1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 2021 Collabora, Ltd.
10 #include <linux/module.h>
12 #include <media/v4l2-vp9.h>
14 const u8 v4l2_vp9_kf_y_mode_prob[10][10][9] = {
17 { 137, 30, 42, 148, 151, 207, 70, 52, 91 }, /*left = dc */
18 { 92, 45, 102, 136, 116, 180, 74, 90, 100 }, /*left = v */
19 { 73, 32, 19, 187, 222, 215, 46, 34, 100 }, /*left = h */
20 { 91, 30, 32, 116, 121, 186, 93, 86, 94 }, /*left = d45 */
21 { 72, 35, 36, 149, 68, 206, 68, 63, 105 }, /*left = d135*/
22 { 73, 31, 28, 138, 57, 124, 55, 122, 151 }, /*left = d117*/
23 { 67, 23, 21, 140, 126, 197, 40, 37, 171 }, /*left = d153*/
24 { 86, 27, 28, 128, 154, 212, 45, 43, 53 }, /*left = d207*/
25 { 74, 32, 27, 107, 86, 160, 63, 134, 102 }, /*left = d63 */
26 { 59, 67, 44, 140, 161, 202, 78, 67, 119 }, /*left = tm */
28 { 63, 36, 126, 146, 123, 158, 60, 90, 96 }, /*left = dc */
29 { 43, 46, 168, 134, 107, 128, 69, 142, 92 }, /*left = v */
30 { 44, 29, 68, 159, 201, 177, 50, 57, 77 }, /*left = h */
31 { 58, 38, 76, 114, 97, 172, 78, 133, 92 }, /*left = d45 */
32 { 46, 41, 76, 140, 63, 184, 69, 112, 57 }, /*left = d135*/
33 { 38, 32, 85, 140, 46, 112, 54, 151, 133 }, /*left = d117*/
34 { 39, 27, 61, 131, 110, 175, 44, 75, 136 }, /*left = d153*/
35 { 52, 30, 74, 113, 130, 175, 51, 64, 58 }, /*left = d207*/
36 { 47, 35, 80, 100, 74, 143, 64, 163, 74 }, /*left = d63 */
37 { 36, 61, 116, 114, 128, 162, 80, 125, 82 }, /*left = tm */
39 { 82, 26, 26, 171, 208, 204, 44, 32, 105 }, /*left = dc */
40 { 55, 44, 68, 166, 179, 192, 57, 57, 108 }, /*left = v */
41 { 42, 26, 11, 199, 241, 228, 23, 15, 85 }, /*left = h */
42 { 68, 42, 19, 131, 160, 199, 55, 52, 83 }, /*left = d45 */
43 { 58, 50, 25, 139, 115, 232, 39, 52, 118 }, /*left = d135*/
44 { 50, 35, 33, 153, 104, 162, 64, 59, 131 }, /*left = d117*/
45 { 44, 24, 16, 150, 177, 202, 33, 19, 156 }, /*left = d153*/
46 { 55, 27, 12, 153, 203, 218, 26, 27, 49 }, /*left = d207*/
47 { 53, 49, 21, 110, 116, 168, 59, 80, 76 }, /*left = d63 */
48 { 38, 72, 19, 168, 203, 212, 50, 50, 107 }, /*left = tm */
49 }, { /* above = d45 */
50 { 103, 26, 36, 129, 132, 201, 83, 80, 93 }, /*left = dc */
51 { 59, 38, 83, 112, 103, 162, 98, 136, 90 }, /*left = v */
52 { 62, 30, 23, 158, 200, 207, 59, 57, 50 }, /*left = h */
53 { 67, 30, 29, 84, 86, 191, 102, 91, 59 }, /*left = d45 */
54 { 60, 32, 33, 112, 71, 220, 64, 89, 104 }, /*left = d135*/
55 { 53, 26, 34, 130, 56, 149, 84, 120, 103 }, /*left = d117*/
56 { 53, 21, 23, 133, 109, 210, 56, 77, 172 }, /*left = d153*/
57 { 77, 19, 29, 112, 142, 228, 55, 66, 36 }, /*left = d207*/
58 { 61, 29, 29, 93, 97, 165, 83, 175, 162 }, /*left = d63 */
59 { 47, 47, 43, 114, 137, 181, 100, 99, 95 }, /*left = tm */
60 }, { /* above = d135 */
61 { 69, 23, 29, 128, 83, 199, 46, 44, 101 }, /*left = dc */
62 { 53, 40, 55, 139, 69, 183, 61, 80, 110 }, /*left = v */
63 { 40, 29, 19, 161, 180, 207, 43, 24, 91 }, /*left = h */
64 { 60, 34, 19, 105, 61, 198, 53, 64, 89 }, /*left = d45 */
65 { 52, 31, 22, 158, 40, 209, 58, 62, 89 }, /*left = d135*/
66 { 44, 31, 29, 147, 46, 158, 56, 102, 198 }, /*left = d117*/
67 { 35, 19, 12, 135, 87, 209, 41, 45, 167 }, /*left = d153*/
68 { 55, 25, 21, 118, 95, 215, 38, 39, 66 }, /*left = d207*/
69 { 51, 38, 25, 113, 58, 164, 70, 93, 97 }, /*left = d63 */
70 { 47, 54, 34, 146, 108, 203, 72, 103, 151 }, /*left = tm */
71 }, { /* above = d117 */
72 { 64, 19, 37, 156, 66, 138, 49, 95, 133 }, /*left = dc */
73 { 46, 27, 80, 150, 55, 124, 55, 121, 135 }, /*left = v */
74 { 36, 23, 27, 165, 149, 166, 54, 64, 118 }, /*left = h */
75 { 53, 21, 36, 131, 63, 163, 60, 109, 81 }, /*left = d45 */
76 { 40, 26, 35, 154, 40, 185, 51, 97, 123 }, /*left = d135*/
77 { 35, 19, 34, 179, 19, 97, 48, 129, 124 }, /*left = d117*/
78 { 36, 20, 26, 136, 62, 164, 33, 77, 154 }, /*left = d153*/
79 { 45, 18, 32, 130, 90, 157, 40, 79, 91 }, /*left = d207*/
80 { 45, 26, 28, 129, 45, 129, 49, 147, 123 }, /*left = d63 */
81 { 38, 44, 51, 136, 74, 162, 57, 97, 121 }, /*left = tm */
82 }, { /* above = d153 */
83 { 75, 17, 22, 136, 138, 185, 32, 34, 166 }, /*left = dc */
84 { 56, 39, 58, 133, 117, 173, 48, 53, 187 }, /*left = v */
85 { 35, 21, 12, 161, 212, 207, 20, 23, 145 }, /*left = h */
86 { 56, 29, 19, 117, 109, 181, 55, 68, 112 }, /*left = d45 */
87 { 47, 29, 17, 153, 64, 220, 59, 51, 114 }, /*left = d135*/
88 { 46, 16, 24, 136, 76, 147, 41, 64, 172 }, /*left = d117*/
89 { 34, 17, 11, 108, 152, 187, 13, 15, 209 }, /*left = d153*/
90 { 51, 24, 14, 115, 133, 209, 32, 26, 104 }, /*left = d207*/
91 { 55, 30, 18, 122, 79, 179, 44, 88, 116 }, /*left = d63 */
92 { 37, 49, 25, 129, 168, 164, 41, 54, 148 }, /*left = tm */
93 }, { /* above = d207 */
94 { 82, 22, 32, 127, 143, 213, 39, 41, 70 }, /*left = dc */
95 { 62, 44, 61, 123, 105, 189, 48, 57, 64 }, /*left = v */
96 { 47, 25, 17, 175, 222, 220, 24, 30, 86 }, /*left = h */
97 { 68, 36, 17, 106, 102, 206, 59, 74, 74 }, /*left = d45 */
98 { 57, 39, 23, 151, 68, 216, 55, 63, 58 }, /*left = d135*/
99 { 49, 30, 35, 141, 70, 168, 82, 40, 115 }, /*left = d117*/
100 { 51, 25, 15, 136, 129, 202, 38, 35, 139 }, /*left = d153*/
101 { 68, 26, 16, 111, 141, 215, 29, 28, 28 }, /*left = d207*/
102 { 59, 39, 19, 114, 75, 180, 77, 104, 42 }, /*left = d63 */
103 { 40, 61, 26, 126, 152, 206, 61, 59, 93 }, /*left = tm */
104 }, { /* above = d63 */
105 { 78, 23, 39, 111, 117, 170, 74, 124, 94 }, /*left = dc */
106 { 48, 34, 86, 101, 92, 146, 78, 179, 134 }, /*left = v */
107 { 47, 22, 24, 138, 187, 178, 68, 69, 59 }, /*left = h */
108 { 56, 25, 33, 105, 112, 187, 95, 177, 129 }, /*left = d45 */
109 { 48, 31, 27, 114, 63, 183, 82, 116, 56 }, /*left = d135*/
110 { 43, 28, 37, 121, 63, 123, 61, 192, 169 }, /*left = d117*/
111 { 42, 17, 24, 109, 97, 177, 56, 76, 122 }, /*left = d153*/
112 { 58, 18, 28, 105, 139, 182, 70, 92, 63 }, /*left = d207*/
113 { 46, 23, 32, 74, 86, 150, 67, 183, 88 }, /*left = d63 */
114 { 36, 38, 48, 92, 122, 165, 88, 137, 91 }, /*left = tm */
115 }, { /* above = tm */
116 { 65, 70, 60, 155, 159, 199, 61, 60, 81 }, /*left = dc */
117 { 44, 78, 115, 132, 119, 173, 71, 112, 93 }, /*left = v */
118 { 39, 38, 21, 184, 227, 206, 42, 32, 64 }, /*left = h */
119 { 58, 47, 36, 124, 137, 193, 80, 82, 78 }, /*left = d45 */
120 { 49, 50, 35, 144, 95, 205, 63, 78, 59 }, /*left = d135*/
121 { 41, 53, 52, 148, 71, 142, 65, 128, 51 }, /*left = d117*/
122 { 40, 36, 28, 143, 143, 202, 40, 55, 137 }, /*left = d153*/
123 { 52, 34, 29, 129, 183, 227, 42, 35, 43 }, /*left = d207*/
124 { 42, 44, 44, 104, 105, 164, 64, 130, 80 }, /*left = d63 */
125 { 43, 81, 53, 140, 169, 204, 68, 84, 72 }, /*left = tm */
128 EXPORT_SYMBOL_GPL(v4l2_vp9_kf_y_mode_prob);
130 const u8 v4l2_vp9_kf_partition_probs[16][3] = {
132 { 158, 97, 94 }, /* a/l both not split */
133 { 93, 24, 99 }, /* a split, l not split */
134 { 85, 119, 44 }, /* l split, a not split */
135 { 62, 59, 67 }, /* a/l both split */
137 { 149, 53, 53 }, /* a/l both not split */
138 { 94, 20, 48 }, /* a split, l not split */
139 { 83, 53, 24 }, /* l split, a not split */
140 { 52, 18, 18 }, /* a/l both split */
142 { 150, 40, 39 }, /* a/l both not split */
143 { 78, 12, 26 }, /* a split, l not split */
144 { 67, 33, 11 }, /* l split, a not split */
145 { 24, 7, 5 }, /* a/l both split */
147 { 174, 35, 49 }, /* a/l both not split */
148 { 68, 11, 27 }, /* a split, l not split */
149 { 57, 15, 9 }, /* l split, a not split */
150 { 12, 3, 3 }, /* a/l both split */
152 EXPORT_SYMBOL_GPL(v4l2_vp9_kf_partition_probs);
154 const u8 v4l2_vp9_kf_uv_mode_prob[10][9] = {
155 { 144, 11, 54, 157, 195, 130, 46, 58, 108 }, /* y = dc */
156 { 118, 15, 123, 148, 131, 101, 44, 93, 131 }, /* y = v */
157 { 113, 12, 23, 188, 226, 142, 26, 32, 125 }, /* y = h */
158 { 120, 11, 50, 123, 163, 135, 64, 77, 103 }, /* y = d45 */
159 { 113, 9, 36, 155, 111, 157, 32, 44, 161 }, /* y = d135 */
160 { 116, 9, 55, 176, 76, 96, 37, 61, 149 }, /* y = d117 */
161 { 115, 9, 28, 141, 161, 167, 21, 25, 193 }, /* y = d153 */
162 { 120, 12, 32, 145, 195, 142, 32, 38, 86 }, /* y = d207 */
163 { 116, 12, 64, 120, 140, 125, 49, 115, 121 }, /* y = d63 */
164 { 102, 19, 66, 162, 182, 122, 35, 59, 128 } /* y = tm */
166 EXPORT_SYMBOL_GPL(v4l2_vp9_kf_uv_mode_prob);
168 const struct v4l2_vp9_frame_context v4l2_vp9_default_probs = {
1008 .skip = { 192, 128, 64 },
1024 .is_inter = { 9, 102, 187, 225 },
1025 .comp_mode = { 239, 183, 119, 96, 41 },
1033 .comp_ref = { 50, 126, 123, 221, 226 },
1035 { 65, 32, 18, 144, 162, 194, 41, 51, 98 },
1036 { 132, 68, 18, 165, 217, 196, 45, 40, 78 },
1037 { 173, 80, 19, 176, 240, 193, 64, 35, 46 },
1038 { 221, 135, 38, 194, 248, 121, 96, 85, 29 },
1041 { 120, 7, 76, 176, 208, 126, 28, 54, 103 } /* y = dc */,
1042 { 48, 12, 154, 155, 139, 90, 34, 117, 119 } /* y = v */,
1043 { 67, 6, 25, 204, 243, 158, 13, 21, 96 } /* y = h */,
1044 { 97, 5, 44, 131, 176, 139, 48, 68, 97 } /* y = d45 */,
1045 { 83, 5, 42, 156, 111, 152, 26, 49, 152 } /* y = d135 */,
1046 { 80, 5, 58, 178, 74, 83, 33, 62, 145 } /* y = d117 */,
1047 { 86, 5, 32, 154, 192, 168, 14, 22, 163 } /* y = d153 */,
1048 { 85, 5, 32, 156, 216, 148, 19, 29, 73 } /* y = d207 */,
1049 { 77, 7, 64, 116, 132, 122, 37, 126, 120 } /* y = d63 */,
1050 { 101, 21, 107, 181, 192, 103, 19, 67, 125 } /* y = tm */
1054 { 199, 122, 141 } /* a/l both not split */,
1055 { 147, 63, 159 } /* a split, l not split */,
1056 { 148, 133, 118 } /* l split, a not split */,
1057 { 121, 104, 114 } /* a/l both split */,
1059 { 174, 73, 87 } /* a/l both not split */,
1060 { 92, 41, 83 } /* a split, l not split */,
1061 { 82, 99, 50 } /* l split, a not split */,
1062 { 53, 39, 39 } /* a/l both split */,
1063 /* 32x32 -> 16x16 */
1064 { 177, 58, 59 } /* a/l both not split */,
1065 { 68, 26, 63 } /* a split, l not split */,
1066 { 52, 79, 25 } /* l split, a not split */,
1067 { 17, 14, 12 } /* a/l both split */,
1068 /* 64x64 -> 32x32 */
1069 { 222, 34, 30 } /* a/l both not split */,
1070 { 72, 16, 44 } /* a split, l not split */,
1071 { 58, 32, 12 } /* l split, a not split */,
1072 { 10, 7, 6 } /* a/l both split */,
1076 .joint = { 32, 64, 96 },
1077 .sign = { 128, 128 },
1079 { 224, 144, 192, 168, 192, 176, 192, 198, 198, 245 },
1080 { 216, 128, 176, 160, 176, 176, 192, 198, 198, 208 },
1082 .class0_bit = { 216, 208 },
1084 { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240},
1085 { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240},
1101 .class0_hp = { 160, 160 },
1105 EXPORT_SYMBOL_GPL(v4l2_vp9_default_probs);
1107 static u32 fastdiv(u32 dividend, u16 divisor)
1109 #define DIV_INV(d) ((u32)(((1ULL << 32) + ((d) - 1)) / (d)))
1110 #define DIVS_INV(d0, d1, d2, d3, d4, d5, d6, d7, d8, d9) \
1111 DIV_INV(d0), DIV_INV(d1), DIV_INV(d2), DIV_INV(d3), \
1112 DIV_INV(d4), DIV_INV(d5), DIV_INV(d6), DIV_INV(d7), \
1113 DIV_INV(d8), DIV_INV(d9)
1115 static const u32 inv[] = {
1116 DIV_INV(2), DIV_INV(3), DIV_INV(4), DIV_INV(5),
1117 DIV_INV(6), DIV_INV(7), DIV_INV(8), DIV_INV(9),
1118 DIVS_INV(10, 11, 12, 13, 14, 15, 16, 17, 18, 19),
1119 DIVS_INV(20, 21, 22, 23, 24, 25, 26, 27, 28, 29),
1120 DIVS_INV(30, 31, 32, 33, 34, 35, 36, 37, 38, 39),
1121 DIVS_INV(40, 41, 42, 43, 44, 45, 46, 47, 48, 49),
1122 DIVS_INV(50, 51, 52, 53, 54, 55, 56, 57, 58, 59),
1123 DIVS_INV(60, 61, 62, 63, 64, 65, 66, 67, 68, 69),
1124 DIVS_INV(70, 71, 72, 73, 74, 75, 76, 77, 78, 79),
1125 DIVS_INV(80, 81, 82, 83, 84, 85, 86, 87, 88, 89),
1126 DIVS_INV(90, 91, 92, 93, 94, 95, 96, 97, 98, 99),
1127 DIVS_INV(100, 101, 102, 103, 104, 105, 106, 107, 108, 109),
1128 DIVS_INV(110, 111, 112, 113, 114, 115, 116, 117, 118, 119),
1129 DIVS_INV(120, 121, 122, 123, 124, 125, 126, 127, 128, 129),
1130 DIVS_INV(130, 131, 132, 133, 134, 135, 136, 137, 138, 139),
1131 DIVS_INV(140, 141, 142, 143, 144, 145, 146, 147, 148, 149),
1132 DIVS_INV(150, 151, 152, 153, 154, 155, 156, 157, 158, 159),
1133 DIVS_INV(160, 161, 162, 163, 164, 165, 166, 167, 168, 169),
1134 DIVS_INV(170, 171, 172, 173, 174, 175, 176, 177, 178, 179),
1135 DIVS_INV(180, 181, 182, 183, 184, 185, 186, 187, 188, 189),
1136 DIVS_INV(190, 191, 192, 193, 194, 195, 196, 197, 198, 199),
1137 DIVS_INV(200, 201, 202, 203, 204, 205, 206, 207, 208, 209),
1138 DIVS_INV(210, 211, 212, 213, 214, 215, 216, 217, 218, 219),
1139 DIVS_INV(220, 221, 222, 223, 224, 225, 226, 227, 228, 229),
1140 DIVS_INV(230, 231, 232, 233, 234, 235, 236, 237, 238, 239),
1141 DIVS_INV(240, 241, 242, 243, 244, 245, 246, 247, 248, 249),
1142 DIV_INV(250), DIV_INV(251), DIV_INV(252), DIV_INV(253),
1143 DIV_INV(254), DIV_INV(255), DIV_INV(256),
1148 else if (divisor == 1)
1151 if (WARN_ON(divisor - 2 >= ARRAY_SIZE(inv)))
1154 return ((u64)dividend * inv[divisor - 2]) >> 32;
1157 /* 6.3.6 inv_recenter_nonneg(v, m) */
1158 static int inv_recenter_nonneg(int v, int m)
1164 return m - ((v + 1) >> 1);
1166 return m + (v >> 1);
1170 * part of 6.3.5 inv_remap_prob(deltaProb, prob)
1171 * delta = inv_map_table[deltaProb] done by userspace
1173 static int update_prob(int delta, int prob)
1178 return prob <= 128 ?
1179 1 + inv_recenter_nonneg(delta, prob - 1) :
1180 255 - inv_recenter_nonneg(delta, 255 - prob);
1183 /* Counterpart to 6.3.2 tx_mode_probs() */
1184 static void update_tx_probs(struct v4l2_vp9_frame_context *probs,
1185 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1189 for (i = 0; i < ARRAY_SIZE(probs->tx8); i++) {
1190 u8 *p8x8 = probs->tx8[i];
1191 u8 *p16x16 = probs->tx16[i];
1192 u8 *p32x32 = probs->tx32[i];
1193 const u8 *d8x8 = deltas->tx8[i];
1194 const u8 *d16x16 = deltas->tx16[i];
1195 const u8 *d32x32 = deltas->tx32[i];
1197 p8x8[0] = update_prob(d8x8[0], p8x8[0]);
1198 p16x16[0] = update_prob(d16x16[0], p16x16[0]);
1199 p16x16[1] = update_prob(d16x16[1], p16x16[1]);
1200 p32x32[0] = update_prob(d32x32[0], p32x32[0]);
1201 p32x32[1] = update_prob(d32x32[1], p32x32[1]);
1202 p32x32[2] = update_prob(d32x32[2], p32x32[2]);
1206 #define BAND_6(band) ((band) == 0 ? 3 : 6)
1208 static void update_coeff(const u8 deltas[6][6][3], u8 probs[6][6][3])
1212 for (l = 0; l < 6; l++)
1213 for (m = 0; m < BAND_6(l); m++) {
1214 u8 *p = probs[l][m];
1215 const u8 *d = deltas[l][m];
1217 for (n = 0; n < 3; n++)
1218 p[n] = update_prob(d[n], p[n]);
1222 /* Counterpart to 6.3.7 read_coef_probs() */
1223 static void update_coef_probs(struct v4l2_vp9_frame_context *probs,
1224 const struct v4l2_ctrl_vp9_compressed_hdr *deltas,
1225 const struct v4l2_ctrl_vp9_frame *dec_params)
1229 for (i = 0; i < ARRAY_SIZE(probs->coef); i++) {
1230 for (j = 0; j < ARRAY_SIZE(probs->coef[0]); j++)
1231 for (k = 0; k < ARRAY_SIZE(probs->coef[0][0]); k++)
1232 update_coeff(deltas->coef[i][j][k], probs->coef[i][j][k]);
1234 if (deltas->tx_mode == i)
1239 /* Counterpart to 6.3.8 read_skip_prob() */
1240 static void update_skip_probs(struct v4l2_vp9_frame_context *probs,
1241 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1245 for (i = 0; i < ARRAY_SIZE(probs->skip); i++)
1246 probs->skip[i] = update_prob(deltas->skip[i], probs->skip[i]);
1249 /* Counterpart to 6.3.9 read_inter_mode_probs() */
1250 static void update_inter_mode_probs(struct v4l2_vp9_frame_context *probs,
1251 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1255 for (i = 0; i < ARRAY_SIZE(probs->inter_mode); i++) {
1256 u8 *p = probs->inter_mode[i];
1257 const u8 *d = deltas->inter_mode[i];
1259 p[0] = update_prob(d[0], p[0]);
1260 p[1] = update_prob(d[1], p[1]);
1261 p[2] = update_prob(d[2], p[2]);
1265 /* Counterpart to 6.3.10 read_interp_filter_probs() */
1266 static void update_interp_filter_probs(struct v4l2_vp9_frame_context *probs,
1267 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1271 for (i = 0; i < ARRAY_SIZE(probs->interp_filter); i++) {
1272 u8 *p = probs->interp_filter[i];
1273 const u8 *d = deltas->interp_filter[i];
1275 p[0] = update_prob(d[0], p[0]);
1276 p[1] = update_prob(d[1], p[1]);
1280 /* Counterpart to 6.3.11 read_is_inter_probs() */
1281 static void update_is_inter_probs(struct v4l2_vp9_frame_context *probs,
1282 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1286 for (i = 0; i < ARRAY_SIZE(probs->is_inter); i++)
1287 probs->is_inter[i] = update_prob(deltas->is_inter[i], probs->is_inter[i]);
1290 /* 6.3.12 frame_reference_mode() done entirely in userspace */
1292 /* Counterpart to 6.3.13 frame_reference_mode_probs() */
1294 update_frame_reference_mode_probs(unsigned int reference_mode,
1295 struct v4l2_vp9_frame_context *probs,
1296 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1300 if (reference_mode == V4L2_VP9_REFERENCE_MODE_SELECT)
1301 for (i = 0; i < ARRAY_SIZE(probs->comp_mode); i++)
1302 probs->comp_mode[i] = update_prob(deltas->comp_mode[i],
1303 probs->comp_mode[i]);
1305 if (reference_mode != V4L2_VP9_REFERENCE_MODE_COMPOUND_REFERENCE)
1306 for (i = 0; i < ARRAY_SIZE(probs->single_ref); i++) {
1307 u8 *p = probs->single_ref[i];
1308 const u8 *d = deltas->single_ref[i];
1310 p[0] = update_prob(d[0], p[0]);
1311 p[1] = update_prob(d[1], p[1]);
1314 if (reference_mode != V4L2_VP9_REFERENCE_MODE_SINGLE_REFERENCE)
1315 for (i = 0; i < ARRAY_SIZE(probs->comp_ref); i++)
1316 probs->comp_ref[i] = update_prob(deltas->comp_ref[i], probs->comp_ref[i]);
1319 /* Counterpart to 6.3.14 read_y_mode_probs() */
1320 static void update_y_mode_probs(struct v4l2_vp9_frame_context *probs,
1321 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1325 for (i = 0; i < ARRAY_SIZE(probs->y_mode); i++)
1326 for (j = 0; j < ARRAY_SIZE(probs->y_mode[0]); ++j)
1327 probs->y_mode[i][j] =
1328 update_prob(deltas->y_mode[i][j], probs->y_mode[i][j]);
1331 /* Counterpart to 6.3.15 read_partition_probs() */
1332 static void update_partition_probs(struct v4l2_vp9_frame_context *probs,
1333 const struct v4l2_ctrl_vp9_compressed_hdr *deltas)
1337 for (i = 0; i < 4; i++)
1338 for (j = 0; j < 4; j++) {
1339 u8 *p = probs->partition[i * 4 + j];
1340 const u8 *d = deltas->partition[i * 4 + j];
1342 p[0] = update_prob(d[0], p[0]);
1343 p[1] = update_prob(d[1], p[1]);
1344 p[2] = update_prob(d[2], p[2]);
1348 static inline int update_mv_prob(int delta, int prob)
1356 /* Counterpart to 6.3.16 mv_probs() */
1357 static void update_mv_probs(struct v4l2_vp9_frame_context *probs,
1358 const struct v4l2_ctrl_vp9_compressed_hdr *deltas,
1359 const struct v4l2_ctrl_vp9_frame *dec_params)
1361 u8 *p = probs->mv.joint;
1362 const u8 *d = deltas->mv.joint;
1365 p[0] = update_mv_prob(d[0], p[0]);
1366 p[1] = update_mv_prob(d[1], p[1]);
1367 p[2] = update_mv_prob(d[2], p[2]);
1369 for (i = 0; i < ARRAY_SIZE(probs->mv.sign); i++) {
1371 d = deltas->mv.sign;
1372 p[i] = update_mv_prob(d[i], p[i]);
1374 p = probs->mv.classes[i];
1375 d = deltas->mv.classes[i];
1376 for (j = 0; j < ARRAY_SIZE(probs->mv.classes[0]); j++)
1377 p[j] = update_mv_prob(d[j], p[j]);
1379 p = probs->mv.class0_bit;
1380 d = deltas->mv.class0_bit;
1381 p[i] = update_mv_prob(d[i], p[i]);
1383 p = probs->mv.bits[i];
1384 d = deltas->mv.bits[i];
1385 for (j = 0; j < ARRAY_SIZE(probs->mv.bits[0]); j++)
1386 p[j] = update_mv_prob(d[j], p[j]);
1388 for (j = 0; j < ARRAY_SIZE(probs->mv.class0_fr[0]); j++) {
1389 p = probs->mv.class0_fr[i][j];
1390 d = deltas->mv.class0_fr[i][j];
1392 p[0] = update_mv_prob(d[0], p[0]);
1393 p[1] = update_mv_prob(d[1], p[1]);
1394 p[2] = update_mv_prob(d[2], p[2]);
1397 p = probs->mv.fr[i];
1398 d = deltas->mv.fr[i];
1399 for (j = 0; j < ARRAY_SIZE(probs->mv.fr[i]); j++)
1400 p[j] = update_mv_prob(d[j], p[j]);
1402 if (dec_params->flags & V4L2_VP9_FRAME_FLAG_ALLOW_HIGH_PREC_MV) {
1403 p = probs->mv.class0_hp;
1404 d = deltas->mv.class0_hp;
1405 p[i] = update_mv_prob(d[i], p[i]);
1409 p[i] = update_mv_prob(d[i], p[i]);
1414 /* Counterpart to 6.3 compressed_header(), but parsing has been done in userspace. */
1415 void v4l2_vp9_fw_update_probs(struct v4l2_vp9_frame_context *probs,
1416 const struct v4l2_ctrl_vp9_compressed_hdr *deltas,
1417 const struct v4l2_ctrl_vp9_frame *dec_params)
1419 if (deltas->tx_mode == V4L2_VP9_TX_MODE_SELECT)
1420 update_tx_probs(probs, deltas);
1422 update_coef_probs(probs, deltas, dec_params);
1424 update_skip_probs(probs, deltas);
1426 if (dec_params->flags & V4L2_VP9_FRAME_FLAG_KEY_FRAME ||
1427 dec_params->flags & V4L2_VP9_FRAME_FLAG_INTRA_ONLY)
1430 update_inter_mode_probs(probs, deltas);
1432 if (dec_params->interpolation_filter == V4L2_VP9_INTERP_FILTER_SWITCHABLE)
1433 update_interp_filter_probs(probs, deltas);
1435 update_is_inter_probs(probs, deltas);
1437 update_frame_reference_mode_probs(dec_params->reference_mode, probs, deltas);
1439 update_y_mode_probs(probs, deltas);
1441 update_partition_probs(probs, deltas);
1443 update_mv_probs(probs, deltas, dec_params);
1445 EXPORT_SYMBOL_GPL(v4l2_vp9_fw_update_probs);
1447 u8 v4l2_vp9_reset_frame_ctx(const struct v4l2_ctrl_vp9_frame *dec_params,
1448 struct v4l2_vp9_frame_context *frame_context)
1452 u8 fctx_idx = dec_params->frame_context_idx;
1454 if (dec_params->flags & V4L2_VP9_FRAME_FLAG_KEY_FRAME ||
1455 dec_params->flags & V4L2_VP9_FRAME_FLAG_INTRA_ONLY ||
1456 dec_params->flags & V4L2_VP9_FRAME_FLAG_ERROR_RESILIENT) {
1458 * setup_past_independence()
1459 * We do nothing here. Instead of storing default probs in some intermediate
1460 * location and then copying from that location to appropriate contexts
1461 * in save_probs() below, we skip that step and save default probs directly
1462 * to appropriate contexts.
1464 if (dec_params->flags & V4L2_VP9_FRAME_FLAG_KEY_FRAME ||
1465 dec_params->flags & V4L2_VP9_FRAME_FLAG_ERROR_RESILIENT ||
1466 dec_params->reset_frame_context == V4L2_VP9_RESET_FRAME_CTX_ALL)
1467 for (i = 0; i < 4; ++i)
1469 memcpy(&frame_context[i], &v4l2_vp9_default_probs,
1470 sizeof(v4l2_vp9_default_probs));
1471 else if (dec_params->reset_frame_context == V4L2_VP9_RESET_FRAME_CTX_SPEC)
1472 /* save_probs(fctx_idx) */
1473 memcpy(&frame_context[fctx_idx], &v4l2_vp9_default_probs,
1474 sizeof(v4l2_vp9_default_probs));
1480 EXPORT_SYMBOL_GPL(v4l2_vp9_reset_frame_ctx);
1482 /* 8.4.1 Merge prob process */
1483 static u8 merge_prob(u8 pre_prob, u32 ct0, u32 ct1, u16 count_sat, u32 max_update_factor)
1485 u32 den, prob, count, factor;
1490 * prob = 128, count = 0, update_factor = 0
1491 * Round2's argument: pre_prob * 256
1492 * (pre_prob * 256 + 128) >> 8 == pre_prob
1497 prob = clamp(((ct0 << 8) + (den >> 1)) / den, (u32)1, (u32)255);
1498 count = min_t(u32, den, count_sat);
1499 factor = fastdiv(max_update_factor * count, count_sat);
1502 * Round2(pre_prob * (256 - factor) + prob * factor, 8)
1503 * Round2(pre_prob * 256 + (prob - pre_prob) * factor, 8)
1504 * (pre_prob * 256 >> 8) + (((prob - pre_prob) * factor + 128) >> 8)
1506 return pre_prob + (((prob - pre_prob) * factor + 128) >> 8);
1509 static inline u8 noncoef_merge_prob(u8 pre_prob, u32 ct0, u32 ct1)
1511 return merge_prob(pre_prob, ct0, ct1, 20, 128);
1514 /* 8.4.2 Merge probs process */
1516 * merge_probs() is a recursive function in the spec. We avoid recursion in the kernel.
1517 * That said, the "tree" parameter of merge_probs() controls how deep the recursion goes.
1518 * It turns out that in all cases the recursive calls boil down to a short-ish series
1519 * of merge_prob() invocations (note no "s").
1523 * merge_probs(small_token_tree, 2):
1524 * merge_prob(p[1], c[0], c[1] + c[2])
1525 * merge_prob(p[2], c[1], c[2])
1529 * merge_probs(binary_tree, 0) or
1530 * merge_probs(tx_size_8_tree, 0):
1531 * merge_prob(p[0], c[0], c[1])
1535 * merge_probs(inter_mode_tree, 0):
1536 * merge_prob(p[0], c[2], c[1] + c[0] + c[3])
1537 * merge_prob(p[1], c[0], c[1] + c[3])
1538 * merge_prob(p[2], c[1], c[3])
1542 * merge_probs(intra_mode_tree, 0):
1543 * merge_prob(p[0], c[0], c[1] + ... + c[9])
1544 * merge_prob(p[1], c[9], c[1] + ... + c[8])
1545 * merge_prob(p[2], c[1], c[2] + ... + c[8])
1546 * merge_prob(p[3], c[2] + c[4] + c[5], c[3] + c[8] + c[6] + c[7])
1547 * merge_prob(p[4], c[2], c[4] + c[5])
1548 * merge_prob(p[5], c[4], c[5])
1549 * merge_prob(p[6], c[3], c[8] + c[6] + c[7])
1550 * merge_prob(p[7], c[8], c[6] + c[7])
1551 * merge_prob(p[8], c[6], c[7])
1555 * merge_probs(partition_tree, 0) or
1556 * merge_probs(tx_size_32_tree, 0) or
1557 * merge_probs(mv_joint_tree, 0) or
1558 * merge_probs(mv_fr_tree, 0):
1559 * merge_prob(p[0], c[0], c[1] + c[2] + c[3])
1560 * merge_prob(p[1], c[1], c[2] + c[3])
1561 * merge_prob(p[2], c[2], c[3])
1565 * merge_probs(interp_filter_tree, 0) or
1566 * merge_probs(tx_size_16_tree, 0):
1567 * merge_prob(p[0], c[0], c[1] + c[2])
1568 * merge_prob(p[1], c[1], c[2])
1572 * merge_probs(mv_class_tree, 0):
1573 * merge_prob(p[0], c[0], c[1] + ... + c[10])
1574 * merge_prob(p[1], c[1], c[2] + ... + c[10])
1575 * merge_prob(p[2], c[2] + c[3], c[4] + ... + c[10])
1576 * merge_prob(p[3], c[2], c[3])
1577 * merge_prob(p[4], c[4] + c[5], c[6] + ... + c[10])
1578 * merge_prob(p[5], c[4], c[5])
1579 * merge_prob(p[6], c[6], c[7] + ... + c[10])
1580 * merge_prob(p[7], c[7] + c[8], c[9] + c[10])
1581 * merge_prob(p[8], c[7], c[8])
1582 * merge_prob(p[9], c[9], [10])
1585 static inline void merge_probs_variant_a(u8 *p, const u32 *c, u16 count_sat, u32 update_factor)
1587 p[1] = merge_prob(p[1], c[0], c[1] + c[2], count_sat, update_factor);
1588 p[2] = merge_prob(p[2], c[1], c[2], count_sat, update_factor);
1591 static inline void merge_probs_variant_b(u8 *p, const u32 *c, u16 count_sat, u32 update_factor)
1593 p[0] = merge_prob(p[0], c[0], c[1], count_sat, update_factor);
1596 static inline void merge_probs_variant_c(u8 *p, const u32 *c)
1598 p[0] = noncoef_merge_prob(p[0], c[2], c[1] + c[0] + c[3]);
1599 p[1] = noncoef_merge_prob(p[1], c[0], c[1] + c[3]);
1600 p[2] = noncoef_merge_prob(p[2], c[1], c[3]);
1603 static void merge_probs_variant_d(u8 *p, const u32 *c)
1607 sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9];
1609 p[0] = noncoef_merge_prob(p[0], c[0], sum);
1611 p[1] = noncoef_merge_prob(p[1], c[9], sum);
1613 p[2] = noncoef_merge_prob(p[2], c[1], sum);
1614 s2 = c[2] + c[4] + c[5];
1616 p[3] = noncoef_merge_prob(p[3], s2, sum);
1618 p[4] = noncoef_merge_prob(p[4], c[2], s2);
1619 p[5] = noncoef_merge_prob(p[5], c[4], c[5]);
1621 p[6] = noncoef_merge_prob(p[6], c[3], sum);
1623 p[7] = noncoef_merge_prob(p[7], c[8], sum);
1624 p[8] = noncoef_merge_prob(p[8], c[6], c[7]);
1627 static inline void merge_probs_variant_e(u8 *p, const u32 *c)
1629 p[0] = noncoef_merge_prob(p[0], c[0], c[1] + c[2] + c[3]);
1630 p[1] = noncoef_merge_prob(p[1], c[1], c[2] + c[3]);
1631 p[2] = noncoef_merge_prob(p[2], c[2], c[3]);
1634 static inline void merge_probs_variant_f(u8 *p, const u32 *c)
1636 p[0] = noncoef_merge_prob(p[0], c[0], c[1] + c[2]);
1637 p[1] = noncoef_merge_prob(p[1], c[1], c[2]);
1640 static void merge_probs_variant_g(u8 *p, const u32 *c)
1644 sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9] + c[10];
1645 p[0] = noncoef_merge_prob(p[0], c[0], sum);
1647 p[1] = noncoef_merge_prob(p[1], c[1], sum);
1649 p[2] = noncoef_merge_prob(p[2], c[2] + c[3], sum);
1650 p[3] = noncoef_merge_prob(p[3], c[2], c[3]);
1652 p[4] = noncoef_merge_prob(p[4], c[4] + c[5], sum);
1653 p[5] = noncoef_merge_prob(p[5], c[4], c[5]);
1655 p[6] = noncoef_merge_prob(p[6], c[6], sum);
1656 p[7] = noncoef_merge_prob(p[7], c[7] + c[8], c[9] + c[10]);
1657 p[8] = noncoef_merge_prob(p[8], c[7], c[8]);
1658 p[9] = noncoef_merge_prob(p[9], c[9], c[10]);
1661 /* 8.4.3 Coefficient probability adaptation process */
1662 static inline void adapt_probs_variant_a_coef(u8 *p, const u32 *c, u32 update_factor)
1664 merge_probs_variant_a(p, c, 24, update_factor);
1667 static inline void adapt_probs_variant_b_coef(u8 *p, const u32 *c, u32 update_factor)
1669 merge_probs_variant_b(p, c, 24, update_factor);
1672 static void _adapt_coeff(unsigned int i, unsigned int j, unsigned int k,
1673 struct v4l2_vp9_frame_context *probs,
1674 const struct v4l2_vp9_frame_symbol_counts *counts,
1679 for (l = 0; l < ARRAY_SIZE(probs->coef[0][0][0]); l++) {
1680 for (m = 0; m < BAND_6(l); m++) {
1681 u8 *p = probs->coef[i][j][k][l][m];
1682 const u32 counts_more_coefs[2] = {
1683 *counts->eob[i][j][k][l][m][1],
1684 *counts->eob[i][j][k][l][m][0] - *counts->eob[i][j][k][l][m][1],
1687 adapt_probs_variant_a_coef(p, *counts->coeff[i][j][k][l][m], uf);
1688 adapt_probs_variant_b_coef(p, counts_more_coefs, uf);
1693 static void _adapt_coef_probs(struct v4l2_vp9_frame_context *probs,
1694 const struct v4l2_vp9_frame_symbol_counts *counts,
1697 unsigned int i, j, k;
1699 for (i = 0; i < ARRAY_SIZE(probs->coef); i++)
1700 for (j = 0; j < ARRAY_SIZE(probs->coef[0]); j++)
1701 for (k = 0; k < ARRAY_SIZE(probs->coef[0][0]); k++)
1702 _adapt_coeff(i, j, k, probs, counts, uf);
1705 void v4l2_vp9_adapt_coef_probs(struct v4l2_vp9_frame_context *probs,
1706 struct v4l2_vp9_frame_symbol_counts *counts,
1708 bool frame_is_intra)
1710 if (frame_is_intra) {
1711 _adapt_coef_probs(probs, counts, 112);
1714 _adapt_coef_probs(probs, counts, 128);
1716 _adapt_coef_probs(probs, counts, 112);
1719 EXPORT_SYMBOL_GPL(v4l2_vp9_adapt_coef_probs);
1721 /* 8.4.4 Non coefficient probability adaptation process, adapt_probs() */
1722 static inline void adapt_probs_variant_b(u8 *p, const u32 *c)
1724 merge_probs_variant_b(p, c, 20, 128);
1727 static inline void adapt_probs_variant_c(u8 *p, const u32 *c)
1729 merge_probs_variant_c(p, c);
1732 static inline void adapt_probs_variant_d(u8 *p, const u32 *c)
1734 merge_probs_variant_d(p, c);
1737 static inline void adapt_probs_variant_e(u8 *p, const u32 *c)
1739 merge_probs_variant_e(p, c);
1742 static inline void adapt_probs_variant_f(u8 *p, const u32 *c)
1744 merge_probs_variant_f(p, c);
1747 static inline void adapt_probs_variant_g(u8 *p, const u32 *c)
1749 merge_probs_variant_g(p, c);
1752 /* 8.4.4 Non coefficient probability adaptation process, adapt_prob() */
1753 static inline u8 adapt_prob(u8 prob, const u32 counts[2])
1755 return noncoef_merge_prob(prob, counts[0], counts[1]);
1758 /* 8.4.4 Non coefficient probability adaptation process */
1759 void v4l2_vp9_adapt_noncoef_probs(struct v4l2_vp9_frame_context *probs,
1760 struct v4l2_vp9_frame_symbol_counts *counts,
1761 u8 reference_mode, u8 interpolation_filter, u8 tx_mode,
1766 for (i = 0; i < ARRAY_SIZE(probs->is_inter); i++)
1767 probs->is_inter[i] = adapt_prob(probs->is_inter[i], (*counts->intra_inter)[i]);
1769 for (i = 0; i < ARRAY_SIZE(probs->comp_mode); i++)
1770 probs->comp_mode[i] = adapt_prob(probs->comp_mode[i], (*counts->comp)[i]);
1772 for (i = 0; i < ARRAY_SIZE(probs->comp_ref); i++)
1773 probs->comp_ref[i] = adapt_prob(probs->comp_ref[i], (*counts->comp_ref)[i]);
1775 if (reference_mode != V4L2_VP9_REFERENCE_MODE_COMPOUND_REFERENCE)
1776 for (i = 0; i < ARRAY_SIZE(probs->single_ref); i++)
1777 for (j = 0; j < ARRAY_SIZE(probs->single_ref[0]); j++)
1778 probs->single_ref[i][j] = adapt_prob(probs->single_ref[i][j],
1779 (*counts->single_ref)[i][j]);
1781 for (i = 0; i < ARRAY_SIZE(probs->inter_mode); i++)
1782 adapt_probs_variant_c(probs->inter_mode[i], (*counts->mv_mode)[i]);
1784 for (i = 0; i < ARRAY_SIZE(probs->y_mode); i++)
1785 adapt_probs_variant_d(probs->y_mode[i], (*counts->y_mode)[i]);
1787 for (i = 0; i < ARRAY_SIZE(probs->uv_mode); i++)
1788 adapt_probs_variant_d(probs->uv_mode[i], (*counts->uv_mode)[i]);
1790 for (i = 0; i < ARRAY_SIZE(probs->partition); i++)
1791 adapt_probs_variant_e(probs->partition[i], (*counts->partition)[i]);
1793 for (i = 0; i < ARRAY_SIZE(probs->skip); i++)
1794 probs->skip[i] = adapt_prob(probs->skip[i], (*counts->skip)[i]);
1796 if (interpolation_filter == V4L2_VP9_INTERP_FILTER_SWITCHABLE)
1797 for (i = 0; i < ARRAY_SIZE(probs->interp_filter); i++)
1798 adapt_probs_variant_f(probs->interp_filter[i], (*counts->filter)[i]);
1800 if (tx_mode == V4L2_VP9_TX_MODE_SELECT)
1801 for (i = 0; i < ARRAY_SIZE(probs->tx8); i++) {
1802 adapt_probs_variant_b(probs->tx8[i], (*counts->tx8p)[i]);
1803 adapt_probs_variant_f(probs->tx16[i], (*counts->tx16p)[i]);
1804 adapt_probs_variant_e(probs->tx32[i], (*counts->tx32p)[i]);
1807 adapt_probs_variant_e(probs->mv.joint, *counts->mv_joint);
1809 for (i = 0; i < ARRAY_SIZE(probs->mv.sign); i++) {
1810 probs->mv.sign[i] = adapt_prob(probs->mv.sign[i], (*counts->sign)[i]);
1812 adapt_probs_variant_g(probs->mv.classes[i], (*counts->classes)[i]);
1814 probs->mv.class0_bit[i] = adapt_prob(probs->mv.class0_bit[i], (*counts->class0)[i]);
1816 for (j = 0; j < ARRAY_SIZE(probs->mv.bits[0]); j++)
1817 probs->mv.bits[i][j] = adapt_prob(probs->mv.bits[i][j],
1818 (*counts->bits)[i][j]);
1820 for (j = 0; j < ARRAY_SIZE(probs->mv.class0_fr[0]); j++)
1821 adapt_probs_variant_e(probs->mv.class0_fr[i][j],
1822 (*counts->class0_fp)[i][j]);
1824 adapt_probs_variant_e(probs->mv.fr[i], (*counts->fp)[i]);
1826 if (!(flags & V4L2_VP9_FRAME_FLAG_ALLOW_HIGH_PREC_MV))
1829 probs->mv.class0_hp[i] = adapt_prob(probs->mv.class0_hp[i],
1830 (*counts->class0_hp)[i]);
1832 probs->mv.hp[i] = adapt_prob(probs->mv.hp[i], (*counts->hp)[i]);
1835 EXPORT_SYMBOL_GPL(v4l2_vp9_adapt_noncoef_probs);
1838 v4l2_vp9_seg_feat_enabled(const u8 *feature_enabled,
1839 unsigned int feature,
1842 u8 mask = V4L2_VP9_SEGMENT_FEATURE_ENABLED(feature);
1844 return !!(feature_enabled[segid] & mask);
1846 EXPORT_SYMBOL_GPL(v4l2_vp9_seg_feat_enabled);
1848 MODULE_LICENSE("GPL");
1849 MODULE_DESCRIPTION("V4L2 VP9 Helpers");