OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. | 2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 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 | 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 | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
11 #include <stdlib.h> | 11 #include <stdlib.h> |
12 #include "vpx_config.h" | 12 #include "vpx_config.h" |
13 #include "vp9/common/vp9_loopfilter.h" | 13 #include "vp9/common/vp9_loopfilter.h" |
14 #include "vp9/common/vp9_onyxc_int.h" | 14 #include "vp9/common/vp9_onyxc_int.h" |
15 | 15 |
16 typedef unsigned char uc; | 16 static __inline int8_t signed_char_clamp(int t) { |
17 | |
18 static __inline signed char signed_char_clamp(int t) { | |
19 t = (t < -128 ? -128 : t); | 17 t = (t < -128 ? -128 : t); |
20 t = (t > 127 ? 127 : t); | 18 t = (t > 127 ? 127 : t); |
21 return (signed char) t; | 19 return (int8_t) t; |
22 } | 20 } |
23 | 21 |
24 | 22 |
25 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */ | 23 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */ |
26 static __inline signed char filter_mask(uc limit, uc blimit, | 24 static __inline int8_t filter_mask(uint8_t limit, uint8_t blimit, |
27 uc p3, uc p2, uc p1, uc p0, | 25 uint8_t p3, uint8_t p2, |
28 uc q0, uc q1, uc q2, uc q3) { | 26 uint8_t p1, uint8_t p0, |
29 signed char mask = 0; | 27 uint8_t q0, uint8_t q1, |
| 28 uint8_t q2, uint8_t q3) { |
| 29 int8_t mask = 0; |
30 mask |= (abs(p3 - p2) > limit) * -1; | 30 mask |= (abs(p3 - p2) > limit) * -1; |
31 mask |= (abs(p2 - p1) > limit) * -1; | 31 mask |= (abs(p2 - p1) > limit) * -1; |
32 mask |= (abs(p1 - p0) > limit) * -1; | 32 mask |= (abs(p1 - p0) > limit) * -1; |
33 mask |= (abs(q1 - q0) > limit) * -1; | 33 mask |= (abs(q1 - q0) > limit) * -1; |
34 mask |= (abs(q2 - q1) > limit) * -1; | 34 mask |= (abs(q2 - q1) > limit) * -1; |
35 mask |= (abs(q3 - q2) > limit) * -1; | 35 mask |= (abs(q3 - q2) > limit) * -1; |
36 mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit) * -1; | 36 mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 > blimit) * -1; |
37 mask = ~mask; | 37 mask = ~mask; |
38 return mask; | 38 return mask; |
39 } | 39 } |
40 | 40 |
41 /* is there high variance internal edge ( 11111111 yes, 00000000 no) */ | 41 /* is there high variance internal edge ( 11111111 yes, 00000000 no) */ |
42 static __inline signed char hevmask(uc thresh, uc p1, uc p0, uc q0, uc q1) { | 42 static __inline int8_t hevmask(uint8_t thresh, uint8_t p1, uint8_t p0, |
43 signed char hev = 0; | 43 uint8_t q0, uint8_t q1) { |
| 44 int8_t hev = 0; |
44 hev |= (abs(p1 - p0) > thresh) * -1; | 45 hev |= (abs(p1 - p0) > thresh) * -1; |
45 hev |= (abs(q1 - q0) > thresh) * -1; | 46 hev |= (abs(q1 - q0) > thresh) * -1; |
46 return hev; | 47 return hev; |
47 } | 48 } |
48 | 49 |
49 static __inline void filter(signed char mask, uc hev, uc *op1, | 50 static __inline void filter(int8_t mask, uint8_t hev, uint8_t *op1, |
50 uc *op0, uc *oq0, uc *oq1) | 51 uint8_t *op0, uint8_t *oq0, uint8_t *oq1) { |
| 52 int8_t ps0, qs0; |
| 53 int8_t ps1, qs1; |
| 54 int8_t filter, Filter1, Filter2; |
| 55 int8_t u; |
51 | 56 |
52 { | 57 ps1 = (int8_t) *op1 ^ 0x80; |
53 signed char ps0, qs0; | 58 ps0 = (int8_t) *op0 ^ 0x80; |
54 signed char ps1, qs1; | 59 qs0 = (int8_t) *oq0 ^ 0x80; |
55 signed char filter, Filter1, Filter2; | 60 qs1 = (int8_t) *oq1 ^ 0x80; |
56 signed char u; | |
57 | |
58 ps1 = (signed char) * op1 ^ 0x80; | |
59 ps0 = (signed char) * op0 ^ 0x80; | |
60 qs0 = (signed char) * oq0 ^ 0x80; | |
61 qs1 = (signed char) * oq1 ^ 0x80; | |
62 | 61 |
63 /* add outer taps if we have high edge variance */ | 62 /* add outer taps if we have high edge variance */ |
64 filter = signed_char_clamp(ps1 - qs1); | 63 filter = signed_char_clamp(ps1 - qs1); |
65 filter &= hev; | 64 filter &= hev; |
66 | 65 |
67 /* inner taps */ | 66 /* inner taps */ |
68 filter = signed_char_clamp(filter + 3 * (qs0 - ps0)); | 67 filter = signed_char_clamp(filter + 3 * (qs0 - ps0)); |
69 filter &= mask; | 68 filter &= mask; |
70 | 69 |
71 /* save bottom 3 bits so that we round one side +4 and the other +3 | 70 /* save bottom 3 bits so that we round one side +4 and the other +3 |
(...skipping 12 matching lines...) Expand all Loading... |
84 | 83 |
85 /* outer tap adjustments */ | 84 /* outer tap adjustments */ |
86 filter += 1; | 85 filter += 1; |
87 filter >>= 1; | 86 filter >>= 1; |
88 filter &= ~hev; | 87 filter &= ~hev; |
89 | 88 |
90 u = signed_char_clamp(qs1 - filter); | 89 u = signed_char_clamp(qs1 - filter); |
91 *oq1 = u ^ 0x80; | 90 *oq1 = u ^ 0x80; |
92 u = signed_char_clamp(ps1 + filter); | 91 u = signed_char_clamp(ps1 + filter); |
93 *op1 = u ^ 0x80; | 92 *op1 = u ^ 0x80; |
94 | |
95 } | 93 } |
96 | 94 |
97 void vp9_loop_filter_horizontal_edge_c | 95 void vp9_loop_filter_horizontal_edge_c(uint8_t *s, |
98 ( | 96 int p, /* pitch */ |
99 unsigned char *s, | 97 const unsigned char *blimit, |
100 int p, /* pitch */ | 98 const unsigned char *limit, |
101 const unsigned char *blimit, | 99 const unsigned char *thresh, |
102 const unsigned char *limit, | 100 int count) { |
103 const unsigned char *thresh, | 101 int hev = 0; /* high edge variance */ |
104 int count | 102 int8_t mask = 0; |
105 ) { | |
106 int hev = 0; /* high edge variance */ | |
107 signed char mask = 0; | |
108 int i = 0; | 103 int i = 0; |
109 | 104 |
110 /* loop filter designed to work using chars so that we can make maximum use | 105 /* loop filter designed to work using chars so that we can make maximum use |
111 * of 8 bit simd instructions. | 106 * of 8 bit simd instructions. |
112 */ | 107 */ |
113 do { | 108 do { |
114 mask = filter_mask(limit[0], blimit[0], | 109 mask = filter_mask(limit[0], blimit[0], |
115 s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], | 110 s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], |
116 s[0 * p], s[1 * p], s[2 * p], s[3 * p]); | 111 s[0 * p], s[1 * p], s[2 * p], s[3 * p]); |
117 | 112 |
118 hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]); | 113 hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]); |
119 | 114 |
120 filter(mask, hev, s - 2 * p, s - 1 * p, s, s + 1 * p); | 115 filter(mask, hev, s - 2 * p, s - 1 * p, s, s + 1 * p); |
121 | 116 |
122 ++s; | 117 ++s; |
123 } while (++i < count * 8); | 118 } while (++i < count * 8); |
124 } | 119 } |
125 | 120 |
126 void vp9_loop_filter_vertical_edge_c(unsigned char *s, | 121 void vp9_loop_filter_vertical_edge_c(uint8_t *s, |
127 int p, | 122 int p, |
128 const unsigned char *blimit, | 123 const unsigned char *blimit, |
129 const unsigned char *limit, | 124 const unsigned char *limit, |
130 const unsigned char *thresh, | 125 const unsigned char *thresh, |
131 int count) { | 126 int count) { |
132 int hev = 0; /* high edge variance */ | 127 int hev = 0; /* high edge variance */ |
133 signed char mask = 0; | 128 int8_t mask = 0; |
134 int i = 0; | 129 int i = 0; |
135 | 130 |
136 /* loop filter designed to work using chars so that we can make maximum use | 131 /* loop filter designed to work using chars so that we can make maximum use |
137 * of 8 bit simd instructions. | 132 * of 8 bit simd instructions. |
138 */ | 133 */ |
139 do { | 134 do { |
140 mask = filter_mask(limit[0], blimit[0], | 135 mask = filter_mask(limit[0], blimit[0], |
141 s[-4], s[-3], s[-2], s[-1], | 136 s[-4], s[-3], s[-2], s[-1], |
142 s[0], s[1], s[2], s[3]); | 137 s[0], s[1], s[2], s[3]); |
143 | 138 |
144 hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]); | 139 hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]); |
145 | 140 |
146 filter(mask, hev, s - 2, s - 1, s, s + 1); | 141 filter(mask, hev, s - 2, s - 1, s, s + 1); |
147 | 142 |
148 s += p; | 143 s += p; |
149 } while (++i < count * 8); | 144 } while (++i < count * 8); |
150 } | 145 } |
151 static __inline signed char flatmask(uc thresh, | 146 static __inline signed char flatmask(uint8_t thresh, |
152 uc p4, uc p3, uc p2, uc p1, uc p0, | 147 uint8_t p4, uint8_t p3, uint8_t p2, |
153 uc q0, uc q1, uc q2, uc q3, uc q4) { | 148 uint8_t p1, uint8_t p0, |
154 signed char flat = 0; | 149 uint8_t q0, uint8_t q1, uint8_t q2, |
155 flat |= (abs(p1 - p0) > 1) * -1; | 150 uint8_t q3, uint8_t q4) { |
156 flat |= (abs(q1 - q0) > 1) * -1; | 151 int8_t flat = 0; |
157 flat |= (abs(p0 - p2) > 1) * -1; | 152 flat |= (abs(p1 - p0) > thresh) * -1; |
158 flat |= (abs(q0 - q2) > 1) * -1; | 153 flat |= (abs(q1 - q0) > thresh) * -1; |
159 flat |= (abs(p3 - p0) > 1) * -1; | 154 flat |= (abs(p0 - p2) > thresh) * -1; |
160 flat |= (abs(q3 - q0) > 1) * -1; | 155 flat |= (abs(q0 - q2) > thresh) * -1; |
161 flat |= (abs(p4 - p0) > 1) * -1; | 156 flat |= (abs(p3 - p0) > thresh) * -1; |
162 flat |= (abs(q4 - q0) > 1) * -1; | 157 flat |= (abs(q3 - q0) > thresh) * -1; |
| 158 flat |= (abs(p4 - p0) > thresh) * -1; |
| 159 flat |= (abs(q4 - q0) > thresh) * -1; |
163 flat = ~flat; | 160 flat = ~flat; |
164 return flat; | 161 return flat; |
165 } | 162 } |
166 | 163 |
167 static __inline void mbfilter(signed char mask, uc hev, uc flat, | 164 static __inline void mbfilter(int8_t mask, uint8_t hev, uint8_t flat, |
168 uc *op4, uc *op3, uc *op2, uc *op1, uc *op0, | 165 uint8_t *op4, uint8_t *op3, uint8_t *op2, |
169 uc *oq0, uc *oq1, uc *oq2, uc *oq3, uc *oq4) { | 166 uint8_t *op1, uint8_t *op0, |
| 167 uint8_t *oq0, uint8_t *oq1, uint8_t *oq2, |
| 168 uint8_t *oq3, uint8_t *oq4) { |
170 /* use a 7 tap filter [1, 1, 1, 2, 1, 1, 1] for flat line */ | 169 /* use a 7 tap filter [1, 1, 1, 2, 1, 1, 1] for flat line */ |
171 if (flat && mask) { | 170 if (flat && mask) { |
172 unsigned char p0, q0; | 171 uint8_t p0, q0; |
173 unsigned char p1, q1; | 172 uint8_t p1, q1; |
174 unsigned char p2, q2; | 173 uint8_t p2, q2; |
175 unsigned char p3, q3; | 174 uint8_t p3, q3; |
176 unsigned char p4, q4; | 175 uint8_t p4, q4; |
177 | 176 |
178 p4 = *op4; | 177 p4 = *op4; |
179 p3 = *op3; | 178 p3 = *op3; |
180 p2 = *op2; | 179 p2 = *op2; |
181 p1 = *op1; | 180 p1 = *op1; |
182 p0 = *op0; | 181 p0 = *op0; |
183 q0 = *oq0; | 182 q0 = *oq0; |
184 q1 = *oq1; | 183 q1 = *oq1; |
185 q2 = *oq2; | 184 q2 = *oq2; |
186 q3 = *oq3; | 185 q3 = *oq3; |
187 q4 = *oq4; | 186 q4 = *oq4; |
188 | 187 |
189 *op2 = (p4 + p4 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3; | 188 *op2 = (p4 + p4 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3; |
190 *op1 = (p4 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3; | 189 *op1 = (p4 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3; |
191 *op0 = (p3 + p2 + p1 + p0 + p0 + q0 + q1 + q2 + 4) >> 3; | 190 *op0 = (p3 + p2 + p1 + p0 + p0 + q0 + q1 + q2 + 4) >> 3; |
192 *oq0 = (p2 + p1 + p0 + q0 + q0 + q1 + q2 + q3 + 4) >> 3; | 191 *oq0 = (p2 + p1 + p0 + q0 + q0 + q1 + q2 + q3 + 4) >> 3; |
193 *oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q4 + 4) >> 3; | 192 *oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q4 + 4) >> 3; |
194 *oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q4 + q4 + 4) >> 3; | 193 *oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q4 + q4 + 4) >> 3; |
195 } else { | 194 } else { |
196 signed char ps0, qs0; | 195 int8_t ps0, qs0; |
197 signed char ps1, qs1; | 196 int8_t ps1, qs1; |
198 signed char filter, Filter1, Filter2; | 197 int8_t filter, Filter1, Filter2; |
199 signed char u; | 198 int8_t u; |
200 | 199 |
201 ps1 = (signed char) * op1 ^ 0x80; | 200 ps1 = (int8_t) *op1 ^ 0x80; |
202 ps0 = (signed char) * op0 ^ 0x80; | 201 ps0 = (int8_t) *op0 ^ 0x80; |
203 qs0 = (signed char) * oq0 ^ 0x80; | 202 qs0 = (int8_t) *oq0 ^ 0x80; |
204 qs1 = (signed char) * oq1 ^ 0x80; | 203 qs1 = (int8_t) *oq1 ^ 0x80; |
205 | 204 |
206 /* add outer taps if we have high edge variance */ | 205 /* add outer taps if we have high edge variance */ |
207 filter = signed_char_clamp(ps1 - qs1); | 206 filter = signed_char_clamp(ps1 - qs1); |
208 filter &= hev; | 207 filter &= hev; |
209 | 208 |
210 /* inner taps */ | 209 /* inner taps */ |
211 filter = signed_char_clamp(filter + 3 * (qs0 - ps0)); | 210 filter = signed_char_clamp(filter + 3 * (qs0 - ps0)); |
212 filter &= mask; | 211 filter &= mask; |
213 | 212 |
214 Filter1 = signed_char_clamp(filter + 4); | 213 Filter1 = signed_char_clamp(filter + 4); |
(...skipping 11 matching lines...) Expand all Loading... |
226 filter += 1; | 225 filter += 1; |
227 filter >>= 1; | 226 filter >>= 1; |
228 filter &= ~hev; | 227 filter &= ~hev; |
229 | 228 |
230 u = signed_char_clamp(qs1 - filter); | 229 u = signed_char_clamp(qs1 - filter); |
231 *oq1 = u ^ 0x80; | 230 *oq1 = u ^ 0x80; |
232 u = signed_char_clamp(ps1 + filter); | 231 u = signed_char_clamp(ps1 + filter); |
233 *op1 = u ^ 0x80; | 232 *op1 = u ^ 0x80; |
234 } | 233 } |
235 } | 234 } |
236 void vp9_mbloop_filter_horizontal_edge_c | 235 |
237 ( | 236 void vp9_mbloop_filter_horizontal_edge_c(uint8_t *s, |
238 unsigned char *s, | 237 int p, |
239 int p, | 238 const unsigned char *blimit, |
240 const unsigned char *blimit, | 239 const unsigned char *limit, |
241 const unsigned char *limit, | 240 const unsigned char *thresh, |
242 const unsigned char *thresh, | 241 int count) { |
243 int count | 242 int8_t hev = 0; /* high edge variance */ |
244 ) { | 243 int8_t mask = 0; |
245 signed char hev = 0; /* high edge variance */ | 244 int8_t flat = 0; |
246 signed char mask = 0; | |
247 signed char flat = 0; | |
248 int i = 0; | 245 int i = 0; |
249 | 246 |
250 /* loop filter designed to work using chars so that we can make maximum use | 247 /* loop filter designed to work using chars so that we can make maximum use |
251 * of 8 bit simd instructions. | 248 * of 8 bit simd instructions. |
252 */ | 249 */ |
253 do { | 250 do { |
254 | |
255 mask = filter_mask(limit[0], blimit[0], | 251 mask = filter_mask(limit[0], blimit[0], |
256 s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], | 252 s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], |
257 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]); | 253 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]); |
258 | 254 |
259 hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]); | 255 hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]); |
260 | 256 |
261 flat = flatmask(thresh[0], | 257 flat = flatmask(1, |
262 s[-5 * p], s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], | 258 s[-5 * p], s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], |
263 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p], s[ 4 * p]); | 259 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p], s[ 4 * p]); |
264 mbfilter(mask, hev, flat, | 260 mbfilter(mask, hev, flat, |
265 s - 5 * p, s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p, | 261 s - 5 * p, s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p, |
266 s, s + 1 * p, s + 2 * p, s + 3 * p, s + 4 * p); | 262 s, s + 1 * p, s + 2 * p, s + 3 * p, s + 4 * p); |
267 | 263 |
268 ++s; | 264 ++s; |
269 } while (++i < count * 8); | 265 } while (++i < count * 8); |
270 | 266 |
271 } | 267 } |
272 void vp9_mbloop_filter_vertical_edge_c | 268 |
273 ( | 269 void vp9_mbloop_filter_vertical_edge_c(uint8_t *s, |
274 unsigned char *s, | 270 int p, |
275 int p, | 271 const unsigned char *blimit, |
276 const unsigned char *blimit, | 272 const unsigned char *limit, |
277 const unsigned char *limit, | 273 const unsigned char *thresh, |
278 const unsigned char *thresh, | 274 int count) { |
279 int count | 275 int8_t hev = 0; /* high edge variance */ |
280 ) { | 276 int8_t mask = 0; |
281 signed char hev = 0; /* high edge variance */ | 277 int8_t flat = 0; |
282 signed char mask = 0; | |
283 signed char flat = 0; | |
284 int i = 0; | 278 int i = 0; |
285 | 279 |
286 do { | 280 do { |
287 | |
288 mask = filter_mask(limit[0], blimit[0], | 281 mask = filter_mask(limit[0], blimit[0], |
289 s[-4], s[-3], s[-2], s[-1], | 282 s[-4], s[-3], s[-2], s[-1], |
290 s[0], s[1], s[2], s[3]); | 283 s[0], s[1], s[2], s[3]); |
291 | 284 |
292 hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]); | 285 hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]); |
293 flat = flatmask(thresh[0], | 286 flat = flatmask(1, |
294 s[-5], s[-4], s[-3], s[-2], s[-1], | 287 s[-5], s[-4], s[-3], s[-2], s[-1], |
295 s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]); | 288 s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]); |
296 mbfilter(mask, hev, flat, | 289 mbfilter(mask, hev, flat, |
297 s - 5, s - 4, s - 3, s - 2, s - 1, | 290 s - 5, s - 4, s - 3, s - 2, s - 1, |
298 s, s + 1, s + 2, s + 3, s + 4); | 291 s, s + 1, s + 2, s + 3, s + 4); |
299 s += p; | 292 s += p; |
300 } while (++i < count * 8); | 293 } while (++i < count * 8); |
301 | 294 |
302 } | 295 } |
303 | 296 |
304 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */ | 297 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */ |
305 static __inline signed char simple_filter_mask(uc blimit, | 298 static __inline int8_t simple_filter_mask(uint8_t blimit, |
306 uc p1, uc p0, | 299 uint8_t p1, uint8_t p0, |
307 uc q0, uc q1) { | 300 uint8_t q0, uint8_t q1) { |
308 /* Why does this cause problems for win32? | 301 /* Why does this cause problems for win32? |
309 * error C2143: syntax error : missing ';' before 'type' | 302 * error C2143: syntax error : missing ';' before 'type' |
310 * (void) limit; | 303 * (void) limit; |
311 */ | 304 */ |
312 signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 <= blimit) * -1; | 305 int8_t mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2 <= blimit) * -1; |
313 return mask; | 306 return mask; |
314 } | 307 } |
315 | 308 |
316 static __inline void simple_filter(signed char mask, | 309 static __inline void simple_filter(int8_t mask, |
317 uc *op1, uc *op0, | 310 uint8_t *op1, uint8_t *op0, |
318 uc *oq0, uc *oq1) { | 311 uint8_t *oq0, uint8_t *oq1) { |
319 signed char filter, Filter1, Filter2; | 312 int8_t filter, Filter1, Filter2; |
320 signed char p1 = (signed char) * op1 ^ 0x80; | 313 int8_t p1 = (int8_t) *op1 ^ 0x80; |
321 signed char p0 = (signed char) * op0 ^ 0x80; | 314 int8_t p0 = (int8_t) *op0 ^ 0x80; |
322 signed char q0 = (signed char) * oq0 ^ 0x80; | 315 int8_t q0 = (int8_t) *oq0 ^ 0x80; |
323 signed char q1 = (signed char) * oq1 ^ 0x80; | 316 int8_t q1 = (int8_t) *oq1 ^ 0x80; |
324 signed char u; | 317 int8_t u; |
325 | 318 |
326 filter = signed_char_clamp(p1 - q1); | 319 filter = signed_char_clamp(p1 - q1); |
327 filter = signed_char_clamp(filter + 3 * (q0 - p0)); | 320 filter = signed_char_clamp(filter + 3 * (q0 - p0)); |
328 filter &= mask; | 321 filter &= mask; |
329 | 322 |
330 /* save bottom 3 bits so that we round one side +4 and the other +3 */ | 323 /* save bottom 3 bits so that we round one side +4 and the other +3 */ |
331 Filter1 = signed_char_clamp(filter + 4); | 324 Filter1 = signed_char_clamp(filter + 4); |
332 Filter1 >>= 3; | 325 Filter1 >>= 3; |
333 u = signed_char_clamp(q0 - Filter1); | 326 u = signed_char_clamp(q0 - Filter1); |
334 *oq0 = u ^ 0x80; | 327 *oq0 = u ^ 0x80; |
335 | 328 |
336 Filter2 = signed_char_clamp(filter + 3); | 329 Filter2 = signed_char_clamp(filter + 3); |
337 Filter2 >>= 3; | 330 Filter2 >>= 3; |
338 u = signed_char_clamp(p0 + Filter2); | 331 u = signed_char_clamp(p0 + Filter2); |
339 *op0 = u ^ 0x80; | 332 *op0 = u ^ 0x80; |
340 } | 333 } |
341 | 334 |
342 void vp9_loop_filter_simple_horizontal_edge_c | 335 void vp9_loop_filter_simple_horizontal_edge_c(uint8_t *s, |
343 ( | 336 int p, |
344 unsigned char *s, | 337 const unsigned char *blimit) { |
345 int p, | 338 int8_t mask = 0; |
346 const unsigned char *blimit | |
347 ) { | |
348 signed char mask = 0; | |
349 int i = 0; | 339 int i = 0; |
350 | 340 |
351 do { | 341 do { |
352 mask = simple_filter_mask(blimit[0], | 342 mask = simple_filter_mask(blimit[0], |
353 s[-2 * p], s[-1 * p], | 343 s[-2 * p], s[-1 * p], |
354 s[0 * p], s[1 * p]); | 344 s[0 * p], s[1 * p]); |
355 simple_filter(mask, | 345 simple_filter(mask, |
356 s - 2 * p, s - 1 * p, | 346 s - 2 * p, s - 1 * p, |
357 s, s + 1 * p); | 347 s, s + 1 * p); |
358 ++s; | 348 ++s; |
359 } while (++i < 16); | 349 } while (++i < 16); |
360 } | 350 } |
361 | 351 |
362 void vp9_loop_filter_simple_vertical_edge_c | 352 void vp9_loop_filter_simple_vertical_edge_c(uint8_t *s, |
363 ( | 353 int p, |
364 unsigned char *s, | 354 const unsigned char *blimit) { |
365 int p, | 355 int8_t mask = 0; |
366 const unsigned char *blimit | |
367 ) { | |
368 signed char mask = 0; | |
369 int i = 0; | 356 int i = 0; |
370 | 357 |
371 do { | 358 do { |
372 mask = simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]); | 359 mask = simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]); |
373 simple_filter(mask, s - 2, s - 1, s, s + 1); | 360 simple_filter(mask, s - 2, s - 1, s, s + 1); |
374 s += p; | 361 s += p; |
375 } while (++i < 16); | 362 } while (++i < 16); |
376 | |
377 } | 363 } |
378 | 364 |
379 /* Vertical MB Filtering */ | 365 /* Vertical MB Filtering */ |
380 void vp9_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr, | 366 void vp9_loop_filter_mbv_c(uint8_t *y_ptr, uint8_t *u_ptr, |
381 unsigned char *v_ptr, int y_stride, int uv_stride, | 367 uint8_t *v_ptr, int y_stride, int uv_stride, |
382 struct loop_filter_info *lfi) { | 368 struct loop_filter_info *lfi) { |
383 vp9_mbloop_filter_vertical_edge_c(y_ptr, y_stride, | 369 vp9_mbloop_filter_vertical_edge_c(y_ptr, y_stride, |
384 lfi->mblim, lfi->lim, lfi->hev_thr, 2); | 370 lfi->mblim, lfi->lim, lfi->hev_thr, 2); |
385 | 371 |
386 if (u_ptr) | 372 if (u_ptr) |
387 vp9_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, | 373 vp9_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, |
388 lfi->mblim, lfi->lim, lfi->hev_thr, 1); | 374 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
389 | 375 |
390 if (v_ptr) | 376 if (v_ptr) |
391 vp9_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, | 377 vp9_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, |
392 lfi->mblim, lfi->lim, lfi->hev_thr, 1); | 378 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
393 } | 379 } |
394 | 380 |
395 /* Vertical B Filtering */ | 381 /* Vertical B Filtering */ |
396 void vp9_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr, | 382 void vp9_loop_filter_bv_c(uint8_t*y_ptr, uint8_t *u_ptr, |
397 unsigned char *v_ptr, int y_stride, int uv_stride, | 383 uint8_t *v_ptr, int y_stride, int uv_stride, |
398 struct loop_filter_info *lfi) { | 384 struct loop_filter_info *lfi) { |
399 vp9_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, | 385 vp9_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, |
400 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 386 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
401 vp9_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, | 387 vp9_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, |
402 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 388 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
403 vp9_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, | 389 vp9_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, |
404 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 390 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
405 | 391 |
406 if (u_ptr) | 392 if (u_ptr) |
407 vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, | 393 vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, |
408 lfi->blim, lfi->lim, lfi->hev_thr, 1); | 394 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
409 | 395 |
410 if (v_ptr) | 396 if (v_ptr) |
411 vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, | 397 vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, |
412 lfi->blim, lfi->lim, lfi->hev_thr, 1); | 398 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
413 } | 399 } |
414 | 400 |
415 /* Horizontal MB filtering */ | 401 /* Horizontal MB filtering */ |
416 void vp9_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr, | 402 void vp9_loop_filter_mbh_c(uint8_t *y_ptr, uint8_t *u_ptr, |
417 unsigned char *v_ptr, int y_stride, int uv_stride, | 403 uint8_t *v_ptr, int y_stride, int uv_stride, |
418 struct loop_filter_info *lfi) { | 404 struct loop_filter_info *lfi) { |
419 vp9_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, | 405 vp9_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, |
420 lfi->mblim, lfi->lim, lfi->hev_thr, 2); | 406 lfi->mblim, lfi->lim, lfi->hev_thr, 2); |
421 | 407 |
422 if (u_ptr) | 408 if (u_ptr) |
423 vp9_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, | 409 vp9_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, |
424 lfi->mblim, lfi->lim, lfi->hev_thr, 1); | 410 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
425 | 411 |
426 if (v_ptr) | 412 if (v_ptr) |
427 vp9_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, | 413 vp9_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, |
428 lfi->mblim, lfi->lim, lfi->hev_thr, 1); | 414 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
429 } | 415 } |
430 | 416 |
431 /* Horizontal B Filtering */ | 417 /* Horizontal B Filtering */ |
432 void vp9_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr, | 418 void vp9_loop_filter_bh_c(uint8_t *y_ptr, uint8_t *u_ptr, |
433 unsigned char *v_ptr, int y_stride, int uv_stride, | 419 uint8_t *v_ptr, int y_stride, int uv_stride, |
434 struct loop_filter_info *lfi) { | 420 struct loop_filter_info *lfi) { |
435 vp9_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, | 421 vp9_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, |
436 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 422 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
437 vp9_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, | 423 vp9_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, |
438 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 424 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
439 vp9_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, | 425 vp9_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, |
440 lfi->blim, lfi->lim, lfi->hev_thr, 2); | 426 lfi->blim, lfi->lim, lfi->hev_thr, 2); |
441 | 427 |
442 if (u_ptr) | 428 if (u_ptr) |
443 vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, | 429 vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, |
444 lfi->blim, lfi->lim, lfi->hev_thr, 1); | 430 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
445 | 431 |
446 if (v_ptr) | 432 if (v_ptr) |
447 vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, | 433 vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, |
448 lfi->blim, lfi->lim, lfi->hev_thr, 1); | 434 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
449 } | 435 } |
450 | 436 |
451 void vp9_loop_filter_bh8x8_c(unsigned char *y_ptr, unsigned char *u_ptr, | 437 void vp9_loop_filter_bh8x8_c(uint8_t *y_ptr, uint8_t *u_ptr, |
452 unsigned char *v_ptr, int y_stride, int uv_stride, | 438 uint8_t *v_ptr, int y_stride, int uv_stride, |
453 struct loop_filter_info *lfi) { | 439 struct loop_filter_info *lfi) { |
454 vp9_mbloop_filter_horizontal_edge_c( | 440 vp9_mbloop_filter_horizontal_edge_c( |
455 y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2); | 441 y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2); |
| 442 |
| 443 if (u_ptr) |
| 444 vp9_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, |
| 445 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
| 446 |
| 447 if (v_ptr) |
| 448 vp9_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, |
| 449 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
456 } | 450 } |
457 | 451 |
458 void vp9_loop_filter_bhs_c(unsigned char *y_ptr, int y_stride, | 452 void vp9_loop_filter_bhs_c(uint8_t *y_ptr, int y_stride, |
459 const unsigned char *blimit) { | 453 const unsigned char *blimit) { |
460 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, | 454 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, |
461 y_stride, blimit); | 455 y_stride, blimit); |
462 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, | 456 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, |
463 y_stride, blimit); | 457 y_stride, blimit); |
464 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, | 458 vp9_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, |
465 y_stride, blimit); | 459 y_stride, blimit); |
466 } | 460 } |
467 | 461 |
468 void vp9_loop_filter_bv8x8_c(unsigned char *y_ptr, unsigned char *u_ptr, | 462 void vp9_loop_filter_bv8x8_c(uint8_t *y_ptr, uint8_t *u_ptr, |
469 unsigned char *v_ptr, int y_stride, int uv_stride, | 463 uint8_t *v_ptr, int y_stride, int uv_stride, |
470 struct loop_filter_info *lfi) { | 464 struct loop_filter_info *lfi) { |
471 vp9_mbloop_filter_vertical_edge_c( | 465 vp9_mbloop_filter_vertical_edge_c( |
472 y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2); | 466 y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2); |
473 } | 467 |
474 | 468 if (u_ptr) |
475 void vp9_loop_filter_bvs_c(unsigned char *y_ptr, int y_stride, | 469 vp9_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, |
| 470 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
| 471 |
| 472 if (v_ptr) |
| 473 vp9_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, |
| 474 lfi->blim, lfi->lim, lfi->hev_thr, 1); |
| 475 } |
| 476 |
| 477 void vp9_loop_filter_bvs_c(uint8_t *y_ptr, int y_stride, |
476 const unsigned char *blimit) { | 478 const unsigned char *blimit) { |
477 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit); | 479 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit); |
478 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit); | 480 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit); |
479 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit); | 481 vp9_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit); |
480 } | 482 } |
| 483 |
| 484 static __inline void wide_mbfilter(int8_t mask, uint8_t hev, |
| 485 uint8_t flat, uint8_t flat2, |
| 486 uint8_t *op7, uint8_t *op6, uint8_t *op5, |
| 487 uint8_t *op4, uint8_t *op3, uint8_t *op2, |
| 488 uint8_t *op1, uint8_t *op0, uint8_t *oq0, |
| 489 uint8_t *oq1, uint8_t *oq2, uint8_t *oq3, |
| 490 uint8_t *oq4, uint8_t *oq5, uint8_t *oq6, |
| 491 uint8_t *oq7) { |
| 492 /* use a 15 tap filter [1,1,1,1,1,1,1,2,1,1,1,1,1,1,1] for flat line */ |
| 493 if (flat2 && flat && mask) { |
| 494 uint8_t p0, q0; |
| 495 uint8_t p1, q1; |
| 496 uint8_t p2, q2; |
| 497 uint8_t p3, q3; |
| 498 uint8_t p4, q4; |
| 499 uint8_t p5, q5; |
| 500 uint8_t p6, q6; |
| 501 uint8_t p7, q7; |
| 502 |
| 503 p7 = *op7; |
| 504 p6 = *op6; |
| 505 p5 = *op5; |
| 506 p4 = *op4; |
| 507 p3 = *op3; |
| 508 p2 = *op2; |
| 509 p1 = *op1; |
| 510 p0 = *op0; |
| 511 q0 = *oq0; |
| 512 q1 = *oq1; |
| 513 q2 = *oq2; |
| 514 q3 = *oq3; |
| 515 q4 = *oq4; |
| 516 q5 = *oq5; |
| 517 q6 = *oq6; |
| 518 q7 = *oq7; |
| 519 |
| 520 *op6 = (p7 * 7 + p6 * 2 + |
| 521 p5 + p4 + p3 + p2 + p1 + p0 + q0 + 8) >> 4; |
| 522 *op5 = (p7 * 6 + p6 + p5 * 2 + |
| 523 p4 + p3 + p2 + p1 + p0 + q0 + q1 + 8) >> 4; |
| 524 *op4 = (p7 * 5 + p6 + p5 + p4 * 2 + |
| 525 p3 + p2 + p1 + p0 + q0 + q1 + q2 + 8) >> 4; |
| 526 *op3 = (p7 * 4 + p6 + p5 + p4 + p3 * 2 + |
| 527 p2 + p1 + p0 + q0 + q1 + q2 + q3 + 8) >> 4; |
| 528 *op2 = (p7 * 3 + p6 + p5 + p4 + p3 + p2 * 2 + |
| 529 p1 + p0 + q0 + q1 + q2 + q3 + q4 + 8) >> 4; |
| 530 *op1 = (p7 * 2 + p6 + p5 + p4 + p3 + p2 + p1 * 2 + |
| 531 p0 + q0 + q1 + q2 + q3 + q4 + q5 + 8) >> 4; |
| 532 *op0 = (p7 + p6 + p5 + p4 + p3 + p2 + p1 + p0 * 2 + |
| 533 q0 + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4; |
| 534 *oq0 = (p6 + p5 + p4 + p3 + p2 + p1 + p0 + q0 * 2 + |
| 535 q1 + q2 + q3 + q4 + q5 + q6 + q7 + 8) >> 4; |
| 536 *oq1 = (p5 + p4 + p3 + p2 + p1 + p0 + q0 + q1 * 2 + |
| 537 q2 + q3 + q4 + q5 + q6 + q7 * 2 + 8) >> 4; |
| 538 *oq2 = (p4 + p3 + p2 + p1 + p0 + q0 + q1 + q2 * 2 + |
| 539 q3 + q4 + q5 + q6 + q7 * 3 + 8) >> 4; |
| 540 *oq3 = (p3 + p2 + p1 + p0 + q0 + q1 + q2 + q3 * 2 + |
| 541 q4 + q5 + q6 + q7 * 4 + 8) >> 4; |
| 542 *oq4 = (p2 + p1 + p0 + q0 + q1 + q2 + q3 + q4 * 2 + |
| 543 q5 + q6 + q7 * 5 + 8) >> 4; |
| 544 *oq5 = (p1 + p0 + q0 + q1 + q2 + q3 + q4 + q5 * 2 + |
| 545 q6 + q7 * 6 + 8) >> 4; |
| 546 *oq6 = (p0 + q0 + q1 + q2 + q3 + q4 + q5 + q6 * 2 + |
| 547 q7 * 7 + 8) >> 4; |
| 548 } else if (flat && mask) { |
| 549 unsigned char p0, q0; |
| 550 unsigned char p1, q1; |
| 551 unsigned char p2, q2; |
| 552 unsigned char p3, q3; |
| 553 unsigned char p4, q4; |
| 554 |
| 555 p4 = *op4; |
| 556 p3 = *op3; |
| 557 p2 = *op2; |
| 558 p1 = *op1; |
| 559 p0 = *op0; |
| 560 q0 = *oq0; |
| 561 q1 = *oq1; |
| 562 q2 = *oq2; |
| 563 q3 = *oq3; |
| 564 q4 = *oq4; |
| 565 |
| 566 *op2 = (p4 + p4 + p3 + p2 + p2 + p1 + p0 + q0 + 4) >> 3; |
| 567 *op1 = (p4 + p3 + p2 + p1 + p1 + p0 + q0 + q1 + 4) >> 3; |
| 568 *op0 = (p3 + p2 + p1 + p0 + p0 + q0 + q1 + q2 + 4) >> 3; |
| 569 *oq0 = (p2 + p1 + p0 + q0 + q0 + q1 + q2 + q3 + 4) >> 3; |
| 570 *oq1 = (p1 + p0 + q0 + q1 + q1 + q2 + q3 + q4 + 4) >> 3; |
| 571 *oq2 = (p0 + q0 + q1 + q2 + q2 + q3 + q4 + q4 + 4) >> 3; |
| 572 } else { |
| 573 signed char ps0, qs0; |
| 574 signed char ps1, qs1; |
| 575 signed char filter, Filter1, Filter2; |
| 576 signed char u; |
| 577 |
| 578 ps1 = (signed char) * op1 ^ 0x80; |
| 579 ps0 = (signed char) * op0 ^ 0x80; |
| 580 qs0 = (signed char) * oq0 ^ 0x80; |
| 581 qs1 = (signed char) * oq1 ^ 0x80; |
| 582 |
| 583 /* add outer taps if we have high edge variance */ |
| 584 filter = signed_char_clamp(ps1 - qs1); |
| 585 filter &= hev; |
| 586 |
| 587 /* inner taps */ |
| 588 filter = signed_char_clamp(filter + 3 * (qs0 - ps0)); |
| 589 filter &= mask; |
| 590 |
| 591 Filter1 = signed_char_clamp(filter + 4); |
| 592 Filter2 = signed_char_clamp(filter + 3); |
| 593 Filter1 >>= 3; |
| 594 Filter2 >>= 3; |
| 595 |
| 596 u = signed_char_clamp(qs0 - Filter1); |
| 597 *oq0 = u ^ 0x80; |
| 598 u = signed_char_clamp(ps0 + Filter2); |
| 599 *op0 = u ^ 0x80; |
| 600 filter = Filter1; |
| 601 |
| 602 /* outer tap adjustments */ |
| 603 filter += 1; |
| 604 filter >>= 1; |
| 605 filter &= ~hev; |
| 606 |
| 607 u = signed_char_clamp(qs1 - filter); |
| 608 *oq1 = u ^ 0x80; |
| 609 u = signed_char_clamp(ps1 + filter); |
| 610 *op1 = u ^ 0x80; |
| 611 } |
| 612 } |
| 613 |
| 614 void vp9_mb_lpf_horizontal_edge_w |
| 615 ( |
| 616 unsigned char *s, |
| 617 int p, |
| 618 const unsigned char *blimit, |
| 619 const unsigned char *limit, |
| 620 const unsigned char *thresh, |
| 621 int count |
| 622 ) { |
| 623 signed char hev = 0; /* high edge variance */ |
| 624 signed char mask = 0; |
| 625 signed char flat = 0; |
| 626 signed char flat2 = 0; |
| 627 int i = 0; |
| 628 |
| 629 /* loop filter designed to work using chars so that we can make maximum use |
| 630 * of 8 bit simd instructions. |
| 631 */ |
| 632 do { |
| 633 mask = filter_mask(limit[0], blimit[0], |
| 634 s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], |
| 635 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p]); |
| 636 |
| 637 hev = hevmask(thresh[0], s[-2 * p], s[-1 * p], s[0 * p], s[1 * p]); |
| 638 |
| 639 flat = flatmask(1, |
| 640 s[-5 * p], s[-4 * p], s[-3 * p], s[-2 * p], s[-1 * p], |
| 641 s[ 0 * p], s[ 1 * p], s[ 2 * p], s[ 3 * p], s[ 4 * p]); |
| 642 |
| 643 flat2 = flatmask(1, |
| 644 s[-8 * p], s[-7 * p], s[-6 * p], s[-5 * p], s[-1 * p], |
| 645 s[ 0 * p], s[ 4 * p], s[ 5 * p], s[ 6 * p], s[ 7 * p]); |
| 646 |
| 647 wide_mbfilter(mask, hev, flat, flat2, |
| 648 s - 8 * p, s - 7 * p, s - 6 * p, s - 5 * p, |
| 649 s - 4 * p, s - 3 * p, s - 2 * p, s - 1 * p, |
| 650 s, s + 1 * p, s + 2 * p, s + 3 * p, |
| 651 s + 4 * p, s + 5 * p, s + 6 * p, s + 7 * p); |
| 652 |
| 653 ++s; |
| 654 } while (++i < count * 8); |
| 655 } |
| 656 void vp9_mb_lpf_vertical_edge_w |
| 657 ( |
| 658 unsigned char *s, |
| 659 int p, |
| 660 const unsigned char *blimit, |
| 661 const unsigned char *limit, |
| 662 const unsigned char *thresh, |
| 663 int count |
| 664 ) { |
| 665 signed char hev = 0; /* high edge variance */ |
| 666 signed char mask = 0; |
| 667 signed char flat = 0; |
| 668 signed char flat2 = 0; |
| 669 int i = 0; |
| 670 |
| 671 do { |
| 672 mask = filter_mask(limit[0], blimit[0], |
| 673 s[-4], s[-3], s[-2], s[-1], |
| 674 s[0], s[1], s[2], s[3]); |
| 675 |
| 676 hev = hevmask(thresh[0], s[-2], s[-1], s[0], s[1]); |
| 677 flat = flatmask(1, |
| 678 s[-5], s[-4], s[-3], s[-2], s[-1], |
| 679 s[ 0], s[ 1], s[ 2], s[ 3], s[ 4]); |
| 680 flat2 = flatmask(1, |
| 681 s[-8], s[-7], s[-6], s[-5], s[-1], |
| 682 s[ 0], s[ 4], s[ 5], s[ 6], s[ 7]); |
| 683 |
| 684 wide_mbfilter(mask, hev, flat, flat2, |
| 685 s - 8, s - 7, s - 6, s - 5, |
| 686 s - 4, s - 3, s - 2, s - 1, |
| 687 s, s + 1, s + 2, s + 3, |
| 688 s + 4, s + 5, s + 6, s + 7); |
| 689 s += p; |
| 690 } while (++i < count * 8); |
| 691 } |
| 692 |
| 693 void vp9_lpf_mbv_w_c(unsigned char *y_ptr, unsigned char *u_ptr, |
| 694 unsigned char *v_ptr, int y_stride, int uv_stride, |
| 695 struct loop_filter_info *lfi) { |
| 696 vp9_mb_lpf_vertical_edge_w(y_ptr, y_stride, |
| 697 lfi->mblim, lfi->lim, lfi->hev_thr, 2); |
| 698 |
| 699 if (u_ptr) |
| 700 vp9_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, |
| 701 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
| 702 |
| 703 if (v_ptr) |
| 704 vp9_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, |
| 705 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
| 706 } |
| 707 void vp9_lpf_mbh_w_c(unsigned char *y_ptr, unsigned char *u_ptr, |
| 708 unsigned char *v_ptr, int y_stride, int uv_stride, |
| 709 struct loop_filter_info *lfi) { |
| 710 vp9_mb_lpf_horizontal_edge_w(y_ptr, y_stride, |
| 711 lfi->mblim, lfi->lim, lfi->hev_thr, 2); |
| 712 |
| 713 if (u_ptr) |
| 714 vp9_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, |
| 715 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
| 716 |
| 717 if (v_ptr) |
| 718 vp9_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, |
| 719 lfi->mblim, lfi->lim, lfi->hev_thr, 1); |
| 720 } |
| 721 |
OLD | NEW |