Alc/mixer_sse41.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 #include <smmintrin.h>
  26
  27 #include "alu.h"
  28 #include "mixer_defs.h"
  29
  30
  31 const ALfloat *Resample_lerp32_SSE41(const ALfloat *src, ALuint frac, ALuint increment,
  32                                      ALfloat *restrict dst, ALuint numsamples)
  33 {
  34     const __m128i increment4 = _mm_set1_epi32(increment*4);
  35     const __m128 fracOne4 = _mm_set1_ps(1.0f/FRACTIONONE);
  36     const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
  37     alignas(16) union { ALuint i[4]; float f[4]; } pos_;
  38     alignas(16) union { ALuint i[4]; float f[4]; } frac_;
  39     __m128i frac4, pos4;
  40     ALuint pos;
  41     ALuint i;
  42
  43     InitiatePositionArrays(frac, increment, frac_.i, pos_.i, 4);
  44
  45     frac4 = _mm_castps_si128(_mm_load_ps(frac_.f));
  46     pos4 = _mm_castps_si128(_mm_load_ps(pos_.f));
  47
  48     for(i = 0;numsamples-i > 3;i += 4)
  49     {
  50         const __m128 val1 = _mm_setr_ps(src[pos_.i[0]], src[pos_.i[1]], src[pos_.i[2]], src[pos_.i[3]]);
  51         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]);
  52
  53         /* val1 + (val2-val1)*mu */
  54         const __m128 r0 = _mm_sub_ps(val2, val1);
  55         const __m128 mu = _mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4);
  56         const __m128 out = _mm_add_ps(val1, _mm_mul_ps(mu, r0));
  57
  58         _mm_store_ps(&dst[i], out);
  59
  60         frac4 = _mm_add_epi32(frac4, increment4);
  61         pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
  62         frac4 = _mm_and_si128(frac4, fracMask4);
  63
  64         pos_.i[0] = _mm_extract_epi32(pos4, 0);
  65         pos_.i[1] = _mm_extract_epi32(pos4, 1);
  66         pos_.i[2] = _mm_extract_epi32(pos4, 2);
  67         pos_.i[3] = _mm_extract_epi32(pos4, 3);
  68     }
  69
  70     pos = pos_.i[0];
  71     frac = _mm_cvtsi128_si32(frac4);
  72
  73     for(;i < numsamples;i++)
  74     {
  75         dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));
  76
  77         frac += increment;
  78         pos  += frac>>FRACTIONBITS;
  79         frac &= FRACTIONMASK;
  80     }
  81     return dst;
  82 }
  83
  84 const ALfloat *Resample_cubic32_SSE41(const ALfloat *src, ALuint frac, ALuint increment,
  85                                       ALfloat *restrict dst, ALuint numsamples)
  86 {
  87     const __m128i increment4 = _mm_set1_epi32(increment*4);
  88     const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
  89     alignas(16) union { ALuint i[4]; float f[4]; } pos_;
  90     alignas(16) union { ALuint i[4]; float f[4]; } frac_;
  91     __m128i frac4, pos4;
  92     ALuint pos;
  93     ALuint i;
  94
  95     InitiatePositionArrays(frac, increment, frac_.i, pos_.i, 4);
  96
  97     frac4 = _mm_castps_si128(_mm_load_ps(frac_.f));
  98     pos4 = _mm_castps_si128(_mm_load_ps(pos_.f));
  99
 100     --src;
 101     for(i = 0;numsamples-i > 3;i += 4)
 102     {
 103         const __m128 val0 = _mm_loadu_ps(&src[pos_.i[0]]);
 104         const __m128 val1 = _mm_loadu_ps(&src[pos_.i[1]]);
 105         const __m128 val2 = _mm_loadu_ps(&src[pos_.i[2]]);
 106         const __m128 val3 = _mm_loadu_ps(&src[pos_.i[3]]);
 107         __m128 k0 = _mm_load_ps(CubicLUT[frac_.i[0]]);
 108         __m128 k1 = _mm_load_ps(CubicLUT[frac_.i[1]]);
 109         __m128 k2 = _mm_load_ps(CubicLUT[frac_.i[2]]);
 110         __m128 k3 = _mm_load_ps(CubicLUT[frac_.i[3]]);
 111         __m128 out;
 112
 113         k0 = _mm_mul_ps(k0, val0);
 114         k1 = _mm_mul_ps(k1, val1);
 115         k2 = _mm_mul_ps(k2, val2);
 116         k3 = _mm_mul_ps(k3, val3);
 117         _MM_TRANSPOSE4_PS(k0, k1, k2, k3);
 118         out = _mm_add_ps(k0, k1);
 119         out = _mm_add_ps(out, k2);
 120         out = _mm_add_ps(out, k3);
 121
 122         _mm_store_ps(&dst[i], out);
 123
 124         frac4 = _mm_add_epi32(frac4, increment4);
 125         pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
 126         frac4 = _mm_and_si128(frac4, fracMask4);
 127
 128         pos_.i[0] = _mm_extract_epi32(pos4, 0);
 129         pos_.i[1] = _mm_extract_epi32(pos4, 1);
 130         pos_.i[2] = _mm_extract_epi32(pos4, 2);
 131         pos_.i[3] = _mm_extract_epi32(pos4, 3);
 132         frac_.i[0] = _mm_extract_epi32(frac4, 0);
 133         frac_.i[1] = _mm_extract_epi32(frac4, 1);
 134         frac_.i[2] = _mm_extract_epi32(frac4, 2);
 135         frac_.i[3] = _mm_extract_epi32(frac4, 3);
 136     }
 137
 138     pos = pos_.i[0];
 139     frac = frac_.i[0];
 140
 141     for(;i < numsamples;i++)
 142     {
 143         dst[i] = cubic(src[pos], src[pos+1], src[pos+2], src[pos+3], frac);
 144
 145         frac += increment;
 146         pos  += frac>>FRACTIONBITS;
 147         frac &= FRACTIONMASK;
 148     }
 149     return dst;
 150 }