Mercurial > libavcodec.hg
annotate common.h @ 494:54b88078c361 libavcodec
fixed one mpeg decoder memory leak
author | bellard |
---|---|
date | Tue, 11 Jun 2002 13:44:14 +0000 |
parents | 90227eb98bf9 |
children | 09dbc9e0e6a5 |
rev | line source |
---|---|
0 | 1 #ifndef COMMON_H |
2 #define COMMON_H | |
3 | |
86 | 4 #define FFMPEG_VERSION_INT 0x000406 |
5 #define FFMPEG_VERSION "0.4.6" | |
64 | 6 |
213
e80ad397d30e
Cygwin's mangling by Felix Buenemann <atmosfear@users.sourceforge.net>
nickols_k
parents:
199
diff
changeset
|
7 #if defined(WIN32) && !defined(__MINGW32__) && !defined(__CYGWIN__) |
64 | 8 #define CONFIG_WIN32 |
9 #endif | |
10 | |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
11 //#define ALT_BITSTREAM_WRITER |
235
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
12 //#define ALIGNED_BITSTREAM_WRITER |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
13 //#define ALT_BITSTREAM_READER |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
14 //#define ALIGNED_BITSTREAM |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
15 #define FAST_GET_FIRST_VLC |
306 | 16 //#define DUMP_STREAM // only works with the ALT_BITSTREAM_READER |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
17 |
10 | 18 #ifdef HAVE_AV_CONFIG_H |
64 | 19 /* only include the following when compiling package */ |
408 | 20 #include "config.h" |
64 | 21 |
22 #include <stdlib.h> | |
23 #include <stdio.h> | |
24 #include <string.h> | |
25 #include <errno.h> | |
390
48e08d9871da
added proper memory handling functions - fixed include paths
glantau
parents:
370
diff
changeset
|
26 #include <math.h> |
64 | 27 |
28 #ifndef ENODATA | |
29 #define ENODATA 61 | |
10 | 30 #endif |
0 | 31 |
390
48e08d9871da
added proper memory handling functions - fixed include paths
glantau
parents:
370
diff
changeset
|
32 #endif /* HAVE_AV_CONFIG_H */ |
64 | 33 |
34 #ifdef CONFIG_WIN32 | |
35 | |
36 /* windows */ | |
37 | |
38 typedef unsigned short UINT16; | |
39 typedef signed short INT16; | |
40 typedef unsigned char UINT8; | |
41 typedef unsigned int UINT32; | |
42 typedef unsigned __int64 UINT64; | |
43 typedef signed char INT8; | |
44 typedef signed int INT32; | |
45 typedef signed __int64 INT64; | |
46 | |
47 typedef UINT8 uint8_t; | |
48 typedef INT8 int8_t; | |
49 typedef UINT16 uint16_t; | |
50 typedef INT16 int16_t; | |
51 typedef UINT32 uint32_t; | |
52 typedef INT32 int32_t; | |
426 | 53 typedef UINT64 uint64_t; |
54 typedef INT64 int64_t; | |
64 | 55 |
76 | 56 #ifndef __MINGW32__ |
64 | 57 #define INT64_C(c) (c ## i64) |
58 #define UINT64_C(c) (c ## i64) | |
59 | |
60 #define inline __inline | |
61 | |
76 | 62 #else |
63 #define INT64_C(c) (c ## LL) | |
64 #define UINT64_C(c) (c ## ULL) | |
65 #endif /* __MINGW32__ */ | |
66 | |
64 | 67 #define M_PI 3.14159265358979323846 |
68 #define M_SQRT2 1.41421356237309504880 /* sqrt(2) */ | |
69 | |
70 #ifdef _DEBUG | |
71 #define DEBUG | |
72 #endif | |
73 | |
74 #define snprintf _snprintf | |
75 | |
426 | 76 #else /* CONFIG_WIN32 */ |
64 | 77 |
78 /* unix */ | |
79 | |
488 | 80 #include <stdint.h> |
64 | 81 |
2 | 82 #ifndef __WINE_WINDEF16_H |
83 /* workaround for typedef conflict in MPlayer (wine typedefs) */ | |
0 | 84 typedef unsigned short UINT16; |
85 typedef signed short INT16; | |
86 #endif | |
87 | |
88 typedef unsigned char UINT8; | |
89 typedef unsigned int UINT32; | |
90 typedef unsigned long long UINT64; | |
91 typedef signed char INT8; | |
92 typedef signed int INT32; | |
93 typedef signed long long INT64; | |
94 | |
64 | 95 #ifdef HAVE_AV_CONFIG_H |
96 | |
97 #ifdef __FreeBSD__ | |
98 #include <sys/param.h> | |
99 #endif | |
100 | |
101 #ifndef INT64_C | |
102 #define INT64_C(c) (c ## LL) | |
103 #define UINT64_C(c) (c ## ULL) | |
104 #endif | |
105 | |
106 #ifdef USE_FASTMEMCPY | |
107 #include "fastmemcpy.h" | |
108 #endif | |
109 | |
76 | 110 #endif /* HAVE_AV_CONFIG_H */ |
111 | |
112 #endif /* !CONFIG_WIN32 */ | |
113 | |
488 | 114 #ifdef HAVE_AV_CONFIG_H |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
115 |
488 | 116 #include "bswap.h" |
426 | 117 |
432 | 118 #if defined(__MINGW32__) || defined(__CYGWIN__) || \ |
119 defined(__OS2__) || defined (__OpenBSD__) | |
120 #define MANGLE(a) "_" #a | |
121 #else | |
122 #define MANGLE(a) #a | |
123 #endif | |
124 | |
76 | 125 /* debug stuff */ |
126 | |
64 | 127 #ifndef DEBUG |
128 #define NDEBUG | |
129 #endif | |
130 #include <assert.h> | |
131 | |
76 | 132 /* dprintf macros */ |
133 #if defined(CONFIG_WIN32) && !defined(__MINGW32__) | |
134 | |
135 inline void dprintf(const char* fmt,...) {} | |
136 | |
137 #else | |
138 | |
139 #ifdef DEBUG | |
140 #define dprintf(fmt,args...) printf(fmt, ## args) | |
141 #else | |
142 #define dprintf(fmt,args...) | |
143 #endif | |
64 | 144 |
145 #endif /* !CONFIG_WIN32 */ | |
146 | |
423
ed6098c4216a
Add an av_abort macro that aborts, but also prints out the location of the abort.
philipjsg
parents:
408
diff
changeset
|
147 #define av_abort() do { fprintf(stderr, "Abort at %s:%d\n", __FILE__, __LINE__); abort(); } while (0) |
ed6098c4216a
Add an av_abort macro that aborts, but also prints out the location of the abort.
philipjsg
parents:
408
diff
changeset
|
148 |
344 | 149 /* assume b>0 */ |
150 #define ROUNDED_DIV(a,b) (((a)>0 ? (a) + ((b)>>1) : (a) - ((b)>>1))/(b)) | |
359 | 151 #define ABS(a) ((a) >= 0 ? (a) : (-(a))) |
451 | 152 #define MAX(a,b) ((a) > (b) ? (a) : (b)) |
153 #define MIN(a,b) ((a) > (b) ? (b) : (a)) | |
344 | 154 |
0 | 155 /* bit output */ |
156 | |
157 struct PutBitContext; | |
158 | |
159 typedef void (*WriteDataFunc)(void *, UINT8 *, int); | |
160 | |
161 typedef struct PutBitContext { | |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
162 #ifdef ALT_BITSTREAM_WRITER |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
163 UINT8 *buf, *buf_end; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
164 int index; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
165 #else |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
166 UINT32 bit_buf; |
238
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
167 int bit_left; |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
168 UINT8 *buf, *buf_ptr, *buf_end; |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
169 #endif |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
170 INT64 data_out_size; /* in bytes */ |
0 | 171 } PutBitContext; |
172 | |
173 void init_put_bits(PutBitContext *s, | |
174 UINT8 *buffer, int buffer_size, | |
175 void *opaque, | |
176 void (*write_data)(void *, UINT8 *, int)); | |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
177 |
85 | 178 INT64 get_bit_count(PutBitContext *s); /* XXX: change function name */ |
0 | 179 void align_put_bits(PutBitContext *s); |
180 void flush_put_bits(PutBitContext *s); | |
358 | 181 void put_string(PutBitContext * pbc, char *s); |
0 | 182 |
183 /* jpeg specific put_bits */ | |
184 void jflush_put_bits(PutBitContext *s); | |
185 | |
186 /* bit input */ | |
187 | |
188 typedef struct GetBitContext { | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
189 #ifdef ALT_BITSTREAM_READER |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
190 int index; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
191 UINT8 *buffer; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
192 #else |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
193 UINT32 bit_buf; |
0 | 194 int bit_cnt; |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
195 UINT8 *buf, *buf_ptr, *buf_end; |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
196 #endif |
277
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
197 int size; |
0 | 198 } GetBitContext; |
199 | |
290 | 200 static inline int get_bits_count(GetBitContext *s); |
201 | |
0 | 202 typedef struct VLC { |
203 int bits; | |
204 INT16 *table_codes; | |
205 INT8 *table_bits; | |
206 int table_size, table_allocated; | |
207 } VLC; | |
208 | |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
209 /* used to avoid missaligned exceptions on some archs (alpha, ...) */ |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
210 #ifdef ARCH_X86 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
211 #define unaligned32(a) (*(UINT32*)(a)) |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
212 #else |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
213 #ifdef __GNUC__ |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
214 static inline uint32_t unaligned32(const void *v) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
215 struct Unaligned { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
216 uint32_t i; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
217 } __attribute__((packed)); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
218 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
219 return ((const struct Unaligned *) v)->i; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
220 } |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
221 #elif defined(__DECC) |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
222 static inline uint32_t unaligned32(const void *v) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
223 return *(const __unaligned uint32_t *) v; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
224 } |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
225 #else |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
226 static inline uint32_t unaligned32(const void *v) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
227 return *(const uint32_t *) v; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
228 } |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
229 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
230 #endif //!ARCH_X86 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
231 |
238
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
232 #ifndef ALT_BITSTREAM_WRITER |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
233 static inline void put_bits(PutBitContext *s, int n, unsigned int value) |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
234 { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
235 unsigned int bit_buf; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
236 int bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
237 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
238 #ifdef STATS |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
239 st_out_bit_counts[st_current_index] += n; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
240 #endif |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
241 // printf("put_bits=%d %x\n", n, value); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
242 assert(n == 32 || value < (1U << n)); |
306 | 243 |
238
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
244 bit_buf = s->bit_buf; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
245 bit_left = s->bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
246 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
247 // printf("n=%d value=%x cnt=%d buf=%x\n", n, value, bit_cnt, bit_buf); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
248 /* XXX: optimize */ |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
249 if (n < bit_left) { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
250 bit_buf = (bit_buf<<n) | value; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
251 bit_left-=n; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
252 } else { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
253 bit_buf<<=bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
254 bit_buf |= value >> (n - bit_left); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
255 *(UINT32 *)s->buf_ptr = be2me_32(bit_buf); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
256 //printf("bitbuf = %08x\n", bit_buf); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
257 s->buf_ptr+=4; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
258 bit_left+=32 - n; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
259 bit_buf = value; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
260 } |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
261 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
262 s->bit_buf = bit_buf; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
263 s->bit_left = bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
264 } |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
265 #endif |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
266 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
267 |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
268 #ifdef ALT_BITSTREAM_WRITER |
235
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
269 static inline void put_bits(PutBitContext *s, int n, unsigned int value) |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
270 { |
235
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
271 #ifdef ALIGNED_BITSTREAM_WRITER |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
272 #ifdef ARCH_X86 |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
273 asm volatile( |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
274 "movl %0, %%ecx \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
275 "xorl %%eax, %%eax \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
276 "shrdl %%cl, %1, %%eax \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
277 "shrl %%cl, %1 \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
278 "movl %0, %%ecx \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
279 "shrl $3, %%ecx \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
280 "andl $0xFFFFFFFC, %%ecx \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
281 "bswapl %1 \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
282 "orl %1, (%2, %%ecx) \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
283 "bswapl %%eax \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
284 "addl %3, %0 \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
285 "movl %%eax, 4(%2, %%ecx) \n\t" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
286 : "=&r" (s->index), "=&r" (value) |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
287 : "r" (s->buf), "r" (n), "0" (s->index), "1" (value<<(-n)) |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
288 : "%eax", "%ecx" |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
289 ); |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
290 #else |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
291 int index= s->index; |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
292 uint32_t *ptr= ((uint32_t *)s->buf)+(index>>5); |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
293 |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
294 value<<= 32-n; |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
295 |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
296 ptr[0] |= be2me_32(value>>(index&31)); |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
297 ptr[1] = be2me_32(value<<(32-(index&31))); |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
298 //if(n>24) printf("%d %d\n", n, value); |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
299 index+= n; |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
300 s->index= index; |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
301 #endif |
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
302 #else //ALIGNED_BITSTREAM_WRITER |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
303 #ifdef ARCH_X86 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
304 asm volatile( |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
305 "movl $7, %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
306 "andl %0, %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
307 "addl %3, %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
308 "negl %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
309 "shll %%cl, %1 \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
310 "bswapl %1 \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
311 "movl %0, %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
312 "shrl $3, %%ecx \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
313 "orl %1, (%%ecx, %2) \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
314 "addl %3, %0 \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
315 "movl $0, 4(%%ecx, %2) \n\t" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
316 : "=&r" (s->index), "=&r" (value) |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
317 : "r" (s->buf), "r" (n), "0" (s->index), "1" (value) |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
318 : "%ecx" |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
319 ); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
320 #else |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
321 int index= s->index; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
322 uint32_t *ptr= (uint32_t*)(((uint8_t *)s->buf)+(index>>3)); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
323 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
324 ptr[0] |= be2me_32(value<<(32-n-(index&7) )); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
325 ptr[1] = 0; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
326 //if(n>24) printf("%d %d\n", n, value); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
327 index+= n; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
328 s->index= index; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
329 #endif |
235
41f0ef2cd942
aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
michaelni
parents:
234
diff
changeset
|
330 #endif //!ALIGNED_BITSTREAM_WRITER |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
331 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
332 #endif |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
333 |
238
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
334 #ifndef ALT_BITSTREAM_WRITER |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
335 /* for jpeg : escape 0xff with 0x00 after it */ |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
336 static inline void jput_bits(PutBitContext *s, int n, unsigned int value) |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
337 { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
338 unsigned int bit_buf, b; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
339 int bit_left, i; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
340 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
341 assert(n == 32 || value < (1U << n)); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
342 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
343 bit_buf = s->bit_buf; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
344 bit_left = s->bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
345 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
346 //printf("n=%d value=%x cnt=%d buf=%x\n", n, value, bit_cnt, bit_buf); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
347 /* XXX: optimize */ |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
348 if (n < bit_left) { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
349 bit_buf = (bit_buf<<n) | value; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
350 bit_left-=n; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
351 } else { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
352 bit_buf<<=bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
353 bit_buf |= value >> (n - bit_left); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
354 /* handle escape */ |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
355 for(i=0;i<4;i++) { |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
356 b = (bit_buf >> 24); |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
357 *(s->buf_ptr++) = b; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
358 if (b == 0xff) |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
359 *(s->buf_ptr++) = 0; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
360 bit_buf <<= 8; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
361 } |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
362 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
363 bit_left+= 32 - n; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
364 bit_buf = value; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
365 } |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
366 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
367 s->bit_buf = bit_buf; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
368 s->bit_left = bit_left; |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
369 } |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
370 #endif |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
371 |
99a9f903f0e3
optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
michaelni
parents:
235
diff
changeset
|
372 |
234
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
373 #ifdef ALT_BITSTREAM_WRITER |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
374 static inline void jput_bits(PutBitContext *s, int n, int value) |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
375 { |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
376 int index= s->index; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
377 uint32_t *ptr= (uint32_t*)(((uint8_t *)s->buf)+(index>>3)); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
378 int v= ptr[0]; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
379 //if(n>24) printf("%d %d\n", n, value); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
380 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
381 v |= be2me_32(value<<(32-n-(index&7) )); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
382 if(((v+0x01010101)^0xFFFFFFFF)&v&0x80808080) |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
383 { |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
384 /* handle idiotic (m)jpeg escapes */ |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
385 uint8_t *bPtr= (uint8_t*)ptr; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
386 int numChecked= ((index+n)>>3) - (index>>3); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
387 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
388 v= be2me_32(v); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
389 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
390 *(bPtr++)= v>>24; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
391 if((v&0xFF000000)==0xFF000000 && numChecked>0){ |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
392 *(bPtr++)= 0x00; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
393 index+=8; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
394 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
395 *(bPtr++)= (v>>16)&0xFF; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
396 if((v&0x00FF0000)==0x00FF0000 && numChecked>1){ |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
397 *(bPtr++)= 0x00; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
398 index+=8; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
399 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
400 *(bPtr++)= (v>>8)&0xFF; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
401 if((v&0x0000FF00)==0x0000FF00 && numChecked>2){ |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
402 *(bPtr++)= 0x00; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
403 index+=8; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
404 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
405 *(bPtr++)= v&0xFF; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
406 if((v&0x000000FF)==0x000000FF && numChecked>3){ |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
407 *(bPtr++)= 0x00; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
408 index+=8; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
409 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
410 *((uint32_t*)bPtr)= 0; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
411 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
412 else |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
413 { |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
414 ptr[0] = v; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
415 ptr[1] = 0; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
416 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
417 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
418 index+= n; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
419 s->index= index; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
420 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
421 #endif |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
422 |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
423 static inline uint8_t* pbBufPtr(PutBitContext *s) |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
424 { |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
425 #ifdef ALT_BITSTREAM_WRITER |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
426 return s->buf + (s->index>>3); |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
427 #else |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
428 return s->buf_ptr; |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
429 #endif |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
430 } |
5fc0c3af3fe4
alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
michaelni
parents:
213
diff
changeset
|
431 |
0 | 432 void init_get_bits(GetBitContext *s, |
433 UINT8 *buffer, int buffer_size); | |
434 | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
435 #ifndef ALT_BITSTREAM_READER |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
436 unsigned int get_bits_long(GetBitContext *s, int n); |
144 | 437 unsigned int show_bits_long(GetBitContext *s, int n); |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
438 #endif |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
439 |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
440 static inline unsigned int get_bits(GetBitContext *s, int n){ |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
441 #ifdef ALT_BITSTREAM_READER |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
442 #ifdef ALIGNED_BITSTREAM |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
443 int index= s->index; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
444 uint32_t result1= be2me_32( ((uint32_t *)s->buffer)[index>>5] ); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
445 uint32_t result2= be2me_32( ((uint32_t *)s->buffer)[(index>>5) + 1] ); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
446 #ifdef ARCH_X86 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
447 asm ("shldl %%cl, %2, %0\n\t" |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
448 : "=r" (result1) |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
449 : "0" (result1), "r" (result2), "c" (index)); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
450 #else |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
451 result1<<= (index&0x1F); |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
452 result2= (result2>>1) >> (31-(index&0x1F)); |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
453 result1|= result2; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
454 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
455 result1>>= 32 - n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
456 index+= n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
457 s->index= index; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
458 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
459 return result1; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
460 #else //ALIGNED_BITSTREAM |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
461 int index= s->index; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
462 uint32_t result= be2me_32( unaligned32( ((uint8_t *)s->buffer)+(index>>3) ) ); |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
463 |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
464 result<<= (index&0x07); |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
465 result>>= 32 - n; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
466 index+= n; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
467 s->index= index; |
306 | 468 #ifdef DUMP_STREAM |
469 while(n){ | |
470 printf("%d", (result>>(n-1))&1); | |
471 n--; | |
472 } | |
473 printf(" "); | |
474 #endif | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
475 return result; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
476 #endif //!ALIGNED_BITSTREAM |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
477 #else //ALT_BITSTREAM_READER |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
478 if(s->bit_cnt>=n){ |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
479 /* most common case here */ |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
480 unsigned int val = s->bit_buf >> (32 - n); |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
481 s->bit_buf <<= n; |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
482 s->bit_cnt -= n; |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
483 #ifdef STATS |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
484 st_bit_counts[st_current_index] += n; |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
485 #endif |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
486 return val; |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
487 } |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
488 return get_bits_long(s,n); |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
489 #endif //!ALT_BITSTREAM_READER |
20
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
490 } |
907b67420d84
inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
arpi_esp
parents:
10
diff
changeset
|
491 |
21 | 492 static inline unsigned int get_bits1(GetBitContext *s){ |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
493 #ifdef ALT_BITSTREAM_READER |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
494 int index= s->index; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
495 uint8_t result= s->buffer[ index>>3 ]; |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
496 result<<= (index&0x07); |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
497 result>>= 8 - 1; |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
498 index++; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
499 s->index= index; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
500 |
306 | 501 #ifdef DUMP_STREAM |
502 printf("%d ", result); | |
503 #endif | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
504 return result; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
505 #else |
21 | 506 if(s->bit_cnt>0){ |
507 /* most common case here */ | |
508 unsigned int val = s->bit_buf >> 31; | |
509 s->bit_buf <<= 1; | |
510 s->bit_cnt--; | |
511 #ifdef STATS | |
512 st_bit_counts[st_current_index]++; | |
513 #endif | |
514 return val; | |
515 } | |
516 return get_bits_long(s,1); | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
517 #endif |
21 | 518 } |
519 | |
144 | 520 /* This function is identical to get_bits(), the only */ |
521 /* diference is that it doesn't touch the buffer */ | |
522 /* it is usefull to see the buffer. */ | |
523 static inline unsigned int show_bits(GetBitContext *s, int n) | |
524 { | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
525 #ifdef ALT_BITSTREAM_READER |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
526 #ifdef ALIGNED_BITSTREAM |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
527 int index= s->index; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
528 uint32_t result1= be2me_32( ((uint32_t *)s->buffer)[index>>5] ); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
529 uint32_t result2= be2me_32( ((uint32_t *)s->buffer)[(index>>5) + 1] ); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
530 #ifdef ARCH_X86 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
531 asm ("shldl %%cl, %2, %0\n\t" |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
532 : "=r" (result1) |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
533 : "0" (result1), "r" (result2), "c" (index)); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
534 #else |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
535 result1<<= (index&0x1F); |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
536 result2= (result2>>1) >> (31-(index&0x1F)); |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
537 result1|= result2; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
538 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
539 result1>>= 32 - n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
540 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
541 return result1; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
542 #else //ALIGNED_BITSTREAM |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
543 int index= s->index; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
544 uint32_t result= be2me_32( unaligned32( ((uint8_t *)s->buffer)+(index>>3) ) ); |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
545 |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
546 result<<= (index&0x07); |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
547 result>>= 32 - n; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
548 |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
549 return result; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
550 #endif //!ALIGNED_BITSTREAM |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
551 #else //ALT_BITSTREAM_READER |
144 | 552 if(s->bit_cnt>=n) { |
553 /* most common case here */ | |
554 unsigned int val = s->bit_buf >> (32 - n); | |
555 return val; | |
556 } | |
557 return show_bits_long(s,n); | |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
558 #endif //!ALT_BITSTREAM_READER |
144 | 559 } |
560 | |
290 | 561 static inline int show_aligned_bits(GetBitContext *s, int offset, int n) |
562 { | |
563 #ifdef ALT_BITSTREAM_READER | |
564 #ifdef ALIGNED_BITSTREAM | |
565 int index= (s->index + offset + 7)&(~7); | |
566 uint32_t result1= be2me_32( ((uint32_t *)s->buffer)[index>>5] ); | |
567 uint32_t result2= be2me_32( ((uint32_t *)s->buffer)[(index>>5) + 1] ); | |
568 #ifdef ARCH_X86 | |
569 asm ("shldl %%cl, %2, %0\n\t" | |
570 : "=r" (result1) | |
571 : "0" (result1), "r" (result2), "c" (index)); | |
572 #else | |
573 result1<<= (index&0x1F); | |
574 result2= (result2>>1) >> (31-(index&0x1F)); | |
575 result1|= result2; | |
576 #endif | |
577 result1>>= 32 - n; | |
578 | |
579 return result1; | |
580 #else //ALIGNED_BITSTREAM | |
581 int index= (s->index + offset + 7)>>3; | |
582 uint32_t result= be2me_32( unaligned32( ((uint8_t *)s->buffer)+index ) ); | |
583 | |
584 result>>= 32 - n; | |
585 | |
586 return result; | |
587 #endif //!ALIGNED_BITSTREAM | |
588 #else //ALT_BITSTREAM_READER | |
589 int index= (get_bits_count(s) + offset + 7)>>3; | |
590 uint32_t result= be2me_32( unaligned32( ((uint8_t *)s->buf)+index ) ); | |
591 | |
592 result>>= 32 - n; | |
593 //printf(" %X %X %d \n", (int)(((uint8_t *)s->buf)+index ), (int)s->buf_ptr, s->bit_cnt); | |
594 return result; | |
595 #endif //!ALT_BITSTREAM_READER | |
596 } | |
597 | |
21 | 598 static inline void skip_bits(GetBitContext *s, int n){ |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
599 #ifdef ALT_BITSTREAM_READER |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
600 s->index+= n; |
306 | 601 #ifdef DUMP_STREAM |
602 { | |
603 int result; | |
604 s->index-= n; | |
605 result= get_bits(s, n); | |
606 } | |
607 #endif | |
608 | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
609 #else |
21 | 610 if(s->bit_cnt>=n){ |
611 /* most common case here */ | |
612 s->bit_buf <<= n; | |
613 s->bit_cnt -= n; | |
614 #ifdef STATS | |
615 st_bit_counts[st_current_index] += n; | |
616 #endif | |
617 } else { | |
618 get_bits_long(s,n); | |
619 } | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
620 #endif |
21 | 621 } |
622 | |
623 static inline void skip_bits1(GetBitContext *s){ | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
624 #ifdef ALT_BITSTREAM_READER |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
625 s->index++; |
306 | 626 #ifdef DUMP_STREAM |
627 s->index--; | |
628 printf("%d ", get_bits1(s)); | |
629 #endif | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
630 #else |
21 | 631 if(s->bit_cnt>0){ |
632 /* most common case here */ | |
633 s->bit_buf <<= 1; | |
634 s->bit_cnt--; | |
635 #ifdef STATS | |
636 st_bit_counts[st_current_index]++; | |
637 #endif | |
638 } else { | |
639 get_bits_long(s,1); | |
640 } | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
641 #endif |
21 | 642 } |
643 | |
85 | 644 static inline int get_bits_count(GetBitContext *s) |
645 { | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
646 #ifdef ALT_BITSTREAM_READER |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
647 return s->index; |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
648 #else |
85 | 649 return (s->buf_ptr - s->buf) * 8 - s->bit_cnt; |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
650 #endif |
85 | 651 } |
21 | 652 |
264
28c5c62b1c4c
support decoding (with mplayer) of 3 .mp4 files from mphq
michaelni
parents:
238
diff
changeset
|
653 int check_marker(GetBitContext *s, char *msg); |
0 | 654 void align_get_bits(GetBitContext *s); |
655 int init_vlc(VLC *vlc, int nb_bits, int nb_codes, | |
656 const void *bits, int bits_wrap, int bits_size, | |
657 const void *codes, int codes_wrap, int codes_size); | |
658 void free_vlc(VLC *vlc); | |
659 | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
660 #ifdef ALT_BITSTREAM_READER |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
661 #ifdef ALIGNED_BITSTREAM |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
662 #ifdef ARCH_X86 |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
663 #define SHOW_BITS(s, val, n) \ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
664 val= be2me_32( ((uint32_t *)(s)->buffer)[bit_cnt>>5] );\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
665 {uint32_t result2= be2me_32( ((uint32_t *)(s)->buffer)[(bit_cnt>>5) + 1] );\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
666 asm ("shldl %%cl, %2, %0\n\t"\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
667 : "=r" (val)\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
668 : "0" (val), "r" (result2), "c" (bit_cnt));\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
669 ((uint32_t)val)>>= 32 - n;} |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
670 #else //ARCH_X86 |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
671 #define SHOW_BITS(s, val, n) \ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
672 val= be2me_32( ((uint32_t *)(s)->buffer)[bit_cnt>>5] );\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
673 {uint32_t result2= be2me_32( ((uint32_t *)(s)->buffer)[(bit_cnt>>5) + 1] );\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
674 val<<= (bit_cnt&0x1F);\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
675 result2= (result2>>1) >> (31-(bit_cnt&0x1F));\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
676 val|= result2;\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
677 ((uint32_t)val)>>= 32 - n;} |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
678 #endif //!ARCH_X86 |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
679 #else //ALIGNED_BITSTREAM |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
680 #define SHOW_BITS(s, val, n) \ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
681 val= be2me_32( unaligned32( ((uint8_t *)(s)->buffer)+(bit_cnt>>3) ) );\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
682 val<<= (bit_cnt&0x07);\ |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
683 ((uint32_t)val)>>= 32 - n; |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
684 #endif // !ALIGNED_BITSTREAM |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
685 #define FLUSH_BITS(n) bit_cnt+=n; |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
686 #define SAVE_BITS(s) bit_cnt= (s)->index; |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
687 #define RESTORE_BITS(s) (s)->index= bit_cnt; |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
688 #else |
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
689 |
0 | 690 /* macro to go faster */ |
691 /* n must be <= 24 */ | |
692 /* XXX: optimize buffer end test */ | |
693 #define SHOW_BITS(s, val, n)\ | |
694 {\ | |
695 if (bit_cnt < n && buf_ptr < (s)->buf_end) {\ | |
696 bit_buf |= *buf_ptr++ << (24 - bit_cnt);\ | |
697 bit_cnt += 8;\ | |
698 if (bit_cnt < n && buf_ptr < (s)->buf_end) {\ | |
699 bit_buf |= *buf_ptr++ << (24 - bit_cnt);\ | |
700 bit_cnt += 8;\ | |
701 if (bit_cnt < n && buf_ptr < (s)->buf_end) {\ | |
702 bit_buf |= *buf_ptr++ << (24 - bit_cnt);\ | |
703 bit_cnt += 8;\ | |
704 }\ | |
705 }\ | |
706 }\ | |
707 val = bit_buf >> (32 - n);\ | |
708 } | |
709 | |
710 /* SHOW_BITS with n1 >= n must be been done before */ | |
711 #define FLUSH_BITS(n)\ | |
712 {\ | |
713 bit_buf <<= n;\ | |
714 bit_cnt -= n;\ | |
715 } | |
716 | |
717 #define SAVE_BITS(s) \ | |
718 {\ | |
719 bit_cnt = (s)->bit_cnt;\ | |
720 bit_buf = (s)->bit_buf;\ | |
721 buf_ptr = (s)->buf_ptr;\ | |
722 } | |
723 | |
724 #define RESTORE_BITS(s) \ | |
725 {\ | |
726 (s)->buf_ptr = buf_ptr;\ | |
727 (s)->bit_buf = bit_buf;\ | |
728 (s)->bit_cnt = bit_cnt;\ | |
729 } | |
192
1e5f64be86fc
another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
151
diff
changeset
|
730 #endif // !ALT_BITSTREAM_READER |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
731 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
732 static inline int get_vlc(GetBitContext *s, VLC *vlc) |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
733 { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
734 int code, n, nb_bits, index; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
735 INT16 *table_codes; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
736 INT8 *table_bits; |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
737 int bit_cnt; |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
738 #ifndef ALT_BITSTREAM_READER |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
739 UINT32 bit_buf; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
740 UINT8 *buf_ptr; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
741 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
742 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
743 SAVE_BITS(s); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
744 nb_bits = vlc->bits; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
745 table_codes = vlc->table_codes; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
746 table_bits = vlc->table_bits; |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
747 |
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
748 #ifdef FAST_GET_FIRST_VLC |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
749 SHOW_BITS(s, index, nb_bits); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
750 code = table_codes[index]; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
751 n = table_bits[index]; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
752 if (n > 0) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
753 /* most common case (90%)*/ |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
754 FLUSH_BITS(n); |
306 | 755 #ifdef DUMP_STREAM |
756 { | |
757 int n= bit_cnt - s->index; | |
758 skip_bits(s, n); | |
759 RESTORE_BITS(s); | |
760 } | |
761 #endif | |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
762 RESTORE_BITS(s); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
763 return code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
764 } else if (n == 0) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
765 return -1; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
766 } else { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
767 FLUSH_BITS(nb_bits); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
768 nb_bits = -n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
769 table_codes = vlc->table_codes + code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
770 table_bits = vlc->table_bits + code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
771 } |
199
0f1dba8fc617
(commited by michael / arpi was crazy enough to give me his password)
arpi_esp
parents:
193
diff
changeset
|
772 #endif |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
773 for(;;) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
774 SHOW_BITS(s, index, nb_bits); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
775 code = table_codes[index]; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
776 n = table_bits[index]; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
777 if (n > 0) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
778 /* most common case */ |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
779 FLUSH_BITS(n); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
780 #ifdef STATS |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
781 st_bit_counts[st_current_index] += n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
782 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
783 break; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
784 } else if (n == 0) { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
785 return -1; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
786 } else { |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
787 FLUSH_BITS(nb_bits); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
788 #ifdef STATS |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
789 st_bit_counts[st_current_index] += nb_bits; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
790 #endif |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
791 nb_bits = -n; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
792 table_codes = vlc->table_codes + code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
793 table_bits = vlc->table_bits + code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
794 } |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
795 } |
306 | 796 #ifdef DUMP_STREAM |
797 { | |
798 int n= bit_cnt - s->index; | |
799 skip_bits(s, n); | |
800 RESTORE_BITS(s); | |
801 } | |
802 #endif | |
193
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
803 RESTORE_BITS(s); |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
804 return code; |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
805 } |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
806 |
b691dd3e9088
aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni@gmx.at>
arpi_esp
parents:
192
diff
changeset
|
807 |
0 | 808 /* define it to include statistics code (useful only for optimizing |
809 codec efficiency */ | |
810 //#define STATS | |
811 | |
812 #ifdef STATS | |
813 | |
814 enum { | |
815 ST_UNKNOWN, | |
816 ST_DC, | |
817 ST_INTRA_AC, | |
818 ST_INTER_AC, | |
819 ST_INTRA_MB, | |
820 ST_INTER_MB, | |
821 ST_MV, | |
822 ST_NB, | |
823 }; | |
824 | |
825 extern int st_current_index; | |
826 extern unsigned int st_bit_counts[ST_NB]; | |
827 extern unsigned int st_out_bit_counts[ST_NB]; | |
828 | |
829 void print_stats(void); | |
830 #endif | |
831 | |
832 /* misc math functions */ | |
833 | |
151
ae0516eadae2
fixed gcc-3.0.x compilation (by Michael Niedermayer)
nickols_k
parents:
144
diff
changeset
|
834 static inline int av_log2(unsigned int v) |
0 | 835 { |
836 int n; | |
837 | |
838 n = 0; | |
839 if (v & 0xffff0000) { | |
840 v >>= 16; | |
841 n += 16; | |
842 } | |
843 if (v & 0xff00) { | |
844 v >>= 8; | |
845 n += 8; | |
846 } | |
847 if (v & 0xf0) { | |
848 v >>= 4; | |
849 n += 4; | |
850 } | |
851 if (v & 0xc) { | |
852 v >>= 2; | |
853 n += 2; | |
854 } | |
855 if (v & 0x2) { | |
856 n++; | |
857 } | |
858 return n; | |
859 } | |
860 | |
277
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
861 /* median of 3 */ |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
862 static inline int mid_pred(int a, int b, int c) |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
863 { |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
864 int vmin, vmax; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
865 vmax = vmin = a; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
866 if (b < vmin) |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
867 vmin = b; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
868 else |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
869 vmax = b; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
870 |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
871 if (c < vmin) |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
872 vmin = c; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
873 else if (c > vmax) |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
874 vmax = c; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
875 |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
876 return a + b + c - vmin - vmax; |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
877 } |
5cb2978e701f
new motion estimation (epzs) not complete yet but allready pretty good :)
michaelni
parents:
264
diff
changeset
|
878 |
327 | 879 static inline int clip(int a, int amin, int amax) |
880 { | |
881 if (a < amin) | |
882 return amin; | |
883 else if (a > amax) | |
884 return amax; | |
885 else | |
886 return a; | |
887 } | |
888 | |
324 | 889 /* math */ |
890 int ff_gcd(int a, int b); | |
891 | |
451 | 892 static inline int ff_sqrt(int a) |
893 { | |
894 int ret=0; | |
895 int s; | |
896 int ret_sq=0; | |
897 | |
898 for(s=15; s>=0; s--){ | |
899 int b= ret_sq + (1<<(s*2)) + (ret<<s)*2; | |
900 if(b<=a){ | |
901 ret_sq=b; | |
902 ret+= 1<<s; | |
903 } | |
904 } | |
905 return ret; | |
906 } | |
907 #if __CPU__ >= 686 && !defined(RUNTIME_CPUDETECT) | |
908 #define COPY3_IF_LT(x,y,a,b,c,d)\ | |
909 asm volatile (\ | |
910 "cmpl %0, %3 \n\t"\ | |
911 "cmovl %3, %0 \n\t"\ | |
912 "cmovl %4, %1 \n\t"\ | |
913 "cmovl %5, %2 \n\t"\ | |
914 : "+r" (x), "+r" (a), "+r" (c)\ | |
915 : "r" (y), "r" (b), "r" (d)\ | |
916 ); | |
917 #else | |
918 #define COPY3_IF_LT(x,y,a,b,c,d)\ | |
919 if((y)<(x)){\ | |
920 (x)=(y);\ | |
921 (a)=(b);\ | |
922 (c)=(d);\ | |
923 } | |
924 #endif | |
925 | |
370
0eca28d16cbd
clamp intra matrix to 8bit for mjpeg (workaround for qscale>=25)
al3x
parents:
359
diff
changeset
|
926 #define CLAMP_TO_8BIT(d) ((d > 0xff) ? 0xff : (d < 0) ? 0 : d) |
0eca28d16cbd
clamp intra matrix to 8bit for mjpeg (workaround for qscale>=25)
al3x
parents:
359
diff
changeset
|
927 |
488 | 928 #endif /* HAVE_AV_CONFIG_H */ |
929 | |
930 #endif /* COMMON_H */ |