Mercurial > libavcodec.hg
annotate snow.h @ 7066:f23e64bcec1e libavcodec
simplify
author | michael |
---|---|
date | Wed, 18 Jun 2008 21:09:36 +0000 |
parents | 1d83e9c34641 |
children | c4a4495715dd |
rev | line source |
---|---|
3198 | 1 /* |
2 * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at> | |
3 * Copyright (C) 2006 Robert Edele <yartrebo@earthlink.net> | |
4 * | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
5 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
6 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
7 * FFmpeg is free software; you can redistribute it and/or |
3198 | 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:
3582
diff
changeset
|
10 * version 2.1 of the License, or (at your option) any later version. |
3198 | 11 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
12 * FFmpeg is distributed in the hope that it will be useful, |
3198 | 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:
3582
diff
changeset
|
18 * License along with FFmpeg; if not, write to the Free Software |
3198 | 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 */ | |
21 | |
5830
1d83e9c34641
Add FFMPEG_ prefix to all multiple inclusion guards.
diego
parents:
5589
diff
changeset
|
22 #ifndef FFMPEG_SNOW_H |
1d83e9c34641
Add FFMPEG_ prefix to all multiple inclusion guards.
diego
parents:
5589
diff
changeset
|
23 #define FFMPEG_SNOW_H |
3198 | 24 |
25 #include "dsputil.h" | |
26 | |
27 #define MID_STATE 128 | |
28 | |
29 #define MAX_DECOMPOSITIONS 8 | |
30 #define MAX_PLANES 4 | |
31 #define QSHIFT 5 | |
32 #define QROOT (1<<QSHIFT) | |
33 #define LOSSLESS_QLOG -128 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
34 #define FRAC_BITS 4 |
3314 | 35 #define MAX_REF_FRAMES 8 |
3198 | 36 |
3206 | 37 #define LOG2_OBMC_MAX 8 |
3198 | 38 #define OBMC_MAX (1<<(LOG2_OBMC_MAX)) |
39 | |
3326 | 40 #define DWT_97 0 |
41 #define DWT_53 1 | |
42 | |
3198 | 43 /** Used to minimize the amount of memory used in order to optimize cache performance. **/ |
44 struct slice_buffer_s { | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
45 IDWTELEM * * line; ///< For use by idwt and predict_slices. |
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
46 IDWTELEM * * data_stack; ///< Used for internal purposes. |
3198 | 47 int data_stack_top; |
48 int line_count; | |
49 int line_width; | |
50 int data_count; | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
51 IDWTELEM * base_buffer; ///< Buffer that this structure is caching. |
3198 | 52 }; |
53 | |
54 #define liftS lift | |
55 #if 1 | |
56 #define W_AM 3 | |
57 #define W_AO 0 | |
58 #define W_AS 1 | |
59 | |
60 #undef liftS | |
61 #define W_BM 1 | |
62 #define W_BO 8 | |
63 #define W_BS 4 | |
64 | |
65 #define W_CM 1 | |
66 #define W_CO 0 | |
67 #define W_CS 0 | |
68 | |
69 #define W_DM 3 | |
70 #define W_DO 4 | |
71 #define W_DS 3 | |
72 #elif 0 | |
73 #define W_AM 55 | |
74 #define W_AO 16 | |
75 #define W_AS 5 | |
76 | |
77 #define W_BM 3 | |
78 #define W_BO 32 | |
79 #define W_BS 6 | |
80 | |
81 #define W_CM 127 | |
82 #define W_CO 64 | |
83 #define W_CS 7 | |
84 | |
85 #define W_DM 7 | |
86 #define W_DO 8 | |
87 #define W_DS 4 | |
88 #elif 0 | |
89 #define W_AM 97 | |
90 #define W_AO 32 | |
91 #define W_AS 6 | |
92 | |
93 #define W_BM 63 | |
94 #define W_BO 512 | |
95 #define W_BS 10 | |
96 | |
97 #define W_CM 13 | |
98 #define W_CO 8 | |
99 #define W_CS 4 | |
100 | |
101 #define W_DM 15 | |
102 #define W_DO 16 | |
103 #define W_DS 5 | |
104 | |
105 #else | |
106 | |
107 #define W_AM 203 | |
108 #define W_AO 64 | |
109 #define W_AS 7 | |
110 | |
111 #define W_BM 217 | |
112 #define W_BO 2048 | |
113 #define W_BS 12 | |
114 | |
115 #define W_CM 113 | |
116 #define W_CO 64 | |
117 #define W_CS 7 | |
118 | |
119 #define W_DM 227 | |
120 #define W_DO 128 | |
121 #define W_DS 9 | |
122 #endif | |
123 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
124 extern void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width); |
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
125 extern void ff_snow_horizontal_compose97i(IDWTELEM *b, int width); |
4436
d3e389536b0a
Add the const specifier as needed to reduce the number of warnings.
takis
parents:
4283
diff
changeset
|
126 extern void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8); |
3198 | 127 |
3582
6310389a9688
Compile fix (for GCC<3.3) when encoders are disabled
gpoirier
parents:
3326
diff
changeset
|
128 #ifdef CONFIG_SNOW_ENCODER |
4197 | 129 int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h); |
130 int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h); | |
3582
6310389a9688
Compile fix (for GCC<3.3) when encoders are disabled
gpoirier
parents:
3326
diff
changeset
|
131 #else |
6310389a9688
Compile fix (for GCC<3.3) when encoders are disabled
gpoirier
parents:
3326
diff
changeset
|
132 static int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {assert (0);} |
6310389a9688
Compile fix (for GCC<3.3) when encoders are disabled
gpoirier
parents:
3326
diff
changeset
|
133 static int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {assert (0);} |
6310389a9688
Compile fix (for GCC<3.3) when encoders are disabled
gpoirier
parents:
3326
diff
changeset
|
134 #endif |
3223 | 135 |
136 /* C bits used by mmx/sse2/altivec */ | |
137 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
138 static av_always_inline void snow_interleave_line_header(int * i, int width, IDWTELEM * low, IDWTELEM * high){ |
3223 | 139 (*i) = (width) - 2; |
140 | |
141 if (width & 1){ | |
142 low[(*i)+1] = low[((*i)+1)>>1]; | |
143 (*i)--; | |
144 } | |
145 } | |
146 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
147 static av_always_inline void snow_interleave_line_footer(int * i, IDWTELEM * low, IDWTELEM * high){ |
3223 | 148 for (; (*i)>=0; (*i)-=2){ |
149 low[(*i)+1] = high[(*i)>>1]; | |
150 low[*i] = low[(*i)>>1]; | |
151 } | |
152 } | |
153 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
154 static av_always_inline void snow_horizontal_compose_lift_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w, int lift_high, int mul, int add, int shift){ |
3223 | 155 for(; i<w; i++){ |
156 dst[i] = src[i] - ((mul * (ref[i] + ref[i + 1]) + add) >> shift); | |
157 } | |
158 | |
159 if((width^lift_high)&1){ | |
160 dst[w] = src[w] - ((mul * 2 * ref[w] + add) >> shift); | |
161 } | |
162 } | |
163 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
164 static av_always_inline void snow_horizontal_compose_liftS_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w){ |
3223 | 165 for(; i<w; i++){ |
5565
93082c591c8b
Change rounding of the horizontal DWT to match the vertical one.
michael
parents:
5552
diff
changeset
|
166 dst[i] = src[i] + ((ref[i] + ref[(i+1)]+W_BO + 4 * src[i]) >> W_BS); |
3223 | 167 } |
168 | |
169 if(width&1){ | |
5565
93082c591c8b
Change rounding of the horizontal DWT to match the vertical one.
michael
parents:
5552
diff
changeset
|
170 dst[w] = src[w] + ((2 * ref[w] + W_BO + 4 * src[w]) >> W_BS); |
3223 | 171 } |
172 } | |
173 | |
5830
1d83e9c34641
Add FFMPEG_ prefix to all multiple inclusion guards.
diego
parents:
5589
diff
changeset
|
174 #endif /* FFMPEG_SNOW_H */ |