1 /* Copyright (c) 2014, Cisco Systems, INC 2 Written by XiangMingZhu WeiZhou MinPeng YanWang 3 4 Redistribution and use in source and binary forms, with or without 5 modification, are permitted provided that the following conditions 6 are met: 7 8 - Redistributions of source code must retain the above copyright 9 notice, this list of conditions and the following disclaimer. 10 11 - Redistributions in binary form must reproduce the above copyright 12 notice, this list of conditions and the following disclaimer in the 13 documentation and/or other materials provided with the distribution. 14 15 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 16 ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 17 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 18 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER 19 OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 20 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 21 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 22 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 23 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 24 NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 25 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #if defined(HAVE_CONFIG_H) 29 #include "config.h" 30 #endif 31 32 #include "x86/x86cpu.h" 33 #include "celt_lpc.h" 34 #include "pitch.h" 35 #include "pitch_sse.h" 36 #include "vq.h" 37 38 #if defined(OPUS_HAVE_RTCD) 39 40 # if defined(FIXED_POINT) 41 42 #if defined(OPUS_X86_MAY_HAVE_SSE4_1) && !defined(OPUS_X86_PRESUME_SSE4_1) 43 44 void (*const CELT_FIR_IMPL[OPUS_ARCHMASK + 1])( 45 const opus_val16 *x, 46 const opus_val16 *num, 47 opus_val16 *y, 48 int N, 49 int ord, 50 int arch 51 ) = { 52 celt_fir_c, /* non-sse */ 53 celt_fir_c, 54 celt_fir_c, 55 MAY_HAVE_SSE4_1(celt_fir), /* sse4.1 */ 56 MAY_HAVE_SSE4_1(celt_fir) /* avx */ 57 }; 58 59 void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])( 60 const opus_val16 *x, 61 const opus_val16 *y, 62 opus_val32 sum[4], 63 int len 64 ) = { 65 xcorr_kernel_c, /* non-sse */ 66 xcorr_kernel_c, 67 xcorr_kernel_c, 68 MAY_HAVE_SSE4_1(xcorr_kernel), /* sse4.1 */ 69 MAY_HAVE_SSE4_1(xcorr_kernel) /* avx */ 70 }; 71 72 #endif 73 74 #if (defined(OPUS_X86_MAY_HAVE_SSE4_1) && !defined(OPUS_X86_PRESUME_SSE4_1)) || \ 75 (!defined(OPUS_X86_MAY_HAVE_SSE_4_1) && defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(OPUS_X86_PRESUME_SSE2)) 76 77 opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])( 78 const opus_val16 *x, 79 const opus_val16 *y, 80 int N 81 ) = { 82 celt_inner_prod_c, /* non-sse */ 83 celt_inner_prod_c, 84 MAY_HAVE_SSE2(celt_inner_prod), 85 MAY_HAVE_SSE4_1(celt_inner_prod), /* sse4.1 */ 86 MAY_HAVE_SSE4_1(celt_inner_prod) /* avx */ 87 }; 88 89 #endif 90 91 # else 92 93 #if defined(OPUS_X86_MAY_HAVE_AVX2) && !defined(OPUS_X86_PRESUME_AVX2) 94 95 void (*const PITCH_XCORR_IMPL[OPUS_ARCHMASK + 1])( 96 const float *_x, 97 const float *_y, 98 float *xcorr, 99 int len, 100 int max_pitch, 101 int arch 102 ) = { 103 celt_pitch_xcorr_c, /* non-sse */ 104 celt_pitch_xcorr_c, 105 celt_pitch_xcorr_c, 106 celt_pitch_xcorr_c, 107 MAY_HAVE_AVX2(celt_pitch_xcorr) 108 }; 109 110 #endif 111 112 113 #if defined(OPUS_X86_MAY_HAVE_SSE) && !defined(OPUS_X86_PRESUME_SSE) 114 115 void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])( 116 const opus_val16 *x, 117 const opus_val16 *y, 118 opus_val32 sum[4], 119 int len 120 ) = { 121 xcorr_kernel_c, /* non-sse */ 122 MAY_HAVE_SSE(xcorr_kernel), 123 MAY_HAVE_SSE(xcorr_kernel), 124 MAY_HAVE_SSE(xcorr_kernel), 125 MAY_HAVE_SSE(xcorr_kernel) 126 }; 127 128 opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])( 129 const opus_val16 *x, 130 const opus_val16 *y, 131 int N 132 ) = { 133 celt_inner_prod_c, /* non-sse */ 134 MAY_HAVE_SSE(celt_inner_prod), 135 MAY_HAVE_SSE(celt_inner_prod), 136 MAY_HAVE_SSE(celt_inner_prod), 137 MAY_HAVE_SSE(celt_inner_prod) 138 }; 139 140 void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])( 141 const opus_val16 *x, 142 const opus_val16 *y01, 143 const opus_val16 *y02, 144 int N, 145 opus_val32 *xy1, 146 opus_val32 *xy2 147 ) = { 148 dual_inner_prod_c, /* non-sse */ 149 MAY_HAVE_SSE(dual_inner_prod), 150 MAY_HAVE_SSE(dual_inner_prod), 151 MAY_HAVE_SSE(dual_inner_prod), 152 MAY_HAVE_SSE(dual_inner_prod) 153 }; 154 155 void (*const COMB_FILTER_CONST_IMPL[OPUS_ARCHMASK + 1])( 156 opus_val32 *y, 157 opus_val32 *x, 158 int T, 159 int N, 160 opus_val16 g10, 161 opus_val16 g11, 162 opus_val16 g12 163 ) = { 164 comb_filter_const_c, /* non-sse */ 165 MAY_HAVE_SSE(comb_filter_const), 166 MAY_HAVE_SSE(comb_filter_const), 167 MAY_HAVE_SSE(comb_filter_const), 168 MAY_HAVE_SSE(comb_filter_const) 169 }; 170 171 172 #endif 173 174 #if defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(OPUS_X86_PRESUME_SSE2) 175 opus_val16 (*const OP_PVQ_SEARCH_IMPL[OPUS_ARCHMASK + 1])( 176 celt_norm *_X, int *iy, int K, int N, int arch 177 ) = { 178 op_pvq_search_c, /* non-sse */ 179 op_pvq_search_c, 180 MAY_HAVE_SSE2(op_pvq_search), 181 MAY_HAVE_SSE2(op_pvq_search), 182 MAY_HAVE_SSE2(op_pvq_search) 183 }; 184 #endif 185 186 #endif 187 #endif 188