Vc/sse/const.h

0001 /*  This file is part of the Vc library. {{{
0002 Copyright © 2009-2015 Matthias Kretz <kretz@kde.org>
0003
0004 Redistribution and use in source and binary forms, with or without
0005 modification, are permitted provided that the following conditions are met:
0006     * Redistributions of source code must retain the above copyright
0007       notice, this list of conditions and the following disclaimer.
0008     * Redistributions in binary form must reproduce the above copyright
0009       notice, this list of conditions and the following disclaimer in the
0010       documentation and/or other materials provided with the distribution.
0011     * Neither the names of contributing organizations nor the
0012       names of its contributors may be used to endorse or promote products
0013       derived from this software without specific prior written permission.
0014
0015 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
0016 ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
0017 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
0018 DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE FOR ANY
0019 DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
0020 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
0021 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
0022 ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
0023 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
0024 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
0025
0026 }}}*/
0027
0028 #ifndef VC_SSE_CONST_H_
0029 #define VC_SSE_CONST_H_
0030
0031 #include "const_data.h"
0032 #include "vector.h"
0033 #include "macros.h"
0034
0035 namespace Vc_VERSIONED_NAMESPACE
0036 {
0037 namespace SSE
0038 {
0039     template<typename T> struct Const
0040     {
0041         typedef Vector<T> V;
0042         typedef Mask<T> M;
0043         enum Constants { Stride = 16 / sizeof(T) };
0044
0045         static Vc_ALWAYS_INLINE Vc_CONST V _pi_4()        { return load(&c_trig<T>::data[0 * Stride]); }
0046         static Vc_ALWAYS_INLINE Vc_CONST V _pi_4_hi()     { return load(&c_trig<T>::data[1 * Stride]); }
0047         static Vc_ALWAYS_INLINE Vc_CONST V _pi_4_rem1()   { return load(&c_trig<T>::data[2 * Stride]); }
0048         static Vc_ALWAYS_INLINE Vc_CONST V _pi_4_rem2()   { return load(&c_trig<T>::data[3 * Stride]); }
0049         static Vc_ALWAYS_INLINE Vc_CONST V _1_16()        { return load(&c_trig<T>::data[4 * Stride]); }
0050         static Vc_ALWAYS_INLINE Vc_CONST V _16()          { return load(&c_trig<T>::data[5 * Stride]); }
0051
0052         static Vc_ALWAYS_INLINE Vc_CONST V atanP(int i)    { return load(&c_trig<T>::data[(12 + i) * Stride]); }
0053         static Vc_ALWAYS_INLINE Vc_CONST V atanQ(int i)    { return load(&c_trig<T>::data[(17 + i) * Stride]); }
0054         static Vc_ALWAYS_INLINE Vc_CONST V atanThrsHi()    { return load(&c_trig<T>::data[22 * Stride]); }
0055         static Vc_ALWAYS_INLINE Vc_CONST V atanThrsLo()    { return load(&c_trig<T>::data[23 * Stride]); }
0056         static Vc_ALWAYS_INLINE Vc_CONST V _pi_2_rem()     { return load(&c_trig<T>::data[24 * Stride]); }
0057         static Vc_ALWAYS_INLINE Vc_CONST V lossThreshold() { return load(&c_trig<T>::data[8 * Stride]); }
0058         static Vc_ALWAYS_INLINE Vc_CONST V _4_pi()         { return load(&c_trig<T>::data[9 * Stride]); }
0059         static Vc_ALWAYS_INLINE Vc_CONST V _pi_2()         { return load(&c_trig<T>::data[10 * Stride]); }
0060         static Vc_ALWAYS_INLINE Vc_CONST V _pi()           { return load(&c_trig<T>::data[11 * Stride]); }
0061         static Vc_ALWAYS_INLINE Vc_CONST V asinCoeff0(int i) { return load(&c_trig<T>::data[(28 + i) * Stride]); }
0062         static Vc_ALWAYS_INLINE Vc_CONST V asinCoeff1(int i) { return load(&c_trig<T>::data[(33 + i) * Stride]); }
0063         static Vc_ALWAYS_INLINE Vc_CONST V asinCoeff2(int i) { return load(&c_trig<T>::data[(37 + i) * Stride]); }
0064         static Vc_ALWAYS_INLINE Vc_CONST V asinCoeff3(int i) { return load(&c_trig<T>::data[(43 + i) * Stride]); }
0065         static Vc_ALWAYS_INLINE Vc_CONST V smallAsinInput()  { return load(&c_trig<T>::data[25 * Stride]); }
0066         static Vc_ALWAYS_INLINE Vc_CONST V largeAsinInput()  { return load(&c_trig<T>::data[26 * Stride]); }
0067
0068         static Vc_ALWAYS_INLINE Vc_CONST M exponentMask() { return M(load(c_log<T>::d(1)).data()); }
0069         static Vc_ALWAYS_INLINE Vc_CONST V _1_2()         { return load(c_log<T>::d(18)); }
0070         static Vc_ALWAYS_INLINE Vc_CONST V _1_sqrt2()     { return load(c_log<T>::d(15)); }
0071         static Vc_ALWAYS_INLINE Vc_CONST V P(int i)       { return load(c_log<T>::d(2 + i)); }
0072         static Vc_ALWAYS_INLINE Vc_CONST V Q(int i)       { return load(c_log<T>::d(8 + i)); }
0073         static Vc_ALWAYS_INLINE Vc_CONST V min()          { return load(c_log<T>::d(14)); }
0074         static Vc_ALWAYS_INLINE Vc_CONST V ln2_small()    { return load(c_log<T>::d(17)); }
0075         static Vc_ALWAYS_INLINE Vc_CONST V ln2_large()    { return load(c_log<T>::d(16)); }
0076         static Vc_ALWAYS_INLINE Vc_CONST V neginf()       { return load(c_log<T>::d(13)); }
0077         static Vc_ALWAYS_INLINE Vc_CONST V log10_e()      { return load(c_log<T>::d(19)); }
0078         static Vc_ALWAYS_INLINE Vc_CONST V log2_e()       { return load(c_log<T>::d(20)); }
0079
0080         static Vc_ALWAYS_INLINE_L Vc_CONST_L V highMask()         Vc_ALWAYS_INLINE_R Vc_CONST_R;
0081         static Vc_ALWAYS_INLINE_L Vc_CONST_L V highMask(int bits) Vc_ALWAYS_INLINE_R Vc_CONST_R;
0082     private:
0083         static Vc_ALWAYS_INLINE_L Vc_CONST_L V load(const T *mem) Vc_ALWAYS_INLINE_R Vc_CONST_R;
0084     };
0085     template<typename T> Vc_ALWAYS_INLINE Vc_CONST Vector<T> Const<T>::load(const T *mem) { return V(mem); }
0086
0087     template <> Vc_ALWAYS_INLINE Vc_CONST Vector<float> Const<float>::highMask()
0088     {
0089         return Vector<float>(reinterpret_cast<const float *>(&c_general::highMaskFloat));
0090     }
0091     template <> Vc_ALWAYS_INLINE Vc_CONST Vector<double> Const<double>::highMask()
0092     {
0093         return Vector<double>(
0094             reinterpret_cast<const double *>(&c_general::highMaskDouble));
0095     }
0096     template <> Vc_ALWAYS_INLINE Vc_CONST Vector<float> Const<float>::highMask(int bits)
0097     {
0098         return _mm_castsi128_ps(_mm_slli_epi32(_mm_setallone_si128(), bits));
0099     }
0100     template <> Vc_ALWAYS_INLINE Vc_CONST Vector<double> Const<double>::highMask(int bits)
0101     {
0102         return _mm_castsi128_pd(_mm_slli_epi64(_mm_setallone_si128(), bits));
0103     }
0104 }  // namespace SSE
0105 }  // namespace Vc
0106
0107 #endif // VC_SSE_CONST_H_