Chris@69: /* Copyright (c) 2014, Cisco Systems, INC Chris@69: Written by XiangMingZhu WeiZhou MinPeng YanWang Chris@69: Chris@69: Redistribution and use in source and binary forms, with or without Chris@69: modification, are permitted provided that the following conditions Chris@69: are met: Chris@69: Chris@69: - Redistributions of source code must retain the above copyright Chris@69: notice, this list of conditions and the following disclaimer. Chris@69: Chris@69: - Redistributions in binary form must reproduce the above copyright Chris@69: notice, this list of conditions and the following disclaimer in the Chris@69: documentation and/or other materials provided with the distribution. Chris@69: Chris@69: THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS Chris@69: ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT Chris@69: LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR Chris@69: A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER Chris@69: OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, Chris@69: EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, Chris@69: PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR Chris@69: PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF Chris@69: LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING Chris@69: NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS Chris@69: SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Chris@69: */ Chris@69: Chris@69: #if defined(HAVE_CONFIG_H) Chris@69: #include "config.h" Chris@69: #endif Chris@69: Chris@69: #include "celt/x86/x86cpu.h" Chris@69: #include "structs.h" Chris@69: #include "SigProc_FIX.h" Chris@69: #include "pitch.h" Chris@69: #include "main.h" Chris@69: Chris@69: #if !defined(OPUS_X86_PRESUME_SSE4_1) Chris@69: Chris@69: #if defined(FIXED_POINT) Chris@69: Chris@69: #include "fixed/main_FIX.h" Chris@69: Chris@69: opus_int64 (*const SILK_INNER_PROD16_ALIGNED_64_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: const opus_int16 *inVec1, Chris@69: const opus_int16 *inVec2, Chris@69: const opus_int len Chris@69: ) = { Chris@69: silk_inner_prod16_aligned_64_c, /* non-sse */ Chris@69: silk_inner_prod16_aligned_64_c, Chris@69: silk_inner_prod16_aligned_64_c, Chris@69: MAY_HAVE_SSE4_1( silk_inner_prod16_aligned_64 ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_inner_prod16_aligned_64 ) /* avx */ Chris@69: }; Chris@69: Chris@69: #endif Chris@69: Chris@69: opus_int (*const SILK_VAD_GETSA_Q8_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: silk_encoder_state *psEncC, Chris@69: const opus_int16 pIn[] Chris@69: ) = { Chris@69: silk_VAD_GetSA_Q8_c, /* non-sse */ Chris@69: silk_VAD_GetSA_Q8_c, Chris@69: silk_VAD_GetSA_Q8_c, Chris@69: MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ) /* avx */ Chris@69: }; Chris@69: Chris@69: #if 0 /* FIXME: SSE disabled until the NSQ code gets updated. */ Chris@69: void (*const SILK_NSQ_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: const silk_encoder_state *psEncC, /* I Encoder State */ Chris@69: silk_nsq_state *NSQ, /* I/O NSQ state */ Chris@69: SideInfoIndices *psIndices, /* I/O Quantization Indices */ Chris@69: const opus_int32 x_Q3[], /* I Prefiltered input signal */ Chris@69: opus_int8 pulses[], /* O Quantized pulse signal */ Chris@69: const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ Chris@69: const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ Chris@69: const opus_int16 AR2_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ Chris@69: const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ Chris@69: const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ Chris@69: const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ Chris@69: const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ Chris@69: const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ Chris@69: const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ Chris@69: const opus_int LTP_scale_Q14 /* I LTP state scaling */ Chris@69: ) = { Chris@69: silk_NSQ_c, /* non-sse */ Chris@69: silk_NSQ_c, Chris@69: silk_NSQ_c, Chris@69: MAY_HAVE_SSE4_1( silk_NSQ ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_NSQ ) /* avx */ Chris@69: }; Chris@69: #endif Chris@69: Chris@69: #if 0 /* FIXME: SSE disabled until silk_VQ_WMat_EC_sse4_1() gets updated. */ Chris@69: void (*const SILK_VQ_WMAT_EC_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: opus_int8 *ind, /* O index of best codebook vector */ Chris@69: opus_int32 *rate_dist_Q14, /* O best weighted quant error + mu * rate */ Chris@69: opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ Chris@69: const opus_int16 *in_Q14, /* I input vector to be quantized */ Chris@69: const opus_int32 *W_Q18, /* I weighting matrix */ Chris@69: const opus_int8 *cb_Q7, /* I codebook */ Chris@69: const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ Chris@69: const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ Chris@69: const opus_int mu_Q9, /* I tradeoff betw. weighted error and rate */ Chris@69: const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ Chris@69: opus_int L /* I number of vectors in codebook */ Chris@69: ) = { Chris@69: silk_VQ_WMat_EC_c, /* non-sse */ Chris@69: silk_VQ_WMat_EC_c, Chris@69: silk_VQ_WMat_EC_c, Chris@69: MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ) /* avx */ Chris@69: }; Chris@69: #endif Chris@69: Chris@69: #if 0 /* FIXME: SSE disabled until the NSQ code gets updated. */ Chris@69: void (*const SILK_NSQ_DEL_DEC_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: const silk_encoder_state *psEncC, /* I Encoder State */ Chris@69: silk_nsq_state *NSQ, /* I/O NSQ state */ Chris@69: SideInfoIndices *psIndices, /* I/O Quantization Indices */ Chris@69: const opus_int32 x_Q3[], /* I Prefiltered input signal */ Chris@69: opus_int8 pulses[], /* O Quantized pulse signal */ Chris@69: const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ Chris@69: const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ Chris@69: const opus_int16 AR2_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ Chris@69: const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ Chris@69: const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ Chris@69: const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ Chris@69: const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ Chris@69: const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ Chris@69: const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ Chris@69: const opus_int LTP_scale_Q14 /* I LTP state scaling */ Chris@69: ) = { Chris@69: silk_NSQ_del_dec_c, /* non-sse */ Chris@69: silk_NSQ_del_dec_c, Chris@69: silk_NSQ_del_dec_c, Chris@69: MAY_HAVE_SSE4_1( silk_NSQ_del_dec ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_NSQ_del_dec ) /* avx */ Chris@69: }; Chris@69: #endif Chris@69: Chris@69: #if defined(FIXED_POINT) Chris@69: Chris@69: void (*const SILK_BURG_MODIFIED_IMPL[ OPUS_ARCHMASK + 1 ] )( Chris@69: opus_int32 *res_nrg, /* O Residual energy */ Chris@69: opus_int *res_nrg_Q, /* O Residual energy Q value */ Chris@69: opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ Chris@69: const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ Chris@69: const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ Chris@69: const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ Chris@69: const opus_int nb_subfr, /* I Number of subframes stacked in x */ Chris@69: const opus_int D, /* I Order */ Chris@69: int arch /* I Run-time architecture */ Chris@69: ) = { Chris@69: silk_burg_modified_c, /* non-sse */ Chris@69: silk_burg_modified_c, Chris@69: silk_burg_modified_c, Chris@69: MAY_HAVE_SSE4_1( silk_burg_modified ), /* sse4.1 */ Chris@69: MAY_HAVE_SSE4_1( silk_burg_modified ) /* avx */ Chris@69: }; Chris@69: Chris@69: #endif Chris@69: #endif