• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 #include "pitch.h"
36 #include "main.h"
37 
38 #if !defined(OPUS_X86_PRESUME_SSE4_1)
39 
40 #if defined(FIXED_POINT)
41 
42 #include "fixed/main_FIX.h"
43 
44 opus_int64 (*const SILK_INNER_PROD16_ALIGNED_64_IMPL[ OPUS_ARCHMASK + 1 ] )(
45     const opus_int16 *inVec1,
46     const opus_int16 *inVec2,
47     const opus_int   len
48 ) = {
49   silk_inner_prod16_aligned_64_c,                  /* non-sse */
50   silk_inner_prod16_aligned_64_c,
51   silk_inner_prod16_aligned_64_c,
52   MAY_HAVE_SSE4_1( silk_inner_prod16_aligned_64 ), /* sse4.1 */
53   MAY_HAVE_SSE4_1( silk_inner_prod16_aligned_64 )  /* avx */
54 };
55 
56 #endif
57 
58 opus_int (*const SILK_VAD_GETSA_Q8_IMPL[ OPUS_ARCHMASK + 1 ] )(
59     silk_encoder_state *psEncC,
60     const opus_int16   pIn[]
61 ) = {
62   silk_VAD_GetSA_Q8_c,                  /* non-sse */
63   silk_VAD_GetSA_Q8_c,
64   silk_VAD_GetSA_Q8_c,
65   MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ), /* sse4.1 */
66   MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 )  /* avx */
67 };
68 
69 #if 0 /* FIXME: SSE disabled until the NSQ code gets updated. */
70 void (*const SILK_NSQ_IMPL[ OPUS_ARCHMASK + 1 ] )(
71     const silk_encoder_state    *psEncC,                                    /* I    Encoder State                   */
72     silk_nsq_state              *NSQ,                                       /* I/O  NSQ state                       */
73     SideInfoIndices             *psIndices,                                 /* I/O  Quantization Indices            */
74     const opus_int32            x_Q3[],                                     /* I    Prefiltered input signal        */
75     opus_int8                   pulses[],                                   /* O    Quantized pulse signal          */
76     const opus_int16            PredCoef_Q12[ 2 * MAX_LPC_ORDER ],          /* I    Short term prediction coefs     */
77     const opus_int16            LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ],    /* I    Long term prediction coefs      */
78     const opus_int16            AR2_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs             */
79     const opus_int              HarmShapeGain_Q14[ MAX_NB_SUBFR ],          /* I    Long term shaping coefs         */
80     const opus_int              Tilt_Q14[ MAX_NB_SUBFR ],                   /* I    Spectral tilt                   */
81     const opus_int32            LF_shp_Q14[ MAX_NB_SUBFR ],                 /* I    Low frequency shaping coefs     */
82     const opus_int32            Gains_Q16[ MAX_NB_SUBFR ],                  /* I    Quantization step sizes         */
83     const opus_int              pitchL[ MAX_NB_SUBFR ],                     /* I    Pitch lags                      */
84     const opus_int              Lambda_Q10,                                 /* I    Rate/distortion tradeoff        */
85     const opus_int              LTP_scale_Q14                               /* I    LTP state scaling               */
86 ) = {
87   silk_NSQ_c,                  /* non-sse */
88   silk_NSQ_c,
89   silk_NSQ_c,
90   MAY_HAVE_SSE4_1( silk_NSQ ), /* sse4.1 */
91   MAY_HAVE_SSE4_1( silk_NSQ )  /* avx */
92 };
93 #endif
94 
95 #if 0 /* FIXME: SSE disabled until silk_VQ_WMat_EC_sse4_1() gets updated. */
96 void (*const SILK_VQ_WMAT_EC_IMPL[ OPUS_ARCHMASK + 1 ] )(
97     opus_int8                   *ind,                           /* O    index of best codebook vector               */
98     opus_int32                  *rate_dist_Q14,                 /* O    best weighted quant error + mu * rate       */
99     opus_int                    *gain_Q7,                       /* O    sum of absolute LTP coefficients            */
100     const opus_int16            *in_Q14,                        /* I    input vector to be quantized                */
101     const opus_int32            *W_Q18,                         /* I    weighting matrix                            */
102     const opus_int8             *cb_Q7,                         /* I    codebook                                    */
103     const opus_uint8            *cb_gain_Q7,                    /* I    codebook effective gain                     */
104     const opus_uint8            *cl_Q5,                         /* I    code length for each codebook vector        */
105     const opus_int              mu_Q9,                          /* I    tradeoff betw. weighted error and rate      */
106     const opus_int32            max_gain_Q7,                    /* I    maximum sum of absolute LTP coefficients    */
107     opus_int                    L                               /* I    number of vectors in codebook               */
108 ) = {
109   silk_VQ_WMat_EC_c,                  /* non-sse */
110   silk_VQ_WMat_EC_c,
111   silk_VQ_WMat_EC_c,
112   MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ), /* sse4.1 */
113   MAY_HAVE_SSE4_1( silk_VQ_WMat_EC )  /* avx */
114 };
115 #endif
116 
117 #if 0 /* FIXME: SSE disabled until the NSQ code gets updated. */
118 void (*const SILK_NSQ_DEL_DEC_IMPL[ OPUS_ARCHMASK + 1 ] )(
119     const silk_encoder_state    *psEncC,                                    /* I    Encoder State                   */
120     silk_nsq_state              *NSQ,                                       /* I/O  NSQ state                       */
121     SideInfoIndices             *psIndices,                                 /* I/O  Quantization Indices            */
122     const opus_int32            x_Q3[],                                     /* I    Prefiltered input signal        */
123     opus_int8                   pulses[],                                   /* O    Quantized pulse signal          */
124     const opus_int16            PredCoef_Q12[ 2 * MAX_LPC_ORDER ],          /* I    Short term prediction coefs     */
125     const opus_int16            LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ],    /* I    Long term prediction coefs      */
126     const opus_int16            AR2_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs             */
127     const opus_int              HarmShapeGain_Q14[ MAX_NB_SUBFR ],          /* I    Long term shaping coefs         */
128     const opus_int              Tilt_Q14[ MAX_NB_SUBFR ],                   /* I    Spectral tilt                   */
129     const opus_int32            LF_shp_Q14[ MAX_NB_SUBFR ],                 /* I    Low frequency shaping coefs     */
130     const opus_int32            Gains_Q16[ MAX_NB_SUBFR ],                  /* I    Quantization step sizes         */
131     const opus_int              pitchL[ MAX_NB_SUBFR ],                     /* I    Pitch lags                      */
132     const opus_int              Lambda_Q10,                                 /* I    Rate/distortion tradeoff        */
133     const opus_int              LTP_scale_Q14                               /* I    LTP state scaling               */
134 ) = {
135   silk_NSQ_del_dec_c,                  /* non-sse */
136   silk_NSQ_del_dec_c,
137   silk_NSQ_del_dec_c,
138   MAY_HAVE_SSE4_1( silk_NSQ_del_dec ), /* sse4.1 */
139   MAY_HAVE_SSE4_1( silk_NSQ_del_dec )  /* avx */
140 };
141 #endif
142 
143 #if defined(FIXED_POINT)
144 
145 void (*const SILK_BURG_MODIFIED_IMPL[ OPUS_ARCHMASK + 1 ] )(
146     opus_int32                  *res_nrg,           /* O    Residual energy                                             */
147     opus_int                    *res_nrg_Q,         /* O    Residual energy Q value                                     */
148     opus_int32                  A_Q16[],            /* O    Prediction coefficients (length order)                      */
149     const opus_int16            x[],                /* I    Input signal, length: nb_subfr * ( D + subfr_length )       */
150     const opus_int32            minInvGain_Q30,     /* I    Inverse of max prediction gain                              */
151     const opus_int              subfr_length,       /* I    Input signal subframe length (incl. D preceding samples)    */
152     const opus_int              nb_subfr,           /* I    Number of subframes stacked in x                            */
153     const opus_int              D,                  /* I    Order                                                       */
154     int                         arch                /* I    Run-time architecture                                       */
155 ) = {
156   silk_burg_modified_c,                  /* non-sse */
157   silk_burg_modified_c,
158   silk_burg_modified_c,
159   MAY_HAVE_SSE4_1( silk_burg_modified ), /* sse4.1 */
160   MAY_HAVE_SSE4_1( silk_burg_modified )  /* avx */
161 };
162 
163 #endif
164 #endif
165