*/
/**
- * @file vp3.c
+ * @file libavcodec/vp3.c
* On2 VP3 Video Decoder
*
* VP3 Video Decoder by Mike Melanson (mike at multimedia.cx)
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <unistd.h>
#include "avcodec.h"
#include "dsputil.h"
-#include "bitstream.h"
+#include "get_bits.h"
#include "vp3data.h"
#include "xiph.h"
#define FRAGMENT_PIXELS 8
-/*
- * Debugging Variables
- *
- * Define one or more of the following compile-time variables to 1 to obtain
- * elaborate information about certain aspects of the decoding process.
- *
- * KEYFRAMES_ONLY: set this to 1 to only see keyframes (VP3 slideshow mode)
- * DEBUG_VP3: high-level decoding flow
- * DEBUG_INIT: initialization parameters
- * DEBUG_DEQUANTIZERS: display how the dequanization tables are built
- * DEBUG_BLOCK_CODING: unpacking the superblock/macroblock/fragment coding
- * DEBUG_MODES: unpacking the coding modes for individual fragments
- * DEBUG_VECTORS: display the motion vectors
- * DEBUG_TOKEN: display exhaustive information about each DCT token
- * DEBUG_VLC: display the VLCs as they are extracted from the stream
- * DEBUG_DC_PRED: display the process of reversing DC prediction
- * DEBUG_IDCT: show every detail of the IDCT process
- */
-
-#define KEYFRAMES_ONLY 0
-
-#define DEBUG_VP3 0
-#define DEBUG_INIT 0
-#define DEBUG_DEQUANTIZERS 0
-#define DEBUG_BLOCK_CODING 0
-#define DEBUG_MODES 0
-#define DEBUG_VECTORS 0
-#define DEBUG_TOKEN 0
-#define DEBUG_VLC 0
-#define DEBUG_DC_PRED 0
-#define DEBUG_IDCT 0
-
-#if DEBUG_VP3
-#define debug_vp3(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_vp3(const char *format, ...) { }
-#endif
-
-#if DEBUG_INIT
-#define debug_init(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_init(const char *format, ...) { }
-#endif
-
-#if DEBUG_DEQUANTIZERS
-#define debug_dequantizers(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_dequantizers(const char *format, ...) { }
-#endif
-
-#if DEBUG_BLOCK_CODING
-#define debug_block_coding(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_block_coding(const char *format, ...) { }
-#endif
-
-#if DEBUG_MODES
-#define debug_modes(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_modes(const char *format, ...) { }
-#endif
-
-#if DEBUG_VECTORS
-#define debug_vectors(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_vectors(const char *format, ...) { }
-#endif
-
-#if DEBUG_TOKEN
-#define debug_token(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_token(const char *format, ...) { }
-#endif
-
-#if DEBUG_VLC
-#define debug_vlc(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_vlc(const char *format, ...) { }
-#endif
-
-#if DEBUG_DC_PRED
-#define debug_dc_pred(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_dc_pred(const char *format, ...) { }
-#endif
-
-#if DEBUG_IDCT
-#define debug_idct(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
-#else
-static inline void debug_idct(const char *format, ...) { }
-#endif
+static av_cold int vp3_decode_end(AVCodecContext *avctx);
typedef struct Coeff {
struct Coeff *next;
uint8_t coding_method;
int8_t motion_x;
int8_t motion_y;
+ uint8_t qpi;
} Vp3Fragment;
#define SB_NOT_CODED 0
DSPContext dsp;
int flipped_image;
- int qis[3];
- int nqis;
- int quality_index;
- int last_quality_index;
+ int qps[3];
+ int nqps;
+ int last_qps[3];
int superblock_count;
- int superblock_width;
- int superblock_height;
int y_superblock_width;
int y_superblock_height;
int c_superblock_width;
int coded_fragment_list_index;
int pixel_addresses_initialized;
+ /* track which fragments have already been decoded; called 'fast'
+ * because this data structure avoids having to iterate through every
+ * fragment in coded_fragment_list; once a fragment has been fully
+ * decoded, it is removed from this list */
+ int *fast_fragment_list;
+ int fragment_list_y_head;
+ int fragment_list_c_head;
+
VLC dc_vlc[16];
VLC ac_vlc_1[16];
VLC ac_vlc_2[16];
/* these arrays need to be on 16-byte boundaries since SSE2 operations
* index into them */
- DECLARE_ALIGNED_16(int16_t, qmat[2][4][64]); //<qmat[is_inter][plane]
+ DECLARE_ALIGNED_16(int16_t, qmat[3][2][3][64]); //<qmat[qpi][is_inter][plane]
/* This table contains superblock_count * 16 entries. Each set of 16
* numbers corresponds to the fragment indexes 0..15 of the superblock.
int huff_code_size;
uint16_t huffman_table[80][32][2];
- uint32_t filter_limit_values[64];
- int bounding_values_array[256];
+ uint8_t filter_limit_values[64];
+ DECLARE_ALIGNED_8(int, bounding_values_array[256+2]);
} Vp3DecodeContext;
/************************************************************************
int right_edge = 0;
int bottom_edge = 0;
int superblock_row_inc = 0;
- int *hilbert = NULL;
int mapping_index = 0;
int current_macroblock;
int c_fragment;
- signed char travel_width[16] = {
+ static const signed char travel_width[16] = {
1, 1, 0, -1,
0, 0, 1, 0,
1, 0, 1, 0,
0, -1, 0, 1
};
- signed char travel_height[16] = {
+ static const signed char travel_height[16] = {
0, 0, 1, 0,
1, 1, 0, -1,
0, 1, 0, -1,
-1, 0, -1, 0
};
- signed char travel_width_mb[4] = {
+ static const signed char travel_width_mb[4] = {
1, 0, 1, 0
};
- signed char travel_height_mb[4] = {
+ static const signed char travel_height_mb[4] = {
0, 1, 0, -1
};
- debug_vp3(" vp3: initialize block mapping tables\n");
-
hilbert_walk_mb[0] = 1;
hilbert_walk_mb[1] = s->macroblock_width;
hilbert_walk_mb[2] = 1;
/* iterate through each superblock (all planes) and map the fragments */
for (i = 0; i < s->superblock_count; i++) {
- debug_init(" superblock %d (u starts @ %d, v starts @ %d)\n",
- i, s->u_superblock_start, s->v_superblock_start);
-
/* time to re-assign the limits? */
if (i == 0) {
if ((current_width < right_edge) &&
(current_height < bottom_edge)) {
s->superblock_fragments[mapping_index] = current_fragment;
- debug_init(" mapping fragment %d to superblock %d, position %d (%d/%d x %d/%d)\n",
- s->superblock_fragments[mapping_index], i, j,
- current_width, right_edge, current_height, bottom_edge);
} else {
s->superblock_fragments[mapping_index] = -1;
- debug_init(" superblock %d, position %d has no fragment (%d/%d x %d/%d)\n",
- i, j,
- current_width, right_edge, current_height, bottom_edge);
}
mapping_index++;
current_height = 0;
superblock_row_inc = s->macroblock_width -
(s->y_superblock_width * 2 - s->macroblock_width);
- hilbert = hilbert_walk_mb;
mapping_index = 0;
current_macroblock = -1;
for (i = 0; i < s->u_superblock_start; i++) {
if ((current_width < right_edge) &&
(current_height < bottom_edge)) {
s->superblock_macroblocks[mapping_index] = current_macroblock;
- debug_init(" mapping macroblock %d to superblock %d, position %d (%d/%d x %d/%d)\n",
- s->superblock_macroblocks[mapping_index], i, j,
- current_width, right_edge, current_height, bottom_edge);
} else {
s->superblock_macroblocks[mapping_index] = -1;
- debug_init(" superblock %d, position %d has no macroblock (%d/%d x %d/%d)\n",
- i, j,
- current_width, right_edge, current_height, bottom_edge);
}
mapping_index++;
for (j = 0; j < s->fragment_width; j += 2) {
- debug_init(" macroblock %d contains fragments: ", current_macroblock);
s->all_fragments[current_fragment].macroblock = current_macroblock;
s->macroblock_fragments[mapping_index++] = current_fragment;
- debug_init("%d ", current_fragment);
if (j + 1 < s->fragment_width) {
s->all_fragments[current_fragment + 1].macroblock = current_macroblock;
s->macroblock_fragments[mapping_index++] = current_fragment + 1;
- debug_init("%d ", current_fragment + 1);
} else
s->macroblock_fragments[mapping_index++] = -1;
current_macroblock;
s->macroblock_fragments[mapping_index++] =
current_fragment + s->fragment_width;
- debug_init("%d ", current_fragment + s->fragment_width);
} else
s->macroblock_fragments[mapping_index++] = -1;
current_macroblock;
s->macroblock_fragments[mapping_index++] =
current_fragment + s->fragment_width + 1;
- debug_init("%d ", current_fragment + s->fragment_width + 1);
} else
s->macroblock_fragments[mapping_index++] = -1;
(i * s->fragment_width / 4) + (j / 2);
s->all_fragments[c_fragment].macroblock = s->macroblock_count;
s->macroblock_fragments[mapping_index++] = c_fragment;
- debug_init("%d ", c_fragment);
c_fragment = s->fragment_start[2] +
(i * s->fragment_width / 4) + (j / 2);
s->all_fragments[c_fragment].macroblock = s->macroblock_count;
s->macroblock_fragments[mapping_index++] = c_fragment;
- debug_init("%d ", c_fragment);
-
- debug_init("\n");
if (j + 2 <= s->fragment_width)
current_fragment += 2;
s->all_fragments[i].motion_x = 127;
s->all_fragments[i].motion_y = 127;
s->all_fragments[i].next_coeff= NULL;
+ s->all_fragments[i].qpi = 0;
s->coeffs[i].index=
s->coeffs[i].coeff=0;
s->coeffs[i].next= NULL;
* This function sets up the dequantization tables used for a particular
* frame.
*/
-static void init_dequantizer(Vp3DecodeContext *s)
+static void init_dequantizer(Vp3DecodeContext *s, int qpi)
{
- int ac_scale_factor = s->coded_ac_scale_factor[s->quality_index];
- int dc_scale_factor = s->coded_dc_scale_factor[s->quality_index];
+ int ac_scale_factor = s->coded_ac_scale_factor[s->qps[qpi]];
+ int dc_scale_factor = s->coded_dc_scale_factor[s->qps[qpi]];
int i, plane, inter, qri, bmi, bmj, qistart;
- debug_vp3(" vp3: initializing dequantization tables\n");
-
for(inter=0; inter<2; inter++){
for(plane=0; plane<3; plane++){
int sum=0;
for(qri=0; qri<s->qr_count[inter][plane]; qri++){
sum+= s->qr_size[inter][plane][qri];
- if(s->quality_index <= sum)
+ if(s->qps[qpi] <= sum)
break;
}
qistart= sum - s->qr_size[inter][plane][qri];
bmi= s->qr_base[inter][plane][qri ];
bmj= s->qr_base[inter][plane][qri+1];
for(i=0; i<64; i++){
- int coeff= ( 2*(sum -s->quality_index)*s->base_matrix[bmi][i]
- - 2*(qistart-s->quality_index)*s->base_matrix[bmj][i]
+ int coeff= ( 2*(sum -s->qps[qpi])*s->base_matrix[bmi][i]
+ - 2*(qistart-s->qps[qpi])*s->base_matrix[bmj][i]
+ s->qr_size[inter][plane][qri])
/ (2*s->qr_size[inter][plane][qri]);
int qmin= 8<<(inter + !i);
int qscale= i ? ac_scale_factor : dc_scale_factor;
- s->qmat[inter][plane][i]= av_clip((qscale * coeff)/100 * 4, qmin, 4096);
+ s->qmat[qpi][inter][plane][s->dsp.idct_permutation[i]]= av_clip((qscale * coeff)/100 * 4, qmin, 4096);
}
+ // all DC coefficients use the same quant so as not to interfere with DC prediction
+ s->qmat[qpi][inter][plane][0] = s->qmat[0][inter][plane][0];
}
}
- memset(s->qscale_table, (FFMAX(s->qmat[0][0][1], s->qmat[0][1][1])+8)/16, 512); //FIXME finetune
+ memset(s->qscale_table, (FFMAX(s->qmat[0][0][0][1], s->qmat[0][0][1][1])+8)/16, 512); //FIXME finetune
}
/*
* This function initializes the loop filter boundary limits if the frame's
* quality index is different from the previous frame's.
+ *
+ * The filter_limit_values may not be larger than 127.
*/
static void init_loop_filter(Vp3DecodeContext *s)
{
int *bounding_values= s->bounding_values_array+127;
int filter_limit;
int x;
+ int value;
- filter_limit = s->filter_limit_values[s->quality_index];
+ filter_limit = s->filter_limit_values[s->qps[0]];
/* set up the bounding values */
memset(s->bounding_values_array, 0, 256 * sizeof(int));
for (x = 0; x < filter_limit; x++) {
- bounding_values[-x - filter_limit] = -filter_limit + x;
bounding_values[-x] = -x;
bounding_values[x] = x;
- bounding_values[x + filter_limit] = filter_limit - x;
}
+ for (x = value = filter_limit; x < 128 && value; x++, value--) {
+ bounding_values[ x] = value;
+ bounding_values[-x] = -value;
+ }
+ if (value)
+ bounding_values[128] = value;
+ bounding_values[129] = bounding_values[130] = filter_limit * 0x02020202;
}
/*
int i, j;
int current_fragment;
- debug_vp3(" vp3: unpacking superblock coding\n");
-
if (s->keyframe) {
-
- debug_vp3(" keyframe-- all superblocks are fully coded\n");
memset(s->superblock_coding, SB_FULLY_CODED, s->superblock_count);
} else {
s->superblock_run_length_vlc.table, 6, 2);
if (current_run == 33)
current_run += get_bits(gb, 12);
- debug_block_coding(" setting superblocks %d..%d to %s\n",
- current_superblock,
- current_superblock + current_run - 1,
- (bit) ? "partially coded" : "not coded");
/* if any of the superblocks are not partially coded, flag
* a boolean to decode the list of fully-coded superblocks */
if (current_run == 33)
current_run += get_bits(gb, 12);
}
-
- debug_block_coding(" setting superblock %d to %s\n",
- current_superblock,
- (bit) ? "fully coded" : "not coded");
s->superblock_coding[current_superblock] = 2*bit;
}
current_superblock++;
}
s->coded_fragment_list_index++;
s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
- debug_block_coding(" superblock %d is partially coded, fragment %d is coded\n",
- i, current_fragment);
} else {
/* not coded; copy this fragment from the prior frame */
s->all_fragments[current_fragment].coding_method =
MODE_COPY;
- debug_block_coding(" superblock %d is partially coded, fragment %d is not coded\n",
- i, current_fragment);
}
} else {
}
s->coded_fragment_list_index++;
s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
- debug_block_coding(" superblock %d is fully coded, fragment %d is coded\n",
- i, current_fragment);
}
}
}
/* end the list of coded C fragments */
s->last_coded_c_fragment = s->coded_fragment_list_index - 1;
- debug_block_coding(" %d total coded fragments, y: %d -> %d, c: %d -> %d\n",
- s->coded_fragment_list_index,
- s->first_coded_y_fragment,
- s->last_coded_y_fragment,
- s->first_coded_c_fragment,
- s->last_coded_c_fragment);
+ for (i = 0; i < s->fragment_count - 1; i++) {
+ s->fast_fragment_list[i] = i + 1;
+ }
+ s->fast_fragment_list[s->fragment_count - 1] = -1;
+
+ if (s->last_coded_y_fragment == -1)
+ s->fragment_list_y_head = -1;
+ else {
+ s->fragment_list_y_head = s->first_coded_y_fragment;
+ s->fast_fragment_list[s->last_coded_y_fragment] = -1;
+ }
+
+ if (s->last_coded_c_fragment == -1)
+ s->fragment_list_c_head = -1;
+ else {
+ s->fragment_list_c_head = s->first_coded_c_fragment;
+ s->fast_fragment_list[s->last_coded_c_fragment] = -1;
+ }
return 0;
}
int coding_mode;
int custom_mode_alphabet[CODING_MODE_COUNT];
- debug_vp3(" vp3: unpacking encoding modes\n");
-
if (s->keyframe) {
- debug_vp3(" keyframe-- all blocks are coded as INTRA\n");
-
for (i = 0; i < s->fragment_count; i++)
s->all_fragments[i].coding_method = MODE_INTRA;
/* fetch the mode coding scheme for this frame */
scheme = get_bits(gb, 3);
- debug_modes(" using mode alphabet %d\n", scheme);
/* is it a custom coding scheme? */
if (scheme == 0) {
- debug_modes(" custom mode alphabet ahead:\n");
+ for (i = 0; i < 8; i++)
+ custom_mode_alphabet[i] = MODE_INTER_NO_MV;
for (i = 0; i < 8; i++)
custom_mode_alphabet[get_bits(gb, 3)] = i;
}
- for (i = 0; i < 8; i++) {
- if(scheme)
- debug_modes(" mode[%d][%d] = %d\n", scheme, i,
- ModeAlphabet[scheme-1][i]);
- else
- debug_modes(" mode[0][%d] = %d\n", i,
- custom_mode_alphabet[i]);
- }
-
/* iterate through all of the macroblocks that contain 1 or more
* coded fragments */
for (i = 0; i < s->u_superblock_start; i++) {
s->all_fragments[current_fragment].coding_method =
coding_mode;
}
-
- debug_modes(" coding method for macroblock starting @ fragment %d = %d\n",
- s->macroblock_fragments[current_macroblock * 6], coding_mode);
}
}
}
int current_macroblock;
int current_fragment;
- debug_vp3(" vp3: unpacking motion vectors\n");
- if (s->keyframe) {
+ if (s->keyframe)
+ return 0;
- debug_vp3(" keyframe-- there are no motion vectors\n");
+ memset(motion_x, 0, 6 * sizeof(int));
+ memset(motion_y, 0, 6 * sizeof(int));
- } else {
+ /* coding mode 0 is the VLC scheme; 1 is the fixed code scheme */
+ coding_mode = get_bits1(gb);
- memset(motion_x, 0, 6 * sizeof(int));
- memset(motion_y, 0, 6 * sizeof(int));
-
- /* coding mode 0 is the VLC scheme; 1 is the fixed code scheme */
- coding_mode = get_bits1(gb);
- debug_vectors(" using %s scheme for unpacking motion vectors\n",
- (coding_mode == 0) ? "VLC" : "fixed-length");
+ /* iterate through all of the macroblocks that contain 1 or more
+ * coded fragments */
+ for (i = 0; i < s->u_superblock_start; i++) {
- /* iterate through all of the macroblocks that contain 1 or more
- * coded fragments */
- for (i = 0; i < s->u_superblock_start; i++) {
+ for (j = 0; j < 4; j++) {
+ current_macroblock = s->superblock_macroblocks[i * 4 + j];
+ if ((current_macroblock == -1) ||
+ (s->macroblock_coding[current_macroblock] == MODE_COPY))
+ continue;
+ if (current_macroblock >= s->macroblock_count) {
+ av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad macroblock number (%d >= %d)\n",
+ current_macroblock, s->macroblock_count);
+ return 1;
+ }
- for (j = 0; j < 4; j++) {
- current_macroblock = s->superblock_macroblocks[i * 4 + j];
- if ((current_macroblock == -1) ||
- (s->macroblock_coding[current_macroblock] == MODE_COPY))
- continue;
- if (current_macroblock >= s->macroblock_count) {
- av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad macroblock number (%d >= %d)\n",
- current_macroblock, s->macroblock_count);
- return 1;
+ current_fragment = s->macroblock_fragments[current_macroblock * 6];
+ if (current_fragment >= s->fragment_count) {
+ av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad fragment number (%d >= %d\n",
+ current_fragment, s->fragment_count);
+ return 1;
+ }
+ switch (s->macroblock_coding[current_macroblock]) {
+
+ case MODE_INTER_PLUS_MV:
+ case MODE_GOLDEN_MV:
+ /* all 6 fragments use the same motion vector */
+ if (coding_mode == 0) {
+ motion_x[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
+ motion_y[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
+ } else {
+ motion_x[0] = fixed_motion_vector_table[get_bits(gb, 6)];
+ motion_y[0] = fixed_motion_vector_table[get_bits(gb, 6)];
}
- current_fragment = s->macroblock_fragments[current_macroblock * 6];
- if (current_fragment >= s->fragment_count) {
- av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad fragment number (%d >= %d\n",
- current_fragment, s->fragment_count);
- return 1;
+ for (k = 1; k < 6; k++) {
+ motion_x[k] = motion_x[0];
+ motion_y[k] = motion_y[0];
}
- switch (s->macroblock_coding[current_macroblock]) {
-
- case MODE_INTER_PLUS_MV:
- case MODE_GOLDEN_MV:
- /* all 6 fragments use the same motion vector */
- if (coding_mode == 0) {
- motion_x[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
- motion_y[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
- } else {
- motion_x[0] = fixed_motion_vector_table[get_bits(gb, 6)];
- motion_y[0] = fixed_motion_vector_table[get_bits(gb, 6)];
- }
- for (k = 1; k < 6; k++) {
- motion_x[k] = motion_x[0];
- motion_y[k] = motion_y[0];
- }
-
- /* vector maintenance, only on MODE_INTER_PLUS_MV */
- if (s->macroblock_coding[current_macroblock] ==
- MODE_INTER_PLUS_MV) {
- prior_last_motion_x = last_motion_x;
- prior_last_motion_y = last_motion_y;
- last_motion_x = motion_x[0];
- last_motion_y = motion_y[0];
- }
- break;
-
- case MODE_INTER_FOURMV:
- /* vector maintenance */
+ /* vector maintenance, only on MODE_INTER_PLUS_MV */
+ if (s->macroblock_coding[current_macroblock] ==
+ MODE_INTER_PLUS_MV) {
prior_last_motion_x = last_motion_x;
prior_last_motion_y = last_motion_y;
+ last_motion_x = motion_x[0];
+ last_motion_y = motion_y[0];
+ }
+ break;
- /* fetch 4 vectors from the bitstream, one for each
- * Y fragment, then average for the C fragment vectors */
- motion_x[4] = motion_y[4] = 0;
- for (k = 0; k < 4; k++) {
- for (l = 0; l < s->coded_fragment_list_index; l++)
- if (s->coded_fragment_list[l] == s->macroblock_fragments[6*current_macroblock + k])
- break;
- if (l < s->coded_fragment_list_index) {
- if (coding_mode == 0) {
- motion_x[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
- motion_y[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
- } else {
- motion_x[k] = fixed_motion_vector_table[get_bits(gb, 6)];
- motion_y[k] = fixed_motion_vector_table[get_bits(gb, 6)];
- }
- last_motion_x = motion_x[k];
- last_motion_y = motion_y[k];
+ case MODE_INTER_FOURMV:
+ /* vector maintenance */
+ prior_last_motion_x = last_motion_x;
+ prior_last_motion_y = last_motion_y;
+
+ /* fetch 4 vectors from the bitstream, one for each
+ * Y fragment, then average for the C fragment vectors */
+ motion_x[4] = motion_y[4] = 0;
+ for (k = 0; k < 4; k++) {
+ for (l = 0; l < s->coded_fragment_list_index; l++)
+ if (s->coded_fragment_list[l] == s->macroblock_fragments[6*current_macroblock + k])
+ break;
+ if (l < s->coded_fragment_list_index) {
+ if (coding_mode == 0) {
+ motion_x[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
+ motion_y[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
} else {
- motion_x[k] = 0;
- motion_y[k] = 0;
+ motion_x[k] = fixed_motion_vector_table[get_bits(gb, 6)];
+ motion_y[k] = fixed_motion_vector_table[get_bits(gb, 6)];
}
- motion_x[4] += motion_x[k];
- motion_y[4] += motion_y[k];
+ last_motion_x = motion_x[k];
+ last_motion_y = motion_y[k];
+ } else {
+ motion_x[k] = 0;
+ motion_y[k] = 0;
}
+ motion_x[4] += motion_x[k];
+ motion_y[4] += motion_y[k];
+ }
- motion_x[5]=
- motion_x[4]= RSHIFT(motion_x[4], 2);
- motion_y[5]=
- motion_y[4]= RSHIFT(motion_y[4], 2);
- break;
+ motion_x[5]=
+ motion_x[4]= RSHIFT(motion_x[4], 2);
+ motion_y[5]=
+ motion_y[4]= RSHIFT(motion_y[4], 2);
+ break;
- case MODE_INTER_LAST_MV:
- /* all 6 fragments use the last motion vector */
- motion_x[0] = last_motion_x;
- motion_y[0] = last_motion_y;
- for (k = 1; k < 6; k++) {
- motion_x[k] = motion_x[0];
- motion_y[k] = motion_y[0];
- }
+ case MODE_INTER_LAST_MV:
+ /* all 6 fragments use the last motion vector */
+ motion_x[0] = last_motion_x;
+ motion_y[0] = last_motion_y;
+ for (k = 1; k < 6; k++) {
+ motion_x[k] = motion_x[0];
+ motion_y[k] = motion_y[0];
+ }
- /* no vector maintenance (last vector remains the
- * last vector) */
- break;
+ /* no vector maintenance (last vector remains the
+ * last vector) */
+ break;
- case MODE_INTER_PRIOR_LAST:
- /* all 6 fragments use the motion vector prior to the
- * last motion vector */
- motion_x[0] = prior_last_motion_x;
- motion_y[0] = prior_last_motion_y;
- for (k = 1; k < 6; k++) {
- motion_x[k] = motion_x[0];
- motion_y[k] = motion_y[0];
- }
+ case MODE_INTER_PRIOR_LAST:
+ /* all 6 fragments use the motion vector prior to the
+ * last motion vector */
+ motion_x[0] = prior_last_motion_x;
+ motion_y[0] = prior_last_motion_y;
+ for (k = 1; k < 6; k++) {
+ motion_x[k] = motion_x[0];
+ motion_y[k] = motion_y[0];
+ }
- /* vector maintenance */
- prior_last_motion_x = last_motion_x;
- prior_last_motion_y = last_motion_y;
- last_motion_x = motion_x[0];
- last_motion_y = motion_y[0];
- break;
+ /* vector maintenance */
+ prior_last_motion_x = last_motion_x;
+ prior_last_motion_y = last_motion_y;
+ last_motion_x = motion_x[0];
+ last_motion_y = motion_y[0];
+ break;
- default:
- /* covers intra, inter without MV, golden without MV */
- memset(motion_x, 0, 6 * sizeof(int));
- memset(motion_y, 0, 6 * sizeof(int));
+ default:
+ /* covers intra, inter without MV, golden without MV */
+ memset(motion_x, 0, 6 * sizeof(int));
+ memset(motion_y, 0, 6 * sizeof(int));
- /* no vector maintenance */
- break;
- }
+ /* no vector maintenance */
+ break;
+ }
- /* assign the motion vectors to the correct fragments */
- debug_vectors(" vectors for macroblock starting @ fragment %d (coding method %d):\n",
- current_fragment,
- s->macroblock_coding[current_macroblock]);
- for (k = 0; k < 6; k++) {
- current_fragment =
- s->macroblock_fragments[current_macroblock * 6 + k];
- if (current_fragment == -1)
- continue;
- if (current_fragment >= s->fragment_count) {
- av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad fragment number (%d >= %d)\n",
- current_fragment, s->fragment_count);
- return 1;
- }
- s->all_fragments[current_fragment].motion_x = motion_x[k];
- s->all_fragments[current_fragment].motion_y = motion_y[k];
- debug_vectors(" vector %d: fragment %d = (%d, %d)\n",
- k, current_fragment, motion_x[k], motion_y[k]);
+ /* assign the motion vectors to the correct fragments */
+ for (k = 0; k < 6; k++) {
+ current_fragment =
+ s->macroblock_fragments[current_macroblock * 6 + k];
+ if (current_fragment == -1)
+ continue;
+ if (current_fragment >= s->fragment_count) {
+ av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vectors(): bad fragment number (%d >= %d)\n",
+ current_fragment, s->fragment_count);
+ return 1;
}
+ s->all_fragments[current_fragment].motion_x = motion_x[k];
+ s->all_fragments[current_fragment].motion_y = motion_y[k];
}
}
}
return 0;
}
+static int unpack_block_qpis(Vp3DecodeContext *s, GetBitContext *gb)
+{
+ int qpi, i, j, bit, run_length, blocks_decoded, num_blocks_at_qpi;
+ int num_blocks = s->coded_fragment_list_index;
+
+ for (qpi = 0; qpi < s->nqps-1 && num_blocks > 0; qpi++) {
+ i = blocks_decoded = num_blocks_at_qpi = 0;
+
+ bit = get_bits1(gb);
+
+ do {
+ run_length = get_vlc2(gb, s->superblock_run_length_vlc.table, 6, 2) + 1;
+ if (run_length == 34)
+ run_length += get_bits(gb, 12);
+ blocks_decoded += run_length;
+
+ if (!bit)
+ num_blocks_at_qpi += run_length;
+
+ for (j = 0; j < run_length; i++) {
+ if (i >= s->coded_fragment_list_index)
+ return -1;
+
+ if (s->all_fragments[s->coded_fragment_list[i]].qpi == qpi) {
+ s->all_fragments[s->coded_fragment_list[i]].qpi += bit;
+ j++;
+ }
+ }
+
+ if (run_length == 4129)
+ bit = get_bits1(gb);
+ else
+ bit ^= 1;
+ } while (blocks_decoded < num_blocks);
+
+ num_blocks -= num_blocks_at_qpi;
+ }
+
+ return 0;
+}
+
/*
* This function is called by unpack_dct_coeffs() to extract the VLCs from
* the bitstream. The VLCs encode tokens which are used to unpack DCT
*/
static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb,
VLC *table, int coeff_index,
- int first_fragment, int last_fragment,
+ int y_plane,
int eob_run)
{
int i;
int zero_run = 0;
DCTELEM coeff = 0;
Vp3Fragment *fragment;
- uint8_t *perm= s->scantable.permutated;
int bits_to_get;
+ int next_fragment;
+ int previous_fragment;
+ int fragment_num;
+ int *list_head;
- if ((first_fragment >= s->fragment_count) ||
- (last_fragment >= s->fragment_count)) {
-
- av_log(s->avctx, AV_LOG_ERROR, " vp3:unpack_vlcs(): bad fragment number (%d -> %d ?)\n",
- first_fragment, last_fragment);
- return 0;
+ /* local references to structure members to avoid repeated deferences */
+ uint8_t *perm= s->scantable.permutated;
+ int *coded_fragment_list = s->coded_fragment_list;
+ Vp3Fragment *all_fragments = s->all_fragments;
+ uint8_t *coeff_counts = s->coeff_counts;
+ VLC_TYPE (*vlc_table)[2] = table->table;
+ int *fast_fragment_list = s->fast_fragment_list;
+
+ if (y_plane) {
+ next_fragment = s->fragment_list_y_head;
+ list_head = &s->fragment_list_y_head;
+ } else {
+ next_fragment = s->fragment_list_c_head;
+ list_head = &s->fragment_list_c_head;
}
- for (i = first_fragment; i <= last_fragment; i++) {
- int fragment_num = s->coded_fragment_list[i];
+ i = next_fragment;
+ previous_fragment = -1; /* this indicates that the previous fragment is actually the list head */
+ while (i != -1) {
+ fragment_num = coded_fragment_list[i];
- if (s->coeff_counts[fragment_num] > coeff_index)
+ if (coeff_counts[fragment_num] > coeff_index) {
+ previous_fragment = i;
+ i = fast_fragment_list[i];
continue;
- fragment = &s->all_fragments[fragment_num];
+ }
+ fragment = &all_fragments[fragment_num];
if (!eob_run) {
/* decode a VLC into a token */
- token = get_vlc2(gb, table->table, 5, 3);
- debug_vlc(" token = %2d, ", token);
+ token = get_vlc2(gb, vlc_table, 5, 3);
/* use the token to get a zero run, a coefficient, and an eob run */
if (token <= 6) {
eob_run = eob_run_base[token];
coeff = zero_run = 0;
} else {
bits_to_get = coeff_get_bits[token];
- if (!bits_to_get)
- coeff = coeff_tables[token][0];
- else
- coeff = coeff_tables[token][get_bits(gb, bits_to_get)];
+ if (bits_to_get)
+ bits_to_get = get_bits(gb, bits_to_get);
+ coeff = coeff_tables[token][bits_to_get];
zero_run = zero_run_base[token];
if (zero_run_get_bits[token])
}
if (!eob_run) {
- s->coeff_counts[fragment_num] += zero_run;
- if (s->coeff_counts[fragment_num] < 64){
+ coeff_counts[fragment_num] += zero_run;
+ if (coeff_counts[fragment_num] < 64){
fragment->next_coeff->coeff= coeff;
- fragment->next_coeff->index= perm[s->coeff_counts[fragment_num]++]; //FIXME perm here already?
+ fragment->next_coeff->index= perm[coeff_counts[fragment_num]++]; //FIXME perm here already?
fragment->next_coeff->next= s->next_coeff;
s->next_coeff->next=NULL;
fragment->next_coeff= s->next_coeff++;
}
- debug_vlc(" fragment %d coeff = %d\n",
- s->coded_fragment_list[i], fragment->next_coeff[coeff_index]);
+ /* previous fragment is now this fragment */
+ previous_fragment = i;
} else {
- s->coeff_counts[fragment_num] |= 128;
- debug_vlc(" fragment %d eob with %d coefficients\n",
- s->coded_fragment_list[i], s->coeff_counts[fragment_num]&127);
+ coeff_counts[fragment_num] |= 128;
eob_run--;
+ /* remove this fragment from the list */
+ if (previous_fragment != -1)
+ fast_fragment_list[previous_fragment] = fast_fragment_list[i];
+ else
+ *list_head = fast_fragment_list[i];
+ /* previous fragment remains unchanged */
}
+
+ i = fast_fragment_list[i];
}
return eob_run;
}
+static void reverse_dc_prediction(Vp3DecodeContext *s,
+ int first_fragment,
+ int fragment_width,
+ int fragment_height);
/*
* This function unpacks all of the DCT coefficient data from the
* bitstream.
int ac_y_table;
int ac_c_table;
int residual_eob_run = 0;
+ VLC *y_tables[64];
+ VLC *c_tables[64];
/* fetch the DC table indexes */
dc_y_table = get_bits(gb, 4);
dc_c_table = get_bits(gb, 4);
/* unpack the Y plane DC coefficients */
- debug_vp3(" vp3: unpacking Y plane DC coefficients using table %d\n",
- dc_y_table);
residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_y_table], 0,
- s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
+ 1, residual_eob_run);
+
+ /* reverse prediction of the Y-plane DC coefficients */
+ reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height);
/* unpack the C plane DC coefficients */
- debug_vp3(" vp3: unpacking C plane DC coefficients using table %d\n",
- dc_c_table);
residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_c_table], 0,
- s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
+ 0, residual_eob_run);
+
+ /* reverse prediction of the C-plane DC coefficients */
+ if (!(s->avctx->flags & CODEC_FLAG_GRAY))
+ {
+ reverse_dc_prediction(s, s->fragment_start[1],
+ s->fragment_width / 2, s->fragment_height / 2);
+ reverse_dc_prediction(s, s->fragment_start[2],
+ s->fragment_width / 2, s->fragment_height / 2);
+ }
/* fetch the AC table indexes */
ac_y_table = get_bits(gb, 4);
ac_c_table = get_bits(gb, 4);
- /* unpack the group 1 AC coefficients (coeffs 1-5) */
+ /* build tables of AC VLC tables */
for (i = 1; i <= 5; i++) {
-
- debug_vp3(" vp3: unpacking level %d Y plane AC coefficients using table %d\n",
- i, ac_y_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_y_table], i,
- s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
-
- debug_vp3(" vp3: unpacking level %d C plane AC coefficients using table %d\n",
- i, ac_c_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_c_table], i,
- s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
+ y_tables[i] = &s->ac_vlc_1[ac_y_table];
+ c_tables[i] = &s->ac_vlc_1[ac_c_table];
}
-
- /* unpack the group 2 AC coefficients (coeffs 6-14) */
for (i = 6; i <= 14; i++) {
-
- debug_vp3(" vp3: unpacking level %d Y plane AC coefficients using table %d\n",
- i, ac_y_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_y_table], i,
- s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
-
- debug_vp3(" vp3: unpacking level %d C plane AC coefficients using table %d\n",
- i, ac_c_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_c_table], i,
- s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
+ y_tables[i] = &s->ac_vlc_2[ac_y_table];
+ c_tables[i] = &s->ac_vlc_2[ac_c_table];
}
-
- /* unpack the group 3 AC coefficients (coeffs 15-27) */
for (i = 15; i <= 27; i++) {
-
- debug_vp3(" vp3: unpacking level %d Y plane AC coefficients using table %d\n",
- i, ac_y_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_y_table], i,
- s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
-
- debug_vp3(" vp3: unpacking level %d C plane AC coefficients using table %d\n",
- i, ac_c_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_c_table], i,
- s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
+ y_tables[i] = &s->ac_vlc_3[ac_y_table];
+ c_tables[i] = &s->ac_vlc_3[ac_c_table];
}
-
- /* unpack the group 4 AC coefficients (coeffs 28-63) */
for (i = 28; i <= 63; i++) {
+ y_tables[i] = &s->ac_vlc_4[ac_y_table];
+ c_tables[i] = &s->ac_vlc_4[ac_c_table];
+ }
- debug_vp3(" vp3: unpacking level %d Y plane AC coefficients using table %d\n",
- i, ac_y_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_y_table], i,
- s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
+ /* decode all AC coefficents */
+ for (i = 1; i <= 63; i++) {
+ if (s->fragment_list_y_head != -1)
+ residual_eob_run = unpack_vlcs(s, gb, y_tables[i], i,
+ 1, residual_eob_run);
- debug_vp3(" vp3: unpacking level %d C plane AC coefficients using table %d\n",
- i, ac_c_table);
- residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_c_table], i,
- s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
+ if (s->fragment_list_c_head != -1)
+ residual_eob_run = unpack_vlcs(s, gb, c_tables[i], i,
+ 0, residual_eob_run);
}
return 0;
*/
#define COMPATIBLE_FRAME(x) \
(compatible_frame[s->all_fragments[x].coding_method] == current_frame_type)
-#define FRAME_CODED(x) (s->all_fragments[x].coding_method != MODE_COPY)
#define DC_COEFF(u) (s->coeffs[u].index ? 0 : s->coeffs[u].coeff) //FIXME do somethin to simplify this
static void reverse_dc_prediction(Vp3DecodeContext *s,
* 2: up-right multiplier
* 3: left multiplier
*/
- int predictor_transform[16][4] = {
+ static const int predictor_transform[16][4] = {
{ 0, 0, 0, 0},
{ 0, 0, 0,128}, // PL
{ 0, 0,128, 0}, // PUR
* from other INTRA blocks. There are 2 golden frame coding types;
* blocks encoding in these modes can only predict from other blocks
* that were encoded with these 1 of these 2 modes. */
- unsigned char compatible_frame[8] = {
+ static const unsigned char compatible_frame[9] = {
1, /* MODE_INTER_NO_MV */
0, /* MODE_INTRA */
1, /* MODE_INTER_PLUS_MV */
1, /* MODE_INTER_PRIOR_MV */
2, /* MODE_USING_GOLDEN */
2, /* MODE_GOLDEN_MV */
- 1 /* MODE_INTER_FOUR_MV */
+ 1, /* MODE_INTER_FOUR_MV */
+ 3 /* MODE_COPY */
};
int current_frame_type;
int transform = 0;
- debug_vp3(" vp3: reversing DC prediction\n");
-
vul = vu = vur = vl = 0;
last_dc[0] = last_dc[1] = last_dc[2] = 0;
current_frame_type =
compatible_frame[s->all_fragments[i].coding_method];
- debug_dc_pred(" frag %d: orig DC = %d, ",
- i, DC_COEFF(i));
transform= 0;
if(x){
l= i-1;
vl = DC_COEFF(l);
- if(FRAME_CODED(l) && COMPATIBLE_FRAME(l))
+ if(COMPATIBLE_FRAME(l))
transform |= PL;
}
if(y){
u= i-fragment_width;
vu = DC_COEFF(u);
- if(FRAME_CODED(u) && COMPATIBLE_FRAME(u))
+ if(COMPATIBLE_FRAME(u))
transform |= PU;
if(x){
ul= i-fragment_width-1;
vul = DC_COEFF(ul);
- if(FRAME_CODED(ul) && COMPATIBLE_FRAME(ul))
+ if(COMPATIBLE_FRAME(ul))
transform |= PUL;
}
if(x + 1 < fragment_width){
ur= i-fragment_width+1;
vur = DC_COEFF(ur);
- if(FRAME_CODED(ur) && COMPATIBLE_FRAME(ur))
+ if(COMPATIBLE_FRAME(ur))
transform |= PUR;
}
}
- debug_dc_pred("transform = %d, ", transform);
-
if (transform == 0) {
/* if there were no fragments to predict from, use last
* DC saved */
predicted_dc = last_dc[current_frame_type];
- debug_dc_pred("from last DC (%d) = %d\n",
- current_frame_type, DC_COEFF(i));
-
} else {
/* apply the appropriate predictor transform */
/* check for outranging on the [ul u l] and
* [ul u ur l] predictors */
- if ((transform == 13) || (transform == 15)) {
+ if ((transform == 15) || (transform == 13)) {
if (FFABS(predicted_dc - vu) > 128)
predicted_dc = vu;
else if (FFABS(predicted_dc - vl) > 128)
else if (FFABS(predicted_dc - vul) > 128)
predicted_dc = vul;
}
-
- debug_dc_pred("from pred DC = %d\n",
- DC_COEFF(i));
}
/* at long last, apply the predictor */
}
}
-
-static void horizontal_filter(unsigned char *first_pixel, int stride,
- int *bounding_values);
-static void vertical_filter(unsigned char *first_pixel, int stride,
- int *bounding_values);
-
/*
* Perform the final rendering for a particular slice of data.
* The slice number ranges from 0..(macroblock_height - 1).
static void render_slice(Vp3DecodeContext *s, int slice)
{
int x;
- int m, n;
int16_t *dequantizer;
DECLARE_ALIGNED_16(DCTELEM, block[64]);
int motion_x = 0xdeadbeef, motion_y = 0xdeadbeef;
motion_source + stride + 1 + d,
stride, 8);
}
- dequantizer = s->qmat[1][plane];
+ dequantizer = s->qmat[s->all_fragments[i].qpi][1][plane];
}else{
- dequantizer = s->qmat[0][plane];
+ dequantizer = s->qmat[s->all_fragments[i].qpi][0][plane];
}
/* dequantize the DCT coefficients */
- debug_idct("fragment %d, coding mode %d, DC = %d, dequant = %d:\n",
- i, s->all_fragments[i].coding_method,
- DC_COEFF(i), dequantizer[0]);
-
if(s->avctx->idct_algo==FF_IDCT_VP3){
Coeff *coeff= s->coeffs + i;
- memset(block, 0, sizeof(block));
+ s->dsp.clear_block(block);
while(coeff->next){
block[coeff->index]= coeff->coeff * dequantizer[coeff->index];
coeff= coeff->next;
}
}else{
Coeff *coeff= s->coeffs + i;
- memset(block, 0, sizeof(block));
+ s->dsp.clear_block(block);
while(coeff->next){
block[coeff->index]= (coeff->coeff * dequantizer[coeff->index] + 2)>>2;
coeff= coeff->next;
stride,
block);
}
-
- debug_idct("block after idct_%s():\n",
- (s->all_fragments[i].coding_method == MODE_INTRA)?
- "put" : "add");
- for (m = 0; m < 8; m++) {
- for (n = 0; n < 8; n++) {
- debug_idct(" %3d", *(output_plane +
- s->all_fragments[i].first_pixel + (m * stride + n)));
- }
- debug_idct("\n");
- }
- debug_idct("\n");
-
} else {
/* copy directly from the previous frame */
emms_c();
}
-static void horizontal_filter(unsigned char *first_pixel, int stride,
- int *bounding_values)
-{
- unsigned char *end;
- int filter_value;
-
- for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
- filter_value =
- (first_pixel[-2] - first_pixel[ 1])
- +3*(first_pixel[ 0] - first_pixel[-1]);
- filter_value = bounding_values[(filter_value + 4) >> 3];
- first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
- first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
- }
-}
-
-static void vertical_filter(unsigned char *first_pixel, int stride,
- int *bounding_values)
-{
- unsigned char *end;
- int filter_value;
- const int nstride= -stride;
-
- for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
- filter_value =
- (first_pixel[2 * nstride] - first_pixel[ stride])
- +3*(first_pixel[0 ] - first_pixel[nstride]);
- filter_value = bounding_values[(filter_value + 4) >> 3];
- first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
- first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
- }
-}
-
static void apply_loop_filter(Vp3DecodeContext *s)
{
int plane;
for (y = 0; y < height; y++) {
for (x = 0; x < width; x++) {
- /* do not perform left edge filter for left columns frags */
- if ((x > 0) &&
- (s->all_fragments[fragment].coding_method != MODE_COPY)) {
- horizontal_filter(
- plane_data + s->all_fragments[fragment].first_pixel,
- stride, bounding_values);
- }
+ /* This code basically just deblocks on the edges of coded blocks.
+ * However, it has to be much more complicated because of the
+ * braindamaged deblock ordering used in VP3/Theora. Order matters
+ * because some pixels get filtered twice. */
+ if( s->all_fragments[fragment].coding_method != MODE_COPY )
+ {
+ /* do not perform left edge filter for left columns frags */
+ if (x > 0) {
+ s->dsp.vp3_h_loop_filter(
+ plane_data + s->all_fragments[fragment].first_pixel,
+ stride, bounding_values);
+ }
- /* do not perform top edge filter for top row fragments */
- if ((y > 0) &&
- (s->all_fragments[fragment].coding_method != MODE_COPY)) {
- vertical_filter(
- plane_data + s->all_fragments[fragment].first_pixel,
- stride, bounding_values);
- }
+ /* do not perform top edge filter for top row fragments */
+ if (y > 0) {
+ s->dsp.vp3_v_loop_filter(
+ plane_data + s->all_fragments[fragment].first_pixel,
+ stride, bounding_values);
+ }
- /* do not perform right edge filter for right column
- * fragments or if right fragment neighbor is also coded
- * in this frame (it will be filtered in next iteration) */
- if ((x < width - 1) &&
- (s->all_fragments[fragment].coding_method != MODE_COPY) &&
- (s->all_fragments[fragment + 1].coding_method == MODE_COPY)) {
- horizontal_filter(
- plane_data + s->all_fragments[fragment + 1].first_pixel,
- stride, bounding_values);
- }
+ /* do not perform right edge filter for right column
+ * fragments or if right fragment neighbor is also coded
+ * in this frame (it will be filtered in next iteration) */
+ if ((x < width - 1) &&
+ (s->all_fragments[fragment + 1].coding_method == MODE_COPY)) {
+ s->dsp.vp3_h_loop_filter(
+ plane_data + s->all_fragments[fragment + 1].first_pixel,
+ stride, bounding_values);
+ }
- /* do not perform bottom edge filter for bottom row
- * fragments or if bottom fragment neighbor is also coded
- * in this frame (it will be filtered in the next row) */
- if ((y < height - 1) &&
- (s->all_fragments[fragment].coding_method != MODE_COPY) &&
- (s->all_fragments[fragment + width].coding_method == MODE_COPY)) {
- vertical_filter(
- plane_data + s->all_fragments[fragment + width].first_pixel,
- stride, bounding_values);
+ /* do not perform bottom edge filter for bottom row
+ * fragments or if bottom fragment neighbor is also coded
+ * in this frame (it will be filtered in the next row) */
+ if ((y < height - 1) &&
+ (s->all_fragments[fragment + width].coding_method == MODE_COPY)) {
+ s->dsp.vp3_v_loop_filter(
+ plane_data + s->all_fragments[fragment + width].first_pixel,
+ stride, bounding_values);
+ }
}
fragment++;
*/
static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s)
{
+#define Y_INITIAL(chroma_shift) s->flipped_image ? 1 : s->fragment_height >> chroma_shift
+#define Y_FINISHED(chroma_shift) s->flipped_image ? y <= s->fragment_height >> chroma_shift : y > 0
int i, x, y;
+ const int y_inc = s->flipped_image ? 1 : -1;
/* figure out the first pixel addresses for each of the fragments */
/* Y plane */
i = 0;
- for (y = s->fragment_height; y > 0; y--) {
- for (x = 0; x < s->fragment_width; x++) {
- s->all_fragments[i++].first_pixel =
- s->golden_frame.linesize[0] * y * FRAGMENT_PIXELS -
- s->golden_frame.linesize[0] +
- x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
- }
- }
-
- /* U plane */
- i = s->fragment_start[1];
- for (y = s->fragment_height / 2; y > 0; y--) {
- for (x = 0; x < s->fragment_width / 2; x++) {
- s->all_fragments[i++].first_pixel =
- s->golden_frame.linesize[1] * y * FRAGMENT_PIXELS -
- s->golden_frame.linesize[1] +
- x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
- }
- }
-
- /* V plane */
- i = s->fragment_start[2];
- for (y = s->fragment_height / 2; y > 0; y--) {
- for (x = 0; x < s->fragment_width / 2; x++) {
- s->all_fragments[i++].first_pixel =
- s->golden_frame.linesize[2] * y * FRAGMENT_PIXELS -
- s->golden_frame.linesize[2] +
- x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
- }
- }
-}
-
-/* FIXME: this should be merged with the above! */
-static void theora_calculate_pixel_addresses(Vp3DecodeContext *s)
-{
-
- int i, x, y;
-
- /* figure out the first pixel addresses for each of the fragments */
- /* Y plane */
- i = 0;
- for (y = 1; y <= s->fragment_height; y++) {
+ for (y = Y_INITIAL(0); Y_FINISHED(0); y += y_inc) {
for (x = 0; x < s->fragment_width; x++) {
s->all_fragments[i++].first_pixel =
s->golden_frame.linesize[0] * y * FRAGMENT_PIXELS -
s->golden_frame.linesize[0] +
x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
}
}
/* U plane */
i = s->fragment_start[1];
- for (y = 1; y <= s->fragment_height / 2; y++) {
+ for (y = Y_INITIAL(1); Y_FINISHED(1); y += y_inc) {
for (x = 0; x < s->fragment_width / 2; x++) {
s->all_fragments[i++].first_pixel =
s->golden_frame.linesize[1] * y * FRAGMENT_PIXELS -
s->golden_frame.linesize[1] +
x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
}
}
/* V plane */
i = s->fragment_start[2];
- for (y = 1; y <= s->fragment_height / 2; y++) {
+ for (y = Y_INITIAL(1); Y_FINISHED(1); y += y_inc) {
for (x = 0; x < s->fragment_width / 2; x++) {
s->all_fragments[i++].first_pixel =
s->golden_frame.linesize[2] * y * FRAGMENT_PIXELS -
s->golden_frame.linesize[2] +
x * FRAGMENT_PIXELS;
- debug_init(" fragment %d, first pixel @ %d\n",
- i-1, s->all_fragments[i-1].first_pixel);
}
}
}
s->version = 1;
s->avctx = avctx;
- s->width = (avctx->width + 15) & 0xFFFFFFF0;
- s->height = (avctx->height + 15) & 0xFFFFFFF0;
+ s->width = FFALIGN(avctx->width, 16);
+ s->height = FFALIGN(avctx->height, 16);
avctx->pix_fmt = PIX_FMT_YUV420P;
+ avctx->chroma_sample_location = AVCHROMA_LOC_CENTER;
if(avctx->idct_algo==FF_IDCT_AUTO)
avctx->idct_algo=FF_IDCT_VP3;
dsputil_init(&s->dsp, avctx);
/* initialize to an impossible value which will force a recalculation
* in the first frame decode */
- s->quality_index = -1;
+ for (i = 0; i < 3; i++)
+ s->qps[i] = -1;
s->y_superblock_width = (s->width + 31) / 32;
s->y_superblock_height = (s->height + 31) / 32;
s->fragment_start[1] = s->fragment_width * s->fragment_height;
s->fragment_start[2] = s->fragment_width * s->fragment_height * 5 / 4;
- debug_init(" Y plane: %d x %d\n", s->width, s->height);
- debug_init(" C plane: %d x %d\n", c_width, c_height);
- debug_init(" Y superblocks: %d x %d, %d total\n",
- s->y_superblock_width, s->y_superblock_height, y_superblock_count);
- debug_init(" C superblocks: %d x %d, %d total\n",
- s->c_superblock_width, s->c_superblock_height, c_superblock_count);
- debug_init(" total superblocks = %d, U starts @ %d, V starts @ %d\n",
- s->superblock_count, s->u_superblock_start, s->v_superblock_start);
- debug_init(" macroblocks: %d x %d, %d total\n",
- s->macroblock_width, s->macroblock_height, s->macroblock_count);
- debug_init(" %d fragments, %d x %d, u starts @ %d, v starts @ %d\n",
- s->fragment_count,
- s->fragment_width,
- s->fragment_height,
- s->fragment_start[1],
- s->fragment_start[2]);
-
s->all_fragments = av_malloc(s->fragment_count * sizeof(Vp3Fragment));
s->coeff_counts = av_malloc(s->fragment_count * sizeof(*s->coeff_counts));
s->coeffs = av_malloc(s->fragment_count * sizeof(Coeff) * 65);
s->coded_fragment_list = av_malloc(s->fragment_count * sizeof(int));
+ s->fast_fragment_list = av_malloc(s->fragment_count * sizeof(int));
s->pixel_addresses_initialized = 0;
+ if (!s->superblock_coding || !s->all_fragments || !s->coeff_counts ||
+ !s->coeffs || !s->coded_fragment_list || !s->fast_fragment_list) {
+ vp3_decode_end(avctx);
+ return -1;
+ }
if (!s->theora_tables)
{
for (i = 0; i < 16; i++) {
/* DC histograms */
- init_vlc(&s->dc_vlc[i], 5, 32,
+ if (init_vlc(&s->dc_vlc[i], 5, 32,
&s->huffman_table[i][0][1], 4, 2,
- &s->huffman_table[i][0][0], 4, 2, 0);
+ &s->huffman_table[i][0][0], 4, 2, 0) < 0)
+ goto vlc_fail;
/* group 1 AC histograms */
- init_vlc(&s->ac_vlc_1[i], 5, 32,
+ if (init_vlc(&s->ac_vlc_1[i], 5, 32,
&s->huffman_table[i+16][0][1], 4, 2,
- &s->huffman_table[i+16][0][0], 4, 2, 0);
+ &s->huffman_table[i+16][0][0], 4, 2, 0) < 0)
+ goto vlc_fail;
/* group 2 AC histograms */
- init_vlc(&s->ac_vlc_2[i], 5, 32,
+ if (init_vlc(&s->ac_vlc_2[i], 5, 32,
&s->huffman_table[i+16*2][0][1], 4, 2,
- &s->huffman_table[i+16*2][0][0], 4, 2, 0);
+ &s->huffman_table[i+16*2][0][0], 4, 2, 0) < 0)
+ goto vlc_fail;
/* group 3 AC histograms */
- init_vlc(&s->ac_vlc_3[i], 5, 32,
+ if (init_vlc(&s->ac_vlc_3[i], 5, 32,
&s->huffman_table[i+16*3][0][1], 4, 2,
- &s->huffman_table[i+16*3][0][0], 4, 2, 0);
+ &s->huffman_table[i+16*3][0][0], 4, 2, 0) < 0)
+ goto vlc_fail;
/* group 4 AC histograms */
- init_vlc(&s->ac_vlc_4[i], 5, 32,
+ if (init_vlc(&s->ac_vlc_4[i], 5, 32,
&s->huffman_table[i+16*4][0][1], 4, 2,
- &s->huffman_table[i+16*4][0][0], 4, 2, 0);
+ &s->huffman_table[i+16*4][0][0], 4, 2, 0) < 0)
+ goto vlc_fail;
}
}
s->superblock_macroblocks = av_malloc(s->superblock_count * 4 * sizeof(int));
s->macroblock_fragments = av_malloc(s->macroblock_count * 6 * sizeof(int));
s->macroblock_coding = av_malloc(s->macroblock_count + 1);
+ if (!s->superblock_fragments || !s->superblock_macroblocks ||
+ !s->macroblock_fragments || !s->macroblock_coding) {
+ vp3_decode_end(avctx);
+ return -1;
+ }
init_block_mapping(s);
for (i = 0; i < 3; i++) {
}
return 0;
+
+vlc_fail:
+ av_log(avctx, AV_LOG_FATAL, "Invalid huffman table\n");
+ return -1;
}
/*
*/
static int vp3_decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
- const uint8_t *buf, int buf_size)
+ AVPacket *avpkt)
{
+ const uint8_t *buf = avpkt->data;
+ int buf_size = avpkt->size;
Vp3DecodeContext *s = avctx->priv_data;
GetBitContext gb;
static int counter = 0;
s->keyframe = !get_bits1(&gb);
if (!s->theora)
skip_bits(&gb, 1);
- s->last_quality_index = s->quality_index;
+ for (i = 0; i < 3; i++)
+ s->last_qps[i] = s->qps[i];
- s->nqis=0;
+ s->nqps=0;
do{
- s->qis[s->nqis++]= get_bits(&gb, 6);
- } while(s->theora >= 0x030200 && s->nqis<3 && get_bits1(&gb));
-
- s->quality_index= s->qis[0];
+ s->qps[s->nqps++]= get_bits(&gb, 6);
+ } while(s->theora >= 0x030200 && s->nqps<3 && get_bits1(&gb));
+ for (i = s->nqps; i < 3; i++)
+ s->qps[i] = -1;
if (s->avctx->debug & FF_DEBUG_PICT_INFO)
av_log(s->avctx, AV_LOG_INFO, " VP3 %sframe #%d: Q index = %d\n",
- s->keyframe?"key":"", counter, s->quality_index);
+ s->keyframe?"key":"", counter, s->qps[0]);
counter++;
- if (s->quality_index != s->last_quality_index) {
- init_dequantizer(s);
+ if (s->qps[0] != s->last_qps[0])
init_loop_filter(s);
- }
+
+ for (i = 0; i < s->nqps; i++)
+ // reinit all dequantizers if the first one changed, because
+ // the DC of the first quantizer must be used for all matrices
+ if (s->qps[i] != s->last_qps[i] || s->qps[0] != s->last_qps[0])
+ init_dequantizer(s, i);
+
+ if (avctx->skip_frame >= AVDISCARD_NONKEY && !s->keyframe)
+ return buf_size;
if (s->keyframe) {
if (!s->theora)
/* time to figure out pixel addresses? */
if (!s->pixel_addresses_initialized)
{
- if (!s->flipped_image)
- vp3_calculate_pixel_addresses(s);
- else
- theora_calculate_pixel_addresses(s);
+ vp3_calculate_pixel_addresses(s);
s->pixel_addresses_initialized = 1;
}
} else {
init_frame(s, &gb);
-#if KEYFRAMES_ONLY
-if (!s->keyframe) {
-
- memcpy(s->current_frame.data[0], s->golden_frame.data[0],
- s->current_frame.linesize[0] * s->height);
- memcpy(s->current_frame.data[1], s->golden_frame.data[1],
- s->current_frame.linesize[1] * s->height / 2);
- memcpy(s->current_frame.data[2], s->golden_frame.data[2],
- s->current_frame.linesize[2] * s->height / 2);
-
-} else {
-#endif
-
if (unpack_superblocks(s, &gb)){
av_log(s->avctx, AV_LOG_ERROR, "error in unpack_superblocks\n");
return -1;
av_log(s->avctx, AV_LOG_ERROR, "error in unpack_vectors\n");
return -1;
}
+ if (unpack_block_qpis(s, &gb)){
+ av_log(s->avctx, AV_LOG_ERROR, "error in unpack_block_qpis\n");
+ return -1;
+ }
if (unpack_dct_coeffs(s, &gb)){
av_log(s->avctx, AV_LOG_ERROR, "error in unpack_dct_coeffs\n");
return -1;
}
- reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height);
- if ((avctx->flags & CODEC_FLAG_GRAY) == 0) {
- reverse_dc_prediction(s, s->fragment_start[1],
- s->fragment_width / 2, s->fragment_height / 2);
- reverse_dc_prediction(s, s->fragment_start[2],
- s->fragment_width / 2, s->fragment_height / 2);
- }
-
for (i = 0; i < s->macroblock_height; i++)
render_slice(s, i);
apply_loop_filter(s);
-#if KEYFRAMES_ONLY
-}
-#endif
*data_size=sizeof(AVFrame);
*(AVFrame*)data= s->current_frame;
av_free(s->coeff_counts);
av_free(s->coeffs);
av_free(s->coded_fragment_list);
+ av_free(s->fast_fragment_list);
av_free(s->superblock_fragments);
av_free(s->superblock_macroblocks);
av_free(s->macroblock_fragments);
}
s->huff_code_size++;
s->hbits <<= 1;
- read_huffman_tree(avctx, gb);
+ if (read_huffman_tree(avctx, gb))
+ return -1;
s->hbits |= 1;
- read_huffman_tree(avctx, gb);
+ if (read_huffman_tree(avctx, gb))
+ return -1;
s->hbits >>= 1;
s->huff_code_size--;
}
return 0;
}
-#ifdef CONFIG_THEORA_DECODER
+#if CONFIG_THEORA_DECODER
static int theora_decode_header(AVCodecContext *avctx, GetBitContext *gb)
{
Vp3DecodeContext *s = avctx->priv_data;
av_log(avctx, AV_LOG_DEBUG, "Old (<alpha3) Theora bitstream, flipped image\n");
}
- s->width = get_bits(gb, 16) << 4;
- s->height = get_bits(gb, 16) << 4;
+ visible_width = s->width = get_bits(gb, 16) << 4;
+ visible_height = s->height = get_bits(gb, 16) << 4;
if(avcodec_check_dimensions(avctx, s->width, s->height)){
av_log(avctx, AV_LOG_ERROR, "Invalid dimensions (%dx%d)\n", s->width, s->height);
skip_bits(gb, 32); /* total number of macroblocks in a frame */
}
- visible_width = get_bits_long(gb, 24);
- visible_height = get_bits_long(gb, 24);
-
if (s->theora >= 0x030200) {
+ visible_width = get_bits_long(gb, 24);
+ visible_height = get_bits_long(gb, 24);
+
skip_bits(gb, 8); /* offset x */
skip_bits(gb, 8); /* offset y */
}
if (s->theora >= 0x030200) {
n = get_bits(gb, 3);
/* loop filter limit values table */
- for (i = 0; i < 64; i++)
+ for (i = 0; i < 64; i++) {
s->filter_limit_values[i] = get_bits(gb, n);
+ if (s->filter_limit_values[i] > 127) {
+ av_log(avctx, AV_LOG_ERROR, "filter limit value too large (%i > 127), clamping\n", s->filter_limit_values[i]);
+ s->filter_limit_values[i] = 127;
+ }
+ }
}
if (s->theora >= 0x030200)
s->huff_code_size = 1;
if (!get_bits1(gb)) {
s->hbits = 0;
- read_huffman_tree(avctx, gb);
+ if(read_huffman_tree(avctx, gb))
+ return -1;
s->hbits = 1;
- read_huffman_tree(avctx, gb);
+ if(read_huffman_tree(avctx, gb))
+ return -1;
}
}
return 0;
}
-static int theora_decode_init(AVCodecContext *avctx)
+static av_cold int theora_decode_init(AVCodecContext *avctx)
{
Vp3DecodeContext *s = avctx->priv_data;
GetBitContext gb;
}
for(i=0;i<3;i++) {
- init_get_bits(&gb, header_start[i], header_len[i]);
+ init_get_bits(&gb, header_start[i], header_len[i] * 8);
ptype = get_bits(&gb, 8);
- debug_vp3("Theora headerpacket type: %x\n", ptype);
if (!(ptype & 0x80))
{
}
// FIXME: Check for this as well.
- skip_bits(&gb, 6*8); /* "theora" */
+ skip_bits_long(&gb, 6*8); /* "theora" */
switch(ptype)
{
// theora_decode_comments(avctx, gb);
break;
case 0x82:
- theora_decode_tables(avctx, &gb);
+ if (theora_decode_tables(avctx, &gb))
+ return -1;
break;
default:
av_log(avctx, AV_LOG_ERROR, "Unknown Theora config packet: %d\n", ptype&~0x80);
break;
}
- if(8*header_len[i] != get_bits_count(&gb))
- av_log(avctx, AV_LOG_ERROR, "%d bits left in packet %X\n", 8*header_len[i] - get_bits_count(&gb), ptype);
+ if(ptype != 0x81 && 8*header_len[i] != get_bits_count(&gb))
+ av_log(avctx, AV_LOG_WARNING, "%d bits left in packet %X\n", 8*header_len[i] - get_bits_count(&gb), ptype);
if (s->theora < 0x030200)
break;
}
- vp3_decode_init(avctx);
- return 0;
+ return vp3_decode_init(avctx);
}
AVCodec theora_decoder = {
NULL,
vp3_decode_end,
vp3_decode_frame,
- 0,
+ CODEC_CAP_DR1,
NULL,
.long_name = NULL_IF_CONFIG_SMALL("Theora"),
};
NULL,
vp3_decode_end,
vp3_decode_frame,
- 0,
+ CODEC_CAP_DR1,
NULL,
.long_name = NULL_IF_CONFIG_SMALL("On2 VP3"),
};
projects / ffmpeg / blobdiff