ffmpeg/libswresample/x86/rematrix_init.c

   1 /*
   2  * Copyright (C) 2012 Michael Niedermayer (michaelni@gmx.at)
   3  *
   4  * This file is part of libswresample
   5  *
   6  * libswresample is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * libswresample is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with libswresample; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include "libavutil/x86/cpu.h"
  22 #include "libswresample/swresample_internal.h"
  23
  24 #define D(type, simd) \
  25 mix_1_1_func_type ff_mix_1_1_a_## type ## _ ## simd;\
  26 mix_2_1_func_type ff_mix_2_1_a_## type ## _ ## simd;
  27
  28 D(float, sse)
  29 D(float, avx)
  30 D(int16, mmx)
  31 D(int16, sse2)
  32
  33 av_cold void swri_rematrix_init_x86(struct SwrContext *s){
  34 #if HAVE_YASM
  35     int mm_flags = av_get_cpu_flags();
  36     int nb_in  = av_get_channel_layout_nb_channels(s->in_ch_layout);
  37     int nb_out = av_get_channel_layout_nb_channels(s->out_ch_layout);
  38     int num    = nb_in * nb_out;
  39     int i,j;
  40
  41     s->mix_1_1_simd = NULL;
  42     s->mix_2_1_simd = NULL;
  43
  44     if (s->midbuf.fmt == AV_SAMPLE_FMT_S16P){
  45         if(EXTERNAL_MMX(mm_flags)) {
  46             s->mix_1_1_simd = ff_mix_1_1_a_int16_mmx;
  47             s->mix_2_1_simd = ff_mix_2_1_a_int16_mmx;
  48         }
  49         if(EXTERNAL_SSE2(mm_flags)) {
  50             s->mix_1_1_simd = ff_mix_1_1_a_int16_sse2;
  51             s->mix_2_1_simd = ff_mix_2_1_a_int16_sse2;
  52         }
  53         s->native_simd_matrix = av_mallocz(2 * num * sizeof(int16_t));
  54         s->native_simd_one    = av_mallocz(2 * sizeof(int16_t));
  55         for(i=0; i<nb_out; i++){
  56             int sh = 0;
  57             for(j=0; j<nb_in; j++)
  58                 sh = FFMAX(sh, FFABS(((int*)s->native_matrix)[i * nb_in + j]));
  59             sh = FFMAX(av_log2(sh) - 14, 0);
  60             for(j=0; j<nb_in; j++) {
  61                 ((int16_t*)s->native_simd_matrix)[2*(i * nb_in + j)+1] = 15 - sh;
  62                 ((int16_t*)s->native_simd_matrix)[2*(i * nb_in + j)] =
  63                     ((((int*)s->native_matrix)[i * nb_in + j]) + (1<<sh>>1)) >> sh;
  64             }
  65         }
  66         ((int16_t*)s->native_simd_one)[1] = 14;
  67         ((int16_t*)s->native_simd_one)[0] = 16384;
  68     } else if(s->midbuf.fmt == AV_SAMPLE_FMT_FLTP){
  69         if(EXTERNAL_SSE(mm_flags)) {
  70             s->mix_1_1_simd = ff_mix_1_1_a_float_sse;
  71             s->mix_2_1_simd = ff_mix_2_1_a_float_sse;
  72         }
  73         if(EXTERNAL_AVX(mm_flags)) {
  74             s->mix_1_1_simd = ff_mix_1_1_a_float_avx;
  75             s->mix_2_1_simd = ff_mix_2_1_a_float_avx;
  76         }
  77         s->native_simd_matrix = av_mallocz(num * sizeof(float));
  78         memcpy(s->native_simd_matrix, s->native_matrix, num * sizeof(float));
  79         s->native_simd_one = av_mallocz(sizeof(float));
  80         memcpy(s->native_simd_one, s->native_one, sizeof(float));
  81     }
  82 #endif
  83 }