• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2008 Gael Guennebaud <gael.guennebaud@inria.fr>
5 // Copyright (C) 2006-2008 Benoit Jacob <jacob.benoit.1@gmail.com>
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_GENERIC_PACKET_MATH_H
12 #define EIGEN_GENERIC_PACKET_MATH_H
13 
14 namespace Eigen {
15 
16 namespace internal {
17 
18 /** \internal
19   * \file GenericPacketMath.h
20   *
21   * Default implementation for types not supported by the vectorization.
22   * In practice these functions are provided to make easier the writing
23   * of generic vectorized code.
24   */
25 
26 #ifndef EIGEN_DEBUG_ALIGNED_LOAD
27 #define EIGEN_DEBUG_ALIGNED_LOAD
28 #endif
29 
30 #ifndef EIGEN_DEBUG_UNALIGNED_LOAD
31 #define EIGEN_DEBUG_UNALIGNED_LOAD
32 #endif
33 
34 #ifndef EIGEN_DEBUG_ALIGNED_STORE
35 #define EIGEN_DEBUG_ALIGNED_STORE
36 #endif
37 
38 #ifndef EIGEN_DEBUG_UNALIGNED_STORE
39 #define EIGEN_DEBUG_UNALIGNED_STORE
40 #endif
41 
42 struct default_packet_traits
43 {
44   enum {
45     HasAdd    = 1,
46     HasSub    = 1,
47     HasMul    = 1,
48     HasNegate = 1,
49     HasAbs    = 1,
50     HasAbs2   = 1,
51     HasMin    = 1,
52     HasMax    = 1,
53     HasConj   = 1,
54     HasSetLinear = 1,
55 
56     HasDiv    = 0,
57     HasSqrt   = 0,
58     HasExp    = 0,
59     HasLog    = 0,
60     HasPow    = 0,
61 
62     HasSin    = 0,
63     HasCos    = 0,
64     HasTan    = 0,
65     HasASin   = 0,
66     HasACos   = 0,
67     HasATan   = 0
68   };
69 };
70 
71 template<typename T> struct packet_traits : default_packet_traits
72 {
73   typedef T type;
74   enum {
75     Vectorizable = 0,
76     size = 1,
77     AlignedOnScalar = 0
78   };
79   enum {
80     HasAdd    = 0,
81     HasSub    = 0,
82     HasMul    = 0,
83     HasNegate = 0,
84     HasAbs    = 0,
85     HasAbs2   = 0,
86     HasMin    = 0,
87     HasMax    = 0,
88     HasConj   = 0,
89     HasSetLinear = 0
90   };
91 };
92 
93 /** \internal \returns a + b (coeff-wise) */
94 template<typename Packet> inline Packet
padd(const Packet & a,const Packet & b)95 padd(const Packet& a,
96         const Packet& b) { return a+b; }
97 
98 /** \internal \returns a - b (coeff-wise) */
99 template<typename Packet> inline Packet
psub(const Packet & a,const Packet & b)100 psub(const Packet& a,
101         const Packet& b) { return a-b; }
102 
103 /** \internal \returns -a (coeff-wise) */
104 template<typename Packet> inline Packet
pnegate(const Packet & a)105 pnegate(const Packet& a) { return -a; }
106 
107 /** \internal \returns conj(a) (coeff-wise) */
108 template<typename Packet> inline Packet
pconj(const Packet & a)109 pconj(const Packet& a) { return numext::conj(a); }
110 
111 /** \internal \returns a * b (coeff-wise) */
112 template<typename Packet> inline Packet
pmul(const Packet & a,const Packet & b)113 pmul(const Packet& a,
114         const Packet& b) { return a*b; }
115 
116 /** \internal \returns a / b (coeff-wise) */
117 template<typename Packet> inline Packet
pdiv(const Packet & a,const Packet & b)118 pdiv(const Packet& a,
119         const Packet& b) { return a/b; }
120 
121 /** \internal \returns the min of \a a and \a b  (coeff-wise) */
122 template<typename Packet> inline Packet
pmin(const Packet & a,const Packet & b)123 pmin(const Packet& a,
124         const Packet& b) { using std::min; return (min)(a, b); }
125 
126 /** \internal \returns the max of \a a and \a b  (coeff-wise) */
127 template<typename Packet> inline Packet
pmax(const Packet & a,const Packet & b)128 pmax(const Packet& a,
129         const Packet& b) { using std::max; return (max)(a, b); }
130 
131 /** \internal \returns the absolute value of \a a */
132 template<typename Packet> inline Packet
pabs(const Packet & a)133 pabs(const Packet& a) { using std::abs; return abs(a); }
134 
135 /** \internal \returns the bitwise and of \a a and \a b */
136 template<typename Packet> inline Packet
pand(const Packet & a,const Packet & b)137 pand(const Packet& a, const Packet& b) { return a & b; }
138 
139 /** \internal \returns the bitwise or of \a a and \a b */
140 template<typename Packet> inline Packet
por(const Packet & a,const Packet & b)141 por(const Packet& a, const Packet& b) { return a | b; }
142 
143 /** \internal \returns the bitwise xor of \a a and \a b */
144 template<typename Packet> inline Packet
pxor(const Packet & a,const Packet & b)145 pxor(const Packet& a, const Packet& b) { return a ^ b; }
146 
147 /** \internal \returns the bitwise andnot of \a a and \a b */
148 template<typename Packet> inline Packet
pandnot(const Packet & a,const Packet & b)149 pandnot(const Packet& a, const Packet& b) { return a & (!b); }
150 
151 /** \internal \returns a packet version of \a *from, from must be 16 bytes aligned */
152 template<typename Packet> inline Packet
pload(const typename unpacket_traits<Packet>::type * from)153 pload(const typename unpacket_traits<Packet>::type* from) { return *from; }
154 
155 /** \internal \returns a packet version of \a *from, (un-aligned load) */
156 template<typename Packet> inline Packet
ploadu(const typename unpacket_traits<Packet>::type * from)157 ploadu(const typename unpacket_traits<Packet>::type* from) { return *from; }
158 
159 /** \internal \returns a packet with elements of \a *from duplicated.
160   * For instance, for a packet of 8 elements, 4 scalar will be read from \a *from and
161   * duplicated to form: {from[0],from[0],from[1],from[1],,from[2],from[2],,from[3],from[3]}
162   * Currently, this function is only used for scalar * complex products.
163  */
164 template<typename Packet> inline Packet
ploaddup(const typename unpacket_traits<Packet>::type * from)165 ploaddup(const typename unpacket_traits<Packet>::type* from) { return *from; }
166 
167 /** \internal \returns a packet with constant coefficients \a a, e.g.: (a,a,a,a) */
168 template<typename Packet> inline Packet
pset1(const typename unpacket_traits<Packet>::type & a)169 pset1(const typename unpacket_traits<Packet>::type& a) { return a; }
170 
171 /** \internal \brief Returns a packet with coefficients (a,a+1,...,a+packet_size-1). */
172 template<typename Scalar> inline typename packet_traits<Scalar>::type
plset(const Scalar & a)173 plset(const Scalar& a) { return a; }
174 
175 /** \internal copy the packet \a from to \a *to, \a to must be 16 bytes aligned */
pstore(Scalar * to,const Packet & from)176 template<typename Scalar, typename Packet> inline void pstore(Scalar* to, const Packet& from)
177 { (*to) = from; }
178 
179 /** \internal copy the packet \a from to \a *to, (un-aligned store) */
pstoreu(Scalar * to,const Packet & from)180 template<typename Scalar, typename Packet> inline void pstoreu(Scalar* to, const Packet& from)
181 { (*to) = from; }
182 
183 /** \internal tries to do cache prefetching of \a addr */
prefetch(const Scalar * addr)184 template<typename Scalar> inline void prefetch(const Scalar* addr)
185 {
186 #if !defined(_MSC_VER)
187 __builtin_prefetch(addr);
188 #endif
189 }
190 
191 /** \internal \returns the first element of a packet */
pfirst(const Packet & a)192 template<typename Packet> inline typename unpacket_traits<Packet>::type pfirst(const Packet& a)
193 { return a; }
194 
195 /** \internal \returns a packet where the element i contains the sum of the packet of \a vec[i] */
196 template<typename Packet> inline Packet
preduxp(const Packet * vecs)197 preduxp(const Packet* vecs) { return vecs[0]; }
198 
199 /** \internal \returns the sum of the elements of \a a*/
predux(const Packet & a)200 template<typename Packet> inline typename unpacket_traits<Packet>::type predux(const Packet& a)
201 { return a; }
202 
203 /** \internal \returns the product of the elements of \a a*/
predux_mul(const Packet & a)204 template<typename Packet> inline typename unpacket_traits<Packet>::type predux_mul(const Packet& a)
205 { return a; }
206 
207 /** \internal \returns the min of the elements of \a a*/
predux_min(const Packet & a)208 template<typename Packet> inline typename unpacket_traits<Packet>::type predux_min(const Packet& a)
209 { return a; }
210 
211 /** \internal \returns the max of the elements of \a a*/
predux_max(const Packet & a)212 template<typename Packet> inline typename unpacket_traits<Packet>::type predux_max(const Packet& a)
213 { return a; }
214 
215 /** \internal \returns the reversed elements of \a a*/
preverse(const Packet & a)216 template<typename Packet> inline Packet preverse(const Packet& a)
217 { return a; }
218 
219 
220 /** \internal \returns \a a with real and imaginary part flipped (for complex type only) */
pcplxflip(const Packet & a)221 template<typename Packet> inline Packet pcplxflip(const Packet& a)
222 {
223   // FIXME: uncomment the following in case we drop the internal imag and real functions.
224 //   using std::imag;
225 //   using std::real;
226   return Packet(imag(a),real(a));
227 }
228 
229 /**************************
230 * Special math functions
231 ***************************/
232 
233 /** \internal \returns the sine of \a a (coeff-wise) */
234 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
psin(const Packet & a)235 Packet psin(const Packet& a) { using std::sin; return sin(a); }
236 
237 /** \internal \returns the cosine of \a a (coeff-wise) */
238 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
pcos(const Packet & a)239 Packet pcos(const Packet& a) { using std::cos; return cos(a); }
240 
241 /** \internal \returns the tan of \a a (coeff-wise) */
242 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
ptan(const Packet & a)243 Packet ptan(const Packet& a) { using std::tan; return tan(a); }
244 
245 /** \internal \returns the arc sine of \a a (coeff-wise) */
246 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
pasin(const Packet & a)247 Packet pasin(const Packet& a) { using std::asin; return asin(a); }
248 
249 /** \internal \returns the arc cosine of \a a (coeff-wise) */
250 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
pacos(const Packet & a)251 Packet pacos(const Packet& a) { using std::acos; return acos(a); }
252 
253 /** \internal \returns the exp of \a a (coeff-wise) */
254 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
pexp(const Packet & a)255 Packet pexp(const Packet& a) { using std::exp; return exp(a); }
256 
257 /** \internal \returns the log of \a a (coeff-wise) */
258 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
plog(const Packet & a)259 Packet plog(const Packet& a) { using std::log; return log(a); }
260 
261 /** \internal \returns the square-root of \a a (coeff-wise) */
262 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
psqrt(const Packet & a)263 Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); }
264 
265 /***************************************************************************
266 * The following functions might not have to be overwritten for vectorized types
267 ***************************************************************************/
268 
269 /** \internal copy a packet with constant coeficient \a a (e.g., [a,a,a,a]) to \a *to. \a to must be 16 bytes aligned */
270 // NOTE: this function must really be templated on the packet type (think about different packet types for the same scalar type)
271 template<typename Packet>
pstore1(typename unpacket_traits<Packet>::type * to,const typename unpacket_traits<Packet>::type & a)272 inline void pstore1(typename unpacket_traits<Packet>::type* to, const typename unpacket_traits<Packet>::type& a)
273 {
274   pstore(to, pset1<Packet>(a));
275 }
276 
277 /** \internal \returns a * b + c (coeff-wise) */
278 template<typename Packet> inline Packet
pmadd(const Packet & a,const Packet & b,const Packet & c)279 pmadd(const Packet&  a,
280          const Packet&  b,
281          const Packet&  c)
282 { return padd(pmul(a, b),c); }
283 
284 /** \internal \returns a packet version of \a *from.
285   * If LoadMode equals #Aligned, \a from must be 16 bytes aligned */
286 template<typename Packet, int LoadMode>
ploadt(const typename unpacket_traits<Packet>::type * from)287 inline Packet ploadt(const typename unpacket_traits<Packet>::type* from)
288 {
289   if(LoadMode == Aligned)
290     return pload<Packet>(from);
291   else
292     return ploadu<Packet>(from);
293 }
294 
295 /** \internal copy the packet \a from to \a *to.
296   * If StoreMode equals #Aligned, \a to must be 16 bytes aligned */
297 template<typename Scalar, typename Packet, int LoadMode>
pstoret(Scalar * to,const Packet & from)298 inline void pstoret(Scalar* to, const Packet& from)
299 {
300   if(LoadMode == Aligned)
301     pstore(to, from);
302   else
303     pstoreu(to, from);
304 }
305 
306 /** \internal default implementation of palign() allowing partial specialization */
307 template<int Offset,typename PacketType>
308 struct palign_impl
309 {
310   // by default data are aligned, so there is nothing to be done :)
runpalign_impl311   static inline void run(PacketType&, const PacketType&) {}
312 };
313 
314 /** \internal update \a first using the concatenation of the packet_size minus \a Offset last elements
315   * of \a first and \a Offset first elements of \a second.
316   *
317   * This function is currently only used to optimize matrix-vector products on unligned matrices.
318   * It takes 2 packets that represent a contiguous memory array, and returns a packet starting
319   * at the position \a Offset. For instance, for packets of 4 elements, we have:
320   *  Input:
321   *  - first = {f0,f1,f2,f3}
322   *  - second = {s0,s1,s2,s3}
323   * Output:
324   *   - if Offset==0 then {f0,f1,f2,f3}
325   *   - if Offset==1 then {f1,f2,f3,s0}
326   *   - if Offset==2 then {f2,f3,s0,s1}
327   *   - if Offset==3 then {f3,s0,s1,s3}
328   */
329 template<int Offset,typename PacketType>
palign(PacketType & first,const PacketType & second)330 inline void palign(PacketType& first, const PacketType& second)
331 {
332   palign_impl<Offset,PacketType>::run(first,second);
333 }
334 
335 /***************************************************************************
336 * Fast complex products (GCC generates a function call which is very slow)
337 ***************************************************************************/
338 
pmul(const std::complex<float> & a,const std::complex<float> & b)339 template<> inline std::complex<float> pmul(const std::complex<float>& a, const std::complex<float>& b)
340 { return std::complex<float>(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); }
341 
pmul(const std::complex<double> & a,const std::complex<double> & b)342 template<> inline std::complex<double> pmul(const std::complex<double>& a, const std::complex<double>& b)
343 { return std::complex<double>(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); }
344 
345 } // end namespace internal
346 
347 } // end namespace Eigen
348 
349 #endif // EIGEN_GENERIC_PACKET_MATH_H
350 
351