diff options
Diffstat (limited to 'lib/rbcodec/codecs/libopus/silk/fixed/x86/prefilter_FIX_sse.c')
-rw-r--r-- | lib/rbcodec/codecs/libopus/silk/fixed/x86/prefilter_FIX_sse.c | 160 |
1 files changed, 160 insertions, 0 deletions
diff --git a/lib/rbcodec/codecs/libopus/silk/fixed/x86/prefilter_FIX_sse.c b/lib/rbcodec/codecs/libopus/silk/fixed/x86/prefilter_FIX_sse.c new file mode 100644 index 0000000000..555432cd96 --- /dev/null +++ b/lib/rbcodec/codecs/libopus/silk/fixed/x86/prefilter_FIX_sse.c | |||
@@ -0,0 +1,160 @@ | |||
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 | #ifdef HAVE_CONFIG_H | ||
29 | #include "config.h" | ||
30 | #endif | ||
31 | |||
32 | #include <xmmintrin.h> | ||
33 | #include <emmintrin.h> | ||
34 | #include <smmintrin.h> | ||
35 | #include "main.h" | ||
36 | #include "celt/x86/x86cpu.h" | ||
37 | |||
38 | void silk_warped_LPC_analysis_filter_FIX_sse4_1( | ||
39 | opus_int32 state[], /* I/O State [order + 1] */ | ||
40 | opus_int32 res_Q2[], /* O Residual signal [length] */ | ||
41 | const opus_int16 coef_Q13[], /* I Coefficients [order] */ | ||
42 | const opus_int16 input[], /* I Input signal [length] */ | ||
43 | const opus_int16 lambda_Q16, /* I Warping factor */ | ||
44 | const opus_int length, /* I Length of input signal */ | ||
45 | const opus_int order /* I Filter order (even) */ | ||
46 | ) | ||
47 | { | ||
48 | opus_int n, i; | ||
49 | opus_int32 acc_Q11, tmp1, tmp2; | ||
50 | |||
51 | /* Order must be even */ | ||
52 | celt_assert( ( order & 1 ) == 0 ); | ||
53 | |||
54 | if (order == 10) | ||
55 | { | ||
56 | if (0 == lambda_Q16) | ||
57 | { | ||
58 | __m128i coef_Q13_3210, coef_Q13_7654; | ||
59 | __m128i coef_Q13_0123, coef_Q13_4567; | ||
60 | __m128i state_0123, state_4567; | ||
61 | __m128i xmm_product1, xmm_product2; | ||
62 | __m128i xmm_tempa, xmm_tempb; | ||
63 | |||
64 | register opus_int32 sum; | ||
65 | register opus_int32 state_8, state_9, state_a; | ||
66 | register opus_int64 coef_Q13_8, coef_Q13_9; | ||
67 | |||
68 | celt_assert( length > 0 ); | ||
69 | |||
70 | coef_Q13_3210 = OP_CVTEPI16_EPI32_M64( &coef_Q13[ 0 ] ); | ||
71 | coef_Q13_7654 = OP_CVTEPI16_EPI32_M64( &coef_Q13[ 4 ] ); | ||
72 | |||
73 | coef_Q13_0123 = _mm_shuffle_epi32( coef_Q13_3210, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
74 | coef_Q13_4567 = _mm_shuffle_epi32( coef_Q13_7654, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
75 | |||
76 | coef_Q13_8 = (opus_int64) coef_Q13[ 8 ]; | ||
77 | coef_Q13_9 = (opus_int64) coef_Q13[ 9 ]; | ||
78 | |||
79 | state_0123 = _mm_loadu_si128( (__m128i *)(&state[ 0 ] ) ); | ||
80 | state_4567 = _mm_loadu_si128( (__m128i *)(&state[ 4 ] ) ); | ||
81 | |||
82 | state_0123 = _mm_shuffle_epi32( state_0123, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
83 | state_4567 = _mm_shuffle_epi32( state_4567, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
84 | |||
85 | state_8 = state[ 8 ]; | ||
86 | state_9 = state[ 9 ]; | ||
87 | state_a = 0; | ||
88 | |||
89 | for( n = 0; n < length; n++ ) | ||
90 | { | ||
91 | xmm_product1 = _mm_mul_epi32( coef_Q13_0123, state_0123 ); /* 64-bit multiply, only 2 pairs */ | ||
92 | xmm_product2 = _mm_mul_epi32( coef_Q13_4567, state_4567 ); | ||
93 | |||
94 | xmm_tempa = _mm_shuffle_epi32( state_0123, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
95 | xmm_tempb = _mm_shuffle_epi32( state_4567, _MM_SHUFFLE( 0, 1, 2, 3 ) ); | ||
96 | |||
97 | xmm_product1 = _mm_srli_epi64( xmm_product1, 16 ); /* >> 16, zero extending works */ | ||
98 | xmm_product2 = _mm_srli_epi64( xmm_product2, 16 ); | ||
99 | |||
100 | xmm_tempa = _mm_mul_epi32( coef_Q13_3210, xmm_tempa ); | ||
101 | xmm_tempb = _mm_mul_epi32( coef_Q13_7654, xmm_tempb ); | ||
102 | |||
103 | xmm_tempa = _mm_srli_epi64( xmm_tempa, 16 ); | ||
104 | xmm_tempb = _mm_srli_epi64( xmm_tempb, 16 ); | ||
105 | |||
106 | xmm_tempa = _mm_add_epi32( xmm_tempa, xmm_product1 ); | ||
107 | xmm_tempb = _mm_add_epi32( xmm_tempb, xmm_product2 ); | ||
108 | xmm_tempa = _mm_add_epi32( xmm_tempa, xmm_tempb ); | ||
109 | |||
110 | sum = (opus_int32)((coef_Q13_8 * state_8) >> 16); | ||
111 | sum += (opus_int32)((coef_Q13_9 * state_9) >> 16); | ||
112 | |||
113 | xmm_tempa = _mm_add_epi32( xmm_tempa, _mm_shuffle_epi32( xmm_tempa, _MM_SHUFFLE( 0, 0, 0, 2 ) ) ); | ||
114 | sum += _mm_cvtsi128_si32( xmm_tempa); | ||
115 | res_Q2[ n ] = silk_LSHIFT( (opus_int32)input[ n ], 2 ) - silk_RSHIFT_ROUND( ( 5 + sum ), 9); | ||
116 | |||
117 | /* move right */ | ||
118 | state_a = state_9; | ||
119 | state_9 = state_8; | ||
120 | state_8 = _mm_cvtsi128_si32( state_4567 ); | ||
121 | state_4567 = _mm_alignr_epi8( state_0123, state_4567, 4 ); | ||
122 | |||
123 | state_0123 = _mm_alignr_epi8( _mm_cvtsi32_si128( silk_LSHIFT( input[ n ], 14 ) ), state_0123, 4 ); | ||
124 | } | ||
125 | |||
126 | _mm_storeu_si128( (__m128i *)( &state[ 0 ] ), _mm_shuffle_epi32( state_0123, _MM_SHUFFLE( 0, 1, 2, 3 ) ) ); | ||
127 | _mm_storeu_si128( (__m128i *)( &state[ 4 ] ), _mm_shuffle_epi32( state_4567, _MM_SHUFFLE( 0, 1, 2, 3 ) ) ); | ||
128 | state[ 8 ] = state_8; | ||
129 | state[ 9 ] = state_9; | ||
130 | state[ 10 ] = state_a; | ||
131 | |||
132 | return; | ||
133 | } | ||
134 | } | ||
135 | |||
136 | for( n = 0; n < length; n++ ) { | ||
137 | /* Output of lowpass section */ | ||
138 | tmp2 = silk_SMLAWB( state[ 0 ], state[ 1 ], lambda_Q16 ); | ||
139 | state[ 0 ] = silk_LSHIFT( input[ n ], 14 ); | ||
140 | /* Output of allpass section */ | ||
141 | tmp1 = silk_SMLAWB( state[ 1 ], state[ 2 ] - tmp2, lambda_Q16 ); | ||
142 | state[ 1 ] = tmp2; | ||
143 | acc_Q11 = silk_RSHIFT( order, 1 ); | ||
144 | acc_Q11 = silk_SMLAWB( acc_Q11, tmp2, coef_Q13[ 0 ] ); | ||
145 | /* Loop over allpass sections */ | ||
146 | for( i = 2; i < order; i += 2 ) { | ||
147 | /* Output of allpass section */ | ||
148 | tmp2 = silk_SMLAWB( state[ i ], state[ i + 1 ] - tmp1, lambda_Q16 ); | ||
149 | state[ i ] = tmp1; | ||
150 | acc_Q11 = silk_SMLAWB( acc_Q11, tmp1, coef_Q13[ i - 1 ] ); | ||
151 | /* Output of allpass section */ | ||
152 | tmp1 = silk_SMLAWB( state[ i + 1 ], state[ i + 2 ] - tmp2, lambda_Q16 ); | ||
153 | state[ i + 1 ] = tmp2; | ||
154 | acc_Q11 = silk_SMLAWB( acc_Q11, tmp2, coef_Q13[ i ] ); | ||
155 | } | ||
156 | state[ order ] = tmp1; | ||
157 | acc_Q11 = silk_SMLAWB( acc_Q11, tmp1, coef_Q13[ order - 1 ] ); | ||
158 | res_Q2[ n ] = silk_LSHIFT( (opus_int32)input[ n ], 2 ) - silk_RSHIFT_ROUND( acc_Q11, 9 ); | ||
159 | } | ||
160 | } | ||