Alc/mixer_sse2.c

   1 /**
   2  * OpenAL cross platform audio library
   3  * Copyright (C) 2014 by Timothy Arceri <t_arceri@yahoo.com.au>.
   4  * This library is free software; you can redistribute it and/or
   5  *  modify it under the terms of the GNU Library General Public
   6  *  License as published by the Free Software Foundation; either
   7  *  version 2 of the License, or (at your option) any later version.
   8  *
   9  * This library 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 GNU
  12  *  Library General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU Library General Public
  15  *  License along with this library; if not, write to the
  16  *  Free Software Foundation, Inc.,
  17  *  51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  18  * Or go to http://www.gnu.org/copyleft/lgpl.html
  19  */
  20
  21 #include "config.h"
  22
  23 #include <xmmintrin.h>
  24 #include <emmintrin.h>
  25
  26 #include "alu.h"
  27 #include "mixer_defs.h"
  28
  29
  30 const ALfloat *Resample_lerp32_SSE2(const ALfloat *src, ALuint frac, ALuint increment,
  31                                     ALfloat *restrict dst, ALuint numsamples)
  32 {
  33     const __m128i increment4 = _mm_set1_epi32(increment*4);
  34     const __m128 fracOne4 = _mm_set1_ps(1.0f/FRACTIONONE);
  35     const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
  36     alignas(16) union { ALuint i[4]; float f[4]; } pos_;
  37     alignas(16) union { ALuint i[4]; float f[4]; } frac_;
  38     __m128i frac4, pos4;
  39     ALuint pos;
  40     ALuint i;
  41
  42     InitiatePositionArrays(frac, increment, frac_.i, pos_.i, 4);
  43
  44     frac4 = _mm_castps_si128(_mm_load_ps(frac_.f));
  45     pos4 = _mm_castps_si128(_mm_load_ps(pos_.f));
  46
  47     for(i = 0;numsamples-i > 3;i += 4)
  48     {
  49         const __m128 val1 = _mm_setr_ps(src[pos_.i[0]], src[pos_.i[1]], src[pos_.i[2]], src[pos_.i[3]]);
  50         const __m128 val2 = _mm_setr_ps(src[pos_.i[0]+1], src[pos_.i[1]+1], src[pos_.i[2]+1], src[pos_.i[3]+1]);
  51
  52         /* val1 + (val2-val1)*mu */
  53         const __m128 r0 = _mm_sub_ps(val2, val1);
  54         const __m128 mu = _mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4);
  55         const __m128 out = _mm_add_ps(val1, _mm_mul_ps(mu, r0));
  56
  57         _mm_store_ps(&dst[i], out);
  58
  59         frac4 = _mm_add_epi32(frac4, increment4);
  60         pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
  61         frac4 = _mm_and_si128(frac4, fracMask4);
  62
  63         _mm_store_ps(pos_.f, _mm_castsi128_ps(pos4));
  64     }
  65
  66     pos = pos_.i[0];
  67     frac = _mm_cvtsi128_si32(frac4);
  68
  69     for(;i < numsamples;i++)
  70     {
  71         dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));
  72
  73         frac += increment;
  74         pos  += frac>>FRACTIONBITS;
  75         frac &= FRACTIONMASK;
  76     }
  77     return dst;
  78 }
  79
  80 const ALfloat *Resample_cubic32_SSE2(const ALfloat *src, ALuint frac, ALuint increment,
  81                                      ALfloat *restrict dst, ALuint numsamples)
  82 {
  83     const __m128i increment4 = _mm_set1_epi32(increment*4);
  84     const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
  85     alignas(16) union { ALuint i[4]; float f[4]; } pos_;
  86     alignas(16) union { ALuint i[4]; float f[4]; } frac_;
  87     __m128i frac4, pos4;
  88     ALuint pos;
  89     ALuint i;
  90
  91     InitiatePositionArrays(frac, increment, frac_.i, pos_.i, 4);
  92
  93     frac4 = _mm_castps_si128(_mm_load_ps(frac_.f));
  94     pos4 = _mm_castps_si128(_mm_load_ps(pos_.f));
  95
  96     --src;
  97     for(i = 0;numsamples-i > 3;i += 4)
  98     {
  99         const __m128 val0 = _mm_loadu_ps(&src[pos_.i[0]]);
 100         const __m128 val1 = _mm_loadu_ps(&src[pos_.i[1]]);
 101         const __m128 val2 = _mm_loadu_ps(&src[pos_.i[2]]);
 102         const __m128 val3 = _mm_loadu_ps(&src[pos_.i[3]]);
 103         __m128 k0 = _mm_load_ps(CubicLUT[frac_.i[0]]);
 104         __m128 k1 = _mm_load_ps(CubicLUT[frac_.i[1]]);
 105         __m128 k2 = _mm_load_ps(CubicLUT[frac_.i[2]]);
 106         __m128 k3 = _mm_load_ps(CubicLUT[frac_.i[3]]);
 107         __m128 out;
 108
 109         k0 = _mm_mul_ps(k0, val0);
 110         k1 = _mm_mul_ps(k1, val1);
 111         k2 = _mm_mul_ps(k2, val2);
 112         k3 = _mm_mul_ps(k3, val3);
 113         _MM_TRANSPOSE4_PS(k0, k1, k2, k3);
 114         out = _mm_add_ps(k0, k1);
 115         out = _mm_add_ps(out, k2);
 116         out = _mm_add_ps(out, k3);
 117
 118         _mm_store_ps(&dst[i], out);
 119
 120         frac4 = _mm_add_epi32(frac4, increment4);
 121         pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
 122         frac4 = _mm_and_si128(frac4, fracMask4);
 123
 124         _mm_store_ps(pos_.f, _mm_castsi128_ps(pos4));
 125         _mm_store_ps(frac_.f, _mm_castsi128_ps(frac4));
 126     }
 127
 128     pos = pos_.i[0];
 129     frac = frac_.i[0];
 130
 131     for(;i < numsamples;i++)
 132     {
 133         dst[i] = cubic(src[pos], src[pos+1], src[pos+2], src[pos+3], frac);
 134
 135         frac += increment;
 136         pos  += frac>>FRACTIONBITS;
 137         frac &= FRACTIONMASK;
 138     }
 139     return dst;
 140 }