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 "celt/x86/x86cpu.h" 33 #include "structs.h" 34 #include "SigProc_FIX.h" 35 #ifndef FIXED_POINT 36 #include "SigProc_FLP.h" 37 #endif 38 #include "pitch.h" 39 #include "main.h" 40 41 #if defined(OPUS_HAVE_RTCD) && !defined(OPUS_X86_PRESUME_AVX2) 42 43 #if defined(FIXED_POINT) 44 45 #include "fixed/main_FIX.h" 46 47 opus_int64 (*const SILK_INNER_PROD16_IMPL[ OPUS_ARCHMASK + 1 ] )( 48 const opus_int16 *inVec1, 49 const opus_int16 *inVec2, 50 const opus_int len 51 ) = { 52 silk_inner_prod16_c, /* non-sse */ 53 silk_inner_prod16_c, 54 silk_inner_prod16_c, 55 MAY_HAVE_SSE4_1( silk_inner_prod16 ), /* sse4.1 */ 56 MAY_HAVE_SSE4_1( silk_inner_prod16 ) /* avx */ 57 }; 58 59 #endif 60 61 opus_int (*const SILK_VAD_GETSA_Q8_IMPL[ OPUS_ARCHMASK + 1 ] )( 62 silk_encoder_state *psEncC, 63 const opus_int16 pIn[] 64 ) = { 65 silk_VAD_GetSA_Q8_c, /* non-sse */ 66 silk_VAD_GetSA_Q8_c, 67 silk_VAD_GetSA_Q8_c, 68 MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ), /* sse4.1 */ 69 MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ) /* avx */ 70 }; 71 72 void (*const SILK_NSQ_IMPL[ OPUS_ARCHMASK + 1 ] )( 73 const silk_encoder_state *psEncC, /* I Encoder State */ 74 silk_nsq_state *NSQ, /* I/O NSQ state */ 75 SideInfoIndices *psIndices, /* I/O Quantization Indices */ 76 const opus_int16 x16[], /* I Input */ 77 opus_int8 pulses[], /* O Quantized pulse signal */ 78 const opus_int16 *PredCoef_Q12, /* I Short term prediction coefs */ 79 const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ 80 const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ 81 const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ 82 const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ 83 const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ 84 const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ 85 const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ 86 const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ 87 const opus_int LTP_scale_Q14 /* I LTP state scaling */ 88 ) = { 89 silk_NSQ_c, /* non-sse */ 90 silk_NSQ_c, 91 silk_NSQ_c, 92 MAY_HAVE_SSE4_1( silk_NSQ ), /* sse4.1 */ 93 MAY_HAVE_SSE4_1( silk_NSQ ) /* avx */ 94 }; 95 96 void (*const SILK_VQ_WMAT_EC_IMPL[ OPUS_ARCHMASK + 1 ] )( 97 opus_int8 *ind, /* O index of best codebook vector */ 98 opus_int32 *res_nrg_Q15, /* O best residual energy */ 99 opus_int32 *rate_dist_Q8, /* O best total bitrate */ 100 opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ 101 const opus_int32 *XX_Q17, /* I correlation matrix */ 102 const opus_int32 *xX_Q17, /* I correlation vector */ 103 const opus_int8 *cb_Q7, /* I codebook */ 104 const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ 105 const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ 106 const opus_int subfr_len, /* I number of samples per subframe */ 107 const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ 108 const opus_int L /* I number of vectors in codebook */ 109 ) = { 110 silk_VQ_WMat_EC_c, /* non-sse */ 111 silk_VQ_WMat_EC_c, 112 silk_VQ_WMat_EC_c, 113 MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ), /* sse4.1 */ 114 MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ) /* avx */ 115 }; 116 117 void (*const SILK_NSQ_DEL_DEC_IMPL[ OPUS_ARCHMASK + 1 ] )( 118 const silk_encoder_state *psEncC, /* I Encoder State */ 119 silk_nsq_state *NSQ, /* I/O NSQ state */ 120 SideInfoIndices *psIndices, /* I/O Quantization Indices */ 121 const opus_int16 x16[], /* I Input */ 122 opus_int8 pulses[], /* O Quantized pulse signal */ 123 const opus_int16 *PredCoef_Q12, /* I Short term prediction coefs */ 124 const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ 125 const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ 126 const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ 127 const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ 128 const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ 129 const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ 130 const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ 131 const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ 132 const opus_int LTP_scale_Q14 /* I LTP state scaling */ 133 ) = { 134 silk_NSQ_del_dec_c, /* non-sse */ 135 silk_NSQ_del_dec_c, 136 silk_NSQ_del_dec_c, 137 MAY_HAVE_SSE4_1( silk_NSQ_del_dec ), /* sse4.1 */ 138 MAY_HAVE_AVX2( silk_NSQ_del_dec ) /* avx */ 139 }; 140 141 #if defined(FIXED_POINT) 142 143 void (*const SILK_BURG_MODIFIED_IMPL[ OPUS_ARCHMASK + 1 ] )( 144 opus_int32 *res_nrg, /* O Residual energy */ 145 opus_int *res_nrg_Q, /* O Residual energy Q value */ 146 opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ 147 const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ 148 const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ 149 const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ 150 const opus_int nb_subfr, /* I Number of subframes stacked in x */ 151 const opus_int D, /* I Order */ 152 int arch /* I Run-time architecture */ 153 ) = { 154 silk_burg_modified_c, /* non-sse */ 155 silk_burg_modified_c, 156 silk_burg_modified_c, 157 MAY_HAVE_SSE4_1( silk_burg_modified ), /* sse4.1 */ 158 MAY_HAVE_SSE4_1( silk_burg_modified ) /* avx */ 159 }; 160 161 #endif 162 163 #ifndef FIXED_POINT 164 165 double (*const SILK_INNER_PRODUCT_FLP_IMPL[ OPUS_ARCHMASK + 1 ] )( 166 const silk_float *data1, 167 const silk_float *data2, 168 opus_int dataSize 169 ) = { 170 silk_inner_product_FLP_c, /* non-sse */ 171 silk_inner_product_FLP_c, 172 silk_inner_product_FLP_c, 173 silk_inner_product_FLP_c, /* sse4.1 */ 174 MAY_HAVE_AVX2( silk_inner_product_FLP ) /* avx */ 175 }; 176 177 #endif 178 179 #endif 180