simd_pan.hpp

   1 //  simd functions for panning
   2 //  Copyright (C) 2009, 2010 Tim Blechmann
   3 //
   4 //  This program is free software; you can redistribute it and/or modify
   5 //  it under the terms of the GNU General Public License as published by
   6 //  the Free Software Foundation; either version 2 of the License, or
   7 //  (at your option) any later version.
   8 //
   9 //  This program is distributed in the hope that it will be useful,
  10 //  but WITHOUT ANY WARRANTY; without even the implied warranty of
  11 //  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12 //  GNU General Public License for more details.
  13 //
  14 //  You should have received a copy of the GNU General Public License
  15 //  along with this program; see the file COPYING.  If not, write to
  16 //  the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  17 //  Boston, MA 02111-1307, USA.
  18
  19
  20 #ifndef SIMD_PAN_HPP
  21 #define SIMD_PAN_HPP
  22
  23 #include "vec.hpp"
  24
  25 #if defined(__GNUC__) && defined(NDEBUG)
  26 #define always_inline inline  __attribute__((always_inline))
  27 #else
  28 #define always_inline inline
  29 #endif
  30
  31 namespace nova
  32 {
  33
  34 template <typename F>
  35 inline void pan2_vec(F * out0, F * out1, const F * in, F factor0, F factor1, unsigned int n)
  36 {
  37     do
  38     {
  39         F sig = *in++;
  40         *out0++ = sig * factor0;
  41         *out1++ = sig * factor1;
  42     } while(--n);
  43 }
  44
  45 template <typename F>
  46 inline void pan2_vec(F * out0, F * out1, const F * in, F factor0, F slope0, F factor1, F slope1, unsigned int n)
  47 {
  48     do
  49     {
  50         F sig = *in++;
  51         *out0++ = sig * factor0;
  52         *out1++ = sig * factor1;
  53         factor0 += slope0;
  54         factor1 += slope1;
  55     } while(--n);
  56 }
  57
  58 namespace detail
  59 {
  60
  61 template <typename F, unsigned int n>
  62 struct pan2
  63 {
  64     static const int offset = vec<F>::size;
  65
  66     static always_inline void mp_iteration(F * out0, F * out1, const F * in, vec<F> const & factor0, vec<F> const & factor1)
  67     {
  68         vec<F> vin, vout0, vout1;
  69         vin.load_aligned(in);
  70
  71         vout0 = vin * factor0;
  72         vout1 = vin * factor1;
  73
  74         vout0.store_aligned(out0);
  75         vout1.store_aligned(out1);
  76
  77         pan2<F, n-offset>::mp_iteration(out0+offset, out1+offset, in+offset, factor0, factor1);
  78     }
  79
  80     static always_inline void mp_iteration(F * out0, F * out1, const F * in, vec<F> & factor0, vec<F> const & slope0,
  81                                            vec<F> & factor1, vec<F> const & slope1)
  82     {
  83         vec<F> vin, vout0, vout1;
  84         vin.load_aligned(in);
  85
  86         vout0 = vin * factor0;
  87         vout1 = vin * factor1;
  88
  89         vout0.store_aligned(out0);
  90         vout1.store_aligned(out1);
  91         factor0 += slope0;
  92         factor1 += slope1;
  93
  94         pan2<F, n-offset>::mp_iteration(out0+offset, out1+offset, in+offset, factor0, slope0, factor1, slope1);
  95     }
  96 };
  97
  98
  99 template <typename F>
 100 struct pan2<F, 0>
 101 {
 102     static always_inline void mp_iteration(F * out0, F * out1, const F * in, vec<F> const & factor0, vec<F> const & factor1)
 103     {}
 104
 105     static always_inline void mp_iteration(F * out0, F * out1, const F * in, vec<F> & factor0, vec<F> const & slope0,
 106                                            vec<F> & factor1, vec<F> const & slope1)
 107     {}
 108 };
 109
 110 } /* namespace detail */
 111
 112 template <typename F>
 113 inline void pan2_vec_simd(F * out0, F * out1, const F * in, F factor0, F factor1, unsigned int n)
 114 {
 115     vec<F> vf0(factor0), vf1(factor1);
 116     const int per_loop = vec<F>::objects_per_cacheline;
 117
 118     n /= per_loop;
 119     do {
 120         detail::pan2<F, per_loop>::mp_iteration(out0, out1, in, vf0, vf1);
 121         out0 += per_loop; out1 += per_loop; in += per_loop;
 122     } while(--n);
 123 }
 124
 125 template <unsigned int n, typename F>
 126 inline void pan2_vec_simd(F * out0, F * out1, const F * in, F factor0, F factor1)
 127 {
 128     vec<F> vf0(factor0), vf1(factor1);
 129
 130     detail::pan2<F, n>::mp_iteration(out0, out1, in, vf0, vf1);
 131 }
 132
 133 template <typename F>
 134 inline void pan2_vec_simd(F * out0, F * out1, const F * in, F factor0, F slope0, F factor1, F slope1, unsigned int n)
 135 {
 136     const int per_loop = vec<F>::objects_per_cacheline;
 137
 138     vec<F> vf0, vf1, vslope0, vslope1;
 139     vslope0.set_vec(vf0.set_slope(factor0, slope0));
 140     vslope1.set_vec(vf1.set_slope(factor1, slope1));
 141
 142     n /= per_loop;
 143     do {
 144         detail::pan2<F, per_loop>::mp_iteration(out0, out1, in, vf0, vslope0, vf1, vslope1);
 145         out0 += per_loop; out1 += per_loop; in += per_loop;
 146     } while(--n);
 147 }
 148
 149 template <unsigned int n, typename F>
 150 inline void pan2_vec_simd(F * out0, F * out1, const F * in, F factor0, F slope0, F factor1, F slope1)
 151 {
 152     vec<F> vf0, vf1, vslope0, vslope1;
 153     vslope0.set_vec(vf0.set_slope(factor0, slope0));
 154     vslope1.set_vec(vf1.set_slope(factor1, slope1));
 155
 156     detail::pan2<F, n>::mp_iteration(out0, out1, in, vf0, vslope0, vf1, vslope1);
 157 }
 158
 159 } /* namespace nova */
 160
 161 #undef always_inline
 162
 163 #endif /* SIMD_PAN_HPP */