Mercurial > libavcodec.hg
annotate vc1dsp.c @ 10015:71ff4286b0d0 libavcodec
Cosmetic: Reindent after last commit
author | heydowns |
---|---|
date | Mon, 03 Aug 2009 17:55:02 +0000 |
parents | 7a116de63777 |
children | bf309c7ce615 |
rev | line source |
---|---|
3526 | 1 /* |
2 * VC-1 and WMV3 decoder - DSP functions | |
3 * Copyright (c) 2006 Konstantin Shishkov | |
4 * | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
5 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
6 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
7 * FFmpeg is free software; you can redistribute it and/or |
3526 | 8 * modify it under the terms of the GNU Lesser General Public |
9 * License as published by the Free Software Foundation; either | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
10 * version 2.1 of the License, or (at your option) any later version. |
3526 | 11 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
12 * FFmpeg is distributed in the hope that it will be useful, |
3526 | 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 * Lesser General Public License for more details. | |
16 * | |
17 * You should have received a copy of the GNU Lesser General Public | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3665
diff
changeset
|
18 * License along with FFmpeg; if not, write to the Free Software |
3526 | 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 */ | |
21 | |
22 /** | |
8718
e9d9d946f213
Use full internal pathname in doxygen @file directives.
diego
parents:
8686
diff
changeset
|
23 * @file libavcodec/vc1dsp.c |
3526 | 24 * VC-1 and WMV3 decoder |
25 * | |
26 */ | |
27 | |
28 #include "dsputil.h" | |
29 | |
30 | |
4210 | 31 /** Apply overlap transform to horizontal edge |
3526 | 32 */ |
4239 | 33 static void vc1_v_overlap_c(uint8_t* src, int stride) |
3526 | 34 { |
35 int i; | |
36 int a, b, c, d; | |
4211 | 37 int d1, d2; |
4239 | 38 int rnd = 1; |
3526 | 39 for(i = 0; i < 8; i++) { |
40 a = src[-2*stride]; | |
41 b = src[-stride]; | |
42 c = src[0]; | |
43 d = src[stride]; | |
4211 | 44 d1 = (a - d + 3 + rnd) >> 3; |
45 d2 = (a - d + b - c + 4 - rnd) >> 3; | |
3526 | 46 |
4211 | 47 src[-2*stride] = a - d1; |
8686 | 48 src[-stride] = av_clip_uint8(b - d2); |
49 src[0] = av_clip_uint8(c + d2); | |
4211 | 50 src[stride] = d + d1; |
3526 | 51 src++; |
4239 | 52 rnd = !rnd; |
3526 | 53 } |
54 } | |
55 | |
4210 | 56 /** Apply overlap transform to vertical edge |
3526 | 57 */ |
4239 | 58 static void vc1_h_overlap_c(uint8_t* src, int stride) |
3526 | 59 { |
60 int i; | |
61 int a, b, c, d; | |
4211 | 62 int d1, d2; |
4239 | 63 int rnd = 1; |
3526 | 64 for(i = 0; i < 8; i++) { |
65 a = src[-2]; | |
66 b = src[-1]; | |
67 c = src[0]; | |
68 d = src[1]; | |
4211 | 69 d1 = (a - d + 3 + rnd) >> 3; |
70 d2 = (a - d + b - c + 4 - rnd) >> 3; | |
3526 | 71 |
4211 | 72 src[-2] = a - d1; |
8686 | 73 src[-1] = av_clip_uint8(b - d2); |
74 src[0] = av_clip_uint8(c + d2); | |
4211 | 75 src[1] = d + d1; |
3526 | 76 src += stride; |
4239 | 77 rnd = !rnd; |
3526 | 78 } |
79 } | |
80 | |
9442 | 81 /** |
82 * VC-1 in-loop deblocking filter for one line | |
83 * @param src source block type | |
84 * @param stride block stride | |
85 * @param pq block quantizer | |
86 * @return whether other 3 pairs should be filtered or not | |
87 * @see 8.6 | |
88 */ | |
89 static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){ | |
90 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; | |
91 | |
92 int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3; | |
93 int a0_sign = a0 >> 31; /* Store sign */ | |
94 a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */ | |
95 if(a0 < pq){ | |
96 int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3); | |
97 int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3); | |
98 if(a1 < a0 || a2 < a0){ | |
99 int clip = src[-1*stride] - src[ 0*stride]; | |
100 int clip_sign = clip >> 31; | |
101 clip = ((clip ^ clip_sign) - clip_sign)>>1; | |
102 if(clip){ | |
103 int a3 = FFMIN(a1, a2); | |
104 int d = 5 * (a3 - a0); | |
105 int d_sign = (d >> 31); | |
106 d = ((d ^ d_sign) - d_sign) >> 3; | |
107 d_sign ^= a0_sign; | |
108 | |
109 if( d_sign ^ clip_sign ) | |
110 d = 0; | |
111 else{ | |
112 d = FFMIN(d, clip); | |
113 d = (d ^ d_sign) - d_sign; /* Restore sign */ | |
114 src[-1*stride] = cm[src[-1*stride] - d]; | |
115 src[ 0*stride] = cm[src[ 0*stride] + d]; | |
116 } | |
117 return 1; | |
118 } | |
119 } | |
120 } | |
121 return 0; | |
122 } | |
123 | |
124 /** | |
125 * VC-1 in-loop deblocking filter | |
126 * @param src source block type | |
127 * @param step distance between horizontally adjacent elements | |
128 * @param stride distance between vertically adjacent elements | |
129 * @param len edge length to filter (4 or 8 pixels) | |
130 * @param pq block quantizer | |
131 * @see 8.6 | |
132 */ | |
9443
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
133 static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq) |
9442 | 134 { |
135 int i; | |
136 int filt3; | |
137 | |
138 for(i = 0; i < len; i += 4){ | |
139 filt3 = vc1_filter_line(src + 2*step, stride, pq); | |
140 if(filt3){ | |
141 vc1_filter_line(src + 0*step, stride, pq); | |
142 vc1_filter_line(src + 1*step, stride, pq); | |
143 vc1_filter_line(src + 3*step, stride, pq); | |
144 } | |
145 src += step * 4; | |
146 } | |
147 } | |
3526 | 148 |
9443
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
149 static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
150 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
151 vc1_loop_filter(src, 1, stride, 4, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
152 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
153 |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
154 static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
155 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
156 vc1_loop_filter(src, stride, 1, 4, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
157 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
158 |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
159 static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
160 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
161 vc1_loop_filter(src, 1, stride, 8, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
162 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
163 |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
164 static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
165 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
166 vc1_loop_filter(src, stride, 1, 8, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
167 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
168 |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
169 static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
170 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
171 vc1_loop_filter(src, 1, stride, 16, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
172 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
173 |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
174 static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq) |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
175 { |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
176 vc1_loop_filter(src, stride, 1, 16, pq); |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
177 } |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
178 |
3526 | 179 /** Do inverse transform on 8x8 block |
180 */ | |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
181 static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block) |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
182 { |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
183 int i; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
184 int dc = block[0]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
185 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
186 dc = (3 * dc + 1) >> 1; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
187 dc = (3 * dc + 16) >> 5; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
188 for(i = 0; i < 8; i++){ |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
189 dest[0] = cm[dest[0]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
190 dest[1] = cm[dest[1]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
191 dest[2] = cm[dest[2]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
192 dest[3] = cm[dest[3]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
193 dest[4] = cm[dest[4]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
194 dest[5] = cm[dest[5]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
195 dest[6] = cm[dest[6]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
196 dest[7] = cm[dest[7]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
197 dest += linesize; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
198 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
199 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
200 |
3526 | 201 static void vc1_inv_trans_8x8_c(DCTELEM block[64]) |
202 { | |
203 int i; | |
204 register int t1,t2,t3,t4,t5,t6,t7,t8; | |
205 DCTELEM *src, *dst; | |
206 | |
207 src = block; | |
208 dst = block; | |
209 for(i = 0; i < 8; i++){ | |
6157 | 210 t1 = 12 * (src[0] + src[4]) + 4; |
211 t2 = 12 * (src[0] - src[4]) + 4; | |
3526 | 212 t3 = 16 * src[2] + 6 * src[6]; |
213 t4 = 6 * src[2] - 16 * src[6]; | |
214 | |
215 t5 = t1 + t3; | |
216 t6 = t2 + t4; | |
217 t7 = t2 - t4; | |
218 t8 = t1 - t3; | |
219 | |
220 t1 = 16 * src[1] + 15 * src[3] + 9 * src[5] + 4 * src[7]; | |
221 t2 = 15 * src[1] - 4 * src[3] - 16 * src[5] - 9 * src[7]; | |
222 t3 = 9 * src[1] - 16 * src[3] + 4 * src[5] + 15 * src[7]; | |
223 t4 = 4 * src[1] - 9 * src[3] + 15 * src[5] - 16 * src[7]; | |
224 | |
6157 | 225 dst[0] = (t5 + t1) >> 3; |
226 dst[1] = (t6 + t2) >> 3; | |
227 dst[2] = (t7 + t3) >> 3; | |
228 dst[3] = (t8 + t4) >> 3; | |
229 dst[4] = (t8 - t4) >> 3; | |
230 dst[5] = (t7 - t3) >> 3; | |
231 dst[6] = (t6 - t2) >> 3; | |
232 dst[7] = (t5 - t1) >> 3; | |
3526 | 233 |
234 src += 8; | |
235 dst += 8; | |
236 } | |
237 | |
238 src = block; | |
239 dst = block; | |
240 for(i = 0; i < 8; i++){ | |
6157 | 241 t1 = 12 * (src[ 0] + src[32]) + 64; |
242 t2 = 12 * (src[ 0] - src[32]) + 64; | |
3526 | 243 t3 = 16 * src[16] + 6 * src[48]; |
244 t4 = 6 * src[16] - 16 * src[48]; | |
245 | |
246 t5 = t1 + t3; | |
247 t6 = t2 + t4; | |
248 t7 = t2 - t4; | |
249 t8 = t1 - t3; | |
250 | |
251 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56]; | |
252 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56]; | |
253 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56]; | |
254 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56]; | |
255 | |
6157 | 256 dst[ 0] = (t5 + t1) >> 7; |
257 dst[ 8] = (t6 + t2) >> 7; | |
258 dst[16] = (t7 + t3) >> 7; | |
259 dst[24] = (t8 + t4) >> 7; | |
260 dst[32] = (t8 - t4 + 1) >> 7; | |
261 dst[40] = (t7 - t3 + 1) >> 7; | |
262 dst[48] = (t6 - t2 + 1) >> 7; | |
263 dst[56] = (t5 - t1 + 1) >> 7; | |
3526 | 264 |
265 src++; | |
266 dst++; | |
267 } | |
268 } | |
269 | |
270 /** Do inverse transform on 8x4 part of block | |
271 */ | |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
272 static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block) |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
273 { |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
274 int i; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
275 int dc = block[0]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
276 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
277 dc = ( 3 * dc + 1) >> 1; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
278 dc = (17 * dc + 64) >> 7; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
279 for(i = 0; i < 4; i++){ |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
280 dest[0] = cm[dest[0]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
281 dest[1] = cm[dest[1]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
282 dest[2] = cm[dest[2]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
283 dest[3] = cm[dest[3]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
284 dest[4] = cm[dest[4]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
285 dest[5] = cm[dest[5]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
286 dest[6] = cm[dest[6]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
287 dest[7] = cm[dest[7]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
288 dest += linesize; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
289 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
290 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
291 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
292 static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block) |
3526 | 293 { |
294 int i; | |
295 register int t1,t2,t3,t4,t5,t6,t7,t8; | |
296 DCTELEM *src, *dst; | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
297 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
3526 | 298 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
299 src = block; |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
300 dst = block; |
3526 | 301 for(i = 0; i < 4; i++){ |
6157 | 302 t1 = 12 * (src[0] + src[4]) + 4; |
303 t2 = 12 * (src[0] - src[4]) + 4; | |
3526 | 304 t3 = 16 * src[2] + 6 * src[6]; |
305 t4 = 6 * src[2] - 16 * src[6]; | |
306 | |
307 t5 = t1 + t3; | |
308 t6 = t2 + t4; | |
309 t7 = t2 - t4; | |
310 t8 = t1 - t3; | |
311 | |
312 t1 = 16 * src[1] + 15 * src[3] + 9 * src[5] + 4 * src[7]; | |
313 t2 = 15 * src[1] - 4 * src[3] - 16 * src[5] - 9 * src[7]; | |
314 t3 = 9 * src[1] - 16 * src[3] + 4 * src[5] + 15 * src[7]; | |
315 t4 = 4 * src[1] - 9 * src[3] + 15 * src[5] - 16 * src[7]; | |
316 | |
6157 | 317 dst[0] = (t5 + t1) >> 3; |
318 dst[1] = (t6 + t2) >> 3; | |
319 dst[2] = (t7 + t3) >> 3; | |
320 dst[3] = (t8 + t4) >> 3; | |
321 dst[4] = (t8 - t4) >> 3; | |
322 dst[5] = (t7 - t3) >> 3; | |
323 dst[6] = (t6 - t2) >> 3; | |
324 dst[7] = (t5 - t1) >> 3; | |
3526 | 325 |
326 src += 8; | |
327 dst += 8; | |
328 } | |
329 | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
330 src = block; |
3526 | 331 for(i = 0; i < 8; i++){ |
6157 | 332 t1 = 17 * (src[ 0] + src[16]) + 64; |
333 t2 = 17 * (src[ 0] - src[16]) + 64; | |
6158 | 334 t3 = 22 * src[ 8] + 10 * src[24]; |
335 t4 = 22 * src[24] - 10 * src[ 8]; | |
3526 | 336 |
6158 | 337 dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)]; |
338 dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)]; | |
339 dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)]; | |
340 dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)]; | |
3526 | 341 |
342 src ++; | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
343 dest++; |
3526 | 344 } |
345 } | |
346 | |
347 /** Do inverse transform on 4x8 parts of block | |
348 */ | |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
349 static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block) |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
350 { |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
351 int i; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
352 int dc = block[0]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
353 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
354 dc = (17 * dc + 4) >> 3; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
355 dc = (12 * dc + 64) >> 7; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
356 for(i = 0; i < 8; i++){ |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
357 dest[0] = cm[dest[0]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
358 dest[1] = cm[dest[1]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
359 dest[2] = cm[dest[2]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
360 dest[3] = cm[dest[3]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
361 dest += linesize; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
362 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
363 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
364 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
365 static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block) |
3526 | 366 { |
367 int i; | |
368 register int t1,t2,t3,t4,t5,t6,t7,t8; | |
369 DCTELEM *src, *dst; | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
370 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
3526 | 371 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
372 src = block; |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
373 dst = block; |
3526 | 374 for(i = 0; i < 8; i++){ |
6157 | 375 t1 = 17 * (src[0] + src[2]) + 4; |
376 t2 = 17 * (src[0] - src[2]) + 4; | |
6158 | 377 t3 = 22 * src[1] + 10 * src[3]; |
378 t4 = 22 * src[3] - 10 * src[1]; | |
3526 | 379 |
6158 | 380 dst[0] = (t1 + t3) >> 3; |
381 dst[1] = (t2 - t4) >> 3; | |
382 dst[2] = (t2 + t4) >> 3; | |
383 dst[3] = (t1 - t3) >> 3; | |
3526 | 384 |
385 src += 8; | |
386 dst += 8; | |
387 } | |
388 | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
389 src = block; |
3526 | 390 for(i = 0; i < 4; i++){ |
6157 | 391 t1 = 12 * (src[ 0] + src[32]) + 64; |
392 t2 = 12 * (src[ 0] - src[32]) + 64; | |
3526 | 393 t3 = 16 * src[16] + 6 * src[48]; |
394 t4 = 6 * src[16] - 16 * src[48]; | |
395 | |
396 t5 = t1 + t3; | |
397 t6 = t2 + t4; | |
398 t7 = t2 - t4; | |
399 t8 = t1 - t3; | |
400 | |
401 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56]; | |
402 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56]; | |
403 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56]; | |
404 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56]; | |
405 | |
6157 | 406 dest[0*linesize] = cm[dest[0*linesize] + ((t5 + t1) >> 7)]; |
407 dest[1*linesize] = cm[dest[1*linesize] + ((t6 + t2) >> 7)]; | |
408 dest[2*linesize] = cm[dest[2*linesize] + ((t7 + t3) >> 7)]; | |
409 dest[3*linesize] = cm[dest[3*linesize] + ((t8 + t4) >> 7)]; | |
410 dest[4*linesize] = cm[dest[4*linesize] + ((t8 - t4 + 1) >> 7)]; | |
411 dest[5*linesize] = cm[dest[5*linesize] + ((t7 - t3 + 1) >> 7)]; | |
412 dest[6*linesize] = cm[dest[6*linesize] + ((t6 - t2 + 1) >> 7)]; | |
413 dest[7*linesize] = cm[dest[7*linesize] + ((t5 - t1 + 1) >> 7)]; | |
3526 | 414 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
415 src ++; |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
416 dest++; |
3526 | 417 } |
418 } | |
419 | |
420 /** Do inverse transform on 4x4 part of block | |
421 */ | |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
422 static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block) |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
423 { |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
424 int i; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
425 int dc = block[0]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
426 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
427 dc = (17 * dc + 4) >> 3; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
428 dc = (17 * dc + 64) >> 7; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
429 for(i = 0; i < 4; i++){ |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
430 dest[0] = cm[dest[0]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
431 dest[1] = cm[dest[1]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
432 dest[2] = cm[dest[2]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
433 dest[3] = cm[dest[3]+dc]; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
434 dest += linesize; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
435 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
436 } |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
437 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
438 static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block) |
3526 | 439 { |
440 int i; | |
6158 | 441 register int t1,t2,t3,t4; |
3526 | 442 DCTELEM *src, *dst; |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
443 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
3526 | 444 |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
445 src = block; |
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
446 dst = block; |
3526 | 447 for(i = 0; i < 4; i++){ |
6157 | 448 t1 = 17 * (src[0] + src[2]) + 4; |
449 t2 = 17 * (src[0] - src[2]) + 4; | |
6158 | 450 t3 = 22 * src[1] + 10 * src[3]; |
451 t4 = 22 * src[3] - 10 * src[1]; | |
3526 | 452 |
6158 | 453 dst[0] = (t1 + t3) >> 3; |
454 dst[1] = (t2 - t4) >> 3; | |
455 dst[2] = (t2 + t4) >> 3; | |
456 dst[3] = (t1 - t3) >> 3; | |
3526 | 457 |
458 src += 8; | |
459 dst += 8; | |
460 } | |
461 | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
462 src = block; |
3526 | 463 for(i = 0; i < 4; i++){ |
6157 | 464 t1 = 17 * (src[ 0] + src[16]) + 64; |
465 t2 = 17 * (src[ 0] - src[16]) + 64; | |
6158 | 466 t3 = 22 * src[ 8] + 10 * src[24]; |
467 t4 = 22 * src[24] - 10 * src[ 8]; | |
3526 | 468 |
6158 | 469 dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)]; |
470 dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)]; | |
471 dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)]; | |
472 dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)]; | |
3526 | 473 |
474 src ++; | |
5997
90de28dfd8d6
Switch VC-1 decoder to output decoded residual immediately.
kostya
parents:
5416
diff
changeset
|
475 dest++; |
3526 | 476 } |
477 } | |
478 | |
479 /* motion compensation functions */ | |
5416 | 480 /** Filter in case of 2 filters */ |
481 #define VC1_MSPEL_FILTER_16B(DIR, TYPE) \ | |
482 static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \ | |
483 { \ | |
484 switch(mode){ \ | |
485 case 0: /* no shift - should not occur */ \ | |
486 return 0; \ | |
487 case 1: /* 1/4 shift */ \ | |
488 return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \ | |
489 case 2: /* 1/2 shift */ \ | |
490 return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \ | |
491 case 3: /* 3/4 shift */ \ | |
492 return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \ | |
493 } \ | |
494 return 0; /* should not occur */ \ | |
495 } | |
496 | |
497 VC1_MSPEL_FILTER_16B(ver, uint8_t); | |
498 VC1_MSPEL_FILTER_16B(hor, int16_t); | |
499 | |
3526 | 500 |
501 /** Filter used to interpolate fractional pel values | |
502 */ | |
4283
d6f83e2f8804
rename always_inline to av_always_inline and move to common.h
mru
parents:
4239
diff
changeset
|
503 static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r) |
3526 | 504 { |
505 switch(mode){ | |
506 case 0: //no shift | |
507 return src[0]; | |
508 case 1: // 1/4 shift | |
509 return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6; | |
510 case 2: // 1/2 shift | |
511 return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4; | |
512 case 3: // 3/4 shift | |
513 return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6; | |
514 } | |
515 return 0; //should not occur | |
516 } | |
517 | |
518 /** Function used to do motion compensation with bicubic interpolation | |
519 */ | |
9437 | 520 #define VC1_MSPEL_MC(OP, OPNAME)\ |
521 static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\ | |
522 {\ | |
523 int i, j;\ | |
524 \ | |
525 if (vmode) { /* Horizontal filter to apply */\ | |
526 int r;\ | |
527 \ | |
528 if (hmode) { /* Vertical filter to apply, output to tmp */\ | |
529 static const int shift_value[] = { 0, 5, 1, 5 };\ | |
530 int shift = (shift_value[hmode]+shift_value[vmode])>>1;\ | |
531 int16_t tmp[11*8], *tptr = tmp;\ | |
532 \ | |
533 r = (1<<(shift-1)) + rnd-1;\ | |
534 \ | |
535 src -= 1;\ | |
536 for(j = 0; j < 8; j++) {\ | |
537 for(i = 0; i < 11; i++)\ | |
538 tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\ | |
539 src += stride;\ | |
540 tptr += 11;\ | |
541 }\ | |
542 \ | |
543 r = 64-rnd;\ | |
544 tptr = tmp+1;\ | |
545 for(j = 0; j < 8; j++) {\ | |
546 for(i = 0; i < 8; i++)\ | |
547 OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\ | |
548 dst += stride;\ | |
549 tptr += 11;\ | |
550 }\ | |
551 \ | |
552 return;\ | |
553 }\ | |
554 else { /* No horizontal filter, output 8 lines to dst */\ | |
555 r = 1-rnd;\ | |
556 \ | |
557 for(j = 0; j < 8; j++) {\ | |
558 for(i = 0; i < 8; i++)\ | |
559 OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\ | |
560 src += stride;\ | |
561 dst += stride;\ | |
562 }\ | |
563 return;\ | |
564 }\ | |
565 }\ | |
566 \ | |
567 /* Horizontal mode with no vertical mode */\ | |
568 for(j = 0; j < 8; j++) {\ | |
569 for(i = 0; i < 8; i++)\ | |
570 OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\ | |
571 dst += stride;\ | |
572 src += stride;\ | |
573 }\ | |
574 } | |
3526 | 575 |
9437 | 576 #define op_put(a, b) a = av_clip_uint8(b) |
577 #define op_avg(a, b) a = (a + av_clip_uint8(b) + 1) >> 1 | |
5416 | 578 |
9437 | 579 VC1_MSPEL_MC(op_put, put_) |
580 VC1_MSPEL_MC(op_avg, avg_) | |
3526 | 581 |
582 /* pixel functions - really are entry points to vc1_mspel_mc */ | |
583 | |
584 /* this one is defined in dsputil.c */ | |
585 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd); | |
9437 | 586 void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd); |
3526 | 587 |
5252 | 588 #define PUT_VC1_MSPEL(a, b)\ |
589 static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \ | |
9437 | 590 put_vc1_mspel_mc(dst, src, stride, a, b, rnd); \ |
591 }\ | |
592 static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \ | |
593 avg_vc1_mspel_mc(dst, src, stride, a, b, rnd); \ | |
3526 | 594 } |
595 | |
5252 | 596 PUT_VC1_MSPEL(1, 0) |
597 PUT_VC1_MSPEL(2, 0) | |
598 PUT_VC1_MSPEL(3, 0) | |
3526 | 599 |
5252 | 600 PUT_VC1_MSPEL(0, 1) |
601 PUT_VC1_MSPEL(1, 1) | |
602 PUT_VC1_MSPEL(2, 1) | |
603 PUT_VC1_MSPEL(3, 1) | |
3526 | 604 |
5252 | 605 PUT_VC1_MSPEL(0, 2) |
606 PUT_VC1_MSPEL(1, 2) | |
607 PUT_VC1_MSPEL(2, 2) | |
608 PUT_VC1_MSPEL(3, 2) | |
3526 | 609 |
5252 | 610 PUT_VC1_MSPEL(0, 3) |
611 PUT_VC1_MSPEL(1, 3) | |
612 PUT_VC1_MSPEL(2, 3) | |
613 PUT_VC1_MSPEL(3, 3) | |
3526 | 614 |
615 void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) { | |
616 dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c; | |
617 dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c; | |
618 dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c; | |
619 dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c; | |
9859
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
620 dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
621 dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
622 dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c; |
7a116de63777
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
darkshikari
parents:
9443
diff
changeset
|
623 dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c; |
3526 | 624 dsp->vc1_h_overlap = vc1_h_overlap_c; |
625 dsp->vc1_v_overlap = vc1_v_overlap_c; | |
9443
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
626 dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c; |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
627 dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c; |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
628 dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c; |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
629 dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c; |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
630 dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c; |
3970fe47fea3
Split VC1 loop filter into separate functions for h/v and size
conrad
parents:
9442
diff
changeset
|
631 dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c; |
3526 | 632 |
633 dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c; | |
5251 | 634 dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c; |
635 dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c; | |
636 dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c; | |
637 dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c; | |
638 dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c; | |
639 dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c; | |
640 dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c; | |
641 dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c; | |
642 dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c; | |
643 dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c; | |
644 dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c; | |
645 dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c; | |
646 dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c; | |
647 dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c; | |
648 dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c; | |
9437 | 649 |
650 dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_c; | |
651 dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c; | |
652 dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c; | |
653 dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c; | |
654 dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c; | |
655 dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c; | |
656 dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c; | |
657 dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c; | |
658 dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c; | |
659 dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c; | |
660 dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c; | |
661 dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c; | |
662 dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c; | |
663 dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c; | |
664 dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c; | |
665 dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c; | |
3526 | 666 } |