NEON/PacketMath.h
Go to the documentation of this file.
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2008-2009 Gael Guennebaud <gael.guennebaud@inria.fr>
5 // Copyright (C) 2010 Konstantinos Margaritis <markos@codex.gr>
6 // Heavily based on Gael's SSE version.
7 //
8 // This Source Code Form is subject to the terms of the Mozilla
9 // Public License v. 2.0. If a copy of the MPL was not distributed
10 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
11 
12 #ifndef EIGEN_PACKET_MATH_NEON_H
13 #define EIGEN_PACKET_MATH_NEON_H
14 
15 namespace Eigen {
16 
17 namespace internal {
18 
19 #ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD
20 #define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8
21 #endif
22 
23 // FIXME NEON has 16 quad registers, but since the current register allocator
24 // is so bad, it is much better to reduce it to 8
25 #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS
26 #define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 8
27 #endif
28 
29 typedef float32x4_t Packet4f;
30 typedef int32x4_t Packet4i;
31 typedef uint32x4_t Packet4ui;
32 
33 #define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \
34  const Packet4f p4f_##NAME = pset1<Packet4f>(X)
35 
36 #define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \
37  const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1<int>(X))
38 
39 #define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \
40  const Packet4i p4i_##NAME = pset1<Packet4i>(X)
41 
42 #if defined(__llvm__) && !defined(__clang__)
43  //Special treatment for Apple's llvm-gcc, its NEON packet types are unions
44  #define EIGEN_INIT_NEON_PACKET2(X, Y) {{X, Y}}
45  #define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W) {{X, Y, Z, W}}
46 #else
47  //Default initializer for packets
48  #define EIGEN_INIT_NEON_PACKET2(X, Y) {X, Y}
49  #define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W) {X, Y, Z, W}
50 #endif
51 
52 #ifndef __pld
53 #define __pld(x) asm volatile ( " pld [%[addr]]\n" :: [addr] "r" (x) : "cc" );
54 #endif
55 
56 template<> struct packet_traits<float> : default_packet_traits
57 {
58  typedef Packet4f type;
59  enum {
60  Vectorizable = 1,
61  AlignedOnScalar = 1,
62  size = 4,
63 
64  HasDiv = 1,
65  // FIXME check the Has*
66  HasSin = 0,
67  HasCos = 0,
68  HasLog = 0,
69  HasExp = 0,
70  HasSqrt = 0
71  };
72 };
73 template<> struct packet_traits<int> : default_packet_traits
74 {
75  typedef Packet4i type;
76  enum {
77  Vectorizable = 1,
78  AlignedOnScalar = 1,
79  size=4
80  // FIXME check the Has*
81  };
82 };
83 
84 #if EIGEN_GNUC_AT_MOST(4,4) && !defined(__llvm__)
85 // workaround gcc 4.2, 4.3 and 4.4 compilatin issue
86 EIGEN_STRONG_INLINE float32x4_t vld1q_f32(const float* x) { return ::vld1q_f32((const float32_t*)x); }
87 EIGEN_STRONG_INLINE float32x2_t vld1_f32 (const float* x) { return ::vld1_f32 ((const float32_t*)x); }
88 EIGEN_STRONG_INLINE void vst1q_f32(float* to, float32x4_t from) { ::vst1q_f32((float32_t*)to,from); }
89 EIGEN_STRONG_INLINE void vst1_f32 (float* to, float32x2_t from) { ::vst1_f32 ((float32_t*)to,from); }
90 #endif
91 
92 template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4}; };
93 template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4}; };
94 
95 template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) { return vdupq_n_f32(from); }
96 template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return vdupq_n_s32(from); }
97 
98 template<> EIGEN_STRONG_INLINE Packet4f plset<float>(const float& a)
99 {
100  Packet4f countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
101  return vaddq_f32(pset1<Packet4f>(a), countdown);
102 }
103 template<> EIGEN_STRONG_INLINE Packet4i plset<int>(const int& a)
104 {
105  Packet4i countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
106  return vaddq_s32(pset1<Packet4i>(a), countdown);
107 }
108 
109 template<> EIGEN_STRONG_INLINE Packet4f padd<Packet4f>(const Packet4f& a, const Packet4f& b) { return vaddq_f32(a,b); }
110 template<> EIGEN_STRONG_INLINE Packet4i padd<Packet4i>(const Packet4i& a, const Packet4i& b) { return vaddq_s32(a,b); }
111 
112 template<> EIGEN_STRONG_INLINE Packet4f psub<Packet4f>(const Packet4f& a, const Packet4f& b) { return vsubq_f32(a,b); }
113 template<> EIGEN_STRONG_INLINE Packet4i psub<Packet4i>(const Packet4i& a, const Packet4i& b) { return vsubq_s32(a,b); }
114 
115 template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) { return vnegq_f32(a); }
116 template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return vnegq_s32(a); }
117 
118 template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; }
119 template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; }
120 
121 template<> EIGEN_STRONG_INLINE Packet4f pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return vmulq_f32(a,b); }
122 template<> EIGEN_STRONG_INLINE Packet4i pmul<Packet4i>(const Packet4i& a, const Packet4i& b) { return vmulq_s32(a,b); }
123 
124 template<> EIGEN_STRONG_INLINE Packet4f pdiv<Packet4f>(const Packet4f& a, const Packet4f& b)
125 {
126  Packet4f inv, restep, div;
127 
128  // NEON does not offer a divide instruction, we have to do a reciprocal approximation
129  // However NEON in contrast to other SIMD engines (AltiVec/SSE), offers
130  // a reciprocal estimate AND a reciprocal step -which saves a few instructions
131  // vrecpeq_f32() returns an estimate to 1/b, which we will finetune with
132  // Newton-Raphson and vrecpsq_f32()
133  inv = vrecpeq_f32(b);
134 
135  // This returns a differential, by which we will have to multiply inv to get a better
136  // approximation of 1/b.
137  restep = vrecpsq_f32(b, inv);
138  inv = vmulq_f32(restep, inv);
139 
140  // Finally, multiply a by 1/b and get the wanted result of the division.
141  div = vmulq_f32(a, inv);
142 
143  return div;
144 }
145 template<> EIGEN_STRONG_INLINE Packet4i pdiv<Packet4i>(const Packet4i& /*a*/, const Packet4i& /*b*/)
146 { eigen_assert(false && "packet integer division are not supported by NEON");
147  return pset1<Packet4i>(0);
148 }
149 
150 // for some weird raisons, it has to be overloaded for packet of integers
151 template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return vmlaq_f32(c,a,b); }
152 template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return vmlaq_s32(c,a,b); }
153 
154 template<> EIGEN_STRONG_INLINE Packet4f pmin<Packet4f>(const Packet4f& a, const Packet4f& b) { return vminq_f32(a,b); }
155 template<> EIGEN_STRONG_INLINE Packet4i pmin<Packet4i>(const Packet4i& a, const Packet4i& b) { return vminq_s32(a,b); }
156 
157 template<> EIGEN_STRONG_INLINE Packet4f pmax<Packet4f>(const Packet4f& a, const Packet4f& b) { return vmaxq_f32(a,b); }
158 template<> EIGEN_STRONG_INLINE Packet4i pmax<Packet4i>(const Packet4i& a, const Packet4i& b) { return vmaxq_s32(a,b); }
159 
160 // Logical Operations are not supported for float, so we have to reinterpret casts using NEON intrinsics
161 template<> EIGEN_STRONG_INLINE Packet4f pand<Packet4f>(const Packet4f& a, const Packet4f& b)
162 {
163  return vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b)));
164 }
165 template<> EIGEN_STRONG_INLINE Packet4i pand<Packet4i>(const Packet4i& a, const Packet4i& b) { return vandq_s32(a,b); }
166 
167 template<> EIGEN_STRONG_INLINE Packet4f por<Packet4f>(const Packet4f& a, const Packet4f& b)
168 {
169  return vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b)));
170 }
171 template<> EIGEN_STRONG_INLINE Packet4i por<Packet4i>(const Packet4i& a, const Packet4i& b) { return vorrq_s32(a,b); }
172 
173 template<> EIGEN_STRONG_INLINE Packet4f pxor<Packet4f>(const Packet4f& a, const Packet4f& b)
174 {
175  return vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b)));
176 }
177 template<> EIGEN_STRONG_INLINE Packet4i pxor<Packet4i>(const Packet4i& a, const Packet4i& b) { return veorq_s32(a,b); }
178 
179 template<> EIGEN_STRONG_INLINE Packet4f pandnot<Packet4f>(const Packet4f& a, const Packet4f& b)
180 {
181  return vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b)));
182 }
183 template<> EIGEN_STRONG_INLINE Packet4i pandnot<Packet4i>(const Packet4i& a, const Packet4i& b) { return vbicq_s32(a,b); }
184 
185 template<> EIGEN_STRONG_INLINE Packet4f pload<Packet4f>(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f32(from); }
186 template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); }
187 
188 template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); }
189 template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); }
190 
191 template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float* from)
192 {
193  float32x2_t lo, hi;
194  lo = vld1_dup_f32(from);
195  hi = vld1_dup_f32(from+1);
196  return vcombine_f32(lo, hi);
197 }
198 template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int* from)
199 {
200  int32x2_t lo, hi;
201  lo = vld1_dup_s32(from);
202  hi = vld1_dup_s32(from+1);
203  return vcombine_s32(lo, hi);
204 }
205 
206 template<> EIGEN_STRONG_INLINE void pstore<float>(float* to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); }
207 template<> EIGEN_STRONG_INLINE void pstore<int>(int* to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); }
208 
209 template<> EIGEN_STRONG_INLINE void pstoreu<float>(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); }
210 template<> EIGEN_STRONG_INLINE void pstoreu<int>(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); }
211 
212 template<> EIGEN_STRONG_INLINE void prefetch<float>(const float* addr) { __pld(addr); }
213 template<> EIGEN_STRONG_INLINE void prefetch<int>(const int* addr) { __pld(addr); }
214 
215 // FIXME only store the 2 first elements ?
216 template<> EIGEN_STRONG_INLINE float pfirst<Packet4f>(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; }
217 template<> EIGEN_STRONG_INLINE int pfirst<Packet4i>(const Packet4i& a) { int EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; }
218 
219 template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) {
220  float32x2_t a_lo, a_hi;
221  Packet4f a_r64;
222 
223  a_r64 = vrev64q_f32(a);
224  a_lo = vget_low_f32(a_r64);
225  a_hi = vget_high_f32(a_r64);
226  return vcombine_f32(a_hi, a_lo);
227 }
228 template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) {
229  int32x2_t a_lo, a_hi;
230  Packet4i a_r64;
231 
232  a_r64 = vrev64q_s32(a);
233  a_lo = vget_low_s32(a_r64);
234  a_hi = vget_high_s32(a_r64);
235  return vcombine_s32(a_hi, a_lo);
236 }
237 template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) { return vabsq_f32(a); }
238 template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vabsq_s32(a); }
239 
240 template<> EIGEN_STRONG_INLINE float predux<Packet4f>(const Packet4f& a)
241 {
242  float32x2_t a_lo, a_hi, sum;
243 
244  a_lo = vget_low_f32(a);
245  a_hi = vget_high_f32(a);
246  sum = vpadd_f32(a_lo, a_hi);
247  sum = vpadd_f32(sum, sum);
248  return vget_lane_f32(sum, 0);
249 }
250 
251 template<> EIGEN_STRONG_INLINE Packet4f preduxp<Packet4f>(const Packet4f* vecs)
252 {
253  float32x4x2_t vtrn1, vtrn2, res1, res2;
254  Packet4f sum1, sum2, sum;
255 
256  // NEON zip performs interleaving of the supplied vectors.
257  // We perform two interleaves in a row to acquire the transposed vector
258  vtrn1 = vzipq_f32(vecs[0], vecs[2]);
259  vtrn2 = vzipq_f32(vecs[1], vecs[3]);
260  res1 = vzipq_f32(vtrn1.val[0], vtrn2.val[0]);
261  res2 = vzipq_f32(vtrn1.val[1], vtrn2.val[1]);
262 
263  // Do the addition of the resulting vectors
264  sum1 = vaddq_f32(res1.val[0], res1.val[1]);
265  sum2 = vaddq_f32(res2.val[0], res2.val[1]);
266  sum = vaddq_f32(sum1, sum2);
267 
268  return sum;
269 }
270 
271 template<> EIGEN_STRONG_INLINE int predux<Packet4i>(const Packet4i& a)
272 {
273  int32x2_t a_lo, a_hi, sum;
274 
275  a_lo = vget_low_s32(a);
276  a_hi = vget_high_s32(a);
277  sum = vpadd_s32(a_lo, a_hi);
278  sum = vpadd_s32(sum, sum);
279  return vget_lane_s32(sum, 0);
280 }
281 
282 template<> EIGEN_STRONG_INLINE Packet4i preduxp<Packet4i>(const Packet4i* vecs)
283 {
284  int32x4x2_t vtrn1, vtrn2, res1, res2;
285  Packet4i sum1, sum2, sum;
286 
287  // NEON zip performs interleaving of the supplied vectors.
288  // We perform two interleaves in a row to acquire the transposed vector
289  vtrn1 = vzipq_s32(vecs[0], vecs[2]);
290  vtrn2 = vzipq_s32(vecs[1], vecs[3]);
291  res1 = vzipq_s32(vtrn1.val[0], vtrn2.val[0]);
292  res2 = vzipq_s32(vtrn1.val[1], vtrn2.val[1]);
293 
294  // Do the addition of the resulting vectors
295  sum1 = vaddq_s32(res1.val[0], res1.val[1]);
296  sum2 = vaddq_s32(res2.val[0], res2.val[1]);
297  sum = vaddq_s32(sum1, sum2);
298 
299  return sum;
300 }
301 
302 // Other reduction functions:
303 // mul
304 template<> EIGEN_STRONG_INLINE float predux_mul<Packet4f>(const Packet4f& a)
305 {
306  float32x2_t a_lo, a_hi, prod;
307 
308  // Get a_lo = |a1|a2| and a_hi = |a3|a4|
309  a_lo = vget_low_f32(a);
310  a_hi = vget_high_f32(a);
311  // Get the product of a_lo * a_hi -> |a1*a3|a2*a4|
312  prod = vmul_f32(a_lo, a_hi);
313  // Multiply prod with its swapped value |a2*a4|a1*a3|
314  prod = vmul_f32(prod, vrev64_f32(prod));
315 
316  return vget_lane_f32(prod, 0);
317 }
318 template<> EIGEN_STRONG_INLINE int predux_mul<Packet4i>(const Packet4i& a)
319 {
320  int32x2_t a_lo, a_hi, prod;
321 
322  // Get a_lo = |a1|a2| and a_hi = |a3|a4|
323  a_lo = vget_low_s32(a);
324  a_hi = vget_high_s32(a);
325  // Get the product of a_lo * a_hi -> |a1*a3|a2*a4|
326  prod = vmul_s32(a_lo, a_hi);
327  // Multiply prod with its swapped value |a2*a4|a1*a3|
328  prod = vmul_s32(prod, vrev64_s32(prod));
329 
330  return vget_lane_s32(prod, 0);
331 }
332 
333 // min
334 template<> EIGEN_STRONG_INLINE float predux_min<Packet4f>(const Packet4f& a)
335 {
336  float32x2_t a_lo, a_hi, min;
337 
338  a_lo = vget_low_f32(a);
339  a_hi = vget_high_f32(a);
340  min = vpmin_f32(a_lo, a_hi);
341  min = vpmin_f32(min, min);
342 
343  return vget_lane_f32(min, 0);
344 }
345 
346 template<> EIGEN_STRONG_INLINE int predux_min<Packet4i>(const Packet4i& a)
347 {
348  int32x2_t a_lo, a_hi, min;
349 
350  a_lo = vget_low_s32(a);
351  a_hi = vget_high_s32(a);
352  min = vpmin_s32(a_lo, a_hi);
353  min = vpmin_s32(min, min);
354 
355  return vget_lane_s32(min, 0);
356 }
357 
358 // max
359 template<> EIGEN_STRONG_INLINE float predux_max<Packet4f>(const Packet4f& a)
360 {
361  float32x2_t a_lo, a_hi, max;
362 
363  a_lo = vget_low_f32(a);
364  a_hi = vget_high_f32(a);
365  max = vpmax_f32(a_lo, a_hi);
366  max = vpmax_f32(max, max);
367 
368  return vget_lane_f32(max, 0);
369 }
370 
371 template<> EIGEN_STRONG_INLINE int predux_max<Packet4i>(const Packet4i& a)
372 {
373  int32x2_t a_lo, a_hi, max;
374 
375  a_lo = vget_low_s32(a);
376  a_hi = vget_high_s32(a);
377  max = vpmax_s32(a_lo, a_hi);
378 
379  return vget_lane_s32(max, 0);
380 }
381 
382 // this PALIGN_NEON business is to work around a bug in LLVM Clang 3.0 causing incorrect compilation errors,
383 // see bug 347 and this LLVM bug: http://llvm.org/bugs/show_bug.cgi?id=11074
384 #define PALIGN_NEON(Offset,Type,Command) \
385 template<>\
386 struct palign_impl<Offset,Type>\
387 {\
388  EIGEN_STRONG_INLINE static void run(Type& first, const Type& second)\
389  {\
390  if (Offset!=0)\
391  first = Command(first, second, Offset);\
392  }\
393 };\
394 
395 PALIGN_NEON(0,Packet4f,vextq_f32)
396 PALIGN_NEON(1,Packet4f,vextq_f32)
397 PALIGN_NEON(2,Packet4f,vextq_f32)
398 PALIGN_NEON(3,Packet4f,vextq_f32)
399 PALIGN_NEON(0,Packet4i,vextq_s32)
400 PALIGN_NEON(1,Packet4i,vextq_s32)
401 PALIGN_NEON(2,Packet4i,vextq_s32)
402 PALIGN_NEON(3,Packet4i,vextq_s32)
403 
404 #undef PALIGN_NEON
405 
406 } // end namespace internal
407 
408 } // end namespace Eigen
409 
410 #endif // EIGEN_PACKET_MATH_NEON_H
EIGEN_STRONG_INLINE Packet4i ploaddup< Packet4i >(const int *from)
#define EIGEN_ALIGN16
EIGEN_STRONG_INLINE Packet4f pxor< Packet4f >(const Packet4f &a, const Packet4f &b)
__vector float Packet4f
EIGEN_STRONG_INLINE int pfirst< Packet4i >(const Packet4i &a)
#define EIGEN_STRONG_INLINE
EIGEN_STRONG_INLINE Packet4i pload< Packet4i >(const int *from)
#define EIGEN_DEBUG_UNALIGNED_LOAD
EIGEN_STRONG_INLINE float pfirst< Packet4f >(const Packet4f &a)
EIGEN_STRONG_INLINE Packet4f ploaddup< Packet4f >(const float *from)
EIGEN_STRONG_INLINE int predux< Packet4i >(const Packet4i &a)
#define EIGEN_DEBUG_ALIGNED_STORE
iterative scaling algorithm to equilibrate rows and column norms in matrices
Definition: matrix.hpp:471
EIGEN_STRONG_INLINE Packet4i ploadu< Packet4i >(const int *from)
EIGEN_STRONG_INLINE float predux_max< Packet4f >(const Packet4f &a)
#define EIGEN_DEBUG_UNALIGNED_STORE
EIGEN_STRONG_INLINE void prefetch< float >(const float *addr)
EIGEN_STRONG_INLINE Packet4i pdiv< Packet4i >(const Packet4i &, const Packet4i &)
EIGEN_STRONG_INLINE int predux_min< Packet4i >(const Packet4i &a)
#define EIGEN_DEBUG_ALIGNED_LOAD
EIGEN_STRONG_INLINE Packet4i plset< int >(const int &a)
EIGEN_STRONG_INLINE Packet4f plset< float >(const float &a)
EIGEN_STRONG_INLINE Packet4i pmul< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet4f pdiv< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE Packet4f pmin< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE void prefetch< int >(const int *addr)
EIGEN_STRONG_INLINE Packet4i pand< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE int predux_max< Packet4i >(const Packet4i &a)
EIGEN_STRONG_INLINE Packet4i pmin< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet4f ploadu< Packet4f >(const float *from)
EIGEN_STRONG_INLINE Packet4i pxor< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE void pstoreu< int >(int *to, const Packet4i &from)
EIGEN_STRONG_INLINE void pstore< float >(float *to, const Packet4f &from)
EIGEN_STRONG_INLINE Packet4f por< Packet4f >(const Packet4f &a, const Packet4f &b)
#define __pld(x)
EIGEN_STRONG_INLINE Packet4i preduxp< Packet4i >(const Packet4i *vecs)
EIGEN_STRONG_INLINE Packet4f pandnot< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE float predux< Packet4f >(const Packet4f &a)
EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf &a)
EIGEN_STRONG_INLINE float predux_min< Packet4f >(const Packet4f &a)
#define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W)
#define PALIGN_NEON(Offset, Type, Command)
EIGEN_STRONG_INLINE void pstoreu< float >(float *to, const Packet4f &from)
EIGEN_STRONG_INLINE int predux_mul< Packet4i >(const Packet4i &a)
EIGEN_STRONG_INLINE Packet4f pmul< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE Packet4f pload< Packet4f >(const float *from)
EIGEN_STRONG_INLINE Packet4i pmax< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf &a)
EIGEN_STRONG_INLINE Packet4i por< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet4i padd< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet4f padd< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE void pstore< int >(int *to, const Packet4i &from)
EIGEN_STRONG_INLINE Packet4i pandnot< Packet4i >(const Packet4i &a, const Packet4i &b)
__vector unsigned int Packet4ui
EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f &a, const Packet4f &b, const Packet4f &c)
EIGEN_STRONG_INLINE Packet4f preduxp< Packet4f >(const Packet4f *vecs)
__vector int Packet4i
#define eigen_assert(x)
EIGEN_STRONG_INLINE Packet4f pmax< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE Packet4f pset1< Packet4f >(const float &from)
EIGEN_STRONG_INLINE Packet4i psub< Packet4i >(const Packet4i &a, const Packet4i &b)
EIGEN_STRONG_INLINE Packet4f psub< Packet4f >(const Packet4f &a, const Packet4f &b)
EIGEN_STRONG_INLINE Packet4i pset1< Packet4i >(const int &from)
EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf &a)
EIGEN_STRONG_INLINE float predux_mul< Packet4f >(const Packet4f &a)
EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f &a)
EIGEN_STRONG_INLINE Packet4f pand< Packet4f >(const Packet4f &a, const Packet4f &b)


acado
Author(s): Milan Vukov, Rien Quirynen
autogenerated on Mon Jun 10 2019 12:34:55