UNALIGNED_STORE pstoreu(reinterpret_cast<float*>(to), from.v); }
247
248template<> EIGEN_DEVICE_FUNC inline Packet1cf pgather<std::complex<float>, Packet1cf>(
249 const std::complex<float>* from, Index stride)
250{
251 const Packet2f tmp = vdup_n_f32(std::real(from[0*stride]));
252 return Packet1cf(vset_lane_f32(std::imag(from[0*stride]), tmp, 1));
253}
254template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather<std::complex<float>, Packet2cf>(
255 const std::complex<float>* from, Index stride)
256{
257 Packet4f res = vdupq_n_f32(std::real(from[0*stride]));
258 res = vsetq_lane_f32(std::imag(from[0*stride]), res, 1);
259 res = vsetq_lane_f32(std::real(from[1*stride]), res, 2);
260 res = vsetq_lane_f32(std::imag(from[1*stride]), res, 3);
261 return Packet2cf(res);
262}
263
264template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet1cf>(
265 std::complex<float>* to, const Packet1cf& from, Index stride)
266{ to[stride*0] = std::complex<float>(vget_lane_f32(from.v, 0), vget_lane_f32(from.v, 1)); }
267template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet2cf>(
268 std::complex<float>* to, const Packet2cf& from, Index stride)
269{
270 to[stride*0] = std::complex<float>(vgetq_lane_f32(from.v, 0), vgetq_lane_f32(from.v, 1));
271 to[stride*1] = std::complex<float>(vgetq_lane_f32(from.v, 2), vgetq_lane_f32(from.v, 3));
272}
273
274template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> *addr)
275{ EIGEN_ARM_PREFETCH(reinterpret_cast<const float*>(addr)); }
276
277template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet1cf>(const Packet1cf& a)
278{
279 EIGEN_ALIGN16 std::complex<float> x;
280 vst1_f32(reinterpret_cast<float*>(&x), a.v);
281 return x;
282}
283template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a)
284{
285 EIGEN_ALIGN16 std::complex<float> x[2];
286 vst1q_f32(reinterpret_cast<float*>(x), a.v);
287 return x[0];
288}
289
290template<> EIGEN_STRONG_INLINE Packet1cf preverse(const Packet1cf& a) { return a; }
291template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a)
292{ return Packet2cf(vcombine_f32(vget_high_f32(a.v), vget_low_f32(a.v))); }
293
294template<> EIGEN_STRONG_INLINE Packet1cf pcplxflip<Packet1cf>(const Packet1cf& a)
295{ return Packet1cf(vrev64_f32(a.v)); }
296template<> EIGEN_STRONG_INLINE Packet2cf pcplxflip<Packet2cf>(const Packet2cf& a)
297{ return Packet2cf(vrev64q_f32(a.v)); }
298
299template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet1cf>(const Packet1cf& a)
300{
301 std::complex<float> s;
302 vst1_f32((float *)&s, a.v);
303 return s;
304}
305template<> EIGEN_STRONG_INLINE std::complex<float> predux<Packet2cf>(const Packet2cf& a)
306{
307 std::complex<float> s;
308 vst1_f32(reinterpret_cast<float*>(&s), vadd_f32(vget_low_f32(a.v), vget_high_f32(a.v)));
309 return s;
310}
311
312template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet1cf>(const Packet1cf& a)
313{
314 std::complex<float> s;
315 vst1_f32((float *)&s, a.v);
316 return s;
317}
318template<> EIGEN_STRONG_INLINE std::complex<float> predux_mul<Packet2cf>(const Packet2cf& a)
319{
320 float32x2_t a1, a2, v1, v2, prod;
321 std::complex<float> s;
322
323 a1 = vget_low_f32(a.v);
324 a2 = vget_high_f32(a.v);
325 // Get the real values of a | a1_re | a1_re | a2_re | a2_re |
326 v1 = vdup_lane_f32(a1, 0);
327 // Get the real values of a | a1_im | a1_im | a2_im | a2_im |
328 v2 = vdup_lane_f32(a1, 1);
329 // Multiply the real a with b
330 v1 = vmul_f32(v1, a2);
331 // Multiply the imag a with b
332 v2 = vmul_f32(v2, a2);
333 // Conjugate v2
334 v2 = vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(v2), p2ui_CONJ_XOR()));
335 // Swap real/imag elements in v2.
336 v2 = vrev64_f32(v2);
337 // Add v1, v2
338 prod = vadd_f32(v1, v2);
339
340 vst1_f32(reinterpret_cast<float*>(&s), prod);
341
342 return s;
343}
344
345EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cf,Packet2f)
346EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f)
347
348template<> EIGEN_STRONG_INLINE Packet1cf pdiv<Packet1cf>(const Packet1cf& a, const Packet1cf& b)
349{
350 // TODO optimize it for NEON
351 Packet1cf res = pmul(a, pconj(b));
352 Packet2f s, rev_s;
353
354 // this computes the norm
355 s = vmul_f32(b.v, b.v);
356 rev_s = vrev64_f32(s);
357
358 return Packet1cf(pdiv<Packet2f>(res.v, vadd_f32(s, rev_s)));
359}
360template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
361{
362 // TODO optimize it for NEON
363 Packet2cf res = pmul(a,pconj(b));
364 Packet4f s, rev_s;
365
366 // this computes the norm
367 s = vmulq_f32(b.v, b.v);
368 rev_s = vrev64q_f32(s);
369
370 return Packet2cf(pdiv<Packet4f>(res.v, vaddq_f32(s, rev_s)));
371}
372
373EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock<Packet1cf, 1>& /*kernel*/) {}
374EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock<Packet2cf, 2>& kernel)
375{
376 Packet4f tmp = vcombine_f32(vget_high_f32(kernel.packet[0].v), vget_high_f32(kernel.packet[1].v));
377 kernel.packet[0].v = vcombine_f32(vget_low_f32(kernel.packet[0].v), vget_low_f32(kernel.packet[1].v));
378 kernel.packet[1].v = tmp;
379}
380
381template<> EIGEN_STRONG_INLINE Packet1cf psqrt<Packet1cf>(const Packet1cf& a) {
382 return psqrt_complex<Packet1cf>(a);
383}
384
385template<> EIGEN_STRONG_INLINE Packet2cf psqrt<Packet2cf>(const Packet2cf& a) {
386 return psqrt_complex<Packet2cf>(a);
387}
388
389//---------- double ----------
390#if EIGEN_ARCH_ARM64 && !EIGEN_APPLE_DOUBLE_NEON_BUG
391
392// See bug 1325, clang fails to call vld1q_u64.
393#if EIGEN_COMP_CLANG || EIGEN_COMP_CASTXML
394 static uint64x2_t p2ul_CONJ_XOR = {0x0, 0x8000000000000000};
395#else
396 const uint64_t p2ul_conj_XOR_DATA[] = { 0x0, 0x8000000000000000 };
397 static uint64x2_t p2ul_CONJ_XOR = vld1q_u64( p2ul_conj_XOR_DATA );
398#endif
399
400struct Packet1cd
401{
402 EIGEN_STRONG_INLINE Packet1cd() {}
403 EIGEN_STRONG_INLINE explicit Packet1cd(const Packet2d& a) : v(a) {}
404 Packet2d v;
405};
406
407template<> struct packet_traits<std::complex<double> > : default_packet_traits
408{
409 typedef Packet1cd type;
410 typedef Packet1cd half;
411 enum
412 {
413 Vectorizable = 1,
414 AlignedOnScalar = 0,
415 size = 1,
416 HasHalfPacket = 0,
417
418 HasAdd = 1,
419 HasSub = 1,
420 HasMul = 1,
421 HasDiv = 1,
422 HasNegate = 1,
423 HasAbs = 0,
424 HasAbs2 = 0,
425 HasMin = 0,
426 HasMax = 0,
427 HasSetLinear = 0
428 };
429};
430
431template<> struct unpacket_traits<Packet1cd>
432{
433 typedef std::complex<double> type;
434 typedef Packet1cd half;
435 typedef Packet2d as_real;
436 enum
437 {
438 size=1,
439 alignment=Aligned16,
440 vectorizable=true,
441 masked_load_available=false,
442 masked_store_available=false
443 };
444};
445
446template<> EIGEN_STRONG_INLINE Packet1cd pload<Packet1cd>(const std::complex<double>* from)
447{ EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>(reinterpret_cast<const double*>(from))); }
448
449template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from)
450{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>(reinterpret_cast<const double*>(from))); }
451
452template<> EIGEN_STRONG_INLINE Packet1cd pset1<Packet1cd>(const std::complex<double>& from)
453{
454 /* here we really have to use unaligned loads :( */
455 return ploadu<Packet1cd>(&from);
456}
457
458template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
459{ return Packet1cd(padd<Packet2d>(a.v, b.v)); }
460
461template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
462{ return Packet1cd(psub<Packet2d>(a.v, b.v)); }
463
464template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a)
465{ return Packet1cd(pnegate<Packet2d>(a.v)); }
466
467template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a)
468{ return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v), p2ul_CONJ_XOR))); }
469
470template<> EIGEN_STRONG_INLINE Packet1cd pmul<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
471{
472 Packet2d v1, v2;
473
474 // Get the real values of a
475 v1 = vdupq_lane_f64(vget_low_f64(a.v), 0);
476 // Get the imag values of a
477 v2 = vdupq_lane_f64(vget_high_f64(a.v), 0);
478 // Multiply the real a with b
479 v1 = vmulq_f64(v1, b.v);
480 // Multiply the imag a with b
481 v2 = vmulq_f64(v2, b.v);
482 // Conjugate v2
483 v2 = vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(v2), p2ul_CONJ_XOR));
484 // Swap real/imag elements in v2.
485 v2 = preverse<Packet2d>(v2);
486 // Add and return the result
487 return Packet1cd(vaddq_f64(v1, v2));
488}
489
490template<> EIGEN_STRONG_INLINE Packet1cd pcmp_eq(const Packet1cd& a, const Packet1cd& b)
491{
492 // Compare real and imaginary parts of a and b to get the mask vector:
493 // [re(a)==re(b), im(a)==im(b)]
494 Packet2d eq = pcmp_eq<Packet2d>(a.v, b.v);
495 // Swap real/imag elements in the mask in to get:
496 // [im(a)==im(b), re(a)==re(b)]
497 Packet2d eq_swapped = vreinterpretq_f64_u32(vrev64q_u32(vreinterpretq_u32_f64(eq)));
498 // Return re(a)==re(b) & im(a)==im(b) by computing bitwise AND of eq and eq_swapped
499 return Packet1cd(pand<Packet2d>(eq, eq_swapped));
500}
501
502template<> EIGEN_STRONG_INLINE Packet1cd pand<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
503{ return Packet1cd(vreinterpretq_f64_u64(vandq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); }
504
505template<> EIGEN_STRONG_INLINE Packet1cd por<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
506{ return Packet1cd(vreinterpretq_f64_u64(vorrq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); }
507
508template<> EIGEN_STRONG_INLINE Packet1cd pxor<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
509{ return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); }
510
511template<> EIGEN_STRONG_INLINE Packet1cd pandnot<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
512{ return Packet1cd(vreinterpretq_f64_u64(vbicq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); }
513
514template<> EIGEN_STRONG_INLINE Packet1cd ploaddup<Packet1cd>(const std::complex<double>* from)
515{ return pset1<Packet1cd>(*from); }
516
517template<> EIGEN_STRONG_INLINE void pstore <std::complex<double> >(std::complex<double> *to, const Packet1cd& from)
518{ EIGEN_DEBUG_ALIGNED_STORE pstore(reinterpret_cast<double*>(to), from.v); }
519
520template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<double> >(std::complex<double> *to, const Packet1cd& from)
521{ EIGEN_DEBUG_UNALIGNED_STORE pstoreu(reinterpret_cast<double*>(to), from.v); }
522
523template<> EIGEN_STRONG_INLINE void prefetch<std::complex<double> >(const std::complex<double> *addr)
524{ EIGEN_ARM_PREFETCH(reinterpret_cast<const double*>(addr)); }
525
526template<> EIGEN_DEVICE_FUNC inline Packet1cd pgather<std::complex<double>, Packet1cd>(
527 const std::complex<double>* from, Index stride)
528{
529 Packet2d res = pset1<Packet2d>(0.0);
530 res = vsetq_lane_f64(std::real(from[0*stride]), res, 0);
531 res = vsetq_lane_f64(std::imag(from[0*stride]), res, 1);
532 return Packet1cd(res);
533}
534
535template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<double>, Packet1cd>(
536 std::complex<double>* to, const Packet1cd& from, Index stride)
537{ to[stride*0] = std::complex<double>(vgetq_lane_f64(from.v, 0), vgetq_lane_f64(from.v, 1)); }
538
539template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet1cd>(const Packet1cd& a)
540{
541 EIGEN_ALIGN16 std::complex<double> res;
542 pstore<std::complex<double> >(&res, a);
543 return res;
544}
545
546template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; }
547
548template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet1cd>(const Packet1cd& a) { return pfirst(a); }
549
550template<> EIGEN_STRONG_INLINE std::complex<double> predux_mul<Packet1cd>(const Packet1cd& a) { return pfirst(a); }
551
552EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d)
553
554template<> EIGEN_STRONG_INLINE Packet1cd pdiv<Packet1cd>(const Packet1cd& a, const Packet1cd& b)
555{
556 // TODO optimize it for NEON
557 Packet1cd res = pmul(a,pconj(b));
558 Packet2d s = pmul<Packet2d>(b.v, b.v);
559 Packet2d rev_s = preverse<Packet2d>(s);
560
561 return Packet1cd(pdiv(res.v, padd<Packet2d>(s,rev_s)));
562}
563
564EIGEN_STRONG_INLINE Packet1cd pcplxflip/*<Packet1cd>*/(const Packet1cd& x)
565{ return Packet1cd(preverse(Packet2d(x.v))); }
566
567EIGEN_STRONG_INLINE void ptranspose(PacketBlock<Packet1cd,2>& kernel)
568{
569 Packet2d tmp = vcombine_f64(vget_high_f64(kernel.packet[0].v), vget_high_f64(kernel.packet[1].v));
570 kernel.packet[0].v = vcombine_f64(vget_low_f64(kernel.packet[0].v), vget_low_f64(kernel.packet[1].v));
571 kernel.packet[1].v = tmp;
572}
573
574template<> EIGEN_STRONG_INLINE Packet1cd psqrt<Packet1cd>(const Packet1cd& a) {
575 return psqrt_complex<Packet1cd>(a);
576}
577
578#endif // EIGEN_ARCH_ARM64
579
580} // end namespace internal
581
582} // end namespace Eigen
583
584#endif // EIGEN_COMPLEX_NEON_H
@ Aligned16
Definition: Constants.h:235
Namespace containing all symbols from the Eigen library.
Definition: Core:141
EIGEN_DEFAULT_DENSE_INDEX_TYPE Index
The Index type as used for the API.
Definition: Meta.h:74