| OLD | NEW |
|---|
| (Empty) | |
| 1 /* |
| 2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
| 3 * |
| 4 * Use of this source code is governed by a BSD-style license |
| 5 * that can be found in the LICENSE file in the root of the source |
| 6 * tree. An additional intellectual property rights grant can be found |
| 7 * in the file PATENTS. All contributing project authors may |
| 8 * be found in the AUTHORS file in the root of the source tree. |
| 9 */ |
| 10 |
| 11 |
| 12 #include <assert.h> |
| 13 #include <math.h> |
| 14 #include "vpx_ports/config.h" |
| 15 #include "vp9/common/vp9_systemdependent.h" |
| 16 |
| 17 #include "vp9/common/vp9_blockd.h" |
| 18 |
| 19 // TODO: these transforms can be converted into integer forms to reduce |
| 20 // the complexity |
| 21 static const float dct_4[16] = { |
| 22 0.500000000000000, 0.500000000000000, 0.500000000000000, 0.500000000000000, |
| 23 0.653281482438188, 0.270598050073099, -0.270598050073099, -0.653281482438188, |
| 24 0.500000000000000, -0.500000000000000, -0.500000000000000, 0.500000000000000, |
| 25 0.270598050073099, -0.653281482438188, 0.653281482438188, -0.270598050073099 |
| 26 }; |
| 27 |
| 28 static const float adst_4[16] = { |
| 29 0.228013428883779, 0.428525073124360, 0.577350269189626, 0.656538502008139, |
| 30 0.577350269189626, 0.577350269189626, 0.000000000000000, -0.577350269189626, |
| 31 0.656538502008139, -0.228013428883779, -0.577350269189626, 0.428525073124359, |
| 32 0.428525073124360, -0.656538502008139, 0.577350269189626, -0.228013428883779 |
| 33 }; |
| 34 |
| 35 static const float dct_8[64] = { |
| 36 0.353553390593274, 0.353553390593274, 0.353553390593274, 0.3535533905932
74, |
| 37 0.353553390593274, 0.353553390593274, 0.353553390593274, 0.3535533905932
74, |
| 38 0.490392640201615, 0.415734806151273, 0.277785116509801, 0.0975451610080
64, |
| 39 -0.097545161008064, -0.277785116509801, -0.415734806151273, -0.4903926402016
15, |
| 40 0.461939766255643, 0.191341716182545, -0.191341716182545, -0.4619397662556
43, |
| 41 -0.461939766255643, -0.191341716182545, 0.191341716182545, 0.4619397662556
43, |
| 42 0.415734806151273, -0.097545161008064, -0.490392640201615, -0.2777851165098
01, |
| 43 0.277785116509801, 0.490392640201615, 0.097545161008064, -0.4157348061512
73, |
| 44 0.353553390593274, -0.353553390593274, -0.353553390593274, 0.3535533905932
74, |
| 45 0.353553390593274, -0.353553390593274, -0.353553390593274, 0.3535533905932
74, |
| 46 0.277785116509801, -0.490392640201615, 0.097545161008064, 0.4157348061512
73, |
| 47 -0.415734806151273, -0.097545161008064, 0.490392640201615, -0.2777851165098
01, |
| 48 0.191341716182545, -0.461939766255643, 0.461939766255643, -0.1913417161825
45, |
| 49 -0.191341716182545, 0.461939766255643, -0.461939766255643, 0.1913417161825
45, |
| 50 0.097545161008064, -0.277785116509801, 0.415734806151273, -0.4903926402016
15, |
| 51 0.490392640201615, -0.415734806151273, 0.277785116509801, -0.0975451610080
64 |
| 52 }; |
| 53 |
| 54 static const float adst_8[64] = { |
| 55 0.089131608307533, 0.175227946595735, 0.255357107325376, 0.3267903880321
45, |
| 56 0.387095214016349, 0.434217976756762, 0.466553967085785, 0.4830020216355
09, |
| 57 0.255357107325376, 0.434217976756762, 0.483002021635509, 0.3870952140163
49, |
| 58 0.175227946595735, -0.089131608307533, -0.326790388032145, -0.4665539670857
85, |
| 59 0.387095214016349, 0.466553967085785, 0.175227946595735, -0.2553571073253
76, |
| 60 -0.483002021635509, -0.326790388032145, 0.089131608307533, 0.4342179767567
62, |
| 61 0.466553967085785, 0.255357107325376, -0.326790388032145, -0.4342179767567
62, |
| 62 0.089131608307533, 0.483002021635509, 0.175227946595735, -0.3870952140163
48, |
| 63 0.483002021635509, -0.089131608307533, -0.466553967085785, 0.1752279465957
35, |
| 64 0.434217976756762, -0.255357107325376, -0.387095214016348, 0.3267903880321
45, |
| 65 0.434217976756762, -0.387095214016348, -0.089131608307533, 0.4665539670857
86, |
| 66 -0.326790388032145, -0.175227946595735, 0.483002021635509, -0.2553571073253
75, |
| 67 0.326790388032145, -0.483002021635509, 0.387095214016349, -0.0891316083075
34, |
| 68 -0.255357107325377, 0.466553967085785, -0.434217976756762, 0.1752279465957
36, |
| 69 0.175227946595735, -0.326790388032145, 0.434217976756762, -0.4830020216355
09, |
| 70 0.466553967085785, -0.387095214016348, 0.255357107325376, -0.0891316083075
32 |
| 71 }; |
| 72 |
| 73 /* Converted the transforms to integers. */ |
| 74 static const int16_t dct_i4[16] = { |
| 75 16384, 16384, 16384, 16384, |
| 76 21407, 8867, -8867, -21407, |
| 77 16384, -16384, -16384, 16384, |
| 78 8867, -21407, 21407, -8867 |
| 79 }; |
| 80 |
| 81 static const int16_t adst_i4[16] = { |
| 82 7472, 14042, 18919, 21513, |
| 83 18919, 18919, 0, -18919, |
| 84 21513, -7472, -18919, 14042, |
| 85 14042, -21513, 18919, -7472 |
| 86 }; |
| 87 |
| 88 static const int16_t dct_i8[64] = { |
| 89 11585, 11585, 11585, 11585, |
| 90 11585, 11585, 11585, 11585, |
| 91 16069, 13623, 9102, 3196, |
| 92 -3196, -9102, -13623, -16069, |
| 93 15137, 6270, -6270, -15137, |
| 94 -15137, -6270, 6270, 15137, |
| 95 13623, -3196, -16069, -9102, |
| 96 9102, 16069, 3196, -13623, |
| 97 11585, -11585, -11585, 11585, |
| 98 11585, -11585, -11585, 11585, |
| 99 9102, -16069, 3196, 13623, |
| 100 -13623, -3196, 16069, -9102, |
| 101 6270, -15137, 15137, -6270, |
| 102 -6270, 15137, -15137, 6270, |
| 103 3196, -9102, 13623, -16069, |
| 104 16069, -13623, 9102, -3196 |
| 105 }; |
| 106 |
| 107 static const int16_t adst_i8[64] = { |
| 108 2921, 5742, 8368, 10708, |
| 109 12684, 14228, 15288, 15827, |
| 110 8368, 14228, 15827, 12684, |
| 111 5742, -2921, -10708, -15288, |
| 112 12684, 15288, 5742, -8368, |
| 113 -15827, -10708, 2921, 14228, |
| 114 15288, 8368, -10708, -14228, |
| 115 2921, 15827, 5742, -12684, |
| 116 15827, -2921, -15288, 5742, |
| 117 14228, -8368, -12684, 10708, |
| 118 14228, -12684, -2921, 15288, |
| 119 -10708, -5742, 15827, -8368, |
| 120 10708, -15827, 12684, -2921, |
| 121 -8368, 15288, -14228, 5742, |
| 122 5742, -10708, 14228, -15827, |
| 123 15288, -12684, 8368, -2921 |
| 124 }; |
| 125 |
| 126 static const float dct_16[256] = { |
| 127 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0
.250000, |
| 128 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0
.250000, |
| 129 0.351851, 0.338330, 0.311806, 0.273300, 0.224292, 0.166664, 0.102631, 0
.034654, |
| 130 -0.034654, -0.102631, -0.166664, -0.224292, -0.273300, -0.311806, -0.338330, -0
.351851, |
| 131 0.346760, 0.293969, 0.196424, 0.068975, -0.068975, -0.196424, -0.293969, -0
.346760, |
| 132 -0.346760, -0.293969, -0.196424, -0.068975, 0.068975, 0.196424, 0.293969, 0
.346760, |
| 133 0.338330, 0.224292, 0.034654, -0.166664, -0.311806, -0.351851, -0.273300, -0
.102631, |
| 134 0.102631, 0.273300, 0.351851, 0.311806, 0.166664, -0.034654, -0.224292, -0
.338330, |
| 135 0.326641, 0.135299, -0.135299, -0.326641, -0.326641, -0.135299, 0.135299, 0
.326641, |
| 136 0.326641, 0.135299, -0.135299, -0.326641, -0.326641, -0.135299, 0.135299, 0
.326641, |
| 137 0.311806, 0.034654, -0.273300, -0.338330, -0.102631, 0.224292, 0.351851, 0
.166664, |
| 138 -0.166664, -0.351851, -0.224292, 0.102631, 0.338330, 0.273300, -0.034654, -0
.311806, |
| 139 0.293969, -0.068975, -0.346760, -0.196424, 0.196424, 0.346760, 0.068975, -0
.293969, |
| 140 -0.293969, 0.068975, 0.346760, 0.196424, -0.196424, -0.346760, -0.068975, 0
.293969, |
| 141 0.273300, -0.166664, -0.338330, 0.034654, 0.351851, 0.102631, -0.311806, -0
.224292, |
| 142 0.224292, 0.311806, -0.102631, -0.351851, -0.034654, 0.338330, 0.166664, -0
.273300, |
| 143 0.250000, -0.250000, -0.250000, 0.250000, 0.250000, -0.250000, -0.250000, 0
.250000, |
| 144 0.250000, -0.250000, -0.250000, 0.250000, 0.250000, -0.250000, -0.250000, 0
.250000, |
| 145 0.224292, -0.311806, -0.102631, 0.351851, -0.034654, -0.338330, 0.166664, 0
.273300, |
| 146 -0.273300, -0.166664, 0.338330, 0.034654, -0.351851, 0.102631, 0.311806, -0
.224292, |
| 147 0.196424, -0.346760, 0.068975, 0.293969, -0.293969, -0.068975, 0.346760, -0
.196424, |
| 148 -0.196424, 0.346760, -0.068975, -0.293969, 0.293969, 0.068975, -0.346760, 0
.196424, |
| 149 0.166664, -0.351851, 0.224292, 0.102631, -0.338330, 0.273300, 0.034654, -0
.311806, |
| 150 0.311806, -0.034654, -0.273300, 0.338330, -0.102631, -0.224292, 0.351851, -0
.166664, |
| 151 0.135299, -0.326641, 0.326641, -0.135299, -0.135299, 0.326641, -0.326641, 0
.135299, |
| 152 0.135299, -0.326641, 0.326641, -0.135299, -0.135299, 0.326641, -0.326641, 0
.135299, |
| 153 0.102631, -0.273300, 0.351851, -0.311806, 0.166664, 0.034654, -0.224292, 0
.338330, |
| 154 -0.338330, 0.224292, -0.034654, -0.166664, 0.311806, -0.351851, 0.273300, -0
.102631, |
| 155 0.068975, -0.196424, 0.293969, -0.346760, 0.346760, -0.293969, 0.196424, -0
.068975, |
| 156 -0.068975, 0.196424, -0.293969, 0.346760, -0.346760, 0.293969, -0.196424, 0
.068975, |
| 157 0.034654, -0.102631, 0.166664, -0.224292, 0.273300, -0.311806, 0.338330, -0
.351851, |
| 158 0.351851, -0.338330, 0.311806, -0.273300, 0.224292, -0.166664, 0.102631, -0
.034654 |
| 159 }; |
| 160 |
| 161 static const float adst_16[256] = { |
| 162 0.033094, 0.065889, 0.098087, 0.129396, 0.159534, 0.188227, 0.215215, 0
.240255, |
| 163 0.263118, 0.283599, 0.301511, 0.316693, 0.329007, 0.338341, 0.344612, 0
.347761, |
| 164 0.098087, 0.188227, 0.263118, 0.316693, 0.344612, 0.344612, 0.316693, 0
.263118, |
| 165 0.188227, 0.098087, 0.000000, -0.098087, -0.188227, -0.263118, -0.316693, -0
.344612, |
| 166 0.159534, 0.283599, 0.344612, 0.329007, 0.240255, 0.098087, -0.065889, -0
.215215, |
| 167 -0.316693, -0.347761, -0.301511, -0.188227, -0.033094, 0.129396, 0.263118, 0
.338341, |
| 168 0.215215, 0.338341, 0.316693, 0.159534, -0.065889, -0.263118, -0.347761, -0
.283599, |
| 169 -0.098087, 0.129396, 0.301511, 0.344612, 0.240255, 0.033094, -0.188227, -0
.329007, |
| 170 0.263118, 0.344612, 0.188227, -0.098087, -0.316693, -0.316693, -0.098087, 0
.188227, |
| 171 0.344612, 0.263118, 0.000000, -0.263118, -0.344612, -0.188227, 0.098087, 0
.316693, |
| 172 0.301511, 0.301511, 0.000000, -0.301511, -0.301511, -0.000000, 0.301511, 0
.301511, |
| 173 0.000000, -0.301511, -0.301511, -0.000000, 0.301511, 0.301511, 0.000000, -0
.301511, |
| 174 0.329007, 0.215215, -0.188227, -0.338341, -0.033094, 0.316693, 0.240255, -0
.159534, |
| 175 -0.344612, -0.065889, 0.301511, 0.263118, -0.129396, -0.347761, -0.098087, 0
.283599, |
| 176 0.344612, 0.098087, -0.316693, -0.188227, 0.263118, 0.263118, -0.188227, -0
.316693, |
| 177 0.098087, 0.344612, 0.000000, -0.344612, -0.098087, 0.316693, 0.188227, -0
.263118, |
| 178 0.347761, -0.033094, -0.344612, 0.065889, 0.338341, -0.098087, -0.329007, 0
.129396, |
| 179 0.316693, -0.159534, -0.301511, 0.188227, 0.283599, -0.215215, -0.263118, 0
.240255, |
| 180 0.338341, -0.159534, -0.263118, 0.283599, 0.129396, -0.344612, 0.033094, 0
.329007, |
| 181 -0.188227, -0.240255, 0.301511, 0.098087, -0.347761, 0.065889, 0.316693, -0
.215215, |
| 182 0.316693, -0.263118, -0.098087, 0.344612, -0.188227, -0.188227, 0.344612, -0
.098087, |
| 183 -0.263118, 0.316693, 0.000000, -0.316693, 0.263118, 0.098087, -0.344612, 0
.188227, |
| 184 0.283599, -0.329007, 0.098087, 0.215215, -0.347761, 0.188227, 0.129396, -0
.338341, |
| 185 0.263118, 0.033094, -0.301511, 0.316693, -0.065889, -0.240255, 0.344612, -0
.159534, |
| 186 0.240255, -0.347761, 0.263118, -0.033094, -0.215215, 0.344612, -0.283599, 0
.065889, |
| 187 0.188227, -0.338341, 0.301511, -0.098087, -0.159534, 0.329007, -0.316693, 0
.129396, |
| 188 0.188227, -0.316693, 0.344612, -0.263118, 0.098087, 0.098087, -0.263118, 0
.344612, |
| 189 -0.316693, 0.188227, 0.000000, -0.188227, 0.316693, -0.344612, 0.263118, -0
.098087, |
| 190 0.129396, -0.240255, 0.316693, -0.347761, 0.329007, -0.263118, 0.159534, -0
.033094, |
| 191 -0.098087, 0.215215, -0.301511, 0.344612, -0.338341, 0.283599, -0.188227, 0
.065889, |
| 192 0.065889, -0.129396, 0.188227, -0.240255, 0.283599, -0.316693, 0.338341, -0
.347761, |
| 193 0.344612, -0.329007, 0.301511, -0.263118, 0.215215, -0.159534, 0.098087, -0
.033094 |
| 194 }; |
| 195 |
| 196 /* Converted the transforms to integers. */ |
| 197 static const int16_t dct_i16[256] = { |
| 198 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, |
| 199 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, |
| 200 11529, 11086, 10217, 8955, 7350, 5461, 3363, 1136, |
| 201 -1136, -3363, -5461, -7350, -8955, -10217, -11086, -11529, |
| 202 11363, 9633, 6436, 2260, -2260, -6436, -9633, -11363, |
| 203 -11363, -9633, -6436, -2260, 2260, 6436, 9633, 11363, |
| 204 11086, 7350, 1136, -5461, -10217, -11529, -8955, -3363, |
| 205 3363, 8955, 11529, 10217, 5461, -1136, -7350, -11086, |
| 206 10703, 4433, -4433, -10703, -10703, -4433, 4433, 10703, |
| 207 10703, 4433, -4433, -10703, -10703, -4433, 4433, 10703, |
| 208 10217, 1136, -8955, -11086, -3363, 7350, 11529, 5461, |
| 209 -5461, -11529, -7350, 3363, 11086, 8955, -1136, -10217, |
| 210 9633, -2260, -11363, -6436, 6436, 11363, 2260, -9633, |
| 211 -9633, 2260, 11363, 6436, -6436, -11363, -2260, 9633, |
| 212 8955, -5461, -11086, 1136, 11529, 3363, -10217, -7350, |
| 213 7350, 10217, -3363, -11529, -1136, 11086, 5461, -8955, |
| 214 8192, -8192, -8192, 8192, 8192, -8192, -8192, 8192, |
| 215 8192, -8192, -8192, 8192, 8192, -8192, -8192, 8192, |
| 216 7350, -10217, -3363, 11529, -1136, -11086, 5461, 8955, |
| 217 -8955, -5461, 11086, 1136, -11529, 3363, 10217, -7350, |
| 218 6436, -11363, 2260, 9633, -9633, -2260, 11363, -6436, |
| 219 -6436, 11363, -2260, -9633, 9633, 2260, -11363, 6436, |
| 220 5461, -11529, 7350, 3363, -11086, 8955, 1136, -10217, |
| 221 10217, -1136, -8955, 11086, -3363, -7350, 11529, -5461, |
| 222 4433, -10703, 10703, -4433, -4433, 10703, -10703, 4433, |
| 223 4433, -10703, 10703, -4433, -4433, 10703, -10703, 4433, |
| 224 3363, -8955, 11529, -10217, 5461, 1136, -7350, 11086, |
| 225 -11086, 7350, -1136, -5461, 10217, -11529, 8955, -3363, |
| 226 2260, -6436, 9633, -11363, 11363, -9633, 6436, -2260, |
| 227 -2260, 6436, -9633, 11363, -11363, 9633, -6436, 2260, |
| 228 1136, -3363, 5461, -7350, 8955, -10217, 11086, -11529, |
| 229 11529, -11086, 10217, -8955, 7350, -5461, 3363, -1136 |
| 230 }; |
| 231 |
| 232 static const int16_t adst_i16[256] = { |
| 233 1084, 2159, 3214, 4240, 5228, 6168, 7052, 7873, |
| 234 8622, 9293, 9880, 10377, 10781, 11087, 11292, 11395, |
| 235 3214, 6168, 8622, 10377, 11292, 11292, 10377, 8622, |
| 236 6168, 3214, 0, -3214, -6168, -8622, -10377, -11292, |
| 237 5228, 9293, 11292, 10781, 7873, 3214, -2159, -7052, |
| 238 -10377, -11395, -9880, -6168, -1084, 4240, 8622, 11087, |
| 239 7052, 11087, 10377, 5228, -2159, -8622, -11395, -9293, |
| 240 -3214, 4240, 9880, 11292, 7873, 1084, -6168, -10781, |
| 241 8622, 11292, 6168, -3214, -10377, -10377, -3214, 6168, |
| 242 11292, 8622, 0, -8622, -11292, -6168, 3214, 10377, |
| 243 9880, 9880, 0, -9880, -9880, 0, 9880, 9880, |
| 244 0, -9880, -9880, 0, 9880, 9880, 0, -9880, |
| 245 10781, 7052, -6168, -11087, -1084, 10377, 7873, -5228, |
| 246 -11292, -2159, 9880, 8622, -4240, -11395, -3214, 9293, |
| 247 11292, 3214, -10377, -6168, 8622, 8622, -6168, -10377, |
| 248 3214, 11292, 0, -11292, -3214, 10377, 6168, -8622, |
| 249 11395, -1084, -11292, 2159, 11087, -3214, -10781, 4240, |
| 250 10377, -5228, -9880, 6168, 9293, -7052, -8622, 7873, |
| 251 11087, -5228, -8622, 9293, 4240, -11292, 1084, 10781, |
| 252 -6168, -7873, 9880, 3214, -11395, 2159, 10377, -7052, |
| 253 10377, -8622, -3214, 11292, -6168, -6168, 11292, -3214, |
| 254 -8622, 10377, 0, -10377, 8622, 3214, -11292, 6168, |
| 255 9293, -10781, 3214, 7052, -11395, 6168, 4240, -11087, |
| 256 8622, 1084, -9880, 10377, -2159, -7873, 11292, -5228, |
| 257 7873, -11395, 8622, -1084, -7052, 11292, -9293, 2159, |
| 258 6168, -11087, 9880, -3214, -5228, 10781, -10377, 4240, |
| 259 6168, -10377, 11292, -8622, 3214, 3214, -8622, 11292, |
| 260 -10377, 6168, 0, -6168, 10377, -11292, 8622, -3214, |
| 261 4240, -7873, 10377, -11395, 10781, -8622, 5228, -1084, |
| 262 -3214, 7052, -9880, 11292, -11087, 9293, -6168, 2159, |
| 263 2159, -4240, 6168, -7873, 9293, -10377, 11087, -11395, |
| 264 11292, -10781, 9880, -8622, 7052, -5228, 3214, -1084 |
| 265 }; |
| 266 |
| 267 static const int xC1S7 = 16069; |
| 268 static const int xC2S6 = 15137; |
| 269 static const int xC3S5 = 13623; |
| 270 static const int xC4S4 = 11585; |
| 271 static const int xC5S3 = 9102; |
| 272 static const int xC6S2 = 6270; |
| 273 static const int xC7S1 = 3196; |
| 274 |
| 275 #define SHIFT_BITS 14 |
| 276 #define DOROUND(X) X += (1<<(SHIFT_BITS-1)); |
| 277 |
| 278 #define FINAL_SHIFT 3 |
| 279 #define FINAL_ROUNDING (1<<(FINAL_SHIFT -1)) |
| 280 #define IN_SHIFT (FINAL_SHIFT+1) |
| 281 |
| 282 |
| 283 void vp9_short_fdct8x8_c(short *InputData, short *OutputData, int pitch) { |
| 284 int loop; |
| 285 int short_pitch = pitch >> 1; |
| 286 int is07, is12, is34, is56; |
| 287 int is0734, is1256; |
| 288 int id07, id12, id34, id56; |
| 289 int irot_input_x, irot_input_y; |
| 290 int icommon_product1; // Re-used product (c4s4 * (s12 - s56)) |
| 291 int icommon_product2; // Re-used product (c4s4 * (d12 + d56)) |
| 292 int temp1, temp2; // intermediate variable for computation |
| 293 |
| 294 int InterData[64]; |
| 295 int *ip = InterData; |
| 296 short *op = OutputData; |
| 297 |
| 298 for (loop = 0; loop < 8; loop++) { |
| 299 // Pre calculate some common sums and differences. |
| 300 is07 = (InputData[0] + InputData[7]) << IN_SHIFT; |
| 301 is12 = (InputData[1] + InputData[2]) << IN_SHIFT; |
| 302 is34 = (InputData[3] + InputData[4]) << IN_SHIFT; |
| 303 is56 = (InputData[5] + InputData[6]) << IN_SHIFT; |
| 304 id07 = (InputData[0] - InputData[7]) << IN_SHIFT; |
| 305 id12 = (InputData[1] - InputData[2]) << IN_SHIFT; |
| 306 id34 = (InputData[3] - InputData[4]) << IN_SHIFT; |
| 307 id56 = (InputData[5] - InputData[6]) << IN_SHIFT; |
| 308 |
| 309 is0734 = is07 + is34; |
| 310 is1256 = is12 + is56; |
| 311 |
| 312 // Pre-Calculate some common product terms. |
| 313 icommon_product1 = xC4S4 * (is12 - is56); |
| 314 DOROUND(icommon_product1) |
| 315 icommon_product1 >>= SHIFT_BITS; |
| 316 |
| 317 icommon_product2 = xC4S4 * (id12 + id56); |
| 318 DOROUND(icommon_product2) |
| 319 icommon_product2 >>= SHIFT_BITS; |
| 320 |
| 321 |
| 322 ip[0] = (xC4S4 * (is0734 + is1256)); |
| 323 DOROUND(ip[0]); |
| 324 ip[0] >>= SHIFT_BITS; |
| 325 |
| 326 ip[4] = (xC4S4 * (is0734 - is1256)); |
| 327 DOROUND(ip[4]); |
| 328 ip[4] >>= SHIFT_BITS; |
| 329 |
| 330 // Define inputs to rotation for outputs 2 and 6 |
| 331 irot_input_x = id12 - id56; |
| 332 irot_input_y = is07 - is34; |
| 333 |
| 334 // Apply rotation for outputs 2 and 6. |
| 335 temp1 = xC6S2 * irot_input_x; |
| 336 DOROUND(temp1); |
| 337 temp1 >>= SHIFT_BITS; |
| 338 temp2 = xC2S6 * irot_input_y; |
| 339 DOROUND(temp2); |
| 340 temp2 >>= SHIFT_BITS; |
| 341 ip[2] = temp1 + temp2; |
| 342 |
| 343 temp1 = xC6S2 * irot_input_y; |
| 344 DOROUND(temp1); |
| 345 temp1 >>= SHIFT_BITS; |
| 346 temp2 = xC2S6 * irot_input_x; |
| 347 DOROUND(temp2); |
| 348 temp2 >>= SHIFT_BITS; |
| 349 ip[6] = temp1 - temp2; |
| 350 |
| 351 // Define inputs to rotation for outputs 1 and 7 |
| 352 irot_input_x = icommon_product1 + id07; |
| 353 irot_input_y = -(id34 + icommon_product2); |
| 354 |
| 355 // Apply rotation for outputs 1 and 7. |
| 356 temp1 = xC1S7 * irot_input_x; |
| 357 DOROUND(temp1); |
| 358 temp1 >>= SHIFT_BITS; |
| 359 temp2 = xC7S1 * irot_input_y; |
| 360 DOROUND(temp2); |
| 361 temp2 >>= SHIFT_BITS; |
| 362 ip[1] = temp1 - temp2; |
| 363 |
| 364 temp1 = xC7S1 * irot_input_x; |
| 365 DOROUND(temp1); |
| 366 temp1 >>= SHIFT_BITS; |
| 367 temp2 = xC1S7 * irot_input_y; |
| 368 DOROUND(temp2); |
| 369 temp2 >>= SHIFT_BITS; |
| 370 ip[7] = temp1 + temp2; |
| 371 |
| 372 // Define inputs to rotation for outputs 3 and 5 |
| 373 irot_input_x = id07 - icommon_product1; |
| 374 irot_input_y = id34 - icommon_product2; |
| 375 |
| 376 // Apply rotation for outputs 3 and 5. |
| 377 temp1 = xC3S5 * irot_input_x; |
| 378 DOROUND(temp1); |
| 379 temp1 >>= SHIFT_BITS; |
| 380 temp2 = xC5S3 * irot_input_y; |
| 381 DOROUND(temp2); |
| 382 temp2 >>= SHIFT_BITS; |
| 383 ip[3] = temp1 - temp2; |
| 384 |
| 385 |
| 386 temp1 = xC5S3 * irot_input_x; |
| 387 DOROUND(temp1); |
| 388 temp1 >>= SHIFT_BITS; |
| 389 temp2 = xC3S5 * irot_input_y; |
| 390 DOROUND(temp2); |
| 391 temp2 >>= SHIFT_BITS; |
| 392 ip[5] = temp1 + temp2; |
| 393 |
| 394 // Increment data pointer for next row |
| 395 InputData += short_pitch; |
| 396 ip += 8; |
| 397 } |
| 398 |
| 399 // Performed DCT on rows, now transform the columns |
| 400 ip = InterData; |
| 401 for (loop = 0; loop < 8; loop++) { |
| 402 // Pre calculate some common sums and differences. |
| 403 is07 = ip[0 * 8] + ip[7 * 8]; |
| 404 is12 = ip[1 * 8] + ip[2 * 8]; |
| 405 is34 = ip[3 * 8] + ip[4 * 8]; |
| 406 is56 = ip[5 * 8] + ip[6 * 8]; |
| 407 |
| 408 id07 = ip[0 * 8] - ip[7 * 8]; |
| 409 id12 = ip[1 * 8] - ip[2 * 8]; |
| 410 id34 = ip[3 * 8] - ip[4 * 8]; |
| 411 id56 = ip[5 * 8] - ip[6 * 8]; |
| 412 |
| 413 is0734 = is07 + is34; |
| 414 is1256 = is12 + is56; |
| 415 |
| 416 // Pre-Calculate some common product terms |
| 417 icommon_product1 = xC4S4 * (is12 - is56); |
| 418 icommon_product2 = xC4S4 * (id12 + id56); |
| 419 DOROUND(icommon_product1) |
| 420 DOROUND(icommon_product2) |
| 421 icommon_product1 >>= SHIFT_BITS; |
| 422 icommon_product2 >>= SHIFT_BITS; |
| 423 |
| 424 |
| 425 temp1 = xC4S4 * (is0734 + is1256); |
| 426 temp2 = xC4S4 * (is0734 - is1256); |
| 427 DOROUND(temp1); |
| 428 DOROUND(temp2); |
| 429 temp1 >>= SHIFT_BITS; |
| 430 |
| 431 temp2 >>= SHIFT_BITS; |
| 432 op[0 * 8] = (temp1 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 433 op[4 * 8] = (temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 434 |
| 435 // Define inputs to rotation for outputs 2 and 6 |
| 436 irot_input_x = id12 - id56; |
| 437 irot_input_y = is07 - is34; |
| 438 |
| 439 // Apply rotation for outputs 2 and 6. |
| 440 temp1 = xC6S2 * irot_input_x; |
| 441 DOROUND(temp1); |
| 442 temp1 >>= SHIFT_BITS; |
| 443 temp2 = xC2S6 * irot_input_y; |
| 444 DOROUND(temp2); |
| 445 temp2 >>= SHIFT_BITS; |
| 446 op[2 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 447 |
| 448 temp1 = xC6S2 * irot_input_y; |
| 449 DOROUND(temp1); |
| 450 temp1 >>= SHIFT_BITS; |
| 451 temp2 = xC2S6 * irot_input_x; |
| 452 DOROUND(temp2); |
| 453 temp2 >>= SHIFT_BITS; |
| 454 op[6 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 455 |
| 456 // Define inputs to rotation for outputs 1 and 7 |
| 457 irot_input_x = icommon_product1 + id07; |
| 458 irot_input_y = -(id34 + icommon_product2); |
| 459 |
| 460 // Apply rotation for outputs 1 and 7. |
| 461 temp1 = xC1S7 * irot_input_x; |
| 462 DOROUND(temp1); |
| 463 temp1 >>= SHIFT_BITS; |
| 464 temp2 = xC7S1 * irot_input_y; |
| 465 DOROUND(temp2); |
| 466 temp2 >>= SHIFT_BITS; |
| 467 op[1 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 468 |
| 469 temp1 = xC7S1 * irot_input_x; |
| 470 DOROUND(temp1); |
| 471 temp1 >>= SHIFT_BITS; |
| 472 temp2 = xC1S7 * irot_input_y; |
| 473 DOROUND(temp2); |
| 474 temp2 >>= SHIFT_BITS; |
| 475 op[7 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 476 |
| 477 // Define inputs to rotation for outputs 3 and 5 |
| 478 irot_input_x = id07 - icommon_product1; |
| 479 irot_input_y = id34 - icommon_product2; |
| 480 |
| 481 // Apply rotation for outputs 3 and 5. |
| 482 temp1 = xC3S5 * irot_input_x; |
| 483 DOROUND(temp1); |
| 484 temp1 >>= SHIFT_BITS; |
| 485 temp2 = xC5S3 * irot_input_y; |
| 486 DOROUND(temp2); |
| 487 temp2 >>= SHIFT_BITS; |
| 488 op[3 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 489 |
| 490 |
| 491 temp1 = xC5S3 * irot_input_x; |
| 492 DOROUND(temp1); |
| 493 temp1 >>= SHIFT_BITS; |
| 494 temp2 = xC3S5 * irot_input_y; |
| 495 DOROUND(temp2); |
| 496 temp2 >>= SHIFT_BITS; |
| 497 op[5 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT; |
| 498 |
| 499 // Increment data pointer for next column. |
| 500 ip++; |
| 501 op++; |
| 502 } |
| 503 } |
| 504 |
| 505 void vp9_short_fhaar2x2_c(short *input, short *output, int pitch) { |
| 506 /* [1 1; 1 -1] orthogonal transform */ |
| 507 /* use position: 0,1, 4, 8 */ |
| 508 int i; |
| 509 short *ip1 = input; |
| 510 short *op1 = output; |
| 511 for (i = 0; i < 16; i++) { |
| 512 op1[i] = 0; |
| 513 } |
| 514 |
| 515 op1[0] = (ip1[0] + ip1[1] + ip1[4] + ip1[8] + 1) >> 1; |
| 516 op1[1] = (ip1[0] - ip1[1] + ip1[4] - ip1[8]) >> 1; |
| 517 op1[4] = (ip1[0] + ip1[1] - ip1[4] - ip1[8]) >> 1; |
| 518 op1[8] = (ip1[0] - ip1[1] - ip1[4] + ip1[8]) >> 1; |
| 519 } |
| 520 |
| 521 /* For test */ |
| 522 #define TEST_INT 1 |
| 523 #if TEST_INT |
| 524 #define vp9_fht_int_c vp9_fht_c |
| 525 #else |
| 526 #define vp9_fht_float_c vp9_fht_c |
| 527 #endif |
| 528 |
| 529 void vp9_fht_float_c(const int16_t *input, int pitch, int16_t *output, |
| 530 TX_TYPE tx_type, int tx_dim) { |
| 531 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 532 { |
| 533 int i, j, k; |
| 534 float bufa[256], bufb[256]; // buffers are for floating-point test purpose |
| 535 // the implementation could be simplified in |
| 536 // conjunction with integer transform |
| 537 const int16_t *ip = input; |
| 538 int16_t *op = output; |
| 539 |
| 540 float *pfa = &bufa[0]; |
| 541 float *pfb = &bufb[0]; |
| 542 |
| 543 // pointers to vertical and horizontal transforms |
| 544 const float *ptv, *pth; |
| 545 |
| 546 assert(tx_type != DCT_DCT); |
| 547 // load and convert residual array into floating-point |
| 548 for (j = 0; j < tx_dim; j++) { |
| 549 for (i = 0; i < tx_dim; i++) { |
| 550 pfa[i] = (float)ip[i]; |
| 551 } |
| 552 pfa += tx_dim; |
| 553 ip += pitch / 2; |
| 554 } |
| 555 |
| 556 // vertical transformation |
| 557 pfa = &bufa[0]; |
| 558 pfb = &bufb[0]; |
| 559 |
| 560 switch (tx_type) { |
| 561 case ADST_ADST : |
| 562 case ADST_DCT : |
| 563 ptv = (tx_dim == 4) ? &adst_4[0] : |
| 564 ((tx_dim == 8) ? &adst_8[0] : &adst_16[0]); |
| 565 break; |
| 566 |
| 567 default : |
| 568 ptv = (tx_dim == 4) ? &dct_4[0] : |
| 569 ((tx_dim == 8) ? &dct_8[0] : &dct_16[0]); |
| 570 break; |
| 571 } |
| 572 |
| 573 for (j = 0; j < tx_dim; j++) { |
| 574 for (i = 0; i < tx_dim; i++) { |
| 575 pfb[i] = 0; |
| 576 for (k = 0; k < tx_dim; k++) { |
| 577 pfb[i] += ptv[k] * pfa[(k * tx_dim)]; |
| 578 } |
| 579 pfa += 1; |
| 580 } |
| 581 pfb += tx_dim; |
| 582 ptv += tx_dim; |
| 583 pfa = &bufa[0]; |
| 584 } |
| 585 |
| 586 // horizontal transformation |
| 587 pfa = &bufa[0]; |
| 588 pfb = &bufb[0]; |
| 589 |
| 590 switch (tx_type) { |
| 591 case ADST_ADST : |
| 592 case DCT_ADST : |
| 593 pth = (tx_dim == 4) ? &adst_4[0] : |
| 594 ((tx_dim == 8) ? &adst_8[0] : &adst_16[0]); |
| 595 break; |
| 596 |
| 597 default : |
| 598 pth = (tx_dim == 4) ? &dct_4[0] : |
| 599 ((tx_dim == 8) ? &dct_8[0] : &dct_16[0]); |
| 600 break; |
| 601 } |
| 602 |
| 603 for (j = 0; j < tx_dim; j++) { |
| 604 for (i = 0; i < tx_dim; i++) { |
| 605 pfa[i] = 0; |
| 606 for (k = 0; k < tx_dim; k++) { |
| 607 pfa[i] += pfb[k] * pth[k]; |
| 608 } |
| 609 pth += tx_dim; |
| 610 } |
| 611 |
| 612 pfa += tx_dim; |
| 613 pfb += tx_dim; |
| 614 // pth -= tx_dim * tx_dim; |
| 615 |
| 616 switch (tx_type) { |
| 617 case ADST_ADST : |
| 618 case DCT_ADST : |
| 619 pth = (tx_dim == 4) ? &adst_4[0] : |
| 620 ((tx_dim == 8) ? &adst_8[0] : &adst_16[0]); |
| 621 break; |
| 622 |
| 623 default : |
| 624 pth = (tx_dim == 4) ? &dct_4[0] : |
| 625 ((tx_dim == 8) ? &dct_8[0] : &dct_16[0]); |
| 626 break; |
| 627 } |
| 628 } |
| 629 |
| 630 // convert to short integer format and load BLOCKD buffer |
| 631 op = output; |
| 632 pfa = &bufa[0]; |
| 633 |
| 634 for (j = 0; j < tx_dim; j++) { |
| 635 for (i = 0; i < tx_dim; i++) { |
| 636 op[i] = (pfa[i] > 0 ) ? (int16_t)( 8 * pfa[i] + 0.49) : |
| 637 -(int16_t)(- 8 * pfa[i] + 0.49); |
| 638 } |
| 639 op += tx_dim; |
| 640 pfa += tx_dim; |
| 641 } |
| 642 } |
| 643 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 644 } |
| 645 |
| 646 /* Converted the transforms to integer form. */ |
| 647 #define VERTICAL_SHIFT 11 |
| 648 #define VERTICAL_ROUNDING ((1 << (VERTICAL_SHIFT - 1)) - 1) |
| 649 #define HORIZONTAL_SHIFT 16 |
| 650 #define HORIZONTAL_ROUNDING ((1 << (HORIZONTAL_SHIFT - 1)) - 1) |
| 651 void vp9_fht_int_c(const int16_t *input, int pitch, int16_t *output, |
| 652 TX_TYPE tx_type, int tx_dim) { |
| 653 int i, j, k; |
| 654 int16_t imbuf[256]; |
| 655 |
| 656 const int16_t *ip = input; |
| 657 int16_t *op = output; |
| 658 int16_t *im = &imbuf[0]; |
| 659 |
| 660 /* pointers to vertical and horizontal transforms. */ |
| 661 const int16_t *ptv = NULL, *pth = NULL; |
| 662 |
| 663 switch (tx_type) { |
| 664 case ADST_ADST : |
| 665 ptv = pth = (tx_dim == 4) ? &adst_i4[0] |
| 666 : ((tx_dim == 8) ? &adst_i8[0] |
| 667 : &adst_i16[0]); |
| 668 break; |
| 669 case ADST_DCT : |
| 670 ptv = (tx_dim == 4) ? &adst_i4[0] |
| 671 : ((tx_dim == 8) ? &adst_i8[0] : &adst_i16[0]); |
| 672 pth = (tx_dim == 4) ? &dct_i4[0] |
| 673 : ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]); |
| 674 break; |
| 675 case DCT_ADST : |
| 676 ptv = (tx_dim == 4) ? &dct_i4[0] |
| 677 : ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]); |
| 678 pth = (tx_dim == 4) ? &adst_i4[0] |
| 679 : ((tx_dim == 8) ? &adst_i8[0] : &adst_i16[0]); |
| 680 break; |
| 681 case DCT_DCT : |
| 682 ptv = pth = (tx_dim == 4) ? &dct_i4[0] |
| 683 : ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]); |
| 684 break; |
| 685 default: |
| 686 assert(0); |
| 687 break; |
| 688 } |
| 689 |
| 690 /* vertical transformation */ |
| 691 for (j = 0; j < tx_dim; j++) { |
| 692 for (i = 0; i < tx_dim; i++) { |
| 693 int temp = 0; |
| 694 |
| 695 for (k = 0; k < tx_dim; k++) { |
| 696 temp += ptv[k] * ip[(k * (pitch >> 1))]; |
| 697 } |
| 698 |
| 699 im[i] = (int16_t)((temp + VERTICAL_ROUNDING) >> VERTICAL_SHIFT); |
| 700 ip++; |
| 701 } |
| 702 im += tx_dim; // 16 |
| 703 ptv += tx_dim; |
| 704 ip = input; |
| 705 } |
| 706 |
| 707 /* horizontal transformation */ |
| 708 im = &imbuf[0]; |
| 709 |
| 710 for (j = 0; j < tx_dim; j++) { |
| 711 const int16_t *pthc = pth; |
| 712 |
| 713 for (i = 0; i < tx_dim; i++) { |
| 714 int temp = 0; |
| 715 |
| 716 for (k = 0; k < tx_dim; k++) { |
| 717 temp += im[k] * pthc[k]; |
| 718 } |
| 719 |
| 720 op[i] = (int16_t)((temp + HORIZONTAL_ROUNDING) >> HORIZONTAL_SHIFT); |
| 721 pthc += tx_dim; |
| 722 } |
| 723 |
| 724 im += tx_dim; // 16 |
| 725 op += tx_dim; |
| 726 } |
| 727 } |
| 728 |
| 729 void vp9_short_fdct4x4_c(short *input, short *output, int pitch) { |
| 730 int i; |
| 731 int a1, b1, c1, d1; |
| 732 short *ip = input; |
| 733 short *op = output; |
| 734 |
| 735 for (i = 0; i < 4; i++) { |
| 736 a1 = ((ip[0] + ip[3]) << 5); |
| 737 b1 = ((ip[1] + ip[2]) << 5); |
| 738 c1 = ((ip[1] - ip[2]) << 5); |
| 739 d1 = ((ip[0] - ip[3]) << 5); |
| 740 |
| 741 op[0] = a1 + b1; |
| 742 op[2] = a1 - b1; |
| 743 |
| 744 op[1] = (c1 * 2217 + d1 * 5352 + 14500) >> 12; |
| 745 op[3] = (d1 * 2217 - c1 * 5352 + 7500) >> 12; |
| 746 |
| 747 ip += pitch / 2; |
| 748 op += 4; |
| 749 |
| 750 } |
| 751 ip = output; |
| 752 op = output; |
| 753 for (i = 0; i < 4; i++) { |
| 754 a1 = ip[0] + ip[12]; |
| 755 b1 = ip[4] + ip[8]; |
| 756 c1 = ip[4] - ip[8]; |
| 757 d1 = ip[0] - ip[12]; |
| 758 |
| 759 op[0] = (a1 + b1 + 7) >> 4; |
| 760 op[8] = (a1 - b1 + 7) >> 4; |
| 761 |
| 762 op[4] = ((c1 * 2217 + d1 * 5352 + 12000) >> 16) + (d1 != 0); |
| 763 op[12] = (d1 * 2217 - c1 * 5352 + 51000) >> 16; |
| 764 |
| 765 ip++; |
| 766 op++; |
| 767 } |
| 768 } |
| 769 |
| 770 void vp9_short_fdct8x4_c(short *input, short *output, int pitch) |
| 771 { |
| 772 vp9_short_fdct4x4_c(input, output, pitch); |
| 773 vp9_short_fdct4x4_c(input + 4, output + 16, pitch); |
| 774 } |
| 775 |
| 776 void vp9_short_walsh4x4_c(short *input, short *output, int pitch) { |
| 777 int i; |
| 778 int a1, b1, c1, d1; |
| 779 short *ip = input; |
| 780 short *op = output; |
| 781 int pitch_short = pitch >> 1; |
| 782 |
| 783 for (i = 0; i < 4; i++) { |
| 784 a1 = ip[0 * pitch_short] + ip[3 * pitch_short]; |
| 785 b1 = ip[1 * pitch_short] + ip[2 * pitch_short]; |
| 786 c1 = ip[1 * pitch_short] - ip[2 * pitch_short]; |
| 787 d1 = ip[0 * pitch_short] - ip[3 * pitch_short]; |
| 788 |
| 789 op[0] = (a1 + b1 + 1) >> 1; |
| 790 op[4] = (c1 + d1) >> 1; |
| 791 op[8] = (a1 - b1) >> 1; |
| 792 op[12] = (d1 - c1) >> 1; |
| 793 |
| 794 ip++; |
| 795 op++; |
| 796 } |
| 797 ip = output; |
| 798 op = output; |
| 799 |
| 800 for (i = 0; i < 4; i++) { |
| 801 a1 = ip[0] + ip[3]; |
| 802 b1 = ip[1] + ip[2]; |
| 803 c1 = ip[1] - ip[2]; |
| 804 d1 = ip[0] - ip[3]; |
| 805 |
| 806 op[0] = (a1 + b1 + 1) >> 1; |
| 807 op[1] = (c1 + d1) >> 1; |
| 808 op[2] = (a1 - b1) >> 1; |
| 809 op[3] = (d1 - c1) >> 1; |
| 810 |
| 811 ip += 4; |
| 812 op += 4; |
| 813 } |
| 814 } |
| 815 |
| 816 #if CONFIG_LOSSLESS |
| 817 void vp9_short_walsh4x4_lossless_c(short *input, short *output, int pitch) { |
| 818 int i; |
| 819 int a1, b1, c1, d1; |
| 820 short *ip = input; |
| 821 short *op = output; |
| 822 int pitch_short = pitch >> 1; |
| 823 |
| 824 for (i = 0; i < 4; i++) { |
| 825 a1 = (ip[0 * pitch_short] + ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; |
| 826 b1 = (ip[1 * pitch_short] + ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; |
| 827 c1 = (ip[1 * pitch_short] - ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; |
| 828 d1 = (ip[0 * pitch_short] - ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR; |
| 829 |
| 830 op[0] = (a1 + b1 + 1) >> 1; |
| 831 op[4] = (c1 + d1) >> 1; |
| 832 op[8] = (a1 - b1) >> 1; |
| 833 op[12] = (d1 - c1) >> 1; |
| 834 |
| 835 ip++; |
| 836 op++; |
| 837 } |
| 838 ip = output; |
| 839 op = output; |
| 840 |
| 841 for (i = 0; i < 4; i++) { |
| 842 a1 = ip[0] + ip[3]; |
| 843 b1 = ip[1] + ip[2]; |
| 844 c1 = ip[1] - ip[2]; |
| 845 d1 = ip[0] - ip[3]; |
| 846 |
| 847 op[0] = ((a1 + b1 + 1) >> 1) << Y2_WHT_UPSCALE_FACTOR; |
| 848 op[1] = ((c1 + d1) >> 1) << Y2_WHT_UPSCALE_FACTOR; |
| 849 op[2] = ((a1 - b1) >> 1) << Y2_WHT_UPSCALE_FACTOR; |
| 850 op[3] = ((d1 - c1) >> 1) << Y2_WHT_UPSCALE_FACTOR; |
| 851 |
| 852 ip += 4; |
| 853 op += 4; |
| 854 } |
| 855 } |
| 856 |
| 857 void vp9_short_walsh4x4_x8_c(short *input, short *output, int pitch) { |
| 858 int i; |
| 859 int a1, b1, c1, d1; |
| 860 short *ip = input; |
| 861 short *op = output; |
| 862 int pitch_short = pitch >> 1; |
| 863 |
| 864 for (i = 0; i < 4; i++) { |
| 865 a1 = ip[0 * pitch_short] + ip[3 * pitch_short]; |
| 866 b1 = ip[1 * pitch_short] + ip[2 * pitch_short]; |
| 867 c1 = ip[1 * pitch_short] - ip[2 * pitch_short]; |
| 868 d1 = ip[0 * pitch_short] - ip[3 * pitch_short]; |
| 869 |
| 870 op[0] = (a1 + b1 + 1) >> 1; |
| 871 op[4] = (c1 + d1) >> 1; |
| 872 op[8] = (a1 - b1) >> 1; |
| 873 op[12] = (d1 - c1) >> 1; |
| 874 |
| 875 ip++; |
| 876 op++; |
| 877 } |
| 878 ip = output; |
| 879 op = output; |
| 880 |
| 881 for (i = 0; i < 4; i++) { |
| 882 a1 = ip[0] + ip[3]; |
| 883 b1 = ip[1] + ip[2]; |
| 884 c1 = ip[1] - ip[2]; |
| 885 d1 = ip[0] - ip[3]; |
| 886 |
| 887 op[0] = ((a1 + b1 + 1) >> 1) << WHT_UPSCALE_FACTOR; |
| 888 op[1] = ((c1 + d1) >> 1) << WHT_UPSCALE_FACTOR; |
| 889 op[2] = ((a1 - b1) >> 1) << WHT_UPSCALE_FACTOR; |
| 890 op[3] = ((d1 - c1) >> 1) << WHT_UPSCALE_FACTOR; |
| 891 |
| 892 ip += 4; |
| 893 op += 4; |
| 894 } |
| 895 } |
| 896 |
| 897 void vp9_short_walsh8x4_x8_c(short *input, short *output, int pitch) { |
| 898 vp9_short_walsh4x4_x8_c(input, output, pitch); |
| 899 vp9_short_walsh4x4_x8_c(input + 4, output + 16, pitch); |
| 900 } |
| 901 #endif |
| 902 |
| 903 #define TEST_INT_16x16_DCT 1 |
| 904 #if !TEST_INT_16x16_DCT |
| 905 static const double C1 = 0.995184726672197; |
| 906 static const double C2 = 0.98078528040323; |
| 907 static const double C3 = 0.956940335732209; |
| 908 static const double C4 = 0.923879532511287; |
| 909 static const double C5 = 0.881921264348355; |
| 910 static const double C6 = 0.831469612302545; |
| 911 static const double C7 = 0.773010453362737; |
| 912 static const double C8 = 0.707106781186548; |
| 913 static const double C9 = 0.634393284163646; |
| 914 static const double C10 = 0.555570233019602; |
| 915 static const double C11 = 0.471396736825998; |
| 916 static const double C12 = 0.38268343236509; |
| 917 static const double C13 = 0.290284677254462; |
| 918 static const double C14 = 0.195090322016128; |
| 919 static const double C15 = 0.098017140329561; |
| 920 |
| 921 static void dct16x16_1d(double input[16], double output[16]) { |
| 922 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 923 { |
| 924 double step[16]; |
| 925 double intermediate[16]; |
| 926 double temp1, temp2; |
| 927 |
| 928 // step 1 |
| 929 step[ 0] = input[0] + input[15]; |
| 930 step[ 1] = input[1] + input[14]; |
| 931 step[ 2] = input[2] + input[13]; |
| 932 step[ 3] = input[3] + input[12]; |
| 933 step[ 4] = input[4] + input[11]; |
| 934 step[ 5] = input[5] + input[10]; |
| 935 step[ 6] = input[6] + input[ 9]; |
| 936 step[ 7] = input[7] + input[ 8]; |
| 937 step[ 8] = input[7] - input[ 8]; |
| 938 step[ 9] = input[6] - input[ 9]; |
| 939 step[10] = input[5] - input[10]; |
| 940 step[11] = input[4] - input[11]; |
| 941 step[12] = input[3] - input[12]; |
| 942 step[13] = input[2] - input[13]; |
| 943 step[14] = input[1] - input[14]; |
| 944 step[15] = input[0] - input[15]; |
| 945 |
| 946 // step 2 |
| 947 output[0] = step[0] + step[7]; |
| 948 output[1] = step[1] + step[6]; |
| 949 output[2] = step[2] + step[5]; |
| 950 output[3] = step[3] + step[4]; |
| 951 output[4] = step[3] - step[4]; |
| 952 output[5] = step[2] - step[5]; |
| 953 output[6] = step[1] - step[6]; |
| 954 output[7] = step[0] - step[7]; |
| 955 |
| 956 temp1 = step[ 8]*C7; |
| 957 temp2 = step[15]*C9; |
| 958 output[ 8] = temp1 + temp2; |
| 959 |
| 960 temp1 = step[ 9]*C11; |
| 961 temp2 = step[14]*C5; |
| 962 output[ 9] = temp1 - temp2; |
| 963 |
| 964 temp1 = step[10]*C3; |
| 965 temp2 = step[13]*C13; |
| 966 output[10] = temp1 + temp2; |
| 967 |
| 968 temp1 = step[11]*C15; |
| 969 temp2 = step[12]*C1; |
| 970 output[11] = temp1 - temp2; |
| 971 |
| 972 temp1 = step[11]*C1; |
| 973 temp2 = step[12]*C15; |
| 974 output[12] = temp2 + temp1; |
| 975 |
| 976 temp1 = step[10]*C13; |
| 977 temp2 = step[13]*C3; |
| 978 output[13] = temp2 - temp1; |
| 979 |
| 980 temp1 = step[ 9]*C5; |
| 981 temp2 = step[14]*C11; |
| 982 output[14] = temp2 + temp1; |
| 983 |
| 984 temp1 = step[ 8]*C9; |
| 985 temp2 = step[15]*C7; |
| 986 output[15] = temp2 - temp1; |
| 987 |
| 988 // step 3 |
| 989 step[ 0] = output[0] + output[3]; |
| 990 step[ 1] = output[1] + output[2]; |
| 991 step[ 2] = output[1] - output[2]; |
| 992 step[ 3] = output[0] - output[3]; |
| 993 |
| 994 temp1 = output[4]*C14; |
| 995 temp2 = output[7]*C2; |
| 996 step[ 4] = temp1 + temp2; |
| 997 |
| 998 temp1 = output[5]*C10; |
| 999 temp2 = output[6]*C6; |
| 1000 step[ 5] = temp1 + temp2; |
| 1001 |
| 1002 temp1 = output[5]*C6; |
| 1003 temp2 = output[6]*C10; |
| 1004 step[ 6] = temp2 - temp1; |
| 1005 |
| 1006 temp1 = output[4]*C2; |
| 1007 temp2 = output[7]*C14; |
| 1008 step[ 7] = temp2 - temp1; |
| 1009 |
| 1010 step[ 8] = output[ 8] + output[11]; |
| 1011 step[ 9] = output[ 9] + output[10]; |
| 1012 step[10] = output[ 9] - output[10]; |
| 1013 step[11] = output[ 8] - output[11]; |
| 1014 |
| 1015 step[12] = output[12] + output[15]; |
| 1016 step[13] = output[13] + output[14]; |
| 1017 step[14] = output[13] - output[14]; |
| 1018 step[15] = output[12] - output[15]; |
| 1019 |
| 1020 // step 4 |
| 1021 output[ 0] = (step[ 0] + step[ 1]); |
| 1022 output[ 8] = (step[ 0] - step[ 1]); |
| 1023 |
| 1024 temp1 = step[2]*C12; |
| 1025 temp2 = step[3]*C4; |
| 1026 temp1 = temp1 + temp2; |
| 1027 output[ 4] = 2*(temp1*C8); |
| 1028 |
| 1029 temp1 = step[2]*C4; |
| 1030 temp2 = step[3]*C12; |
| 1031 temp1 = temp2 - temp1; |
| 1032 output[12] = 2*(temp1*C8); |
| 1033 |
| 1034 output[ 2] = 2*((step[4] + step[ 5])*C8); |
| 1035 output[14] = 2*((step[7] - step[ 6])*C8); |
| 1036 |
| 1037 temp1 = step[4] - step[5]; |
| 1038 temp2 = step[6] + step[7]; |
| 1039 output[ 6] = (temp1 + temp2); |
| 1040 output[10] = (temp1 - temp2); |
| 1041 |
| 1042 intermediate[8] = step[8] + step[14]; |
| 1043 intermediate[9] = step[9] + step[15]; |
| 1044 |
| 1045 temp1 = intermediate[8]*C12; |
| 1046 temp2 = intermediate[9]*C4; |
| 1047 temp1 = temp1 - temp2; |
| 1048 output[3] = 2*(temp1*C8); |
| 1049 |
| 1050 temp1 = intermediate[8]*C4; |
| 1051 temp2 = intermediate[9]*C12; |
| 1052 temp1 = temp2 + temp1; |
| 1053 output[13] = 2*(temp1*C8); |
| 1054 |
| 1055 output[ 9] = 2*((step[10] + step[11])*C8); |
| 1056 |
| 1057 intermediate[11] = step[10] - step[11]; |
| 1058 intermediate[12] = step[12] + step[13]; |
| 1059 intermediate[13] = step[12] - step[13]; |
| 1060 intermediate[14] = step[ 8] - step[14]; |
| 1061 intermediate[15] = step[ 9] - step[15]; |
| 1062 |
| 1063 output[15] = (intermediate[11] + intermediate[12]); |
| 1064 output[ 1] = -(intermediate[11] - intermediate[12]); |
| 1065 |
| 1066 output[ 7] = 2*(intermediate[13]*C8); |
| 1067 |
| 1068 temp1 = intermediate[14]*C12; |
| 1069 temp2 = intermediate[15]*C4; |
| 1070 temp1 = temp1 - temp2; |
| 1071 output[11] = -2*(temp1*C8); |
| 1072 |
| 1073 temp1 = intermediate[14]*C4; |
| 1074 temp2 = intermediate[15]*C12; |
| 1075 temp1 = temp2 + temp1; |
| 1076 output[ 5] = 2*(temp1*C8); |
| 1077 } |
| 1078 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 1079 } |
| 1080 |
| 1081 void vp9_short_fdct16x16_c(short *input, short *out, int pitch) { |
| 1082 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 1083 { |
| 1084 int shortpitch = pitch >> 1; |
| 1085 int i, j; |
| 1086 double output[256]; |
| 1087 // First transform columns |
| 1088 for (i = 0; i < 16; i++) { |
| 1089 double temp_in[16], temp_out[16]; |
| 1090 for (j = 0; j < 16; j++) |
| 1091 temp_in[j] = input[j*shortpitch + i]; |
| 1092 dct16x16_1d(temp_in, temp_out); |
| 1093 for (j = 0; j < 16; j++) |
| 1094 output[j*16 + i] = temp_out[j]; |
| 1095 } |
| 1096 // Then transform rows |
| 1097 for (i = 0; i < 16; ++i) { |
| 1098 double temp_in[16], temp_out[16]; |
| 1099 for (j = 0; j < 16; ++j) |
| 1100 temp_in[j] = output[j + i*16]; |
| 1101 dct16x16_1d(temp_in, temp_out); |
| 1102 for (j = 0; j < 16; ++j) |
| 1103 output[j + i*16] = temp_out[j]; |
| 1104 } |
| 1105 // Scale by some magic number |
| 1106 for (i = 0; i < 256; i++) |
| 1107 out[i] = (short)round(output[i]/2); |
| 1108 } |
| 1109 vp9_clear_system_state(); // Make it simd safe : __asm emms; |
| 1110 } |
| 1111 |
| 1112 #else |
| 1113 static const int16_t C1 = 16305; |
| 1114 static const int16_t C2 = 16069; |
| 1115 static const int16_t C3 = 15679; |
| 1116 static const int16_t C4 = 15137; |
| 1117 static const int16_t C5 = 14449; |
| 1118 static const int16_t C6 = 13623; |
| 1119 static const int16_t C7 = 12665; |
| 1120 static const int16_t C8 = 11585; |
| 1121 static const int16_t C9 = 10394; |
| 1122 static const int16_t C10 = 9102; |
| 1123 static const int16_t C11 = 7723; |
| 1124 static const int16_t C12 = 6270; |
| 1125 static const int16_t C13 = 4756; |
| 1126 static const int16_t C14 = 3196; |
| 1127 static const int16_t C15 = 1606; |
| 1128 |
| 1129 #define RIGHT_SHIFT 14 |
| 1130 #define ROUNDING (1 << (RIGHT_SHIFT - 1)) |
| 1131 |
| 1132 static void dct16x16_1d(int16_t input[16], int16_t output[16], |
| 1133 int last_shift_bits) { |
| 1134 int16_t step[16]; |
| 1135 int intermediate[16]; |
| 1136 int temp1, temp2; |
| 1137 int final_shift = RIGHT_SHIFT; |
| 1138 int final_rounding = ROUNDING; |
| 1139 int output_shift = 0; |
| 1140 int output_rounding = 0; |
| 1141 |
| 1142 final_shift += last_shift_bits; |
| 1143 if (final_shift > 0) |
| 1144 final_rounding = 1 << (final_shift - 1); |
| 1145 |
| 1146 output_shift += last_shift_bits; |
| 1147 if (output_shift > 0) |
| 1148 output_rounding = 1 << (output_shift - 1); |
| 1149 |
| 1150 // step 1 |
| 1151 step[ 0] = input[0] + input[15]; |
| 1152 step[ 1] = input[1] + input[14]; |
| 1153 step[ 2] = input[2] + input[13]; |
| 1154 step[ 3] = input[3] + input[12]; |
| 1155 step[ 4] = input[4] + input[11]; |
| 1156 step[ 5] = input[5] + input[10]; |
| 1157 step[ 6] = input[6] + input[ 9]; |
| 1158 step[ 7] = input[7] + input[ 8]; |
| 1159 step[ 8] = input[7] - input[ 8]; |
| 1160 step[ 9] = input[6] - input[ 9]; |
| 1161 step[10] = input[5] - input[10]; |
| 1162 step[11] = input[4] - input[11]; |
| 1163 step[12] = input[3] - input[12]; |
| 1164 step[13] = input[2] - input[13]; |
| 1165 step[14] = input[1] - input[14]; |
| 1166 step[15] = input[0] - input[15]; |
| 1167 |
| 1168 // step 2 |
| 1169 output[0] = step[0] + step[7]; |
| 1170 output[1] = step[1] + step[6]; |
| 1171 output[2] = step[2] + step[5]; |
| 1172 output[3] = step[3] + step[4]; |
| 1173 output[4] = step[3] - step[4]; |
| 1174 output[5] = step[2] - step[5]; |
| 1175 output[6] = step[1] - step[6]; |
| 1176 output[7] = step[0] - step[7]; |
| 1177 |
| 1178 temp1 = step[ 8] * C7; |
| 1179 temp2 = step[15] * C9; |
| 1180 output[ 8] = (temp1 + temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1181 |
| 1182 temp1 = step[ 9] * C11; |
| 1183 temp2 = step[14] * C5; |
| 1184 output[ 9] = (temp1 - temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1185 |
| 1186 temp1 = step[10] * C3; |
| 1187 temp2 = step[13] * C13; |
| 1188 output[10] = (temp1 + temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1189 |
| 1190 temp1 = step[11] * C15; |
| 1191 temp2 = step[12] * C1; |
| 1192 output[11] = (temp1 - temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1193 |
| 1194 temp1 = step[11] * C1; |
| 1195 temp2 = step[12] * C15; |
| 1196 output[12] = (temp2 + temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1197 |
| 1198 temp1 = step[10] * C13; |
| 1199 temp2 = step[13] * C3; |
| 1200 output[13] = (temp2 - temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1201 |
| 1202 temp1 = step[ 9] * C5; |
| 1203 temp2 = step[14] * C11; |
| 1204 output[14] = (temp2 + temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1205 |
| 1206 temp1 = step[ 8] * C9; |
| 1207 temp2 = step[15] * C7; |
| 1208 output[15] = (temp2 - temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1209 |
| 1210 // step 3 |
| 1211 step[ 0] = output[0] + output[3]; |
| 1212 step[ 1] = output[1] + output[2]; |
| 1213 step[ 2] = output[1] - output[2]; |
| 1214 step[ 3] = output[0] - output[3]; |
| 1215 |
| 1216 temp1 = output[4] * C14; |
| 1217 temp2 = output[7] * C2; |
| 1218 step[ 4] = (temp1 + temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1219 |
| 1220 temp1 = output[5] * C10; |
| 1221 temp2 = output[6] * C6; |
| 1222 step[ 5] = (temp1 + temp2 + ROUNDING) >> RIGHT_SHIFT; |
| 1223 |
| 1224 temp1 = output[5] * C6; |
| 1225 temp2 = output[6] * C10; |
| 1226 step[ 6] = (temp2 - temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1227 |
| 1228 temp1 = output[4] * C2; |
| 1229 temp2 = output[7] * C14; |
| 1230 step[ 7] = (temp2 - temp1 + ROUNDING) >> RIGHT_SHIFT; |
| 1231 |
| 1232 step[ 8] = output[ 8] + output[11]; |
| 1233 step[ 9] = output[ 9] + output[10]; |
| 1234 step[10] = output[ 9] - output[10]; |
| 1235 step[11] = output[ 8] - output[11]; |
| 1236 |
| 1237 step[12] = output[12] + output[15]; |
| 1238 step[13] = output[13] + output[14]; |
| 1239 step[14] = output[13] - output[14]; |
| 1240 step[15] = output[12] - output[15]; |
| 1241 |
| 1242 // step 4 |
| 1243 output[ 0] = (step[ 0] + step[ 1] + output_rounding) >> output_shift; |
| 1244 output[ 8] = (step[ 0] - step[ 1] + output_rounding) >> output_shift; |
| 1245 |
| 1246 temp1 = step[2] * C12; |
| 1247 temp2 = step[3] * C4; |
| 1248 temp1 = (temp1 + temp2 + final_rounding) >> final_shift; |
| 1249 output[ 4] = (2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1250 |
| 1251 temp1 = step[2] * C4; |
| 1252 temp2 = step[3] * C12; |
| 1253 temp1 = (temp2 - temp1 + final_rounding) >> final_shift; |
| 1254 output[12] = (2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1255 |
| 1256 output[ 2] = (2 * ((step[4] + step[ 5]) * C8) + final_rounding) |
| 1257 >> final_shift; |
| 1258 output[14] = (2 * ((step[7] - step[ 6]) * C8) + final_rounding) |
| 1259 >> final_shift; |
| 1260 |
| 1261 temp1 = step[4] - step[5]; |
| 1262 temp2 = step[6] + step[7]; |
| 1263 output[ 6] = (temp1 + temp2 + output_rounding) >> output_shift; |
| 1264 output[10] = (temp1 - temp2 + output_rounding) >> output_shift; |
| 1265 |
| 1266 intermediate[8] = step[8] + step[14]; |
| 1267 intermediate[9] = step[9] + step[15]; |
| 1268 |
| 1269 temp1 = intermediate[8] * C12; |
| 1270 temp2 = intermediate[9] * C4; |
| 1271 temp1 = (temp1 - temp2 + final_rounding) >> final_shift; |
| 1272 output[3] = (2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1273 |
| 1274 temp1 = intermediate[8] * C4; |
| 1275 temp2 = intermediate[9] * C12; |
| 1276 temp1 = (temp2 + temp1 + final_rounding) >> final_shift; |
| 1277 output[13] = (2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1278 |
| 1279 output[ 9] = (2 * ((step[10] + step[11]) * C8) + final_rounding) |
| 1280 >> final_shift; |
| 1281 |
| 1282 intermediate[11] = step[10] - step[11]; |
| 1283 intermediate[12] = step[12] + step[13]; |
| 1284 intermediate[13] = step[12] - step[13]; |
| 1285 intermediate[14] = step[ 8] - step[14]; |
| 1286 intermediate[15] = step[ 9] - step[15]; |
| 1287 |
| 1288 output[15] = (intermediate[11] + intermediate[12] + output_rounding) |
| 1289 >> output_shift; |
| 1290 output[ 1] = -(intermediate[11] - intermediate[12] + output_rounding) |
| 1291 >> output_shift; |
| 1292 |
| 1293 output[ 7] = (2 * (intermediate[13] * C8) + final_rounding) >> final_shift; |
| 1294 |
| 1295 temp1 = intermediate[14] * C12; |
| 1296 temp2 = intermediate[15] * C4; |
| 1297 temp1 = (temp1 - temp2 + final_rounding) >> final_shift; |
| 1298 output[11] = (-2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1299 |
| 1300 temp1 = intermediate[14] * C4; |
| 1301 temp2 = intermediate[15] * C12; |
| 1302 temp1 = (temp2 + temp1 + final_rounding) >> final_shift; |
| 1303 output[ 5] = (2 * (temp1 * C8) + ROUNDING) >> RIGHT_SHIFT; |
| 1304 } |
| 1305 |
| 1306 void vp9_short_fdct16x16_c(int16_t *input, int16_t *out, int pitch) { |
| 1307 int shortpitch = pitch >> 1; |
| 1308 int i, j; |
| 1309 int16_t output[256]; |
| 1310 int16_t *outptr = &output[0]; |
| 1311 |
| 1312 // First transform columns |
| 1313 for (i = 0; i < 16; i++) { |
| 1314 int16_t temp_in[16]; |
| 1315 int16_t temp_out[16]; |
| 1316 for (j = 0; j < 16; j++) |
| 1317 temp_in[j] = input[j * shortpitch + i]; |
| 1318 dct16x16_1d(temp_in, temp_out, 0); |
| 1319 for (j = 0; j < 16; j++) |
| 1320 output[j * 16 + i] = temp_out[j]; |
| 1321 } |
| 1322 |
| 1323 // Then transform rows |
| 1324 for (i = 0; i < 16; ++i) { |
| 1325 dct16x16_1d(outptr, out, 1); |
| 1326 outptr += 16; |
| 1327 out += 16; |
| 1328 } |
| 1329 } |
| 1330 #undef RIGHT_SHIFT |
| 1331 #undef ROUNDING |
| 1332 #endif |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 11555023:
libvpx: Add VP9 decoder. (Closed)
Base URL: svn://chrome-svn/chrome/trunk/deps/third_party/libvpx/