WarpZones:
[divverent/darkplaces.git] / cap_ogg.c
1 #ifndef _MSC_VER
2 #include <stdint.h>
3 #endif
4 #include <sys/types.h>
5
6 #include "quakedef.h"
7 #include "client.h"
8 #include "cap_ogg.h"
9
10 // video capture cvars
11 static cvar_t cl_capturevideo_ogg_theora_quality = {CVAR_SAVE, "cl_capturevideo_ogg_theora_quality", "32", "video quality factor (0 to 63), or -1 to use bitrate only; higher is better"};
12 static cvar_t cl_capturevideo_ogg_theora_bitrate = {CVAR_SAVE, "cl_capturevideo_ogg_theora_bitrate", "-1", "video bitrate (45 to 2000 kbps), or -1 to use quality only; higher is better"};
13 static cvar_t cl_capturevideo_ogg_theora_keyframe_bitrate_multiplier = {CVAR_SAVE, "cl_capturevideo_ogg_theora_keyframe_bitrate_multiplier", "1.5", "how much more bit rate to use for keyframes, specified as a factor of at least 1"};
14 static cvar_t cl_capturevideo_ogg_theora_keyframe_maxinterval = {CVAR_SAVE, "cl_capturevideo_ogg_theora_keyframe_maxinterval", "64", "maximum keyframe interval (1 to 1000)"};
15 static cvar_t cl_capturevideo_ogg_theora_keyframe_mininterval = {CVAR_SAVE, "cl_capturevideo_ogg_theora_keyframe_mininterval", "8", "minimum keyframe interval (1 to 1000)"};
16 static cvar_t cl_capturevideo_ogg_theora_keyframe_auto_threshold = {CVAR_SAVE, "cl_capturevideo_ogg_theora_keyframe_auto_threshold", "80", "threshold for key frame decision (0 to 100)"};
17 static cvar_t cl_capturevideo_ogg_theora_noise_sensitivity = {CVAR_SAVE, "cl_capturevideo_ogg_theora_noise_sensitivity", "1", "video noise sensitivity (0 to 6); lower is better"};
18 static cvar_t cl_capturevideo_ogg_theora_sharpness = {CVAR_SAVE, "cl_capturevideo_ogg_theora_sharpness", "0", "sharpness (0 to 2); lower is sharper"};
19 static cvar_t cl_capturevideo_ogg_vorbis_quality = {CVAR_SAVE, "cl_capturevideo_ogg_vorbis_quality", "3", "audio quality (-1 to 10); higher is better"};
20
21 // ogg.h stuff
22 #ifdef _MSC_VER
23 typedef __int16 ogg_int16_t;
24 typedef unsigned __int16 ogg_uint16_t;
25 typedef __int32 ogg_int32_t;
26 typedef unsigned __int32 ogg_uint32_t;
27 typedef __int64 ogg_int64_t;
28 #else
29 typedef int16_t ogg_int16_t;
30 typedef uint16_t ogg_uint16_t;
31 typedef int32_t ogg_int32_t;
32 typedef uint32_t ogg_uint32_t;
33 typedef int64_t ogg_int64_t;
34 #endif
35
36 typedef struct {
37   long endbyte;
38   int  endbit;
39
40   unsigned char *buffer;
41   unsigned char *ptr;
42   long storage;
43 } oggpack_buffer;
44
45 /* ogg_page is used to encapsulate the data in one Ogg bitstream page *****/
46
47 typedef struct {
48   unsigned char *header;
49   long header_len;
50   unsigned char *body;
51   long body_len;
52 } ogg_page;
53
54 /* ogg_stream_state contains the current encode/decode state of a logical
55    Ogg bitstream **********************************************************/
56
57 typedef struct {
58   unsigned char   *body_data;    /* bytes from packet bodies */
59   long    body_storage;          /* storage elements allocated */
60   long    body_fill;             /* elements stored; fill mark */
61   long    body_returned;         /* elements of fill returned */
62
63
64   int     *lacing_vals;      /* The values that will go to the segment table */
65   ogg_int64_t *granule_vals; /* granulepos values for headers. Not compact
66                                 this way, but it is simple coupled to the
67                                 lacing fifo */
68   long    lacing_storage;
69   long    lacing_fill;
70   long    lacing_packet;
71   long    lacing_returned;
72
73   unsigned char    header[282];      /* working space for header encode */
74   int              header_fill;
75
76   int     e_o_s;          /* set when we have buffered the last packet in the
77                              logical bitstream */
78   int     b_o_s;          /* set after we've written the initial page
79                              of a logical bitstream */
80   long    serialno;
81   long    pageno;
82   ogg_int64_t  packetno;      /* sequence number for decode; the framing
83                              knows where there's a hole in the data,
84                              but we need coupling so that the codec
85                              (which is in a seperate abstraction
86                              layer) also knows about the gap */
87   ogg_int64_t   granulepos;
88
89 } ogg_stream_state;
90
91 /* ogg_packet is used to encapsulate the data and metadata belonging
92    to a single raw Ogg/Vorbis packet *************************************/
93
94 typedef struct {
95   unsigned char *packet;
96   long  bytes;
97   long  b_o_s;
98   long  e_o_s;
99
100   ogg_int64_t  granulepos;
101   
102   ogg_int64_t  packetno;     /* sequence number for decode; the framing
103                                 knows where there's a hole in the data,
104                                 but we need coupling so that the codec
105                                 (which is in a seperate abstraction
106                                 layer) also knows about the gap */
107 } ogg_packet;
108
109 typedef struct {
110   unsigned char *data;
111   int storage;
112   int fill;
113   int returned;
114
115   int unsynced;
116   int headerbytes;
117   int bodybytes;
118 } ogg_sync_state;
119
120 /* Ogg BITSTREAM PRIMITIVES: encoding **************************/
121
122 static int      (*qogg_stream_packetin) (ogg_stream_state *os, ogg_packet *op);
123 static int      (*qogg_stream_pageout) (ogg_stream_state *os, ogg_page *og);
124 static int      (*qogg_stream_flush) (ogg_stream_state *os, ogg_page *og);
125
126 /* Ogg BITSTREAM PRIMITIVES: general ***************************/
127
128 static int      (*qogg_stream_init) (ogg_stream_state *os,int serialno);
129 static int      (*qogg_stream_clear) (ogg_stream_state *os);
130 static ogg_int64_t  (*qogg_page_granulepos) (ogg_page *og);
131
132 // end of ogg.h stuff
133
134 // vorbis/codec.h stuff
135 typedef struct vorbis_info{
136   int version;
137   int channels;
138   long rate;
139
140   /* The below bitrate declarations are *hints*.
141      Combinations of the three values carry the following implications:
142
143      all three set to the same value:
144        implies a fixed rate bitstream
145      only nominal set:
146        implies a VBR stream that averages the nominal bitrate.  No hard
147        upper/lower limit
148      upper and or lower set:
149        implies a VBR bitstream that obeys the bitrate limits. nominal
150        may also be set to give a nominal rate.
151      none set:
152        the coder does not care to speculate.
153   */
154
155   long bitrate_upper;
156   long bitrate_nominal;
157   long bitrate_lower;
158   long bitrate_window;
159
160   void *codec_setup;
161 } vorbis_info;
162
163 /* vorbis_dsp_state buffers the current vorbis audio
164    analysis/synthesis state.  The DSP state belongs to a specific
165    logical bitstream ****************************************************/
166 typedef struct vorbis_dsp_state{
167   int analysisp;
168   vorbis_info *vi;
169
170   float **pcm;
171   float **pcmret;
172   int      pcm_storage;
173   int      pcm_current;
174   int      pcm_returned;
175
176   int  preextrapolate;
177   int  eofflag;
178
179   long lW;
180   long W;
181   long nW;
182   long centerW;
183
184   ogg_int64_t granulepos;
185   ogg_int64_t sequence;
186
187   ogg_int64_t glue_bits;
188   ogg_int64_t time_bits;
189   ogg_int64_t floor_bits;
190   ogg_int64_t res_bits;
191
192   void       *backend_state;
193 } vorbis_dsp_state;
194
195 typedef struct vorbis_block{
196   /* necessary stream state for linking to the framing abstraction */
197   float  **pcm;       /* this is a pointer into local storage */
198   oggpack_buffer opb;
199
200   long  lW;
201   long  W;
202   long  nW;
203   int   pcmend;
204   int   mode;
205
206   int         eofflag;
207   ogg_int64_t granulepos;
208   ogg_int64_t sequence;
209   vorbis_dsp_state *vd; /* For read-only access of configuration */
210
211   /* local storage to avoid remallocing; it's up to the mapping to
212      structure it */
213   void               *localstore;
214   long                localtop;
215   long                localalloc;
216   long                totaluse;
217   struct alloc_chain *reap;
218
219   /* bitmetrics for the frame */
220   long glue_bits;
221   long time_bits;
222   long floor_bits;
223   long res_bits;
224
225   void *internal;
226
227 } vorbis_block;
228
229 /* vorbis_block is a single block of data to be processed as part of
230 the analysis/synthesis stream; it belongs to a specific logical
231 bitstream, but is independant from other vorbis_blocks belonging to
232 that logical bitstream. *************************************************/
233
234 struct alloc_chain{
235   void *ptr;
236   struct alloc_chain *next;
237 };
238
239 /* vorbis_info contains all the setup information specific to the
240    specific compression/decompression mode in progress (eg,
241    psychoacoustic settings, channel setup, options, codebook
242    etc). vorbis_info and substructures are in backends.h.
243 *********************************************************************/
244
245 /* the comments are not part of vorbis_info so that vorbis_info can be
246    static storage */
247 typedef struct vorbis_comment{
248   /* unlimited user comment fields.  libvorbis writes 'libvorbis'
249      whatever vendor is set to in encode */
250   char **user_comments;
251   int   *comment_lengths;
252   int    comments;
253   char  *vendor;
254
255 } vorbis_comment;
256
257
258 /* libvorbis encodes in two abstraction layers; first we perform DSP
259    and produce a packet (see docs/analysis.txt).  The packet is then
260    coded into a framed OggSquish bitstream by the second layer (see
261    docs/framing.txt).  Decode is the reverse process; we sync/frame
262    the bitstream and extract individual packets, then decode the
263    packet back into PCM audio.
264
265    The extra framing/packetizing is used in streaming formats, such as
266    files.  Over the net (such as with UDP), the framing and
267    packetization aren't necessary as they're provided by the transport
268    and the streaming layer is not used */
269
270 /* Vorbis PRIMITIVES: general ***************************************/
271
272 static void     (*qvorbis_info_init) (vorbis_info *vi);
273 static void     (*qvorbis_info_clear) (vorbis_info *vi);
274 static void     (*qvorbis_comment_init) (vorbis_comment *vc);
275 static void     (*qvorbis_comment_clear) (vorbis_comment *vc);
276
277 static int      (*qvorbis_block_init) (vorbis_dsp_state *v, vorbis_block *vb);
278 static int      (*qvorbis_block_clear) (vorbis_block *vb);
279 static void     (*qvorbis_dsp_clear) (vorbis_dsp_state *v);
280 static double   (*qvorbis_granule_time) (vorbis_dsp_state *v,
281                                     ogg_int64_t granulepos);
282
283 /* Vorbis PRIMITIVES: analysis/DSP layer ****************************/
284
285 static int      (*qvorbis_analysis_init) (vorbis_dsp_state *v,vorbis_info *vi);
286 static int      (*qvorbis_commentheader_out) (vorbis_comment *vc, ogg_packet *op);
287 static int      (*qvorbis_analysis_headerout) (vorbis_dsp_state *v,
288                                           vorbis_comment *vc,
289                                           ogg_packet *op,
290                                           ogg_packet *op_comm,
291                                           ogg_packet *op_code);
292 static float ** (*qvorbis_analysis_buffer) (vorbis_dsp_state *v,int vals);
293 static int      (*qvorbis_analysis_wrote) (vorbis_dsp_state *v,int vals);
294 static int      (*qvorbis_analysis_blockout) (vorbis_dsp_state *v,vorbis_block *vb);
295 static int      (*qvorbis_analysis) (vorbis_block *vb,ogg_packet *op);
296
297 static int      (*qvorbis_bitrate_addblock) (vorbis_block *vb);
298 static int      (*qvorbis_bitrate_flushpacket) (vorbis_dsp_state *vd,
299                                            ogg_packet *op);
300
301 // end of vorbis/codec.h stuff
302
303 // vorbisenc.h stuff
304 static int (*qvorbis_encode_init_vbr) (vorbis_info *vi,
305                                   long channels,
306                                   long rate,
307
308                                   float base_quality /* quality level from 0. (lo) to 1. (hi) */
309                                   );
310 // end of vorbisenc.h stuff
311
312 // theora.h stuff
313 typedef struct {
314     int   y_width;      /**< Width of the Y' luminance plane */
315     int   y_height;     /**< Height of the luminance plane */
316     int   y_stride;     /**< Offset in bytes between successive rows */
317
318     int   uv_width;     /**< Width of the Cb and Cr chroma planes */
319     int   uv_height;    /**< Height of the chroma planes */
320     int   uv_stride;    /**< Offset between successive chroma rows */
321     unsigned char *y;   /**< Pointer to start of luminance data */
322     unsigned char *u;   /**< Pointer to start of Cb data */
323     unsigned char *v;   /**< Pointer to start of Cr data */
324
325 } yuv_buffer;
326
327 /**
328  * A Colorspace.
329  */
330 typedef enum {
331   OC_CS_UNSPECIFIED,    /**< The colorspace is unknown or unspecified */
332   OC_CS_ITU_REC_470M,   /**< This is the best option for 'NTSC' content */
333   OC_CS_ITU_REC_470BG,  /**< This is the best option for 'PAL' content */
334   OC_CS_NSPACES         /**< This marks the end of the defined colorspaces */
335 } theora_colorspace;
336
337 /**
338  * A Chroma subsampling
339  *
340  * These enumerate the available chroma subsampling options supported
341  * by the theora format. See Section 4.4 of the specification for
342  * exact definitions.
343  */
344 typedef enum {
345   OC_PF_420,    /**< Chroma subsampling by 2 in each direction (4:2:0) */
346   OC_PF_RSVD,   /**< Reserved value */
347   OC_PF_422,    /**< Horizonatal chroma subsampling by 2 (4:2:2) */
348   OC_PF_444,    /**< No chroma subsampling at all (4:4:4) */
349 } theora_pixelformat;
350 /**
351  * Theora bitstream info.
352  * Contains the basic playback parameters for a stream,
353  * corresponding to the initial 'info' header packet.
354  * 
355  * Encoded theora frames must be a multiple of 16 in width and height.
356  * To handle other frame sizes, a crop rectangle is specified in
357  * frame_height and frame_width, offset_x and * offset_y. The offset
358  * and size should still be a multiple of 2 to avoid chroma sampling
359  * shifts. Offset values in this structure are measured from the
360  * upper left of the image.
361  *
362  * Frame rate, in frames per second, is stored as a rational
363  * fraction. Aspect ratio is also stored as a rational fraction, and
364  * refers to the aspect ratio of the frame pixels, not of the
365  * overall frame itself.
366  * 
367  * See <a href="http://svn.xiph.org/trunk/theora/examples/encoder_example.c">
368  * examples/encoder_example.c</a> for usage examples of the
369  * other paramters and good default settings for the encoder parameters.
370  */
371 typedef struct {
372   ogg_uint32_t  width;      /**< encoded frame width  */
373   ogg_uint32_t  height;     /**< encoded frame height */
374   ogg_uint32_t  frame_width;    /**< display frame width  */
375   ogg_uint32_t  frame_height;   /**< display frame height */
376   ogg_uint32_t  offset_x;   /**< horizontal offset of the displayed frame */
377   ogg_uint32_t  offset_y;   /**< vertical offset of the displayed frame */
378   ogg_uint32_t  fps_numerator;      /**< frame rate numerator **/
379   ogg_uint32_t  fps_denominator;    /**< frame rate denominator **/
380   ogg_uint32_t  aspect_numerator;   /**< pixel aspect ratio numerator */
381   ogg_uint32_t  aspect_denominator; /**< pixel aspect ratio denominator */
382   theora_colorspace colorspace;     /**< colorspace */
383   int           target_bitrate;     /**< nominal bitrate in bits per second */
384   int           quality;  /**< Nominal quality setting, 0-63 */
385   int           quick_p;  /**< Quick encode/decode */
386
387   /* decode only */
388   unsigned char version_major;
389   unsigned char version_minor;
390   unsigned char version_subminor;
391
392   void *codec_setup;
393
394   /* encode only */
395   int           dropframes_p;
396   int           keyframe_auto_p;
397   ogg_uint32_t  keyframe_frequency;
398   ogg_uint32_t  keyframe_frequency_force;  /* also used for decode init to
399                                               get granpos shift correct */
400   ogg_uint32_t  keyframe_data_target_bitrate;
401   ogg_int32_t   keyframe_auto_threshold;
402   ogg_uint32_t  keyframe_mindistance;
403   ogg_int32_t   noise_sensitivity;
404   ogg_int32_t   sharpness;
405
406   theora_pixelformat pixelformat;   /**< chroma subsampling mode to expect */
407
408 } theora_info;
409
410 /** Codec internal state and context.
411  */
412 typedef struct{
413   theora_info *i;
414   ogg_int64_t granulepos;
415
416   void *internal_encode;
417   void *internal_decode;
418
419 } theora_state;
420
421 /** 
422  * Comment header metadata.
423  *
424  * This structure holds the in-stream metadata corresponding to
425  * the 'comment' header packet.
426  *
427  * Meta data is stored as a series of (tag, value) pairs, in
428  * length-encoded string vectors. The first occurence of the 
429  * '=' character delimits the tag and value. A particular tag
430  * may occur more than once. The character set encoding for
431  * the strings is always UTF-8, but the tag names are limited
432  * to case-insensitive ASCII. See the spec for details.
433  *
434  * In filling in this structure, qtheora_decode_header() will
435  * null-terminate the user_comment strings for safety. However,
436  * the bitstream format itself treats them as 8-bit clean,
437  * and so the length array should be treated as authoritative
438  * for their length.
439  */
440 typedef struct theora_comment{
441   char **user_comments;         /**< An array of comment string vectors */
442   int   *comment_lengths;       /**< An array of corresponding string vector lengths in bytes */
443   int    comments;              /**< The total number of comment string vectors */
444   char  *vendor;                /**< The vendor string identifying the encoder, null terminated */
445
446 } theora_comment;
447 static int (*qtheora_encode_init) (theora_state *th, theora_info *ti);
448 static int (*qtheora_encode_YUVin) (theora_state *t, yuv_buffer *yuv);
449 static int (*qtheora_encode_packetout) ( theora_state *t, int last_p,
450                                     ogg_packet *op);
451 static int (*qtheora_encode_header) (theora_state *t, ogg_packet *op);
452 static int (*qtheora_encode_comment) (theora_comment *tc, ogg_packet *op);
453 static int (*qtheora_encode_tables) (theora_state *t, ogg_packet *op);
454 static void (*qtheora_info_init) (theora_info *c);
455 static void (*qtheora_info_clear) (theora_info *c);
456 static void (*qtheora_clear) (theora_state *t);
457 static void (*qtheora_comment_init) (theora_comment *tc);
458 static void  (*qtheora_comment_clear) (theora_comment *tc);
459 static double (*qtheora_granule_time) (theora_state *th,ogg_int64_t granulepos);
460 // end of theora.h stuff
461
462 static dllfunction_t oggfuncs[] =
463 {
464         {"ogg_stream_packetin", (void **) &qogg_stream_packetin},
465         {"ogg_stream_pageout", (void **) &qogg_stream_pageout},
466         {"ogg_stream_flush", (void **) &qogg_stream_flush},
467         {"ogg_stream_init", (void **) &qogg_stream_init},
468         {"ogg_stream_clear", (void **) &qogg_stream_clear},
469         {"ogg_page_granulepos", (void **) &qogg_page_granulepos},
470         {NULL, NULL}
471 };
472
473 static dllfunction_t vorbisencfuncs[] =
474 {
475         {"vorbis_encode_init_vbr", (void **) &qvorbis_encode_init_vbr},
476         {NULL, NULL}
477 };
478
479 static dllfunction_t vorbisfuncs[] =
480 {
481         {"vorbis_info_init", (void **) &qvorbis_info_init},
482         {"vorbis_info_clear", (void **) &qvorbis_info_clear},
483         {"vorbis_comment_init", (void **) &qvorbis_comment_init},
484         {"vorbis_comment_clear", (void **) &qvorbis_comment_clear},
485         {"vorbis_block_init", (void **) &qvorbis_block_init},
486         {"vorbis_block_clear", (void **) &qvorbis_block_clear},
487         {"vorbis_dsp_clear", (void **) &qvorbis_dsp_clear},
488         {"vorbis_analysis_init", (void **) &qvorbis_analysis_init},
489         {"vorbis_commentheader_out", (void **) &qvorbis_commentheader_out},
490         {"vorbis_analysis_headerout", (void **) &qvorbis_analysis_headerout},
491         {"vorbis_analysis_buffer", (void **) &qvorbis_analysis_buffer},
492         {"vorbis_analysis_wrote", (void **) &qvorbis_analysis_wrote},
493         {"vorbis_analysis_blockout", (void **) &qvorbis_analysis_blockout},
494         {"vorbis_analysis", (void **) &qvorbis_analysis},
495         {"vorbis_bitrate_addblock", (void **) &qvorbis_bitrate_addblock},
496         {"vorbis_bitrate_flushpacket", (void **) &qvorbis_bitrate_flushpacket},
497         {"vorbis_granule_time", (void **) &qvorbis_granule_time},
498         {NULL, NULL}
499 };
500
501 static dllfunction_t theorafuncs[] =
502 {
503         {"theora_info_init", (void **) &qtheora_info_init},
504         {"theora_info_clear", (void **) &qtheora_info_clear},
505         {"theora_comment_init", (void **) &qtheora_comment_init},
506         {"theora_comment_clear", (void **) &qtheora_comment_clear},
507         {"theora_encode_init", (void **) &qtheora_encode_init},
508         {"theora_encode_YUVin", (void **) &qtheora_encode_YUVin},
509         {"theora_encode_packetout", (void **) &qtheora_encode_packetout},
510         {"theora_encode_header", (void **) &qtheora_encode_header},
511         {"theora_encode_comment", (void **) &qtheora_encode_comment},
512         {"theora_encode_tables", (void **) &qtheora_encode_tables},
513         {"theora_clear", (void **) &qtheora_clear},
514         {"theora_granule_time", (void **) &qtheora_granule_time},
515         {NULL, NULL}
516 };
517
518 static dllhandle_t og_dll = NULL, vo_dll = NULL, ve_dll = NULL, th_dll = NULL;
519
520 qboolean SCR_CaptureVideo_Ogg_OpenLibrary(void)
521 {
522         const char* dllnames_og [] =
523         {
524 #if defined(WIN32)
525                 "libogg.dll",
526                 "ogg.dll",
527 #elif defined(MACOSX)
528                 "libogg.dylib",
529 #else
530                 "libogg.so.0",
531                 "libogg.so",
532 #endif
533                 NULL
534         };
535         const char* dllnames_vo [] =
536         {
537 #if defined(WIN32)
538                 "libvorbis.dll",
539                 "vorbis.dll",
540 #elif defined(MACOSX)
541                 "libvorbis.dylib",
542 #else
543                 "libvorbis.so.0",
544                 "libvorbis.so",
545 #endif
546                 NULL
547         };
548         const char* dllnames_ve [] =
549         {
550 #if defined(WIN32)
551                 "libvorbisenc.dll",
552                 "vorbisenc.dll",
553 #elif defined(MACOSX)
554                 "libvorbisenc.dylib",
555 #else
556                 "libvorbisenc.so.2",
557                 "libvorbisenc.so",
558 #endif
559                 NULL
560         };
561         const char* dllnames_th [] =
562         {
563 #if defined(WIN32)
564                 "libtheora.dll",
565                 "theora.dll",
566 #elif defined(MACOSX)
567                 "libtheora.dylib",
568 #else
569                 "libtheora.so.0",
570                 "libtheora.so",
571 #endif
572                 NULL
573         };
574
575         return
576                 Sys_LoadLibrary (dllnames_og, &og_dll, oggfuncs)
577                 &&
578                 Sys_LoadLibrary (dllnames_th, &th_dll, theorafuncs)
579                 &&
580                 Sys_LoadLibrary (dllnames_vo, &vo_dll, vorbisfuncs)
581                 &&
582                 Sys_LoadLibrary (dllnames_ve, &ve_dll, vorbisencfuncs);
583 }
584
585 void SCR_CaptureVideo_Ogg_Init(void)
586 {
587         SCR_CaptureVideo_Ogg_OpenLibrary();
588
589         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_quality);
590         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_bitrate);
591         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_keyframe_bitrate_multiplier);
592         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_keyframe_maxinterval);
593         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_keyframe_mininterval);
594         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_keyframe_auto_threshold);
595         Cvar_RegisterVariable(&cl_capturevideo_ogg_theora_noise_sensitivity);
596         Cvar_RegisterVariable(&cl_capturevideo_ogg_vorbis_quality);
597 }
598
599 qboolean SCR_CaptureVideo_Ogg_Available(void)
600 {
601         return og_dll && th_dll && vo_dll && ve_dll;
602 }
603
604 void SCR_CaptureVideo_Ogg_CloseDLL(void)
605 {
606         Sys_UnloadLibrary (&ve_dll);
607         Sys_UnloadLibrary (&vo_dll);
608         Sys_UnloadLibrary (&th_dll);
609         Sys_UnloadLibrary (&og_dll);
610 }
611
612 // this struct should not be needed
613 // however, libogg appears to pull the ogg_page's data element away from our
614 // feet before we get to write the data due to interleaving
615 // so this struct is used to keep the page data around until it actually gets
616 // written
617 typedef struct allocatedoggpage_s
618 {
619         size_t len;
620         double time;
621         unsigned char data[65307];
622         // this number is from RFC 3533. In case libogg writes more, we'll have to increase this
623         // but we'll get a Host_Error in this case so we can track it down
624 }
625 allocatedoggpage_t;
626
627 typedef struct capturevideostate_ogg_formatspecific_s
628 {
629         ogg_stream_state to, vo;
630         int serial1, serial2;
631         theora_state ts;
632         vorbis_dsp_state vd;
633         vorbis_block vb;
634         vorbis_info vi;
635         yuv_buffer yuv[2];
636         int yuvi;
637         int lastnum;
638         int channels;
639
640         allocatedoggpage_t videopage, audiopage;
641 }
642 capturevideostate_ogg_formatspecific_t;
643 #define LOAD_FORMATSPECIFIC_OGG() capturevideostate_ogg_formatspecific_t *format = (capturevideostate_ogg_formatspecific_t *) cls.capturevideo.formatspecific
644
645 static void SCR_CaptureVideo_Ogg_Interleave(void)
646 {
647         LOAD_FORMATSPECIFIC_OGG();
648         ogg_page pg;
649
650         if(!cls.capturevideo.soundrate)
651         {
652                 while(qogg_stream_pageout(&format->to, &pg) > 0)
653                 {
654                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
655                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
656                 }
657                 return;
658         }
659
660         for(;;)
661         {
662                 // first: make sure we have a page of both types
663                 if(!format->videopage.len)
664                         if(qogg_stream_pageout(&format->to, &pg) > 0)
665                         {
666                                 format->videopage.len = pg.header_len + pg.body_len;
667                                 format->videopage.time = qtheora_granule_time(&format->ts, qogg_page_granulepos(&pg));
668                                 if(format->videopage.len > sizeof(format->videopage.data))
669                                         Host_Error("video page too long");
670                                 memcpy(format->videopage.data, pg.header, pg.header_len);
671                                 memcpy(format->videopage.data + pg.header_len, pg.body, pg.body_len);
672                         }
673                 if(!format->audiopage.len)
674                         if(qogg_stream_pageout(&format->vo, &pg) > 0)
675                         {
676                                 format->audiopage.len = pg.header_len + pg.body_len;
677                                 format->audiopage.time = qvorbis_granule_time(&format->vd, qogg_page_granulepos(&pg));
678                                 if(format->audiopage.len > sizeof(format->audiopage.data))
679                                         Host_Error("audio page too long");
680                                 memcpy(format->audiopage.data, pg.header, pg.header_len);
681                                 memcpy(format->audiopage.data + pg.header_len, pg.body, pg.body_len);
682                         }
683
684                 if(format->videopage.len && format->audiopage.len)
685                 {
686                         // output the page that ends first
687                         if(format->videopage.time < format->audiopage.time)
688                         {
689                                 FS_Write(cls.capturevideo.videofile, format->videopage.data, format->videopage.len);
690                                 format->videopage.len = 0;
691                         }
692                         else
693                         {
694                                 FS_Write(cls.capturevideo.videofile, format->audiopage.data, format->audiopage.len);
695                                 format->audiopage.len = 0;
696                         }
697                 }
698                 else
699                         break;
700         }
701 }
702
703 static void SCR_CaptureVideo_Ogg_FlushInterleaving(void)
704 {
705         LOAD_FORMATSPECIFIC_OGG();
706
707         if(cls.capturevideo.soundrate)
708         if(format->audiopage.len)
709         {
710                 FS_Write(cls.capturevideo.videofile, format->audiopage.data, format->audiopage.len);
711                 format->audiopage.len = 0;
712         }
713
714         if(format->videopage.len)
715         {
716                 FS_Write(cls.capturevideo.videofile, format->videopage.data, format->videopage.len);
717                 format->videopage.len = 0;
718         }
719 }
720
721 static void SCR_CaptureVideo_Ogg_EndVideo(void)
722 {
723         LOAD_FORMATSPECIFIC_OGG();
724         ogg_page pg;
725         ogg_packet pt;
726
727         if(format->yuvi >= 0)
728         {
729                 // send the previous (and last) frame
730                 while(format->lastnum-- > 0)
731                 {
732                         qtheora_encode_YUVin(&format->ts, &format->yuv[format->yuvi]);
733
734                         while(qtheora_encode_packetout(&format->ts, !format->lastnum, &pt))
735                                 qogg_stream_packetin(&format->to, &pt);
736
737                         SCR_CaptureVideo_Ogg_Interleave();
738                 }
739         }
740
741         if(cls.capturevideo.soundrate)
742         {
743                 qvorbis_analysis_wrote(&format->vd, 0);
744                 while(qvorbis_analysis_blockout(&format->vd, &format->vb) == 1)
745                 {
746                         qvorbis_analysis(&format->vb, NULL);
747                         qvorbis_bitrate_addblock(&format->vb);
748                         while(qvorbis_bitrate_flushpacket(&format->vd, &pt))
749                                 qogg_stream_packetin(&format->vo, &pt);
750                         SCR_CaptureVideo_Ogg_Interleave();
751                 }
752         }
753
754         SCR_CaptureVideo_Ogg_FlushInterleaving();
755
756         while(qogg_stream_pageout(&format->to, &pg) > 0)
757         {
758                 FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
759                 FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
760         }
761
762         if(cls.capturevideo.soundrate)
763         {
764                 while(qogg_stream_pageout(&format->vo, &pg) > 0)
765                 {
766                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
767                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
768                 }
769         }
770                 
771         while (1) {
772                 int result = qogg_stream_flush (&format->to, &pg);
773                 if (result < 0)
774                         fprintf (stderr, "Internal Ogg library error.\n"); // TODO Host_Error
775                 if (result <= 0)
776                         break;
777                 FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
778                 FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
779         }
780
781         if(cls.capturevideo.soundrate)
782         {
783                 while (1) {
784                         int result = qogg_stream_flush (&format->vo, &pg);
785                         if (result < 0)
786                                 fprintf (stderr, "Internal Ogg library error.\n"); // TODO Host_Error
787                         if (result <= 0)
788                                 break;
789                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
790                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
791                 }
792
793                 qogg_stream_clear(&format->vo);
794                 qvorbis_block_clear(&format->vb);
795                 qvorbis_dsp_clear(&format->vd);
796         }
797
798         qogg_stream_clear(&format->to);
799         qtheora_clear(&format->ts);
800         qvorbis_info_clear(&format->vi);
801
802         Mem_Free(format->yuv[0].y);
803         Mem_Free(format->yuv[0].u);
804         Mem_Free(format->yuv[0].v);
805         Mem_Free(format->yuv[1].y);
806         Mem_Free(format->yuv[1].u);
807         Mem_Free(format->yuv[1].v);
808         Mem_Free(format);
809
810         FS_Close(cls.capturevideo.videofile);
811         cls.capturevideo.videofile = NULL;
812 }
813
814 static void SCR_CaptureVideo_Ogg_ConvertFrame_BGRA_to_YUV(void)
815 {
816         LOAD_FORMATSPECIFIC_OGG();
817         yuv_buffer *yuv;
818         int x, y;
819         int blockr, blockg, blockb;
820         unsigned char *b = cls.capturevideo.outbuffer;
821         int w = cls.capturevideo.width;
822         int h = cls.capturevideo.height;
823         int inpitch = w*4;
824
825         yuv = &format->yuv[format->yuvi];
826
827         for(y = 0; y < h; ++y)
828         {
829                 for(b = cls.capturevideo.outbuffer + (h-1-y)*w*4, x = 0; x < w; ++x)
830                 {
831                         blockr = b[2];
832                         blockg = b[1];
833                         blockb = b[0];
834                         yuv->y[x + yuv->y_stride * y] =
835                                 cls.capturevideo.yuvnormalizetable[0][cls.capturevideo.rgbtoyuvscaletable[0][0][blockr] + cls.capturevideo.rgbtoyuvscaletable[0][1][blockg] + cls.capturevideo.rgbtoyuvscaletable[0][2][blockb]];
836                         b += 4;
837                 }
838
839                 if((y & 1) == 0)
840                 {
841                         for(b = cls.capturevideo.outbuffer + (h-2-y)*w*4, x = 0; x < w/2; ++x)
842                         {
843                                 blockr = (b[2] + b[6] + b[inpitch+2] + b[inpitch+6]) >> 2;
844                                 blockg = (b[1] + b[5] + b[inpitch+1] + b[inpitch+5]) >> 2;
845                                 blockb = (b[0] + b[4] + b[inpitch+0] + b[inpitch+4]) >> 2;
846                                 yuv->u[x + yuv->uv_stride * (y/2)] =
847                                         cls.capturevideo.yuvnormalizetable[1][cls.capturevideo.rgbtoyuvscaletable[1][0][blockr] + cls.capturevideo.rgbtoyuvscaletable[1][1][blockg] + cls.capturevideo.rgbtoyuvscaletable[1][2][blockb] + 128];
848                                 yuv->v[x + yuv->uv_stride * (y/2)] =
849                                         cls.capturevideo.yuvnormalizetable[2][cls.capturevideo.rgbtoyuvscaletable[2][0][blockr] + cls.capturevideo.rgbtoyuvscaletable[2][1][blockg] + cls.capturevideo.rgbtoyuvscaletable[2][2][blockb] + 128];
850                                 b += 8;
851                         }
852                 }
853         }
854 }
855
856 static void SCR_CaptureVideo_Ogg_VideoFrames(int num)
857 {
858         LOAD_FORMATSPECIFIC_OGG();
859         ogg_packet pt;
860
861         // data is in cls.capturevideo.outbuffer as BGRA and has size width*height
862
863         if(format->yuvi >= 0)
864         {
865                 // send the previous frame
866                 while(format->lastnum-- > 0)
867                 {
868                         qtheora_encode_YUVin(&format->ts, &format->yuv[format->yuvi]);
869
870                         while(qtheora_encode_packetout(&format->ts, false, &pt))
871                                 qogg_stream_packetin(&format->to, &pt);
872
873                         SCR_CaptureVideo_Ogg_Interleave();
874                 }
875         }
876
877         format->yuvi = (format->yuvi + 1) % 2;
878         SCR_CaptureVideo_Ogg_ConvertFrame_BGRA_to_YUV();
879         format->lastnum = num;
880
881         // TODO maybe send num-1 frames from here already
882 }
883
884 typedef int channelmapping_t[8];
885 channelmapping_t mapping[8] =
886 {
887         { 0, -1, -1, -1, -1, -1, -1, -1 }, // mono
888         { 0, 1, -1, -1, -1, -1, -1, -1 }, // stereo
889         { 0, 1, 2, -1, -1, -1, -1, -1 }, // L C R
890         { 0, 1, 2, 3, -1, -1, -1, -1 }, // surround40
891         { 0, 2, 3, 4, 1, -1, -1, -1 }, // FL FC FR RL RR
892         { 0, 2, 3, 4, 1, 5, -1, -1 }, // surround51
893         { 0, 2, 3, 4, 1, 5, 6, -1 }, // (not defined by vorbis spec)
894         { 0, 2, 3, 4, 1, 5, 6, 7 } // surround71 (not defined by vorbis spec)
895 };
896
897 static void SCR_CaptureVideo_Ogg_SoundFrame(const portable_sampleframe_t *paintbuffer, size_t length)
898 {
899         LOAD_FORMATSPECIFIC_OGG();
900         float **vorbis_buffer;
901         size_t i;
902         int j;
903         ogg_packet pt;
904         int *map = mapping[bound(1, cls.capturevideo.soundchannels, 8) - 1];
905
906         vorbis_buffer = qvorbis_analysis_buffer(&format->vd, length);
907         for(j = 0; j < cls.capturevideo.soundchannels; ++j)
908         {
909                 float *b = vorbis_buffer[map[j]];
910                 for(i = 0; i < length; ++i)
911                         b[i] = paintbuffer[i].sample[j] / 32768.0f;
912         }
913         qvorbis_analysis_wrote(&format->vd, length);
914
915         while(qvorbis_analysis_blockout(&format->vd, &format->vb) == 1)
916         {
917                 qvorbis_analysis(&format->vb, NULL);
918                 qvorbis_bitrate_addblock(&format->vb);
919
920                 while(qvorbis_bitrate_flushpacket(&format->vd, &pt))
921                         qogg_stream_packetin(&format->vo, &pt);
922         }
923
924         SCR_CaptureVideo_Ogg_Interleave();
925 }
926
927 void SCR_CaptureVideo_Ogg_BeginVideo(void)
928 {
929         cls.capturevideo.format = CAPTUREVIDEOFORMAT_OGG_VORBIS_THEORA;
930         cls.capturevideo.formatextension = "ogv";
931         cls.capturevideo.videofile = FS_OpenRealFile(va("%s.%s", cls.capturevideo.basename, cls.capturevideo.formatextension), "wb", false);
932         cls.capturevideo.endvideo = SCR_CaptureVideo_Ogg_EndVideo;
933         cls.capturevideo.videoframes = SCR_CaptureVideo_Ogg_VideoFrames;
934         cls.capturevideo.soundframe = SCR_CaptureVideo_Ogg_SoundFrame;
935         cls.capturevideo.formatspecific = Mem_Alloc(tempmempool, sizeof(capturevideostate_ogg_formatspecific_t));
936         {
937                 LOAD_FORMATSPECIFIC_OGG();
938                 int num, denom, i;
939                 ogg_page pg;
940                 ogg_packet pt, pt2, pt3;
941                 theora_comment tc;
942                 vorbis_comment vc;
943                 theora_info ti;
944
945                 format->serial1 = rand();
946                 qogg_stream_init(&format->to, format->serial1);
947
948                 if(cls.capturevideo.soundrate)
949                 {
950                         do
951                         {
952                                 format->serial2 = rand();
953                         }
954                         while(format->serial1 == format->serial2);
955                         qogg_stream_init(&format->vo, format->serial2);
956                 }
957
958                 format->videopage.len = format->audiopage.len = 0;
959
960                 qtheora_info_init(&ti);
961                 ti.frame_width = cls.capturevideo.width;
962                 ti.frame_height = cls.capturevideo.height;
963                 ti.width = (ti.frame_width + 15) & ~15;
964                 ti.height = (ti.frame_height + 15) & ~15;
965                 //ti.offset_x = ((ti.width - ti.frame_width) / 2) & ~1;
966                 //ti.offset_y = ((ti.height - ti.frame_height) / 2) & ~1;
967
968                 for(i = 0; i < 2; ++i)
969                 {
970                         format->yuv[i].y_width = ti.width;
971                         format->yuv[i].y_height = ti.height;
972                         format->yuv[i].y_stride = ti.width;
973                         format->yuv[i].uv_width = ti.width / 2;
974                         format->yuv[i].uv_height = ti.height / 2;
975                         format->yuv[i].uv_stride = ti.width / 2;
976                         format->yuv[i].y = (unsigned char *) Mem_Alloc(tempmempool, format->yuv[i].y_stride * format->yuv[i].y_height);
977                         format->yuv[i].u = (unsigned char *) Mem_Alloc(tempmempool, format->yuv[i].uv_stride * format->yuv[i].uv_height);
978                         format->yuv[i].v = (unsigned char *) Mem_Alloc(tempmempool, format->yuv[i].uv_stride * format->yuv[i].uv_height);
979                 }
980                 format->yuvi = -1; // -1: no frame valid yet, write into 0
981
982                 FindFraction(cls.capturevideo.framerate / cls.capturevideo.framestep, &num, &denom, 1001);
983                 ti.fps_numerator = num;
984                 ti.fps_denominator = denom;
985
986                 FindFraction(1 / vid_pixelheight.value, &num, &denom, 1000);
987                 ti.aspect_numerator = num;
988                 ti.aspect_denominator = denom;
989
990                 ti.colorspace = OC_CS_UNSPECIFIED;
991                 ti.pixelformat = OC_PF_420;
992
993                 ti.quick_p = true; // http://mlblog.osdir.com/multimedia.ogg.theora.general/2004-07/index.shtml
994                 ti.dropframes_p = false;
995
996                 ti.target_bitrate = cl_capturevideo_ogg_theora_bitrate.integer * 1000;
997                 ti.quality = cl_capturevideo_ogg_theora_quality.integer;
998
999                 if(ti.target_bitrate <= 0)
1000                 {
1001                         if(ti.quality < 0)
1002                         {
1003                                 ti.target_bitrate = -1;
1004                                 ti.keyframe_data_target_bitrate = (unsigned int)-1;
1005                                 ti.quality = 63;
1006                         }
1007                         else
1008                         {
1009                                 ti.target_bitrate = -1;
1010                                 ti.keyframe_data_target_bitrate = (unsigned int)-1;
1011                                 ti.quality = bound(0, ti.quality, 63);
1012                         }
1013                 }
1014                 else
1015                 {
1016                         if(ti.quality < 0)
1017                         {
1018                                 ti.target_bitrate = bound(45000, ti.target_bitrate, 2000000);
1019                                 ti.keyframe_data_target_bitrate = (int) (ti.target_bitrate * max(1, cl_capturevideo_ogg_theora_keyframe_bitrate_multiplier.value));
1020                                 ti.quality = -1;
1021                         }
1022                         else
1023                         {
1024                                 ti.target_bitrate = bound(45000, ti.target_bitrate, 2000000);
1025                                 ti.keyframe_data_target_bitrate = (int) (ti.target_bitrate * max(1, cl_capturevideo_ogg_theora_keyframe_bitrate_multiplier.value));
1026                                 ti.quality = -1;
1027                         }
1028                 }
1029
1030                 // this -1 magic is because ti.keyframe_frequency and ti.keyframe_mindistance use different metrics
1031                 ti.keyframe_frequency = bound(1, cl_capturevideo_ogg_theora_keyframe_maxinterval.integer, 1000);
1032                 ti.keyframe_mindistance = bound(1, cl_capturevideo_ogg_theora_keyframe_mininterval.integer, (int) ti.keyframe_frequency) - 1;
1033                 ti.noise_sensitivity = bound(0, cl_capturevideo_ogg_theora_noise_sensitivity.integer, 6);
1034                 ti.sharpness = bound(0, cl_capturevideo_ogg_theora_sharpness.integer, 2);
1035                 ti.keyframe_auto_threshold = bound(0, cl_capturevideo_ogg_theora_keyframe_auto_threshold.integer, 100);
1036
1037                 ti.keyframe_frequency_force = ti.keyframe_frequency;
1038                 ti.keyframe_auto_p = (ti.keyframe_frequency != ti.keyframe_mindistance + 1);
1039
1040                 qtheora_encode_init(&format->ts, &ti);
1041                 qtheora_info_clear(&ti);
1042
1043                 // vorbis?
1044                 if(cls.capturevideo.soundrate)
1045                 {
1046                         qvorbis_info_init(&format->vi);
1047                         qvorbis_encode_init_vbr(&format->vi, cls.capturevideo.soundchannels, cls.capturevideo.soundrate, bound(-1, cl_capturevideo_ogg_vorbis_quality.value, 10) * 0.099);
1048                         qvorbis_comment_init(&vc);
1049                         qvorbis_analysis_init(&format->vd, &format->vi);
1050                         qvorbis_block_init(&format->vd, &format->vb);
1051                 }
1052
1053                 qtheora_comment_init(&tc);
1054
1055                 /* create the remaining theora headers */
1056                 qtheora_encode_header(&format->ts, &pt);
1057                 qogg_stream_packetin(&format->to, &pt);
1058                 if (qogg_stream_pageout (&format->to, &pg) != 1)
1059                         fprintf (stderr, "Internal Ogg library error.\n");
1060                 FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
1061                 FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
1062
1063                 qtheora_encode_comment(&tc, &pt);
1064                 qogg_stream_packetin(&format->to, &pt);
1065                 qtheora_encode_tables(&format->ts, &pt);
1066                 qogg_stream_packetin (&format->to, &pt);
1067
1068                 qtheora_comment_clear(&tc);
1069
1070                 if(cls.capturevideo.soundrate)
1071                 {
1072                         qvorbis_analysis_headerout(&format->vd, &vc, &pt, &pt2, &pt3);
1073                         qogg_stream_packetin(&format->vo, &pt);
1074                         if (qogg_stream_pageout (&format->vo, &pg) != 1)
1075                                 fprintf (stderr, "Internal Ogg library error.\n");
1076                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
1077                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
1078
1079                         qogg_stream_packetin(&format->vo, &pt2);
1080                         qogg_stream_packetin(&format->vo, &pt3);
1081
1082                         qvorbis_comment_clear(&vc);
1083                 }
1084
1085                 for(;;)
1086                 {
1087                         int result = qogg_stream_flush (&format->to, &pg);
1088                         if (result < 0)
1089                                 fprintf (stderr, "Internal Ogg library error.\n"); // TODO Host_Error
1090                         if (result <= 0)
1091                                 break;
1092                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
1093                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
1094                 }
1095
1096                 if(cls.capturevideo.soundrate)
1097                 for(;;)
1098                 {
1099                         int result = qogg_stream_flush (&format->vo, &pg);
1100                         if (result < 0)
1101                                 fprintf (stderr, "Internal Ogg library error.\n"); // TODO Host_Error
1102                         if (result <= 0)
1103                                 break;
1104                         FS_Write(cls.capturevideo.videofile, pg.header, pg.header_len);
1105                         FS_Write(cls.capturevideo.videofile, pg.body, pg.body_len);
1106                 }
1107         }
1108 }