00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00027 #include "common.h"
00028 #include "dsputil.h"
00029 #include "avcodec.h"
00030 #include "mpegvideo.h"
00031 #include "h264data.h"
00032 #include "golomb.h"
00033
00034 #include "cabac.h"
00035
00036 #undef NDEBUG
00037 #include <assert.h>
00038
00039 #define interlaced_dct interlaced_dct_is_a_bad_name
00040 #define mb_intra mb_intra_isnt_initalized_see_mb_type
00041
00042 #define LUMA_DC_BLOCK_INDEX 25
00043 #define CHROMA_DC_BLOCK_INDEX 26
00044
00045 #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
00046 #define COEFF_TOKEN_VLC_BITS 8
00047 #define TOTAL_ZEROS_VLC_BITS 9
00048 #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
00049 #define RUN_VLC_BITS 3
00050 #define RUN7_VLC_BITS 6
00051
00052 #define MAX_SPS_COUNT 32
00053 #define MAX_PPS_COUNT 256
00054
00055 #define MAX_MMCO_COUNT 66
00056
00060 typedef struct SPS{
00061
00062 int profile_idc;
00063 int level_idc;
00064 int transform_bypass;
00065 int log2_max_frame_num;
00066 int poc_type;
00067 int log2_max_poc_lsb;
00068 int delta_pic_order_always_zero_flag;
00069 int offset_for_non_ref_pic;
00070 int offset_for_top_to_bottom_field;
00071 int poc_cycle_length;
00072 int ref_frame_count;
00073 int gaps_in_frame_num_allowed_flag;
00074 int mb_width;
00075 int mb_height;
00076 int frame_mbs_only_flag;
00077 int mb_aff;
00078 int direct_8x8_inference_flag;
00079 int crop;
00080 int crop_left;
00081 int crop_right;
00082 int crop_top;
00083 int crop_bottom;
00084 int vui_parameters_present_flag;
00085 AVRational sar;
00086 int timing_info_present_flag;
00087 uint32_t num_units_in_tick;
00088 uint32_t time_scale;
00089 int fixed_frame_rate_flag;
00090 short offset_for_ref_frame[256];
00091 int bitstream_restriction_flag;
00092 int num_reorder_frames;
00093 }SPS;
00094
00098 typedef struct PPS{
00099 int sps_id;
00100 int cabac;
00101 int pic_order_present;
00102 int slice_group_count;
00103 int mb_slice_group_map_type;
00104 int ref_count[2];
00105 int weighted_pred;
00106 int weighted_bipred_idc;
00107 int init_qp;
00108 int init_qs;
00109 int chroma_qp_index_offset;
00110 int deblocking_filter_parameters_present;
00111 int constrained_intra_pred;
00112 int redundant_pic_cnt_present;
00113 int transform_8x8_mode;
00114 }PPS;
00115
00119 typedef enum MMCOOpcode{
00120 MMCO_END=0,
00121 MMCO_SHORT2UNUSED,
00122 MMCO_LONG2UNUSED,
00123 MMCO_SHORT2LONG,
00124 MMCO_SET_MAX_LONG,
00125 MMCO_RESET,
00126 MMCO_LONG,
00127 } MMCOOpcode;
00128
00132 typedef struct MMCO{
00133 MMCOOpcode opcode;
00134 int short_frame_num;
00135 int long_index;
00136 } MMCO;
00137
00141 typedef struct H264Context{
00142 MpegEncContext s;
00143 int nal_ref_idc;
00144 int nal_unit_type;
00145 #define NAL_SLICE 1
00146 #define NAL_DPA 2
00147 #define NAL_DPB 3
00148 #define NAL_DPC 4
00149 #define NAL_IDR_SLICE 5
00150 #define NAL_SEI 6
00151 #define NAL_SPS 7
00152 #define NAL_PPS 8
00153 #define NAL_PICTURE_DELIMITER 9
00154 #define NAL_FILTER_DATA 10
00155 uint8_t *rbsp_buffer;
00156 int rbsp_buffer_size;
00157
00161 int is_avc;
00162 int got_avcC;
00163 int nal_length_size;
00164
00165 int chroma_qp;
00166
00167 int prev_mb_skipped;
00168
00169
00170 int chroma_pred_mode;
00171 int intra16x16_pred_mode;
00172
00173 int top_mb_xy;
00174 int left_mb_xy[2];
00175
00176 int8_t intra4x4_pred_mode_cache[5*8];
00177 int8_t (*intra4x4_pred_mode)[8];
00178 void (*pred4x4 [9+3])(uint8_t *src, uint8_t *topright, int stride);
00179 void (*pred8x8l [9+3])(uint8_t *src, int topleft, int topright, int stride);
00180 void (*pred8x8 [4+3])(uint8_t *src, int stride);
00181 void (*pred16x16[4+3])(uint8_t *src, int stride);
00182 unsigned int topleft_samples_available;
00183 unsigned int top_samples_available;
00184 unsigned int topright_samples_available;
00185 unsigned int left_samples_available;
00186 uint8_t (*top_borders[2])[16+2*8];
00187 uint8_t left_border[2*(17+2*9)];
00188
00193 uint8_t non_zero_count_cache[6*8] __align8;
00194 uint8_t (*non_zero_count)[16];
00195
00199 int16_t mv_cache[2][5*8][2] __align8;
00200 int8_t ref_cache[2][5*8] __align8;
00201 #define LIST_NOT_USED -1 //FIXME rename?
00202 #define PART_NOT_AVAILABLE -2
00203
00207 int mv_cache_clean[2];
00208
00212 int neighbor_transform_size;
00213
00218 int block_offset[2*(16+8)];
00219
00220 uint32_t *mb2b_xy;
00221 uint32_t *mb2b8_xy;
00222 int b_stride;
00223 int b8_stride;
00224
00225 int halfpel_flag;
00226 int thirdpel_flag;
00227
00228 int unknown_svq3_flag;
00229 int next_slice_index;
00230
00231 SPS sps_buffer[MAX_SPS_COUNT];
00232 SPS sps;
00233
00234 PPS pps_buffer[MAX_PPS_COUNT];
00238 PPS pps;
00239
00240 uint16_t (*dequant4_coeff)[16];
00241 uint16_t (*dequant8_coeff)[64];
00242
00243 int slice_num;
00244 uint8_t *slice_table_base;
00245 uint8_t *slice_table;
00246 int slice_type;
00247 int slice_type_fixed;
00248
00249
00250 int mb_aff_frame;
00251 int mb_field_decoding_flag;
00252
00253 int sub_mb_type[4];
00254
00255
00256 int poc_lsb;
00257 int poc_msb;
00258 int delta_poc_bottom;
00259 int delta_poc[2];
00260 int frame_num;
00261 int prev_poc_msb;
00262 int prev_poc_lsb;
00263 int frame_num_offset;
00264 int prev_frame_num_offset;
00265 int prev_frame_num;
00266
00270 int curr_pic_num;
00271
00275 int max_pic_num;
00276
00277
00278 int use_weight;
00279 int use_weight_chroma;
00280 int luma_log2_weight_denom;
00281 int chroma_log2_weight_denom;
00282 int luma_weight[2][16];
00283 int luma_offset[2][16];
00284 int chroma_weight[2][16][2];
00285 int chroma_offset[2][16][2];
00286 int implicit_weight[16][16];
00287
00288
00289 int deblocking_filter;
00290 int slice_alpha_c0_offset;
00291 int slice_beta_offset;
00292
00293 int redundant_pic_count;
00294
00295 int direct_spatial_mv_pred;
00296 int dist_scale_factor[16];
00297 int map_col_to_list0[2][16];
00298
00302 int ref_count[2];
00303 Picture *short_ref[32];
00304 Picture *long_ref[32];
00305 Picture default_ref_list[2][32];
00306 Picture ref_list[2][32];
00307 Picture field_ref_list[2][32];
00308 Picture *delayed_pic[16];
00309 Picture *delayed_output_pic;
00310
00314 MMCO mmco[MAX_MMCO_COUNT];
00315 int mmco_index;
00316
00317 int long_ref_count;
00318 int short_ref_count;
00319
00320
00321 GetBitContext intra_gb;
00322 GetBitContext inter_gb;
00323 GetBitContext *intra_gb_ptr;
00324 GetBitContext *inter_gb_ptr;
00325
00326 DCTELEM mb[16*24] __align8;
00327
00331 CABACContext cabac;
00332 uint8_t cabac_state[460];
00333 int cabac_init_idc;
00334
00335
00336 uint16_t *cbp_table;
00337 int top_cbp;
00338 int left_cbp;
00339
00340 uint8_t *chroma_pred_mode_table;
00341 int last_qscale_diff;
00342 int16_t (*mvd_table[2])[2];
00343 int16_t mvd_cache[2][5*8][2] __align8;
00344 uint8_t *direct_table;
00345 uint8_t direct_cache[5*8];
00346
00347 uint8_t zigzag_scan[16];
00348 uint8_t field_scan[16];
00349 const uint8_t *zigzag_scan_q0;
00350 const uint8_t *field_scan_q0;
00351 }H264Context;
00352
00353 static VLC coeff_token_vlc[4];
00354 static VLC chroma_dc_coeff_token_vlc;
00355
00356 static VLC total_zeros_vlc[15];
00357 static VLC chroma_dc_total_zeros_vlc[3];
00358
00359 static VLC run_vlc[6];
00360 static VLC run7_vlc;
00361
00362 static void svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
00363 static void svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
00364 static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
00365
00366 static inline uint32_t pack16to32(int a, int b){
00367 #ifdef WORDS_BIGENDIAN
00368 return (b&0xFFFF) + (a<<16);
00369 #else
00370 return (a&0xFFFF) + (b<<16);
00371 #endif
00372 }
00373
00380 static inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){
00381 uint8_t *p= (uint8_t*)vp;
00382 assert(size==1 || size==4);
00383
00384 w *= size;
00385 stride *= size;
00386
00387 assert((((int)vp)&(FFMIN(w, STRIDE_ALIGN)-1)) == 0);
00388 assert((stride&(w-1))==0);
00389
00390 if(w==2 && h==2){
00391 *(uint16_t*)(p + 0)=
00392 *(uint16_t*)(p + stride)= size==4 ? val : val*0x0101;
00393 }else if(w==2 && h==4){
00394 *(uint16_t*)(p + 0*stride)=
00395 *(uint16_t*)(p + 1*stride)=
00396 *(uint16_t*)(p + 2*stride)=
00397 *(uint16_t*)(p + 3*stride)= size==4 ? val : val*0x0101;
00398 }else if(w==4 && h==1){
00399 *(uint32_t*)(p + 0*stride)= size==4 ? val : val*0x01010101;
00400 }else if(w==4 && h==2){
00401 *(uint32_t*)(p + 0*stride)=
00402 *(uint32_t*)(p + 1*stride)= size==4 ? val : val*0x01010101;
00403 }else if(w==4 && h==4){
00404 *(uint32_t*)(p + 0*stride)=
00405 *(uint32_t*)(p + 1*stride)=
00406 *(uint32_t*)(p + 2*stride)=
00407 *(uint32_t*)(p + 3*stride)= size==4 ? val : val*0x01010101;
00408 }else if(w==8 && h==1){
00409 *(uint32_t*)(p + 0)=
00410 *(uint32_t*)(p + 4)= size==4 ? val : val*0x01010101;
00411 }else if(w==8 && h==2){
00412 *(uint32_t*)(p + 0 + 0*stride)=
00413 *(uint32_t*)(p + 4 + 0*stride)=
00414 *(uint32_t*)(p + 0 + 1*stride)=
00415 *(uint32_t*)(p + 4 + 1*stride)= size==4 ? val : val*0x01010101;
00416 }else if(w==8 && h==4){
00417 *(uint64_t*)(p + 0*stride)=
00418 *(uint64_t*)(p + 1*stride)=
00419 *(uint64_t*)(p + 2*stride)=
00420 *(uint64_t*)(p + 3*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
00421 }else if(w==16 && h==2){
00422 *(uint64_t*)(p + 0+0*stride)=
00423 *(uint64_t*)(p + 8+0*stride)=
00424 *(uint64_t*)(p + 0+1*stride)=
00425 *(uint64_t*)(p + 8+1*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
00426 }else if(w==16 && h==4){
00427 *(uint64_t*)(p + 0+0*stride)=
00428 *(uint64_t*)(p + 8+0*stride)=
00429 *(uint64_t*)(p + 0+1*stride)=
00430 *(uint64_t*)(p + 8+1*stride)=
00431 *(uint64_t*)(p + 0+2*stride)=
00432 *(uint64_t*)(p + 8+2*stride)=
00433 *(uint64_t*)(p + 0+3*stride)=
00434 *(uint64_t*)(p + 8+3*stride)= size==4 ? val*0x0100000001ULL : val*0x0101010101010101ULL;
00435 }else
00436 assert(0);
00437 }
00438
00439 static inline void fill_caches(H264Context *h, int mb_type, int for_deblock){
00440 MpegEncContext * const s = &h->s;
00441 const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
00442 int topleft_xy, top_xy, topright_xy, left_xy[2];
00443 int topleft_type, top_type, topright_type, left_type[2];
00444 int left_block[8];
00445 int i;
00446
00447
00448
00449
00450 if(for_deblock && h->slice_num == 1)
00451 return;
00452
00453
00454
00455 top_xy = mb_xy - s->mb_stride;
00456 topleft_xy = top_xy - 1;
00457 topright_xy= top_xy + 1;
00458 left_xy[1] = left_xy[0] = mb_xy-1;
00459 left_block[0]= 0;
00460 left_block[1]= 1;
00461 left_block[2]= 2;
00462 left_block[3]= 3;
00463 left_block[4]= 7;
00464 left_block[5]= 10;
00465 left_block[6]= 8;
00466 left_block[7]= 11;
00467 if(h->mb_aff_frame){
00468 const int pair_xy = s->mb_x + (s->mb_y & ~1)*s->mb_stride;
00469 const int top_pair_xy = pair_xy - s->mb_stride;
00470 const int topleft_pair_xy = top_pair_xy - 1;
00471 const int topright_pair_xy = top_pair_xy + 1;
00472 const int topleft_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[topleft_pair_xy]);
00473 const int top_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[top_pair_xy]);
00474 const int topright_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[topright_pair_xy]);
00475 const int left_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[pair_xy-1]);
00476 const int curr_mb_frame_flag = !IS_INTERLACED(mb_type);
00477 const int bottom = (s->mb_y & 1);
00478 tprintf("fill_caches: curr_mb_frame_flag:%d, left_mb_frame_flag:%d, topleft_mb_frame_flag:%d, top_mb_frame_flag:%d, topright_mb_frame_flag:%d\n", curr_mb_frame_flag, left_mb_frame_flag, topleft_mb_frame_flag, top_mb_frame_flag, topright_mb_frame_flag);
00479 if (bottom
00480 ? !curr_mb_frame_flag
00481 : (!curr_mb_frame_flag && !top_mb_frame_flag)
00482 ) {
00483 top_xy -= s->mb_stride;
00484 }
00485 if (bottom
00486 ? !curr_mb_frame_flag
00487 : (!curr_mb_frame_flag && !topleft_mb_frame_flag)
00488 ) {
00489 topleft_xy -= s->mb_stride;
00490 }
00491 if (bottom
00492 ? !curr_mb_frame_flag
00493 : (!curr_mb_frame_flag && !topright_mb_frame_flag)
00494 ) {
00495 topright_xy -= s->mb_stride;
00496 }
00497 if (left_mb_frame_flag != curr_mb_frame_flag) {
00498 left_xy[1] = left_xy[0] = pair_xy - 1;
00499 if (curr_mb_frame_flag) {
00500 if (bottom) {
00501 left_block[0]= 2;
00502 left_block[1]= 2;
00503 left_block[2]= 3;
00504 left_block[3]= 3;
00505 left_block[4]= 8;
00506 left_block[5]= 11;
00507 left_block[6]= 8;
00508 left_block[7]= 11;
00509 } else {
00510 left_block[0]= 0;
00511 left_block[1]= 0;
00512 left_block[2]= 1;
00513 left_block[3]= 1;
00514 left_block[4]= 7;
00515 left_block[5]= 10;
00516 left_block[6]= 7;
00517 left_block[7]= 10;
00518 }
00519 } else {
00520 left_xy[1] += s->mb_stride;
00521
00522 left_block[1]= 2;
00523 left_block[2]= 0;
00524 left_block[3]= 2;
00525
00526 left_block[5]= 10;
00527 left_block[6]= 7;
00528 left_block[7]= 10;
00529 }
00530 }
00531 }
00532
00533 h->top_mb_xy = top_xy;
00534 h->left_mb_xy[0] = left_xy[0];
00535 h->left_mb_xy[1] = left_xy[1];
00536 if(for_deblock){
00537 topleft_type = h->slice_table[topleft_xy ] < 255 ? s->current_picture.mb_type[topleft_xy] : 0;
00538 top_type = h->slice_table[top_xy ] < 255 ? s->current_picture.mb_type[top_xy] : 0;
00539 topright_type= h->slice_table[topright_xy] < 255 ? s->current_picture.mb_type[topright_xy]: 0;
00540 left_type[0] = h->slice_table[left_xy[0] ] < 255 ? s->current_picture.mb_type[left_xy[0]] : 0;
00541 left_type[1] = h->slice_table[left_xy[1] ] < 255 ? s->current_picture.mb_type[left_xy[1]] : 0;
00542 }else{
00543 topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
00544 top_type = h->slice_table[top_xy ] == h->slice_num ? s->current_picture.mb_type[top_xy] : 0;
00545 topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
00546 left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
00547 left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
00548 }
00549
00550 if(IS_INTRA(mb_type)){
00551 h->topleft_samples_available=
00552 h->top_samples_available=
00553 h->left_samples_available= 0xFFFF;
00554 h->topright_samples_available= 0xEEEA;
00555
00556 if(!IS_INTRA(top_type) && (top_type==0 || h->pps.constrained_intra_pred)){
00557 h->topleft_samples_available= 0xB3FF;
00558 h->top_samples_available= 0x33FF;
00559 h->topright_samples_available= 0x26EA;
00560 }
00561 for(i=0; i<2; i++){
00562 if(!IS_INTRA(left_type[i]) && (left_type[i]==0 || h->pps.constrained_intra_pred)){
00563 h->topleft_samples_available&= 0xDF5F;
00564 h->left_samples_available&= 0x5F5F;
00565 }
00566 }
00567
00568 if(!IS_INTRA(topleft_type) && (topleft_type==0 || h->pps.constrained_intra_pred))
00569 h->topleft_samples_available&= 0x7FFF;
00570
00571 if(!IS_INTRA(topright_type) && (topright_type==0 || h->pps.constrained_intra_pred))
00572 h->topright_samples_available&= 0xFBFF;
00573
00574 if(IS_INTRA4x4(mb_type)){
00575 if(IS_INTRA4x4(top_type)){
00576 h->intra4x4_pred_mode_cache[4+8*0]= h->intra4x4_pred_mode[top_xy][4];
00577 h->intra4x4_pred_mode_cache[5+8*0]= h->intra4x4_pred_mode[top_xy][5];
00578 h->intra4x4_pred_mode_cache[6+8*0]= h->intra4x4_pred_mode[top_xy][6];
00579 h->intra4x4_pred_mode_cache[7+8*0]= h->intra4x4_pred_mode[top_xy][3];
00580 }else{
00581 int pred;
00582 if(!top_type || (IS_INTER(top_type) && h->pps.constrained_intra_pred))
00583 pred= -1;
00584 else{
00585 pred= 2;
00586 }
00587 h->intra4x4_pred_mode_cache[4+8*0]=
00588 h->intra4x4_pred_mode_cache[5+8*0]=
00589 h->intra4x4_pred_mode_cache[6+8*0]=
00590 h->intra4x4_pred_mode_cache[7+8*0]= pred;
00591 }
00592 for(i=0; i<2; i++){
00593 if(IS_INTRA4x4(left_type[i])){
00594 h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[0+2*i]];
00595 h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[1+2*i]];
00596 }else{
00597 int pred;
00598 if(!left_type[i] || (IS_INTER(left_type[i]) && h->pps.constrained_intra_pred))
00599 pred= -1;
00600 else{
00601 pred= 2;
00602 }
00603 h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
00604 h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= pred;
00605 }
00606 }
00607 }
00608 }
00609
00610
00611
00612
00613
00614
00615
00616
00617
00618
00619
00620 if(top_type){
00621 h->non_zero_count_cache[4+8*0]= h->non_zero_count[top_xy][4];
00622 h->non_zero_count_cache[5+8*0]= h->non_zero_count[top_xy][5];
00623 h->non_zero_count_cache[6+8*0]= h->non_zero_count[top_xy][6];
00624 h->non_zero_count_cache[7+8*0]= h->non_zero_count[top_xy][3];
00625
00626 h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][9];
00627 h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][8];
00628
00629 h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][12];
00630 h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][11];
00631
00632 }else{
00633 h->non_zero_count_cache[4+8*0]=
00634 h->non_zero_count_cache[5+8*0]=
00635 h->non_zero_count_cache[6+8*0]=
00636 h->non_zero_count_cache[7+8*0]=
00637
00638 h->non_zero_count_cache[1+8*0]=
00639 h->non_zero_count_cache[2+8*0]=
00640
00641 h->non_zero_count_cache[1+8*3]=
00642 h->non_zero_count_cache[2+8*3]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
00643
00644 }
00645
00646 for (i=0; i<2; i++) {
00647 if(left_type[i]){
00648 h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[0+2*i]];
00649 h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[1+2*i]];
00650 h->non_zero_count_cache[0+8*1 + 8*i]= h->non_zero_count[left_xy[i]][left_block[4+2*i]];
00651 h->non_zero_count_cache[0+8*4 + 8*i]= h->non_zero_count[left_xy[i]][left_block[5+2*i]];
00652 }else{
00653 h->non_zero_count_cache[3+8*1 + 2*8*i]=
00654 h->non_zero_count_cache[3+8*2 + 2*8*i]=
00655 h->non_zero_count_cache[0+8*1 + 8*i]=
00656 h->non_zero_count_cache[0+8*4 + 8*i]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
00657 }
00658 }
00659
00660 if( h->pps.cabac ) {
00661
00662 if(top_type) {
00663 h->top_cbp = h->cbp_table[top_xy];
00664 } else if(IS_INTRA(mb_type)) {
00665 h->top_cbp = 0x1C0;
00666 } else {
00667 h->top_cbp = 0;
00668 }
00669
00670 if (left_type[0]) {
00671 h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
00672 } else if(IS_INTRA(mb_type)) {
00673 h->left_cbp = 0x1C0;
00674 } else {
00675 h->left_cbp = 0;
00676 }
00677 if (left_type[0]) {
00678 h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
00679 }
00680 if (left_type[1]) {
00681 h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
00682 }
00683 }
00684
00685 #if 1
00686
00687 if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
00688 int list;
00689 for(list=0; list<1+(h->slice_type==B_TYPE); list++){
00690 if(!USES_LIST(mb_type, list) && !IS_DIRECT(mb_type) && !h->deblocking_filter){
00691
00692
00693
00694
00695
00696 continue;
00697 }
00698 h->mv_cache_clean[list]= 0;
00699
00700 if(IS_INTER(top_type)){
00701 const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
00702 const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
00703 *(uint32_t*)h->mv_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 0];
00704 *(uint32_t*)h->mv_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 1];
00705 *(uint32_t*)h->mv_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 2];
00706 *(uint32_t*)h->mv_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 3];
00707 h->ref_cache[list][scan8[0] + 0 - 1*8]=
00708 h->ref_cache[list][scan8[0] + 1 - 1*8]= s->current_picture.ref_index[list][b8_xy + 0];
00709 h->ref_cache[list][scan8[0] + 2 - 1*8]=
00710 h->ref_cache[list][scan8[0] + 3 - 1*8]= s->current_picture.ref_index[list][b8_xy + 1];
00711 }else{
00712 *(uint32_t*)h->mv_cache [list][scan8[0] + 0 - 1*8]=
00713 *(uint32_t*)h->mv_cache [list][scan8[0] + 1 - 1*8]=
00714 *(uint32_t*)h->mv_cache [list][scan8[0] + 2 - 1*8]=
00715 *(uint32_t*)h->mv_cache [list][scan8[0] + 3 - 1*8]= 0;
00716 *(uint32_t*)&h->ref_cache[list][scan8[0] + 0 - 1*8]= ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101;
00717 }
00718
00719
00720 if(IS_INTER(left_type[0])){
00721 const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
00722 const int b8_xy= h->mb2b8_xy[left_xy[0]] + 1;
00723 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0]];
00724 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1]];
00725 h->ref_cache[list][scan8[0] - 1 + 0*8]=
00726 h->ref_cache[list][scan8[0] - 1 + 1*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0]>>1)];
00727 }else{
00728 *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 0*8]=
00729 *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 1*8]= 0;
00730 h->ref_cache[list][scan8[0] - 1 + 0*8]=
00731 h->ref_cache[list][scan8[0] - 1 + 1*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
00732 }
00733
00734 if(IS_INTER(left_type[1])){
00735 const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
00736 const int b8_xy= h->mb2b8_xy[left_xy[1]] + 1;
00737 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[2]];
00738 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[3]];
00739 h->ref_cache[list][scan8[0] - 1 + 2*8]=
00740 h->ref_cache[list][scan8[0] - 1 + 3*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[2]>>1)];
00741 }else{
00742 *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 2*8]=
00743 *(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 3*8]= 0;
00744 h->ref_cache[list][scan8[0] - 1 + 2*8]=
00745 h->ref_cache[list][scan8[0] - 1 + 3*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
00746 assert((!left_type[0]) == (!left_type[1]));
00747 }
00748
00749 if(for_deblock || (IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred))
00750 continue;
00751
00752 if(IS_INTER(topleft_type)){
00753 const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
00754 const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + h->b8_stride;
00755 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
00756 h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
00757 }else{
00758 *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= 0;
00759 h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
00760 }
00761
00762 if(IS_INTER(topright_type)){
00763 const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
00764 const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
00765 *(uint32_t*)h->mv_cache[list][scan8[0] + 4 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
00766 h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
00767 }else{
00768 *(uint32_t*)h->mv_cache [list][scan8[0] + 4 - 1*8]= 0;
00769 h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
00770 }
00771
00772
00773 h->ref_cache[list][scan8[5 ]+1] =
00774 h->ref_cache[list][scan8[7 ]+1] =
00775 h->ref_cache[list][scan8[13]+1] =
00776 h->ref_cache[list][scan8[4 ]] =
00777 h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
00778 *(uint32_t*)h->mv_cache [list][scan8[5 ]+1]=
00779 *(uint32_t*)h->mv_cache [list][scan8[7 ]+1]=
00780 *(uint32_t*)h->mv_cache [list][scan8[13]+1]=
00781 *(uint32_t*)h->mv_cache [list][scan8[4 ]]=
00782 *(uint32_t*)h->mv_cache [list][scan8[12]]= 0;
00783
00784 if( h->pps.cabac ) {
00785
00786 if(IS_INTER(topleft_type)){
00787 const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
00788 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy];
00789 }else{
00790 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= 0;
00791 }
00792
00793 if(IS_INTER(top_type)){
00794 const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
00795 *(uint32_t*)h->mvd_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 0];
00796 *(uint32_t*)h->mvd_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 1];
00797 *(uint32_t*)h->mvd_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 2];
00798 *(uint32_t*)h->mvd_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 3];
00799 }else{
00800 *(uint32_t*)h->mvd_cache [list][scan8[0] + 0 - 1*8]=
00801 *(uint32_t*)h->mvd_cache [list][scan8[0] + 1 - 1*8]=
00802 *(uint32_t*)h->mvd_cache [list][scan8[0] + 2 - 1*8]=
00803 *(uint32_t*)h->mvd_cache [list][scan8[0] + 3 - 1*8]= 0;
00804 }
00805 if(IS_INTER(left_type[0])){
00806 const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
00807 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[0]];
00808 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[1]];
00809 }else{
00810 *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 0*8]=
00811 *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 1*8]= 0;
00812 }
00813 if(IS_INTER(left_type[1])){
00814 const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
00815 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[2]];
00816 *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[3]];
00817 }else{
00818 *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 2*8]=
00819 *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 3*8]= 0;
00820 }
00821 *(uint32_t*)h->mvd_cache [list][scan8[5 ]+1]=
00822 *(uint32_t*)h->mvd_cache [list][scan8[7 ]+1]=
00823 *(uint32_t*)h->mvd_cache [list][scan8[13]+1]=
00824 *(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
00825 *(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;
00826
00827 if(h->slice_type == B_TYPE){
00828 fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, 0, 1);
00829
00830 if(IS_DIRECT(top_type)){
00831 *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0x01010101;
00832 }else if(IS_8X8(top_type)){
00833 int b8_xy = h->mb2b8_xy[top_xy] + h->b8_stride;
00834 h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy];
00835 h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 1];
00836 }else{
00837 *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0;
00838 }
00839
00840
00841 if(IS_DIRECT(left_type[0])){
00842 h->direct_cache[scan8[0] - 1 + 0*8]=
00843 h->direct_cache[scan8[0] - 1 + 2*8]= 1;
00844 }else if(IS_8X8(left_type[0])){
00845 int b8_xy = h->mb2b8_xy[left_xy[0]] + 1;
00846 h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[b8_xy];
00847 h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[b8_xy + h->b8_stride];
00848 }else{
00849 h->direct_cache[scan8[0] - 1 + 0*8]=
00850 h->direct_cache[scan8[0] - 1 + 2*8]= 0;
00851 }
00852 }
00853 }
00854 }
00855 }
00856 #endif
00857
00858 h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
00859 }
00860
00861 static inline void write_back_intra_pred_mode(H264Context *h){
00862 MpegEncContext * const s = &h->s;
00863 const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
00864
00865 h->intra4x4_pred_mode[mb_xy][0]= h->intra4x4_pred_mode_cache[7+8*1];
00866 h->intra4x4_pred_mode[mb_xy][1]= h->intra4x4_pred_mode_cache[7+8*2];
00867 h->intra4x4_pred_mode[mb_xy][2]= h->intra4x4_pred_mode_cache[7+8*3];
00868 h->intra4x4_pred_mode[mb_xy][3]= h->intra4x4_pred_mode_cache[7+8*4];
00869 h->intra4x4_pred_mode[mb_xy][4]= h->intra4x4_pred_mode_cache[4+8*4];
00870 h->intra4x4_pred_mode[mb_xy][5]= h->intra4x4_pred_mode_cache[5+8*4];
00871 h->intra4x4_pred_mode[mb_xy][6]= h->intra4x4_pred_mode_cache[6+8*4];
00872 }
00873
00877 static inline int check_intra4x4_pred_mode(H264Context *h){
00878 MpegEncContext * const s = &h->s;
00879 static const int8_t top [12]= {-1, 0,LEFT_DC_PRED,-1,-1,-1,-1,-1, 0};
00880 static const int8_t left[12]= { 0,-1, TOP_DC_PRED, 0,-1,-1,-1, 0,-1,DC_128_PRED};
00881 int i;
00882
00883 if(!(h->top_samples_available&0x8000)){
00884 for(i=0; i<4; i++){
00885 int status= top[ h->intra4x4_pred_mode_cache[scan8[0] + i] ];
00886 if(status<0){
00887 av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
00888 return -1;
00889 } else if(status){
00890 h->intra4x4_pred_mode_cache[scan8[0] + i]= status;
00891 }
00892 }
00893 }
00894
00895 if(!(h->left_samples_available&0x8000)){
00896 for(i=0; i<4; i++){
00897 int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ];
00898 if(status<0){
00899 av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
00900 return -1;
00901 } else if(status){
00902 h->intra4x4_pred_mode_cache[scan8[0] + 8*i]= status;
00903 }
00904 }
00905 }
00906
00907 return 0;
00908 }
00909
00913 static inline int check_intra_pred_mode(H264Context *h, int mode){
00914 MpegEncContext * const s = &h->s;
00915 static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
00916 static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
00917
00918 if(mode < 0 || mode > 6) {
00919 av_log(h->s.avctx, AV_LOG_ERROR, "out of range intra chroma pred mode at %d %d\n", s->mb_x, s->mb_y);
00920 return -1;
00921 }
00922
00923 if(!(h->top_samples_available&0x8000)){
00924 mode= top[ mode ];
00925 if(mode<0){
00926 av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
00927 return -1;
00928 }
00929 }
00930
00931 if(!(h->left_samples_available&0x8000)){
00932 mode= left[ mode ];
00933 if(mode<0){
00934 av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
00935 return -1;
00936 }
00937 }
00938
00939 return mode;
00940 }
00941
00945 static inline int pred_intra_mode(H264Context *h, int n){
00946 const int index8= scan8[n];
00947 const int left= h->intra4x4_pred_mode_cache[index8 - 1];
00948 const int top = h->intra4x4_pred_mode_cache[index8 - 8];
00949 const int min= FFMIN(left, top);
00950
00951 tprintf("mode:%d %d min:%d\n", left ,top, min);
00952
00953 if(min<0) return DC_PRED;
00954 else return min;
00955 }
00956
00957 static inline void write_back_non_zero_count(H264Context *h){
00958 MpegEncContext * const s = &h->s;
00959 const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
00960
00961 h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[7+8*1];
00962 h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[7+8*2];
00963 h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[7+8*3];
00964 h->non_zero_count[mb_xy][3]= h->non_zero_count_cache[7+8*4];
00965 h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[4+8*4];
00966 h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[5+8*4];
00967 h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[6+8*4];
00968
00969 h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[1+8*2];
00970 h->non_zero_count[mb_xy][8]= h->non_zero_count_cache[2+8*2];
00971 h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[2+8*1];
00972
00973 h->non_zero_count[mb_xy][12]=h->non_zero_count_cache[1+8*5];
00974 h->non_zero_count[mb_xy][11]=h->non_zero_count_cache[2+8*5];
00975 h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[2+8*4];
00976 }
00977
00982 static inline int pred_non_zero_count(H264Context *h, int n){
00983 const int index8= scan8[n];
00984 const int left= h->non_zero_count_cache[index8 - 1];
00985 const int top = h->non_zero_count_cache[index8 - 8];
00986 int i= left + top;
00987
00988 if(i<64) i= (i+1)>>1;
00989
00990 tprintf("pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
00991
00992 return i&31;
00993 }
00994
00995 static inline int fetch_diagonal_mv(H264Context *h, const int16_t **C, int i, int list, int part_width){
00996 const int topright_ref= h->ref_cache[list][ i - 8 + part_width ];
00997
00998 if(topright_ref != PART_NOT_AVAILABLE){
00999 *C= h->mv_cache[list][ i - 8 + part_width ];
01000 return topright_ref;
01001 }else{
01002 tprintf("topright MV not available\n");
01003
01004 *C= h->mv_cache[list][ i - 8 - 1 ];
01005 return h->ref_cache[list][ i - 8 - 1 ];
01006 }
01007 }
01008
01016 static inline void pred_motion(H264Context * const h, int n, int part_width, int list, int ref, int * const mx, int * const my){
01017 const int index8= scan8[n];
01018 const int top_ref= h->ref_cache[list][ index8 - 8 ];
01019 const int left_ref= h->ref_cache[list][ index8 - 1 ];
01020 const int16_t * const A= h->mv_cache[list][ index8 - 1 ];
01021 const int16_t * const B= h->mv_cache[list][ index8 - 8 ];
01022 const int16_t * C;
01023 int diagonal_ref, match_count;
01024
01025 assert(part_width==1 || part_width==2 || part_width==4);
01026
01027
01028
01029
01030
01031
01032
01033
01034
01035 diagonal_ref= fetch_diagonal_mv(h, &C, index8, list, part_width);
01036 match_count= (diagonal_ref==ref) + (top_ref==ref) + (left_ref==ref);
01037 tprintf("pred_motion match_count=%d\n", match_count);
01038 if(match_count > 1){
01039 *mx= mid_pred(A[0], B[0], C[0]);
01040 *my= mid_pred(A[1], B[1], C[1]);
01041 }else if(match_count==1){
01042 if(left_ref==ref){
01043 *mx= A[0];
01044 *my= A[1];
01045 }else if(top_ref==ref){
01046 *mx= B[0];
01047 *my= B[1];
01048 }else{
01049 *mx= C[0];
01050 *my= C[1];
01051 }
01052 }else{
01053 if(top_ref == PART_NOT_AVAILABLE && diagonal_ref == PART_NOT_AVAILABLE && left_ref != PART_NOT_AVAILABLE){
01054 *mx= A[0];
01055 *my= A[1];
01056 }else{
01057 *mx= mid_pred(A[0], B[0], C[0]);
01058 *my= mid_pred(A[1], B[1], C[1]);
01059 }
01060 }
01061
01062 tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, h->s.mb_x, h->s.mb_y, n, list);
01063 }
01064
01071 static inline void pred_16x8_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
01072 if(n==0){
01073 const int top_ref= h->ref_cache[list][ scan8[0] - 8 ];
01074 const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];
01075
01076 tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
01077
01078 if(top_ref == ref){
01079 *mx= B[0];
01080 *my= B[1];
01081 return;
01082 }
01083 }else{
01084 const int left_ref= h->ref_cache[list][ scan8[8] - 1 ];
01085 const int16_t * const A= h->mv_cache[list][ scan8[8] - 1 ];
01086
01087 tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
01088
01089 if(left_ref == ref){
01090 *mx= A[0];
01091 *my= A[1];
01092 return;
01093 }
01094 }
01095
01096
01097 pred_motion(h, n, 4, list, ref, mx, my);
01098 }
01099
01106 static inline void pred_8x16_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
01107 if(n==0){
01108 const int left_ref= h->ref_cache[list][ scan8[0] - 1 ];
01109 const int16_t * const A= h->mv_cache[list][ scan8[0] - 1 ];
01110
01111 tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
01112
01113 if(left_ref == ref){
01114 *mx= A[0];
01115 *my= A[1];
01116 return;
01117 }
01118 }else{
01119 const int16_t * C;
01120 int diagonal_ref;
01121
01122 diagonal_ref= fetch_diagonal_mv(h, &C, scan8[4], list, 2);
01123
01124 tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n", diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);
01125
01126 if(diagonal_ref == ref){
01127 *mx= C[0];
01128 *my= C[1];
01129 return;
01130 }
01131 }
01132
01133
01134 pred_motion(h, n, 2, list, ref, mx, my);
01135 }
01136
01137 static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my){
01138 const int top_ref = h->ref_cache[0][ scan8[0] - 8 ];
01139 const int left_ref= h->ref_cache[0][ scan8[0] - 1 ];
01140
01141 tprintf("pred_pskip: (%d) (%d) at %2d %2d\n", top_ref, left_ref, h->s.mb_x, h->s.mb_y);
01142
01143 if(top_ref == PART_NOT_AVAILABLE || left_ref == PART_NOT_AVAILABLE
01144 || (top_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 8 ] == 0)
01145 || (left_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 1 ] == 0)){
01146
01147 *mx = *my = 0;
01148 return;
01149 }
01150
01151 pred_motion(h, 0, 4, 0, 0, mx, my);
01152
01153 return;
01154 }
01155
01156 static inline void direct_dist_scale_factor(H264Context * const h){
01157 const int poc = h->s.current_picture_ptr->poc;
01158 const int poc1 = h->ref_list[1][0].poc;
01159 int i;
01160 for(i=0; i<h->ref_count[0]; i++){
01161 int poc0 = h->ref_list[0][i].poc;
01162 int td = clip(poc1 - poc0, -128, 127);
01163 if(td == 0 ){
01164 h->dist_scale_factor[i] = 256;
01165 }else{
01166 int tb = clip(poc - poc0, -128, 127);
01167 int tx = (16384 + (ABS(td) >> 1)) / td;
01168 h->dist_scale_factor[i] = clip((tb*tx + 32) >> 6, -1024, 1023);
01169 }
01170 }
01171 }
01172 static inline void direct_ref_list_init(H264Context * const h){
01173 MpegEncContext * const s = &h->s;
01174 Picture * const ref1 = &h->ref_list[1][0];
01175 Picture * const cur = s->current_picture_ptr;
01176 int list, i, j;
01177 if(cur->pict_type == I_TYPE)
01178 cur->ref_count[0] = 0;
01179 if(cur->pict_type != B_TYPE)
01180 cur->ref_count[1] = 0;
01181 for(list=0; list<2; list++){
01182 cur->ref_count[list] = h->ref_count[list];
01183 for(j=0; j<h->ref_count[list]; j++)
01184 cur->ref_poc[list][j] = h->ref_list[list][j].poc;
01185 }
01186 if(cur->pict_type != B_TYPE || h->direct_spatial_mv_pred)
01187 return;
01188 for(list=0; list<2; list++){
01189 for(i=0; i<ref1->ref_count[list]; i++){
01190 const int poc = ref1->ref_poc[list][i];
01191 h->map_col_to_list0[list][i] = PART_NOT_AVAILABLE;
01192 for(j=0; j<h->ref_count[list]; j++)
01193 if(h->ref_list[list][j].poc == poc){
01194 h->map_col_to_list0[list][i] = j;
01195 break;
01196 }
01197 }
01198 }
01199 }
01200
01201 static inline void pred_direct_motion(H264Context * const h, int *mb_type){
01202 MpegEncContext * const s = &h->s;
01203 const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
01204 const int b8_xy = 2*s->mb_x + 2*s->mb_y*h->b8_stride;
01205 const int b4_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
01206 const int mb_type_col = h->ref_list[1][0].mb_type[mb_xy];
01207 const int16_t (*l1mv0)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[0][b4_xy];
01208 const int16_t (*l1mv1)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[1][b4_xy];
01209 const int8_t *l1ref0 = &h->ref_list[1][0].ref_index[0][b8_xy];
01210 const int8_t *l1ref1 = &h->ref_list[1][0].ref_index[1][b8_xy];
01211 const int is_b8x8 = IS_8X8(*mb_type);
01212 int sub_mb_type;
01213 int i8, i4;
01214
01215 if(IS_8X8(mb_type_col) && !h->sps.direct_8x8_inference_flag){
01216
01217
01218 sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2;
01219 *mb_type = MB_TYPE_8x8|MB_TYPE_L0L1;
01220 }else if(!is_b8x8 && (IS_16X16(mb_type_col) || IS_INTRA(mb_type_col))){
01221 sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2;
01222 *mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2;
01223 }else{
01224 sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2;
01225 *mb_type = MB_TYPE_8x8|MB_TYPE_L0L1;
01226 }
01227 if(!is_b8x8)
01228 *mb_type |= MB_TYPE_DIRECT2;
01229
01230 tprintf("mb_type = %08x, sub_mb_type = %08x, is_b8x8 = %d, mb_type_col = %08x\n", *mb_type, sub_mb_type, is_b8x8, mb_type_col);
01231
01232 if(h->direct_spatial_mv_pred){
01233 int ref[2];
01234 int mv[2][2];
01235 int list;
01236
01237
01238 for(list=0; list<2; list++){
01239 int refa = h->ref_cache[list][scan8[0] - 1];
01240 int refb = h->ref_cache[list][scan8[0] - 8];
01241 int refc = h->ref_cache[list][scan8[0] - 8 + 4];
01242 if(refc == -2)
01243 refc = h->ref_cache[list][scan8[0] - 8 - 1];
01244 ref[list] = refa;
01245 if(ref[list] < 0 || (refb < ref[list] && refb >= 0))
01246 ref[list] = refb;
01247 if(ref[list] < 0 || (refc < ref[list] && refc >= 0))
01248 ref[list] = refc;
01249 if(ref[list] < 0)
01250 ref[list] = -1;
01251 }
01252
01253 if(ref[0] < 0 && ref[1] < 0){
01254 ref[0] = ref[1] = 0;
01255 mv[0][0] = mv[0][1] =
01256 mv[1][0] = mv[1][1] = 0;
01257 }else{
01258 for(list=0; list<2; list++){
01259 if(ref[list] >= 0)
01260 pred_motion(h, 0, 4, list, ref[list], &mv[list][0], &mv[list][1]);
01261 else
01262 mv[list][0] = mv[list][1] = 0;
01263 }
01264 }
01265
01266 if(ref[1] < 0){
01267 *mb_type &= ~MB_TYPE_P0L1;
01268 sub_mb_type &= ~MB_TYPE_P0L1;
01269 }else if(ref[0] < 0){
01270 *mb_type &= ~MB_TYPE_P0L0;
01271 sub_mb_type &= ~MB_TYPE_P0L0;
01272 }
01273
01274 if(IS_16X16(*mb_type)){
01275 fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref[0], 1);
01276 fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, ref[1], 1);
01277 if(!IS_INTRA(mb_type_col)
01278 && ( l1ref0[0] == 0 && ABS(l1mv0[0][0]) <= 1 && ABS(l1mv0[0][1]) <= 1
01279 || l1ref0[0] < 0 && l1ref1[0] == 0 && ABS(l1mv1[0][0]) <= 1 && ABS(l1mv1[0][1]) <= 1)){
01280 if(ref[0] > 0)
01281 fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv[0][0],mv[0][1]), 4);
01282 else
01283 fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
01284 if(ref[1] > 0)
01285 fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv[1][0],mv[1][1]), 4);
01286 else
01287 fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
01288 }else{
01289 fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv[0][0],mv[0][1]), 4);
01290 fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv[1][0],mv[1][1]), 4);
01291 }
01292 }else{
01293 for(i8=0; i8<4; i8++){
01294 const int x8 = i8&1;
01295 const int y8 = i8>>1;
01296
01297 if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
01298 continue;
01299 h->sub_mb_type[i8] = sub_mb_type;
01300
01301 fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mv[0][0],mv[0][1]), 4);
01302 fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mv[1][0],mv[1][1]), 4);
01303 fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref[0], 1);
01304 fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, ref[1], 1);
01305
01306
01307 if(!IS_INTRA(mb_type_col) && ( l1ref0[x8 + y8*h->b8_stride] == 0
01308 || l1ref0[x8 + y8*h->b8_stride] < 0 && l1ref1[x8 + y8*h->b8_stride] == 0)){
01309 const int16_t (*l1mv)[2]= l1ref0[x8 + y8*h->b8_stride] == 0 ? l1mv0 : l1mv1;
01310 for(i4=0; i4<4; i4++){
01311 const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
01312 if(ABS(mv_col[0]) <= 1 && ABS(mv_col[1]) <= 1){
01313 if(ref[0] == 0)
01314 *(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
01315 if(ref[1] == 0)
01316 *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] = 0;
01317 }
01318 }
01319 }
01320 }
01321 }
01322 }else{
01323 if(IS_16X16(*mb_type)){
01324 fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
01325 if(IS_INTRA(mb_type_col)){
01326 fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
01327 fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
01328 fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
01329 }else{
01330 const int ref0 = l1ref0[0] >= 0 ? h->map_col_to_list0[0][l1ref0[0]]
01331 : h->map_col_to_list0[1][l1ref1[0]];
01332 const int dist_scale_factor = h->dist_scale_factor[ref0];
01333 const int16_t *mv_col = l1mv0[0];
01334 int mv_l0[2];
01335 mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
01336 mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
01337 fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref0, 1);
01338 fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv_l0[0],mv_l0[1]), 4);
01339 fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]), 4);
01340 }
01341 }else{
01342 for(i8=0; i8<4; i8++){
01343 const int x8 = i8&1;
01344 const int y8 = i8>>1;
01345 int ref0, dist_scale_factor;
01346
01347 if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
01348 continue;
01349 h->sub_mb_type[i8] = sub_mb_type;
01350 if(IS_INTRA(mb_type_col)){
01351 fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
01352 fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
01353 fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
01354 fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
01355 continue;
01356 }
01357
01358 ref0 = l1ref0[x8 + y8*h->b8_stride];
01359 if(ref0 >= 0)
01360 ref0 = h->map_col_to_list0[0][ref0];
01361 else
01362 ref0 = h->map_col_to_list0[1][l1ref1[x8 + y8*h->b8_stride]];
01363 dist_scale_factor = h->dist_scale_factor[ref0];
01364
01365 fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
01366 fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
01367 for(i4=0; i4<4; i4++){
01368 const int16_t *mv_col = l1mv0[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
01369 int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
01370 mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
01371 mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
01372 *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] =
01373 pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
01374 }
01375 }
01376 }
01377 }
01378 }
01379
01380 static inline void write_back_motion(H264Context *h, int mb_type){
01381 MpegEncContext * const s = &h->s;
01382 const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
01383 const int b8_xy= 2*s->mb_x + 2*s->mb_y*h->b8_stride;
01384 int list;
01385
01386 for(list=0; list<2; list++){
01387 int y;
01388 if(!USES_LIST(mb_type, list)){
01389 if(1){
01390 for(y=0; y<4; y++){
01391 *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]=
01392 *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= 0;
01393 }
01394 if( h->pps.cabac ) {
01395
01396 for(y=0; y<4; y++){
01397 *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]=
01398 *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= 0;
01399 }
01400 }
01401 for(y=0; y<2; y++){
01402 s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]=
01403 s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= LIST_NOT_USED;
01404 }
01405 }
01406 continue;
01407 }
01408
01409 for(y=0; y<4; y++){
01410 *(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+0 + 8*y];
01411 *(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+2 + 8*y];
01412 }
01413 if( h->pps.cabac ) {
01414 for(y=0; y<4; y++){
01415 *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+0 + 8*y];
01416 *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+2 + 8*y];
01417 }
01418 }
01419 for(y=0; y<2; y++){
01420 s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+0 + 16*y];
01421 s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+2 + 16*y];
01422 }
01423 }
01424
01425 if(h->slice_type == B_TYPE && h->pps.cabac){
01426 if(IS_8X8(mb_type)){
01427 h->direct_table[b8_xy+1+0*h->b8_stride] = IS_DIRECT(h->sub_mb_type[1]) ? 1 : 0;
01428 h->direct_table[b8_xy+0+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[2]) ? 1 : 0;
01429 h->direct_table[b8_xy+1+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[3]) ? 1 : 0;
01430 }
01431 }
01432 }
01433
01441 static uint8_t *decode_nal(H264Context *h, uint8_t *src, int *dst_length, int *consumed, int length){
01442 int i, si, di;
01443 uint8_t *dst;
01444
01445
01446 h->nal_ref_idc= src[0]>>5;
01447 h->nal_unit_type= src[0]&0x1F;
01448
01449 src++; length--;
01450 #if 0
01451 for(i=0; i<length; i++)
01452 printf("%2X ", src[i]);
01453 #endif
01454 for(i=0; i+1<length; i+=2){
01455 if(src[i]) continue;
01456 if(i>0 && src[i-1]==0) i--;
01457 if(i+2<length && src[i+1]==0 && src[i+2]<=3){
01458 if(src[i+2]!=3){
01459
01460 length=i;
01461 }
01462 break;
01463 }
01464 }
01465
01466 if(i>=length-1){
01467 *dst_length= length;
01468 *consumed= length+1;
01469 return src;
01470 }
01471
01472 h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length);
01473 dst= h->rbsp_buffer;
01474
01475
01476 si=di=0;
01477 while(si<length){
01478
01479 if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
01480 if(src[si+2]==3){
01481 dst[di++]= 0;
01482 dst[di++]= 0;
01483 si+=3;
01484 continue;
01485 }else
01486 break;
01487 }
01488
01489 dst[di++]= src[si++];
01490 }
01491
01492 *dst_length= di;
01493 *consumed= si + 1;
01494
01495 return dst;
01496 }
01497
01498 #if 0
01499
01506 static int encode_nal(H264Context *h, uint8_t *dst, uint8_t *src, int length, int dst_length){
01507 int i, escape_count, si, di;
01508 uint8_t *temp;
01509
01510 assert(length>=0);
01511 assert(dst_length>0);
01512
01513 dst[0]= (h->nal_ref_idc<<5) + h->nal_unit_type;
01514
01515 if(length==0) return 1;
01516
01517 escape_count= 0;
01518 for(i=0; i<length; i+=2){
01519 if(src[i]) continue;
01520 if(i>0 && src[i-1]==0)
01521 i--;
01522 if(i+2<length && src[i+1]==0 && src[i+2]<=3){
01523 escape_count++;
01524 i+=2;
01525 }
01526 }
01527
01528 if(escape_count==0){
01529 if(dst+1 != src)
01530 memcpy(dst+1, src, length);
01531 return length + 1;
01532 }
01533
01534 if(length + escape_count + 1> dst_length)
01535 return -1;
01536
01537
01538
01539 h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length + escape_count);
01540 temp= h->rbsp_buffer;
01541
01542
01543 si= 0;
01544 di= 0;
01545 while(si < length){
01546 if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
01547 temp[di++]= 0; si++;
01548 temp[di++]= 0; si++;
01549 temp[di++]= 3;
01550 temp[di++]= src[si++];
01551 }
01552 else
01553 temp[di++]= src[si++];
01554 }
01555 memcpy(dst+1, temp, length+escape_count);
01556
01557 assert(di == length+escape_count);
01558
01559 return di + 1;
01560 }
01561
01565 static void encode_rbsp_trailing(PutBitContext *pb){
01566 int length;
01567 put_bits(pb, 1, 1);
01568 length= (-put_bits_count(pb))&7;
01569 if(length) put_bits(pb, length, 0);
01570 }
01571 #endif
01572
01577 static int decode_rbsp_trailing(uint8_t *src){
01578 int v= *src;
01579 int r;
01580
01581 tprintf("rbsp trailing %X\n", v);
01582
01583 for(r=1; r<9; r++){
01584 if(v&1) return r;
01585 v>>=1;
01586 }
01587 return 0;
01588 }
01589
01594 static void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qp){
01595 const int qmul= dequant_coeff[qp][0];
01596 #define stride 16
01597 int i;
01598 int temp[16];
01599 static const int x_offset[4]={0, 1*stride, 4* stride, 5*stride};
01600 static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};
01601
01602
01603
01604 for(i=0; i<4; i++){
01605 const int offset= y_offset[i];
01606 const int z0= block[offset+stride*0] + block[offset+stride*4];
01607 const int z1= block[offset+stride*0] - block[offset+stride*4];
01608 const int z2= block[offset+stride*1] - block[offset+stride*5];
01609 const int z3= block[offset+stride*1] + block[offset+stride*5];
01610
01611 temp[4*i+0]= z0+z3;
01612 temp[4*i+1]= z1+z2;
01613 temp[4*i+2]= z1-z2;
01614 temp[4*i+3]= z0-z3;
01615 }
01616
01617 for(i=0; i<4; i++){
01618 const int offset= x_offset[i];
01619 const int z0= temp[4*0+i] + temp[4*2+i];
01620 const int z1= temp[4*0+i] - temp[4*2+i];
01621 const int z2= temp[4*1+i] - temp[4*3+i];
01622 const int z3= temp[4*1+i] + temp[4*3+i];
01623
01624 block[stride*0 +offset]= ((z0 + z3)*qmul + 2)>>2;
01625 block[stride*2 +offset]= ((z1 + z2)*qmul + 2)>>2;
01626 block[stride*8 +offset]= ((z1 - z2)*qmul + 2)>>2;
01627 block[stride*10+offset]= ((z0 - z3)*qmul + 2)>>2;
01628 }
01629 }
01630
01631 #if 0
01632
01636 static void h264_luma_dc_dct_c(DCTELEM *block){
01637
01638 int i;
01639 int temp[16];
01640 static const int x_offset[4]={0, 1*stride, 4* stride, 5*stride};
01641 static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};
01642
01643 for(i=0; i<4; i++){
01644 const int offset= y_offset[i];
01645 const int z0= block[offset+stride*0] + block[offset+stride*4];
01646 const int z1= block[offset+stride*0] - block[offset+stride*4];
01647 const int z2= block[offset+stride*1] - block[offset+stride*5];
01648 const int z3= block[offset+stride*1] + block[offset+stride*5];
01649
01650 temp[4*i+0]= z0+z3;
01651 temp[4*i+1]= z1+z2;
01652 temp[4*i+2]= z1-z2;
01653 temp[4*i+3]= z0-z3;
01654 }
01655
01656 for(i=0; i<4; i++){
01657 const int offset= x_offset[i];
01658 const int z0= temp[4*0+i] + temp[4*2+i];
01659 const int z1= temp[4*0+i] - temp[4*2+i];
01660 const int z2= temp[4*1+i] - temp[4*3+i];
01661 const int z3= temp[4*1+i] + temp[4*3+i];
01662
01663 block[stride*0 +offset]= (z0 + z3)>>1;
01664 block[stride*2 +offset]= (z1 + z2)>>1;
01665 block[stride*8 +offset]= (z1 - z2)>>1;
01666 block[stride*10+offset]= (z0 - z3)>>1;
01667 }
01668 }
01669 #endif
01670
01671 #undef xStride
01672 #undef stride
01673
01674 static void chroma_dc_dequant_idct_c(DCTELEM *block, int qp){
01675 const int qmul= dequant_coeff[qp][0];
01676 const int stride= 16*2;
01677 const int xStride= 16;
01678 int a,b,c,d,e;
01679
01680 a= block[stride*0 + xStride*0];
01681 b= block[stride*0 + xStride*1];
01682 c= block[stride*1 + xStride*0];
01683 d= block[stride*1 + xStride*1];
01684
01685 e= a-b;
01686 a= a+b;
01687 b= c-d;
01688 c= c+d;
01689
01690 block[stride*0 + xStride*0]= ((a+c)*qmul + 0)>>1;
01691 block[stride*0 + xStride*1]= ((e+b)*qmul + 0)>>1;
01692 block[stride*1 + xStride*0]= ((a-c)*qmul + 0)>>1;
01693 block[stride*1 + xStride*1]= ((e-b)*qmul + 0)>>1;
01694 }
01695
01696 #if 0
01697 static void chroma_dc_dct_c(DCTELEM *block){
01698 const int stride= 16*2;
01699 const int xStride= 16;
01700 int a,b,c,d,e;
01701
01702 a= block[stride*0 + xStride*0];
01703 b= block[stride*0 + xStride*1];
01704 c= block[stride*1 + xStride*0];
01705 d= block[stride*1 + xStride*1];
01706
01707 e= a-b;
01708 a= a+b;
01709 b= c-d;
01710 c= c+d;
01711
01712 block[stride*0 + xStride*0]= (a+c);
01713 block[stride*0 + xStride*1]= (e+b);
01714 block[stride*1 + xStride*0]= (a-c);
01715 block[stride*1 + xStride*1]= (e-b);
01716 }
01717 #endif
01718
01722 static inline int get_chroma_qp(int chroma_qp_index_offset, int qscale){
01723
01724 return chroma_qp[clip(qscale + chroma_qp_index_offset, 0, 51)];
01725 }
01726
01727
01728 #if 0
01729 static void h264_diff_dct_c(DCTELEM *block, uint8_t *src1, uint8_t *src2, int stride){
01730 int i;
01731
01732
01733 for(i=0; i<4; i++){
01734 const int d0= src1[0 + i*stride] - src2[0 + i*stride];
01735 const int d1= src1[1 + i*stride] - src2[1 + i*stride];
01736 const int d2= src1[2 + i*stride] - src2[2 + i*stride];
01737 const int d3= src1[3 + i*stride] - src2[3 + i*stride];
01738 const int z0= d0 + d3;
01739 const int z3= d0 - d3;
01740 const int z1= d1 + d2;
01741 const int z2= d1 - d2;
01742
01743 block[0 + 4*i]= z0 + z1;
01744 block[1 + 4*i]= 2*z3 + z2;
01745 block[2 + 4*i]= z0 - z1;
01746 block[3 + 4*i]= z3 - 2*z2;
01747 }
01748
01749 for(i=0; i<4; i++){
01750 const int z0= block[0*4 + i] + block[3*4 + i];
01751 const int z3= block[0*4 + i] - block[3*4 + i];
01752 const int z1= block[1*4 + i] + block[2*4 + i];
01753 const int z2= block[1*4 + i] - block[2*4 + i];
01754
01755 block[0*4 + i]= z0 + z1;
01756 block[1*4 + i]= 2*z3 + z2;
01757 block[2*4 + i]= z0 - z1;
01758 block[3*4 + i]= z3 - 2*z2;
01759 }
01760 }
01761 #endif
01762
01763
01764
01765 static inline int quantize_c(DCTELEM *block, uint8_t *scantable, int qscale, int intra, int seperate_dc){
01766 int i;
01767 const int * const quant_table= quant_coeff[qscale];
01768 const int bias= intra ? (1<<QUANT_SHIFT)/3 : (1<<QUANT_SHIFT)/6;
01769 const unsigned int threshold1= (1<<QUANT_SHIFT) - bias - 1;
01770 const unsigned int threshold2= (threshold1<<1);
01771 int last_non_zero;
01772
01773 if(seperate_dc){
01774 if(qscale<=18){
01775
01776 const int dc_bias= intra ? (1<<(QUANT_SHIFT-2))/3 : (1<<(QUANT_SHIFT-2))/6;
01777 const unsigned int dc_threshold1= (1<<(QUANT_SHIFT-2)) - dc_bias - 1;
01778 const unsigned int dc_threshold2= (dc_threshold1<<1);
01779
01780 int level= block[0]*quant_coeff[qscale+18][0];
01781 if(((unsigned)(level+dc_threshold1))>dc_threshold2){
01782 if(level>0){
01783 level= (dc_bias + level)>>(QUANT_SHIFT-2);
01784 block[0]= level;
01785 }else{
01786 level= (dc_bias - level)>>(QUANT_SHIFT-2);
01787 block[0]= -level;
01788 }
01789
01790 }else{
01791 block[0]=0;
01792 }
01793 }else{
01794 const int dc_bias= intra ? (1<<(QUANT_SHIFT+1))/3 : (1<<(QUANT_SHIFT+1))/6;
01795 const unsigned int dc_threshold1= (1<<(QUANT_SHIFT+1)) - dc_bias - 1;
01796 const unsigned int dc_threshold2= (dc_threshold1<<1);
01797
01798 int level= block[0]*quant_table[0];
01799 if(((unsigned)(level+dc_threshold1))>dc_threshold2){
01800 if(level>0){
01801 level= (dc_bias + level)>>(QUANT_SHIFT+1);
01802 block[0]= level;
01803 }else{
01804 level= (dc_bias - level)>>(QUANT_SHIFT+1);
01805 block[0]= -level;
01806 }
01807
01808 }else{
01809 block[0]=0;
01810 }
01811 }
01812 last_non_zero= 0;
01813 i=1;
01814 }else{
01815 last_non_zero= -1;
01816 i=0;
01817 }
01818
01819 for(; i<16; i++){
01820 const int j= scantable[i];
01821 int level= block[j]*quant_table[j];
01822
01823
01824
01825 if(((unsigned)(level+threshold1))>threshold2){
01826 if(level>0){
01827 level= (bias + level)>>QUANT_SHIFT;
01828 block[j]= level;
01829 }else{
01830 level= (bias - level)>>QUANT_SHIFT;
01831 block[j]= -level;
01832 }
01833 last_non_zero = i;
01834 }else{
01835 block[j]=0;
01836 }
01837 }
01838
01839 return last_non_zero;
01840 }
01841
01842 static void pred4x4_vertical_c(uint8_t *src, uint8_t *topright, int stride){
01843 const uint32_t a= ((uint32_t*)(src-stride))[0];
01844 ((uint32_t*)(src+0*stride))[0]= a;
01845 ((uint32_t*)(src+1*stride))[0]= a;
01846 ((uint32_t*)(src+2*stride))[0]= a;
01847 ((uint32_t*)(src+3*stride))[0]= a;
01848 }
01849
01850 static void pred4x4_horizontal_c(uint8_t *src, uint8_t *topright, int stride){
01851 ((uint32_t*)(src+0*stride))[0]= src[-1+0*stride]*0x01010101;
01852 ((uint32_t*)(src+1*stride))[0]= src[-1+1*stride]*0x01010101;
01853 ((uint32_t*)(src+2*stride))[0]= src[-1+2*stride]*0x01010101;
01854 ((uint32_t*)(src+3*stride))[0]= src[-1+3*stride]*0x01010101;
01855 }
01856
01857 static void pred4x4_dc_c(uint8_t *src, uint8_t *topright, int stride){
01858 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
01859 + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
01860
01861 ((uint32_t*)(src+0*stride))[0]=
01862 ((uint32_t*)(src+1*stride))[0]=
01863 ((uint32_t*)(src+2*stride))[0]=
01864 ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
01865 }
01866
01867 static void pred4x4_left_dc_c(uint8_t *src, uint8_t *topright, int stride){
01868 const int dc= ( src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
01869
01870 ((uint32_t*)(src+0*stride))[0]=
01871 ((uint32_t*)(src+1*stride))[0]=
01872 ((uint32_t*)(src+2*stride))[0]=
01873 ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
01874 }
01875
01876 static void pred4x4_top_dc_c(uint8_t *src, uint8_t *topright, int stride){
01877 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
01878
01879 ((uint32_t*)(src+0*stride))[0]=
01880 ((uint32_t*)(src+1*stride))[0]=
01881 ((uint32_t*)(src+2*stride))[0]=
01882 ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
01883 }
01884
01885 static void pred4x4_128_dc_c(uint8_t *src, uint8_t *topright, int stride){
01886 ((uint32_t*)(src+0*stride))[0]=
01887 ((uint32_t*)(src+1*stride))[0]=
01888 ((uint32_t*)(src+2*stride))[0]=
01889 ((uint32_t*)(src+3*stride))[0]= 128U*0x01010101U;
01890 }
01891
01892
01893 #define LOAD_TOP_RIGHT_EDGE\
01894 const int t4= topright[0];\
01895 const int t5= topright[1];\
01896 const int t6= topright[2];\
01897 const int t7= topright[3];\
01898
01899 #define LOAD_LEFT_EDGE\
01900 const int l0= src[-1+0*stride];\
01901 const int l1= src[-1+1*stride];\
01902 const int l2= src[-1+2*stride];\
01903 const int l3= src[-1+3*stride];\
01904
01905 #define LOAD_TOP_EDGE\
01906 const int t0= src[ 0-1*stride];\
01907 const int t1= src[ 1-1*stride];\
01908 const int t2= src[ 2-1*stride];\
01909 const int t3= src[ 3-1*stride];\
01910
01911 static void pred4x4_down_right_c(uint8_t *src, uint8_t *topright, int stride){
01912 const int lt= src[-1-1*stride];
01913 LOAD_TOP_EDGE
01914 LOAD_LEFT_EDGE
01915
01916 src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
01917 src[0+2*stride]=
01918 src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
01919 src[0+1*stride]=
01920 src[1+2*stride]=
01921 src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
01922 src[0+0*stride]=
01923 src[1+1*stride]=
01924 src[2+2*stride]=
01925 src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
01926 src[1+0*stride]=
01927 src[2+1*stride]=
01928 src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
01929 src[2+0*stride]=
01930 src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
01931 src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
01932 }
01933
01934 static void pred4x4_down_left_c(uint8_t *src, uint8_t *topright, int stride){
01935 LOAD_TOP_EDGE
01936 LOAD_TOP_RIGHT_EDGE
01937
01938
01939 src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
01940 src[1+0*stride]=
01941 src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
01942 src[2+0*stride]=
01943 src[1+1*stride]=
01944 src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
01945 src[3+0*stride]=
01946 src[2+1*stride]=
01947 src[1+2*stride]=
01948 src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
01949 src[3+1*stride]=
01950 src[2+2*stride]=
01951 src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
01952 src[3+2*stride]=
01953 src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
01954 src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
01955 }
01956
01957 static void pred4x4_vertical_right_c(uint8_t *src, uint8_t *topright, int stride){
01958 const int lt= src[-1-1*stride];
01959 LOAD_TOP_EDGE
01960 LOAD_LEFT_EDGE
01961 const __attribute__((unused)) int unu= l3;
01962
01963 src[0+0*stride]=
01964 src[1+2*stride]=(lt + t0 + 1)>>1;
01965 src[1+0*stride]=
01966 src[2+2*stride]=(t0 + t1 + 1)>>1;
01967 src[2+0*stride]=
01968 src[3+2*stride]=(t1 + t2 + 1)>>1;
01969 src[3+0*stride]=(t2 + t3 + 1)>>1;
01970 src[0+1*stride]=
01971 src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
01972 src[1+1*stride]=
01973 src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
01974 src[2+1*stride]=
01975 src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
01976 src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
01977 src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
01978 src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
01979 }
01980
01981 static void pred4x4_vertical_left_c(uint8_t *src, uint8_t *topright, int stride){
01982 LOAD_TOP_EDGE
01983 LOAD_TOP_RIGHT_EDGE
01984 const __attribute__((unused)) int unu= t7;
01985
01986 src[0+0*stride]=(t0 + t1 + 1)>>1;
01987 src[1+0*stride]=
01988 src[0+2*stride]=(t1 + t2 + 1)>>1;
01989 src[2+0*stride]=
01990 src[1+2*stride]=(t2 + t3 + 1)>>1;
01991 src[3+0*stride]=
01992 src[2+2*stride]=(t3 + t4+ 1)>>1;
01993 src[3+2*stride]=(t4 + t5+ 1)>>1;
01994 src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
01995 src[1+1*stride]=
01996 src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
01997 src[2+1*stride]=
01998 src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
01999 src[3+1*stride]=
02000 src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
02001 src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
02002 }
02003
02004 static void pred4x4_horizontal_up_c(uint8_t *src, uint8_t *topright, int stride){
02005 LOAD_LEFT_EDGE
02006
02007 src[0+0*stride]=(l0 + l1 + 1)>>1;
02008 src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
02009 src[2+0*stride]=
02010 src[0+1*stride]=(l1 + l2 + 1)>>1;
02011 src[3+0*stride]=
02012 src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
02013 src[2+1*stride]=
02014 src[0+2*stride]=(l2 + l3 + 1)>>1;
02015 src[3+1*stride]=
02016 src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
02017 src[3+2*stride]=
02018 src[1+3*stride]=
02019 src[0+3*stride]=
02020 src[2+2*stride]=
02021 src[2+3*stride]=
02022 src[3+3*stride]=l3;
02023 }
02024
02025 static void pred4x4_horizontal_down_c(uint8_t *src, uint8_t *topright, int stride){
02026 const int lt= src[-1-1*stride];
02027 LOAD_TOP_EDGE
02028 LOAD_LEFT_EDGE
02029 const __attribute__((unused)) int unu= t3;
02030
02031 src[0+0*stride]=
02032 src[2+1*stride]=(lt + l0 + 1)>>1;
02033 src[1+0*stride]=
02034 src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
02035 src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
02036 src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
02037 src[0+1*stride]=
02038 src[2+2*stride]=(l0 + l1 + 1)>>1;
02039 src[1+1*stride]=
02040 src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
02041 src[0+2*stride]=
02042 src[2+3*stride]=(l1 + l2+ 1)>>1;
02043 src[1+2*stride]=
02044 src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
02045 src[0+3*stride]=(l2 + l3 + 1)>>1;
02046 src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
02047 }
02048
02049 static void pred16x16_vertical_c(uint8_t *src, int stride){
02050 int i;
02051 const uint32_t a= ((uint32_t*)(src-stride))[0];
02052 const uint32_t b= ((uint32_t*)(src-stride))[1];
02053 const uint32_t c= ((uint32_t*)(src-stride))[2];
02054 const uint32_t d= ((uint32_t*)(src-stride))[3];
02055
02056 for(i=0; i<16; i++){
02057 ((uint32_t*)(src+i*stride))[0]= a;
02058 ((uint32_t*)(src+i*stride))[1]= b;
02059 ((uint32_t*)(src+i*stride))[2]= c;
02060 ((uint32_t*)(src+i*stride))[3]= d;
02061 }
02062 }
02063
02064 static void pred16x16_horizontal_c(uint8_t *src, int stride){
02065 int i;
02066
02067 for(i=0; i<16; i++){
02068 ((uint32_t*)(src+i*stride))[0]=
02069 ((uint32_t*)(src+i*stride))[1]=
02070 ((uint32_t*)(src+i*stride))[2]=
02071 ((uint32_t*)(src+i*stride))[3]= src[-1+i*stride]*0x01010101;
02072 }
02073 }
02074
02075 static void pred16x16_dc_c(uint8_t *src, int stride){
02076 int i, dc=0;
02077
02078 for(i=0;i<16; i++){
02079 dc+= src[-1+i*stride];
02080 }
02081
02082 for(i=0;i<16; i++){
02083 dc+= src[i-stride];
02084 }
02085
02086 dc= 0x01010101*((dc + 16)>>5);
02087
02088 for(i=0; i<16; i++){
02089 ((uint32_t*)(src+i*stride))[0]=
02090 ((uint32_t*)(src+i*stride))[1]=
02091 ((uint32_t*)(src+i*stride))[2]=
02092 ((uint32_t*)(src+i*stride))[3]= dc;
02093 }
02094 }
02095
02096 static void pred16x16_left_dc_c(uint8_t *src, int stride){
02097 int i, dc=0;
02098
02099 for(i=0;i<16; i++){
02100 dc+= src[-1+i*stride];
02101 }
02102
02103 dc= 0x01010101*((dc + 8)>>4);
02104
02105 for(i=0; i<16; i++){
02106 ((uint32_t*)(src+i*stride))[0]=
02107 ((uint32_t*)(src+i*stride))[1]=
02108 ((uint32_t*)(src+i*stride))[2]=
02109 ((uint32_t*)(src+i*stride))[3]= dc;
02110 }
02111 }
02112
02113 static void pred16x16_top_dc_c(uint8_t *src, int stride){
02114 int i, dc=0;
02115
02116 for(i=0;i<16; i++){
02117 dc+= src[i-stride];
02118 }
02119 dc= 0x01010101*((dc + 8)>>4);
02120
02121 for(i=0; i<16; i++){
02122 ((uint32_t*)(src+i*stride))[0]=
02123 ((uint32_t*)(src+i*stride))[1]=
02124 ((uint32_t*)(src+i*stride))[2]=
02125 ((uint32_t*)(src+i*stride))[3]= dc;
02126 }
02127 }
02128
02129 static void pred16x16_128_dc_c(uint8_t *src, int stride){
02130 int i;
02131
02132 for(i=0; i<16; i++){
02133 ((uint32_t*)(src+i*stride))[0]=
02134 ((uint32_t*)(src+i*stride))[1]=
02135 ((uint32_t*)(src+i*stride))[2]=
02136 ((uint32_t*)(src+i*stride))[3]= 0x01010101U*128U;
02137 }
02138 }
02139
02140 static inline void pred16x16_plane_compat_c(uint8_t *src, int stride, const int svq3){
02141 int i, j, k;
02142 int a;
02143 uint8_t *cm = cropTbl + MAX_NEG_CROP;
02144 const uint8_t * const src0 = src+7-stride;
02145 const uint8_t *src1 = src+8*stride-1;
02146 const uint8_t *src2 = src1-2*stride;
02147 int H = src0[1] - src0[-1];
02148 int V = src1[0] - src2[ 0];
02149 for(k=2; k<=8; ++k) {
02150 src1 += stride; src2 -= stride;
02151 H += k*(src0[k] - src0[-k]);
02152 V += k*(src1[0] - src2[ 0]);
02153 }
02154 if(svq3){
02155 H = ( 5*(H/4) ) / 16;
02156 V = ( 5*(V/4) ) / 16;
02157
02158
02159 i = H; H = V; V = i;
02160 }else{
02161 H = ( 5*H+32 ) >> 6;
02162 V = ( 5*V+32 ) >> 6;
02163 }
02164
02165 a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
02166 for(j=16; j>0; --j) {
02167 int b = a;
02168 a += V;
02169 for(i=-16; i<0; i+=4) {
02170 src[16+i] = cm[ (b ) >> 5 ];
02171 src[17+i] = cm[ (b+ H) >> 5 ];
02172 src[18+i] = cm[ (b+2*H) >> 5 ];
02173 src[19+i] = cm[ (b+3*H) >> 5 ];
02174 b += 4*H;
02175 }
02176 src += stride;
02177 }
02178 }
02179
02180 static void pred16x16_plane_c(uint8_t *src, int stride){
02181 pred16x16_plane_compat_c(src, stride, 0);
02182 }
02183
02184 static void pred8x8_vertical_c(uint8_t *src, int stride){
02185 int i;
02186 const uint32_t a= ((uint32_t*)(src-stride))[0];
02187 const uint32_t b= ((uint32_t*)(src-stride))[1];
02188
02189 for(i=0; i<8; i++){
02190 ((uint32_t*)(src+i*stride))[0]= a;
02191 ((uint32_t*)(src+i*stride))[1]= b;
02192 }
02193 }
02194
02195 static void pred8x8_horizontal_c(uint8_t *src, int stride){
02196 int i;
02197
02198 for(i=0; i<8; i++){
02199 ((uint32_t*)(src+i*stride))[0]=
02200 ((uint32_t*)(src+i*stride))[1]= src[-1+i*stride]*0x01010101;
02201 }
02202 }
02203
02204 static void pred8x8_128_dc_c(uint8_t *src, int stride){
02205 int i;
02206
02207 for(i=0; i<8; i++){
02208 ((uint32_t*)(src+i*stride))[0]=
02209 ((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
02210 }
02211 }
02212
02213 static void pred8x8_left_dc_c(uint8_t *src, int stride){
02214 int i;
02215 int dc0, dc2;
02216
02217 dc0=dc2=0;
02218 for(i=0;i<4; i++){
02219 dc0+= src[-1+i*stride];
02220 dc2+= src[-1+(i+4)*stride];
02221 }
02222 dc0= 0x01010101*((dc0 + 2)>>2);
02223 dc2= 0x01010101*((dc2 + 2)>>2);
02224
02225 for(i=0; i<4; i++){
02226 ((uint32_t*)(src+i*stride))[0]=
02227 ((uint32_t*)(src+i*stride))[1]= dc0;
02228 }
02229 for(i=4; i<8; i++){
02230 ((uint32_t*)(src+i*stride))[0]=
02231 ((uint32_t*)(src+i*stride))[1]= dc2;
02232 }
02233 }
02234
02235 static void pred8x8_top_dc_c(uint8_t *src, int stride){
02236 int i;
02237 int dc0, dc1;
02238
02239 dc0=dc1=0;
02240 for(i=0;i<4; i++){
02241 dc0+= src[i-stride];
02242 dc1+= src[4+i-stride];
02243 }
02244 dc0= 0x01010101*((dc0 + 2)>>2);
02245 dc1= 0x01010101*((dc1 + 2)>>2);
02246
02247 for(i=0; i<4; i++){
02248 ((uint32_t*)(src+i*stride))[0]= dc0;
02249 ((uint32_t*)(src+i*stride))[1]= dc1;
02250 }
02251 for(i=4; i<8; i++){
02252 ((uint32_t*)(src+i*stride))[0]= dc0;
02253 ((uint32_t*)(src+i*stride))[1]= dc1;
02254 }
02255 }
02256
02257
02258 static void pred8x8_dc_c(uint8_t *src, int stride){
02259 int i;
02260 int dc0, dc1, dc2, dc3;
02261
02262 dc0=dc1=dc2=0;
02263 for(i=0;i<4; i++){
02264 dc0+= src[-1+i*stride] + src[i-stride];
02265 dc1+= src[4+i-stride];
02266 dc2+= src[-1+(i+4)*stride];
02267 }
02268 dc3= 0x01010101*((dc1 + dc2 + 4)>>3);
02269 dc0= 0x01010101*((dc0 + 4)>>3);
02270 dc1= 0x01010101*((dc1 + 2)>>2);
02271 dc2= 0x01010101*((dc2 + 2)>>2);
02272
02273 for(i=0; i<4; i++){
02274 ((uint32_t*)(src+i*stride))[0]= dc0;
02275 ((uint32_t*)(src+i*stride))[1]= dc1;
02276 }
02277 for(i=4; i<8; i++){
02278 ((uint32_t*)(src+i*stride))[0]= dc2;
02279 ((uint32_t*)(src+i*stride))[1]= dc3;
02280 }
02281 }
02282
02283 static void pred8x8_plane_c(uint8_t *src, int stride){
02284 int j, k;
02285 int a;
02286 uint8_t *cm = cropTbl + MAX_NEG_CROP;
02287 const uint8_t * const src0 = src+3-stride;
02288 const uint8_t *src1 = src+4*stride-1;
02289 const uint8_t *src2 = src1-2*stride;
02290 int H = src0[1] - src0[-1];
02291 int V = src1[0] - src2[ 0];
02292 for(k=2; k<=4; ++k) {
02293 src1 += stride; src2 -= stride;
02294 H += k*(src0[k] - src0[-k]);
02295 V += k*(src1[0] - src2[ 0]);
02296 }
02297 H = ( 17*H+16 ) >> 5;
02298 V = ( 17*V+16 ) >> 5;
02299
02300 a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
02301 for(j=8; j>0; --j) {
02302 int b = a;
02303 a += V;
02304 src[0] = cm[ (b ) >> 5 ];
02305 src[1] = cm[ (b+ H) >> 5 ];
02306 src[2] = cm[ (b+2*H) >> 5 ];
02307 src[3] = cm[ (b+3*H) >> 5 ];
02308 src[4] = cm[ (b+4*H) >> 5 ];
02309 src[5] = cm[ (b+5*H) >> 5 ];
02310 src[6] = cm[ (b+6*H) >> 5 ];
02311 src[7] = cm[ (b+7*H) >> 5 ];
02312 src += stride;
02313 }
02314 }
02315
02316 #define SRC(x,y) src[(x)+(y)*stride]
02317 #define PL(y) \
02318 const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
02319 #define PREDICT_8x8_LOAD_LEFT \
02320 const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
02321 + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
02322 PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
02323 const int l7 = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
02324
02325 #define PT(x) \
02326 const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
02327 #define PREDICT_8x8_LOAD_TOP \
02328 const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
02329 + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
02330 PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
02331 const int t7 = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
02332 + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
02333
02334 #define PTR(x) \
02335 t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
02336 #define PREDICT_8x8_LOAD_TOPRIGHT \
02337 int t8, t9, t10, t11, t12, t13, t14, t15; \
02338 if(has_topright) { \
02339 PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
02340 t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
02341 } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
02342
02343 #define PREDICT_8x8_LOAD_TOPLEFT \
02344 const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
02345
02346 #define PREDICT_8x8_DC(v) \
02347 int y; \
02348 for( y = 0; y < 8; y++ ) { \
02349 ((uint32_t*)src)[0] = \
02350 ((uint32_t*)src)[1] = v; \
02351 src += stride; \
02352 }
02353
02354 static void pred8x8l_128_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02355 {
02356 PREDICT_8x8_DC(0x80808080);
02357 }
02358 static void pred8x8l_left_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02359 {
02360 PREDICT_8x8_LOAD_LEFT;
02361 const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3) * 0x01010101;
02362 PREDICT_8x8_DC(dc);
02363 }
02364 static void pred8x8l_top_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02365 {
02366 PREDICT_8x8_LOAD_TOP;
02367 const uint32_t dc = ((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3) * 0x01010101;
02368 PREDICT_8x8_DC(dc);
02369 }
02370 static void pred8x8l_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02371 {
02372 PREDICT_8x8_LOAD_LEFT;
02373 PREDICT_8x8_LOAD_TOP;
02374 const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7
02375 +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4) * 0x01010101;
02376 PREDICT_8x8_DC(dc);
02377 }
02378 static void pred8x8l_horizontal_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02379 {
02380 PREDICT_8x8_LOAD_LEFT;
02381 #define ROW(y) ((uint32_t*)(src+y*stride))[0] =\
02382 ((uint32_t*)(src+y*stride))[1] = 0x01010101 * l##y
02383 ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
02384 #undef ROW
02385 }
02386 static void pred8x8l_vertical_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02387 {
02388 int y;
02389 PREDICT_8x8_LOAD_TOP;
02390 src[0] = t0;
02391 src[1] = t1;
02392 src[2] = t2;
02393 src[3] = t3;
02394 src[4] = t4;
02395 src[5] = t5;
02396 src[6] = t6;
02397 src[7] = t7;
02398 for( y = 1; y < 8; y++ )
02399 *(uint64_t*)(src+y*stride) = *(uint64_t*)src;
02400 }
02401 static void pred8x8l_down_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02402 {
02403 PREDICT_8x8_LOAD_TOP;
02404 PREDICT_8x8_LOAD_TOPRIGHT;
02405 SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
02406 SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
02407 SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
02408 SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
02409 SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
02410 SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
02411 SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= (t6 + 2*t7 + t8 + 2) >> 2;
02412 SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= (t7 + 2*t8 + t9 + 2) >> 2;
02413 SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= (t8 + 2*t9 + t10 + 2) >> 2;
02414 SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
02415 SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
02416 SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
02417 SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
02418 SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
02419 SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
02420 }
02421 static void pred8x8l_down_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02422 {
02423 PREDICT_8x8_LOAD_TOP;
02424 PREDICT_8x8_LOAD_LEFT;
02425 PREDICT_8x8_LOAD_TOPLEFT;
02426 SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
02427 SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
02428 SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
02429 SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
02430 SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
02431 SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
02432 SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= (l1 + 2*l0 + lt + 2) >> 2;
02433 SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= (l0 + 2*lt + t0 + 2) >> 2;
02434 SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= (lt + 2*t0 + t1 + 2) >> 2;
02435 SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
02436 SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
02437 SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
02438 SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
02439 SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
02440 SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
02441
02442 }
02443 static void pred8x8l_vertical_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02444 {
02445 PREDICT_8x8_LOAD_TOP;
02446 PREDICT_8x8_LOAD_LEFT;
02447 PREDICT_8x8_LOAD_TOPLEFT;
02448 SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
02449 SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
02450 SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
02451 SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
02452 SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
02453 SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
02454 SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
02455 SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
02456 SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
02457 SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
02458 SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
02459 SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
02460 SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
02461 SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
02462 SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
02463 SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
02464 SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
02465 SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
02466 SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
02467 SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
02468 SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
02469 SRC(7,0)= (t6 + t7 + 1) >> 1;
02470 }
02471 static void pred8x8l_horizontal_down_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02472 {
02473 PREDICT_8x8_LOAD_TOP;
02474 PREDICT_8x8_LOAD_LEFT;
02475 PREDICT_8x8_LOAD_TOPLEFT;
02476 SRC(0,7)= (l6 + l7 + 1) >> 1;
02477 SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
02478 SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
02479 SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
02480 SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
02481 SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
02482 SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
02483 SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
02484 SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
02485 SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
02486 SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
02487 SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
02488 SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
02489 SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
02490 SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
02491 SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
02492 SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
02493 SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
02494 SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
02495 SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
02496 SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
02497 SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
02498 }
02499 static void pred8x8l_vertical_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02500 {
02501 PREDICT_8x8_LOAD_TOP;
02502 PREDICT_8x8_LOAD_TOPRIGHT;
02503 SRC(0,0)= (t0 + t1 + 1) >> 1;
02504 SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
02505 SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
02506 SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
02507 SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
02508 SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
02509 SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
02510 SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
02511 SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
02512 SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
02513 SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
02514 SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
02515 SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
02516 SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
02517 SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
02518 SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
02519 SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
02520 SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
02521 SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
02522 SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
02523 SRC(7,6)= (t10 + t11 + 1) >> 1;
02524 SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
02525 }
02526 static void pred8x8l_horizontal_up_c(uint8_t *src, int has_topleft, int has_topright, int stride)
02527 {
02528 PREDICT_8x8_LOAD_LEFT;
02529 SRC(0,0)= (l0 + l1 + 1) >> 1;
02530 SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
02531 SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
02532 SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
02533 SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
02534 SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
02535 SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
02536 SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
02537 SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
02538 SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
02539 SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
02540 SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
02541 SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
02542 SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
02543 SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
02544 SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
02545 SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
02546 SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
02547 }
02548 #undef PREDICT_8x8_LOAD_LEFT
02549 #undef PREDICT_8x8_LOAD_TOP
02550 #undef PREDICT_8x8_LOAD_TOPLEFT
02551 #undef PREDICT_8x8_LOAD_TOPRIGHT
02552 #undef PREDICT_8x8_DC
02553 #undef PTR
02554 #undef PT
02555 #undef PL
02556 #undef SRC
02557
02558 static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, int chroma_height, int delta, int list,
02559 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
02560 int src_x_offset, int src_y_offset,
02561 qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op){
02562 MpegEncContext * const s = &h->s;
02563 const int mx= h->mv_cache[list][ scan8[n] ][0] + src_x_offset*8;
02564 const int my= h->mv_cache[list][ scan8[n] ][1] + src_y_offset*8;
02565 const int luma_xy= (mx&3) + ((my&3)<<2);
02566 uint8_t * src_y = pic->data[0] + (mx>>2) + (my>>2)*s->linesize;
02567 uint8_t * src_cb= pic->data[1] + (mx>>3) + (my>>3)*s->uvlinesize;
02568 uint8_t * src_cr= pic->data[2] + (mx>>3) + (my>>3)*s->uvlinesize;
02569 int extra_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
02570 int extra_height= extra_width;
02571 int emu=0;
02572 const int full_mx= mx>>2;
02573 const int full_my= my>>2;
02574
02575 assert(pic->data[0]);
02576
02577 if(mx&7) extra_width -= 3;
02578 if(my&7) extra_height -= 3;
02579
02580 if( full_mx < 0-extra_width
02581 || full_my < 0-extra_height
02582 || full_mx + 16 > s->width + extra_width
02583 || full_my + 16 > s->height + extra_height){
02584 ff_emulated_edge_mc(s->edge_emu_buffer, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5, full_mx-2, full_my-2, s->width, s->height);
02585 src_y= s->edge_emu_buffer + 2 + 2*s->linesize;
02586 emu=1;
02587 }
02588
02589 qpix_op[luma_xy](dest_y, src_y, s->linesize);
02590 if(!square){
02591 qpix_op[luma_xy](dest_y + delta, src_y + delta, s->linesize);
02592 }
02593
02594 if(s->flags&CODEC_FLAG_GRAY) return;
02595
02596 if(emu){
02597 ff_emulated_edge_mc(s->edge_emu_buffer, src_cb, s->uvlinesize, 9, 9, (mx>>3), (my>>3), s->width>>1, s->height>>1);
02598 src_cb= s->edge_emu_buffer;
02599 }
02600 chroma_op(dest_cb, src_cb, s->uvlinesize, chroma_height, mx&7, my&7);
02601
02602 if(emu){
02603 ff_emulated_edge_mc(s->edge_emu_buffer, src_cr, s->uvlinesize, 9, 9, (mx>>3), (my>>3), s->width>>1, s->height>>1);
02604 src_cr= s->edge_emu_buffer;
02605 }
02606 chroma_op(dest_cr, src_cr, s->uvlinesize, chroma_height, mx&7, my&7);
02607 }
02608
02609 static inline void mc_part_std(H264Context *h, int n, int square, int chroma_height, int delta,
02610 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
02611 int x_offset, int y_offset,
02612 qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
02613 qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
02614 int list0, int list1){
02615 MpegEncContext * const s = &h->s;
02616 qpel_mc_func *qpix_op= qpix_put;
02617 h264_chroma_mc_func chroma_op= chroma_put;
02618
02619 dest_y += 2*x_offset + 2*y_offset*s-> linesize;
02620 dest_cb += x_offset + y_offset*s->uvlinesize;
02621 dest_cr += x_offset + y_offset*s->uvlinesize;
02622 x_offset += 8*s->mb_x;
02623 y_offset += 8*s->mb_y;
02624
02625 if(list0){
02626 Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ];
02627 mc_dir_part(h, ref, n, square, chroma_height, delta, 0,
02628 dest_y, dest_cb, dest_cr, x_offset, y_offset,
02629 qpix_op, chroma_op);
02630
02631 qpix_op= qpix_avg;
02632 chroma_op= chroma_avg;
02633 }
02634
02635 if(list1){
02636 Picture *ref= &h->ref_list[1][ h->ref_cache[1][ scan8[n] ] ];
02637 mc_dir_part(h, ref, n, square, chroma_height, delta, 1,
02638 dest_y, dest_cb, dest_cr, x_offset, y_offset,
02639 qpix_op, chroma_op);
02640 }
02641 }
02642
02643 static inline void mc_part_weighted(H264Context *h, int n, int square, int chroma_height, int delta,
02644 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
02645 int x_offset, int y_offset,
02646 qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
02647 h264_weight_func luma_weight_op, h264_weight_func chroma_weight_op,
02648 h264_biweight_func luma_weight_avg, h264_biweight_func chroma_weight_avg,
02649 int list0, int list1){
02650 MpegEncContext * const s = &h->s;
02651
02652 dest_y += 2*x_offset + 2*y_offset*s-> linesize;
02653 dest_cb += x_offset + y_offset*s->uvlinesize;
02654 dest_cr += x_offset + y_offset*s->uvlinesize;
02655 x_offset += 8*s->mb_x;
02656 y_offset += 8*s->mb_y;
02657
02658 if(list0 && list1){
02659
02660
02661 uint8_t *tmp_cb = s->obmc_scratchpad;
02662 uint8_t *tmp_cr = tmp_cb + 8*s->uvlinesize;
02663 uint8_t *tmp_y = tmp_cr + 8*s->uvlinesize;
02664 int refn0 = h->ref_cache[0][ scan8[n] ];
02665 int refn1 = h->ref_cache[1][ scan8[n] ];
02666
02667 mc_dir_part(h, &h->ref_list[0][refn0], n, square, chroma_height, delta, 0,
02668 dest_y, dest_cb, dest_cr,
02669 x_offset, y_offset, qpix_put, chroma_put);
02670 mc_dir_part(h, &h->ref_list[1][refn1], n, square, chroma_height, delta, 1,
02671 tmp_y, tmp_cb, tmp_cr,
02672 x_offset, y_offset, qpix_put, chroma_put);
02673
02674 if(h->use_weight == 2){
02675 int weight0 = h->implicit_weight[refn0][refn1];
02676 int weight1 = 64 - weight0;
02677 luma_weight_avg( dest_y, tmp_y, s-> linesize, 5, weight0, weight1, 0, 0);
02678 chroma_weight_avg(dest_cb, tmp_cb, s->uvlinesize, 5, weight0, weight1, 0, 0);
02679 chroma_weight_avg(dest_cr, tmp_cr, s->uvlinesize, 5, weight0, weight1, 0, 0);
02680 }else{
02681 luma_weight_avg(dest_y, tmp_y, s->linesize, h->luma_log2_weight_denom,
02682 h->luma_weight[0][refn0], h->luma_weight[1][refn1],
02683 h->luma_offset[0][refn0], h->luma_offset[1][refn1]);
02684 chroma_weight_avg(dest_cb, tmp_cb, s->uvlinesize, h->chroma_log2_weight_denom,
02685 h->chroma_weight[0][refn0][0], h->chroma_weight[1][refn1][0],
02686 h->chroma_offset[0][refn0][0], h->chroma_offset[1][refn1][0]);
02687 chroma_weight_avg(dest_cr, tmp_cr, s->uvlinesize, h->chroma_log2_weight_denom,
02688 h->chroma_weight[0][refn0][1], h->chroma_weight[1][refn1][1],
02689 h->chroma_offset[0][refn0][1], h->chroma_offset[1][refn1][1]);
02690 }
02691 }else{
02692 int list = list1 ? 1 : 0;
02693 int refn = h->ref_cache[list][ scan8[n] ];
02694 Picture *ref= &h->ref_list[list][refn];
02695 mc_dir_part(h, ref, n, square, chroma_height, delta, list,
02696 dest_y, dest_cb, dest_cr, x_offset, y_offset,
02697 qpix_put, chroma_put);
02698
02699 luma_weight_op(dest_y, s->linesize, h->luma_log2_weight_denom,
02700 h->luma_weight[list][refn], h->luma_offset[list][refn]);
02701 if(h->use_weight_chroma){
02702 chroma_weight_op(dest_cb, s->uvlinesize, h->chroma_log2_weight_denom,
02703 h->chroma_weight[list][refn][0], h->chroma_offset[list][refn][0]);
02704 chroma_weight_op(dest_cr, s->uvlinesize, h->chroma_log2_weight_denom,
02705 h->chroma_weight[list][refn][1], h->chroma_offset[list][refn][1]);
02706 }
02707 }
02708 }
02709
02710 static inline void mc_part(H264Context *h, int n, int square, int chroma_height, int delta,
02711 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
02712 int x_offset, int y_offset,
02713 qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
02714 qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
02715 h264_weight_func *weight_op, h264_biweight_func *weight_avg,
02716 int list0, int list1){
02717 if((h->use_weight==2 && list0 && list1
02718 && (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ] != 32))
02719 || h->use_weight==1)
02720 mc_part_weighted(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr,
02721 x_offset, y_offset, qpix_put, chroma_put,
02722 weight_op[0], weight_op[3], weight_avg[0], weight_avg[3], list0, list1);
02723 else
02724 mc_part_std(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr,
02725 x_offset, y_offset, qpix_put, chroma_put, qpix_avg, chroma_avg, list0, list1);
02726 }
02727
02728 static void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
02729 qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
02730 qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
02731 h264_weight_func *weight_op, h264_biweight_func *weight_avg){
02732 MpegEncContext * const s = &h->s;
02733 const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
02734 const int mb_type= s->current_picture.mb_type[mb_xy];
02735
02736 assert(IS_INTER(mb_type));
02737
02738 if(IS_16X16(mb_type)){
02739 mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
02740 qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
02741 &weight_op[0], &weight_avg[0],
02742 IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
02743 }else if(IS_16X8(mb_type)){
02744 mc_part(h, 0, 0, 4, 8, dest_y, dest_cb, dest_cr, 0, 0,
02745 qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
02746 &weight_op[1], &weight_avg[1],
02747 IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
02748 mc_part(h, 8, 0, 4, 8, dest_y, dest_cb, dest_cr, 0, 4,
02749 qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
02750 &weight_op[1], &weight_avg[1],
02751 IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
02752 }else if(IS_8X16(mb_type)){
02753 mc_part(h, 0, 0, 8, 8*s->linesize, dest_y, dest_cb, dest_cr, 0, 0,
02754 qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
02755 &weight_op[2], &weight_avg[2],
02756 IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
02757 mc_part(h, 4, 0, 8, 8*s->linesize, dest_y, dest_cb, dest_cr, 4, 0,
02758 qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
02759 &weight_op[2], &weight_avg[2],
02760 IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
02761 }else{
02762 int i;
02763
02764 assert(IS_8X8(mb_type));
02765
02766 for(i=0; i<4; i++){
02767 const int sub_mb_type= h->sub_mb_type[i];
02768 const int n= 4*i;
02769 int x_offset= (i&1)<<2;
02770 int y_offset= (i&2)<<1;
02771
02772 if(IS_SUB_8X8(sub_mb_type)){
02773 mc_part(h, n, 1, 4, 0, dest_y, dest_cb, dest_cr, x_offset, y_offset,
02774 qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
02775 &weight_op[3], &weight_avg[3],
02776 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02777 }else if(IS_SUB_8X4(sub_mb_type)){
02778 mc_part(h, n , 0, 2, 4, dest_y, dest_cb, dest_cr, x_offset, y_offset,
02779 qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
02780 &weight_op[4], &weight_avg[4],
02781 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02782 mc_part(h, n+2, 0, 2, 4, dest_y, dest_cb, dest_cr, x_offset, y_offset+2,
02783 qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
02784 &weight_op[4], &weight_avg[4],
02785 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02786 }else if(IS_SUB_4X8(sub_mb_type)){
02787 mc_part(h, n , 0, 4, 4*s->linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset,
02788 qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
02789 &weight_op[5], &weight_avg[5],
02790 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02791 mc_part(h, n+1, 0, 4, 4*s->linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset,
02792 qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
02793 &weight_op[5], &weight_avg[5],
02794 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02795 }else{
02796 int j;
02797 assert(IS_SUB_4X4(sub_mb_type));
02798 for(j=0; j<4; j++){
02799 int sub_x_offset= x_offset + 2*(j&1);
02800 int sub_y_offset= y_offset + (j&2);
02801 mc_part(h, n+j, 1, 2, 0, dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset,
02802 qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
02803 &weight_op[6], &weight_avg[6],
02804 IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
02805 }
02806 }
02807 }
02808 }
02809 }
02810
02811 static void decode_init_vlc(H264Context *h){
02812 static int done = 0;
02813
02814 if (!done) {
02815 int i;
02816 done = 1;
02817
02818 init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
02819 &chroma_dc_coeff_token_len [0], 1, 1,
02820 &chroma_dc_coeff_token_bits[0], 1, 1, 1);
02821
02822 for(i=0; i<4; i++){
02823 init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
02824 &coeff_token_len [i][0], 1, 1,
02825 &coeff_token_bits[i][0], 1, 1, 1);
02826 }
02827
02828 for(i=0; i<3; i++){
02829 init_vlc(&chroma_dc_total_zeros_vlc[i], CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
02830 &chroma_dc_total_zeros_len [i][0], 1, 1,
02831 &chroma_dc_total_zeros_bits[i][0], 1, 1, 1);
02832 }
02833 for(i=0; i<15; i++){
02834 init_vlc(&total_zeros_vlc[i], TOTAL_ZEROS_VLC_BITS, 16,
02835 &total_zeros_len [i][0], 1, 1,
02836 &total_zeros_bits[i][0], 1, 1, 1);
02837 }
02838
02839 for(i=0; i<6; i++){
02840 init_vlc(&run_vlc[i], RUN_VLC_BITS, 7,
02841 &run_len [i][0], 1, 1,
02842 &run_bits[i][0], 1, 1, 1);
02843 }
02844 init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
02845 &run_len [6][0], 1, 1,
02846 &run_bits[6][0], 1, 1, 1);
02847 }
02848 }
02849
02853 static void init_pred_ptrs(H264Context *h){
02854
02855
02856 h->pred4x4[VERT_PRED ]= pred4x4_vertical_c;
02857 h->pred4x4[HOR_PRED ]= pred4x4_horizontal_c;
02858 h->pred4x4[DC_PRED ]= pred4x4_dc_c;
02859 h->pred4x4[DIAG_DOWN_LEFT_PRED ]= pred4x4_down_left_c;
02860 h->pred4x4[DIAG_DOWN_RIGHT_PRED]= pred4x4_down_right_c;
02861 h->pred4x4[VERT_RIGHT_PRED ]= pred4x4_vertical_right_c;
02862 h->pred4x4[HOR_DOWN_PRED ]= pred4x4_horizontal_down_c;
02863 h->pred4x4[VERT_LEFT_PRED ]= pred4x4_vertical_left_c;
02864 h->pred4x4[HOR_UP_PRED ]= pred4x4_horizontal_up_c;
02865 h->pred4x4[LEFT_DC_PRED ]= pred4x4_left_dc_c;
02866 h->pred4x4[TOP_DC_PRED ]= pred4x4_top_dc_c;
02867 h->pred4x4[DC_128_PRED ]= pred4x4_128_dc_c;
02868
02869 h->pred8x8l[VERT_PRED ]= pred8x8l_vertical_c;
02870 h->pred8x8l[HOR_PRED ]= pred8x8l_horizontal_c;
02871 h->pred8x8l[DC_PRED ]= pred8x8l_dc_c;
02872 h->pred8x8l[DIAG_DOWN_LEFT_PRED ]= pred8x8l_down_left_c;
02873 h->pred8x8l[DIAG_DOWN_RIGHT_PRED]= pred8x8l_down_right_c;
02874 h->pred8x8l[VERT_RIGHT_PRED ]= pred8x8l_vertical_right_c;
02875 h->pred8x8l[HOR_DOWN_PRED ]= pred8x8l_horizontal_down_c;
02876 h->pred8x8l[VERT_LEFT_PRED ]= pred8x8l_vertical_left_c;
02877 h->pred8x8l[HOR_UP_PRED ]= pred8x8l_horizontal_up_c;
02878 h->pred8x8l[LEFT_DC_PRED ]= pred8x8l_left_dc_c;
02879 h->pred8x8l[TOP_DC_PRED ]= pred8x8l_top_dc_c;
02880 h->pred8x8l[DC_128_PRED ]= pred8x8l_128_dc_c;
02881
02882 h->pred8x8[DC_PRED8x8 ]= pred8x8_dc_c;
02883 h->pred8x8[VERT_PRED8x8 ]= pred8x8_vertical_c;
02884 h->pred8x8[HOR_PRED8x8 ]= pred8x8_horizontal_c;
02885 h->pred8x8[PLANE_PRED8x8 ]= pred8x8_plane_c;
02886 h->pred8x8[LEFT_DC_PRED8x8]= pred8x8_left_dc_c;
02887 h->pred8x8[TOP_DC_PRED8x8 ]= pred8x8_top_dc_c;
02888 h->pred8x8[DC_128_PRED8x8 ]= pred8x8_128_dc_c;
02889
02890 h->pred16x16[DC_PRED8x8 ]= pred16x16_dc_c;
02891 h->pred16x16[VERT_PRED8x8 ]= pred16x16_vertical_c;
02892 h->pred16x16[HOR_PRED8x8 ]= pred16x16_horizontal_c;
02893 h->pred16x16[PLANE_PRED8x8 ]= pred16x16_plane_c;
02894 h->pred16x16[LEFT_DC_PRED8x8]= pred16x16_left_dc_c;
02895 h->pred16x16[TOP_DC_PRED8x8 ]= pred16x16_top_dc_c;
02896 h->pred16x16[DC_128_PRED8x8 ]= pred16x16_128_dc_c;
02897 }
02898
02899 static void free_tables(H264Context *h){
02900 av_freep(&h->intra4x4_pred_mode);
02901 av_freep(&h->chroma_pred_mode_table);
02902 av_freep(&h->cbp_table);
02903 av_freep(&h->mvd_table[0]);
02904 av_freep(&h->mvd_table[1]);
02905 av_freep(&h->direct_table);
02906 av_freep(&h->non_zero_count);
02907 av_freep(&h->slice_table_base);
02908 av_freep(&h->top_borders[1]);
02909 av_freep(&h->top_borders[0]);
02910 h->slice_table= NULL;
02911
02912 av_freep(&h->mb2b_xy);
02913 av_freep(&h->mb2b8_xy);
02914
02915 av_freep(&h->dequant4_coeff);
02916 av_freep(&h->dequant8_coeff);
02917
02918 av_freep(&h->s.obmc_scratchpad);
02919 }
02920
02925 static int alloc_tables(H264Context *h){
02926 MpegEncContext * const s = &h->s;
02927 const int big_mb_num= s->mb_stride * (s->mb_height+1);
02928 int x,y,q;
02929
02930 CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8 * sizeof(uint8_t))
02931
02932 CHECKED_ALLOCZ(h->non_zero_count , big_mb_num * 16 * sizeof(uint8_t))
02933 CHECKED_ALLOCZ(h->slice_table_base , big_mb_num * sizeof(uint8_t))
02934 CHECKED_ALLOCZ(h->top_borders[0] , s->mb_width * (16+8+8) * sizeof(uint8_t))
02935 CHECKED_ALLOCZ(h->top_borders[1] , s->mb_width * (16+8+8) * sizeof(uint8_t))
02936 CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))
02937
02938 if( h->pps.cabac ) {
02939 CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
02940 CHECKED_ALLOCZ(h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t));
02941 CHECKED_ALLOCZ(h->mvd_table[1], 32*big_mb_num * sizeof(uint16_t));
02942 CHECKED_ALLOCZ(h->direct_table, 32*big_mb_num * sizeof(uint8_t));
02943 }
02944
02945 memset(h->slice_table_base, -1, big_mb_num * sizeof(uint8_t));
02946 h->slice_table= h->slice_table_base + s->mb_stride + 1;
02947
02948 CHECKED_ALLOCZ(h->mb2b_xy , big_mb_num * sizeof(uint32_t));
02949 CHECKED_ALLOCZ(h->mb2b8_xy , big_mb_num * sizeof(uint32_t));
02950 for(y=0; y<s->mb_height; y++){
02951 for(x=0; x<s->mb_width; x++){
02952 const int mb_xy= x + y*s->mb_stride;
02953 const int b_xy = 4*x + 4*y*h->b_stride;
02954 const int b8_xy= 2*x + 2*y*h->b8_stride;
02955
02956 h->mb2b_xy [mb_xy]= b_xy;
02957 h->mb2b8_xy[mb_xy]= b8_xy;
02958 }
02959 }
02960
02961 CHECKED_ALLOCZ(h->dequant4_coeff, 52*16 * sizeof(uint16_t));
02962 CHECKED_ALLOCZ(h->dequant8_coeff, 52*64 * sizeof(uint16_t));
02963 memcpy(h->dequant4_coeff, dequant_coeff, 52*16 * sizeof(uint16_t));
02964 for(q=0; q<52; q++){
02965 int shift = div6[q];
02966 int idx = rem6[q];
02967 if(shift >= 2)
02968 shift -= 2;
02969 for(x=0; x<64; x++)
02970 h->dequant8_coeff[q][x] = dequant8_coeff_init[idx][
02971 dequant8_coeff_init_scan[((x>>1)&12) | (x&3)] ] << shift;
02972 }
02973 if(h->sps.transform_bypass){
02974 for(x=0; x<16; x++)
02975 h->dequant4_coeff[0][x] = 1;
02976 for(x=0; x<64; x++)
02977 h->dequant8_coeff[0][x] = 1<<2;
02978 }
02979
02980 s->obmc_scratchpad = NULL;
02981
02982 return 0;
02983 fail:
02984 free_tables(h);
02985 return -1;
02986 }
02987
02988 static void common_init(H264Context *h){
02989 MpegEncContext * const s = &h->s;
02990
02991 s->width = s->avctx->width;
02992 s->height = s->avctx->height;
02993 s->codec_id= s->avctx->codec->id;
02994
02995 init_pred_ptrs(h);
02996
02997 s->unrestricted_mv=1;
02998 s->decode=1;
02999 }
03000
03001 static int decode_init(AVCodecContext *avctx){
03002 H264Context *h= avctx->priv_data;
03003 MpegEncContext * const s = &h->s;
03004
03005 MPV_decode_defaults(s);
03006
03007 s->avctx = avctx;
03008 common_init(h);
03009
03010 s->out_format = FMT_H264;
03011 s->workaround_bugs= avctx->workaround_bugs;
03012
03013
03014
03015 s->low_delay= 1;
03016 avctx->pix_fmt= PIX_FMT_YUV420P;
03017
03018 decode_init_vlc(h);
03019
03020 if(avctx->extradata_size > 0 && avctx->extradata &&
03021 *(char *)avctx->extradata == 1){
03022 h->is_avc = 1;
03023 h->got_avcC = 0;
03024 } else {
03025 h->is_avc = 0;
03026 }
03027
03028 return 0;
03029 }
03030
03031 static void frame_start(H264Context *h){
03032 MpegEncContext * const s = &h->s;
03033 int i;
03034
03035 MPV_frame_start(s, s->avctx);
03036 ff_er_frame_start(s);
03037
03038 assert(s->linesize && s->uvlinesize);
03039
03040 for(i=0; i<16; i++){
03041 h->block_offset[i]= 4*((scan8[i] - scan8[0])&7) + 4*s->linesize*((scan8[i] - scan8[0])>>3);
03042 h->block_offset[24+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->linesize*((scan8[i] - scan8[0])>>3);
03043 }
03044 for(i=0; i<4; i++){
03045 h->block_offset[16+i]=
03046 h->block_offset[20+i]= 4*((scan8[i] - scan8[0])&7) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3);
03047 h->block_offset[24+16+i]=
03048 h->block_offset[24+20+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3);
03049 }
03050
03051
03052
03053 if(!s->obmc_scratchpad)
03054 s->obmc_scratchpad = av_malloc(16*s->linesize + 2*8*s->uvlinesize);
03055
03056
03057 }
03058
03059 static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize){
03060 MpegEncContext * const s = &h->s;
03061 int i;
03062
03063 src_y -= linesize;
03064 src_cb -= uvlinesize;
03065 src_cr -= uvlinesize;
03066
03067
03068
03069 h->left_border[0]= h->top_borders[0][s->mb_x][15];
03070 for(i=1; i<17; i++){
03071 h->left_border[i]= src_y[15+i* linesize];
03072 }
03073
03074 *(uint64_t*)(h->top_borders[0][s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
03075 *(uint64_t*)(h->top_borders[0][s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);
03076
03077 if(!(s->flags&CODEC_FLAG_GRAY)){
03078 h->left_border[17 ]= h->top_borders[0][s->mb_x][16+7];
03079 h->left_border[17+9]= h->top_borders[0][s->mb_x][24+7];
03080 for(i=1; i<9; i++){
03081 h->left_border[i+17 ]= src_cb[7+i*uvlinesize];
03082 h->left_border[i+17+9]= src_cr[7+i*uvlinesize];
03083 }
03084 *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
03085 *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
03086 }
03087 }
03088
03089 static inline void xchg_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int xchg){
03090 MpegEncContext * const s = &h->s;
03091 int temp8, i;
03092 uint64_t temp64;
03093 int deblock_left = (s->mb_x > 0);
03094 int deblock_top = (s->mb_y > 0);
03095
03096 src_y -= linesize + 1;
03097 src_cb -= uvlinesize + 1;
03098 src_cr -= uvlinesize + 1;
03099
03100 #define XCHG(a,b,t,xchg)\
03101 t= a;\
03102 if(xchg)\
03103 a= b;\
03104 b= t;
03105
03106 if(deblock_left){
03107 for(i = !deblock_top; i<17; i++){
03108 XCHG(h->left_border[i ], src_y [i* linesize], temp8, xchg);
03109 }
03110 }
03111
03112 if(deblock_top){
03113 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
03114 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
03115 if(s->mb_x+1 < s->mb_width){
03116 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x+1]), *(uint64_t*)(src_y +17), temp64, 1);
03117 }
03118 }
03119
03120 if(!(s->flags&CODEC_FLAG_GRAY)){
03121 if(deblock_left){
03122 for(i = !deblock_top; i<9; i++){
03123 XCHG(h->left_border[i+17 ], src_cb[i*uvlinesize], temp8, xchg);
03124 XCHG(h->left_border[i+17+9], src_cr[i*uvlinesize], temp8, xchg);
03125 }
03126 }
03127 if(deblock_top){
03128 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
03129 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
03130 }
03131 }
03132 }
03133
03134 static inline void backup_pair_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize){
03135 MpegEncContext * const s = &h->s;
03136 int i;
03137
03138 src_y -= 2 * linesize;
03139 src_cb -= 2 * uvlinesize;
03140 src_cr -= 2 * uvlinesize;
03141
03142
03143
03144 h->left_border[0]= h->top_borders[0][s->mb_x][15];
03145 h->left_border[1]= h->top_borders[1][s->mb_x][15];
03146 for(i=2; i<34; i++){
03147 h->left_border[i]= src_y[15+i* linesize];
03148 }
03149
03150 *(uint64_t*)(h->top_borders[0][s->mb_x]+0)= *(uint64_t*)(src_y + 32*linesize);
03151 *(uint64_t*)(h->top_borders[0][s->mb_x]+8)= *(uint64_t*)(src_y +8+32*linesize);
03152 *(uint64_t*)(h->top_borders[1][s->mb_x]+0)= *(uint64_t*)(src_y + 33*linesize);
03153 *(uint64_t*)(h->top_borders[1][s->mb_x]+8)= *(uint64_t*)(src_y +8+33*linesize);
03154
03155 if(!(s->flags&CODEC_FLAG_GRAY)){
03156 h->left_border[34 ]= h->top_borders[0][s->mb_x][16+7];
03157 h->left_border[34+ 1]= h->top_borders[1][s->mb_x][16+7];
03158 h->left_border[34+18 ]= h->top_borders[0][s->mb_x][24+7];
03159 h->left_border[34+18+1]= h->top_borders[1][s->mb_x][24+7];
03160 for(i=2; i<18; i++){
03161 h->left_border[i+34 ]= src_cb[7+i*uvlinesize];
03162 h->left_border[i+34+18]= src_cr[7+i*uvlinesize];
03163 }
03164 *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+16*uvlinesize);
03165 *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+16*uvlinesize);
03166 *(uint64_t*)(h->top_borders[1][s->mb_x]+16)= *(uint64_t*)(src_cb+17*uvlinesize);
03167 *(uint64_t*)(h->top_borders[1][s->mb_x]+24)= *(uint64_t*)(src_cr+17*uvlinesize);
03168 }
03169 }
03170
03171 static inline void xchg_pair_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int xchg){
03172 MpegEncContext * const s = &h->s;
03173 int temp8, i;
03174 uint64_t temp64;
03175 int deblock_left = (s->mb_x > 0);
03176 int deblock_top = (s->mb_y > 0);
03177
03178 tprintf("xchg_pair_border: src_y:%p src_cb:%p src_cr:%p ls:%d uvls:%d\n", src_y, src_cb, src_cr, linesize, uvlinesize);
03179
03180 src_y -= 2 * linesize + 1;
03181 src_cb -= 2 * uvlinesize + 1;
03182 src_cr -= 2 * uvlinesize + 1;
03183
03184 #define XCHG(a,b,t,xchg)\
03185 t= a;\
03186 if(xchg)\
03187 a= b;\
03188 b= t;
03189
03190 if(deblock_left){
03191 for(i = (!deblock_top)<<1; i<34; i++){
03192 XCHG(h->left_border[i ], src_y [i* linesize], temp8, xchg);
03193 }
03194 }
03195
03196 if(deblock_top){
03197 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
03198 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
03199 XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+0), *(uint64_t*)(src_y +1 +linesize), temp64, xchg);
03200 XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+8), *(uint64_t*)(src_y +9 +linesize), temp64, 1);
03201 }
03202
03203 if(!(s->flags&CODEC_FLAG_GRAY)){
03204 if(deblock_left){
03205 for(i = (!deblock_top) << 1; i<18; i++){
03206 XCHG(h->left_border[i+34 ], src_cb[i*uvlinesize], temp8, xchg);
03207 XCHG(h->left_border[i+34+18], src_cr[i*uvlinesize], temp8, xchg);
03208 }
03209 }
03210 if(deblock_top){
03211 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
03212 XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
03213 XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+16), *(uint64_t*)(src_cb+1 +uvlinesize), temp64, 1);
03214 XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+24), *(uint64_t*)(src_cr+1 +uvlinesize), temp64, 1);
03215 }
03216 }
03217 }
03218
03219 static void hl_decode_mb(H264Context *h){
03220 MpegEncContext * const s = &h->s;
03221 const int mb_x= s->mb_x;
03222 const int mb_y= s->mb_y;
03223 const int mb_xy= mb_x + mb_y*s->mb_stride;
03224 const int mb_type= s->current_picture.mb_type[mb_xy];
03225 uint8_t *dest_y, *dest_cb, *dest_cr;
03226 int linesize, uvlinesize ;
03227 int i;
03228 int *block_offset = &h->block_offset[0];
03229 const unsigned int bottom = mb_y & 1;
03230 const int transform_bypass = (s->qscale == 0 && h->sps.transform_bypass);
03231 void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
03232
03233 if(!s->decode)
03234 return;
03235
03236 dest_y = s->current_picture.data[0] + (mb_y * 16* s->linesize ) + mb_x * 16;
03237 dest_cb = s->current_picture.data[1] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
03238 dest_cr = s->current_picture.data[2] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
03239
03240 if (h->mb_field_decoding_flag) {
03241 linesize = s->linesize * 2;
03242 uvlinesize = s->uvlinesize * 2;
03243 block_offset = &h->block_offset[24];
03244 if(mb_y&1){
03245 dest_y -= s->linesize*15;
03246 dest_cb-= s->uvlinesize*7;
03247 dest_cr-= s->uvlinesize*7;
03248 }
03249 } else {
03250 linesize = s->linesize;
03251 uvlinesize = s->uvlinesize;
03252
03253 }
03254
03255 idct_add = transform_bypass
03256 ? IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 : s->dsp.add_pixels4
03257 : IS_8x8DCT(mb_type) ? s->dsp.h264_idct8_add : s->dsp.h264_idct_add;
03258
03259 if (IS_INTRA_PCM(mb_type)) {
03260 unsigned int x, y;
03261
03262
03263
03264 for(i=0; i<16; i++) {
03265 for (y=0; y<4; y++) {
03266 for (x=0; x<4; x++) {
03267 *(dest_y + block_offset[i] + y*linesize + x) = h->mb[i*16+y*4+x];
03268 }
03269 }
03270 }
03271 for(i=16; i<16+4; i++) {
03272 for (y=0; y<4; y++) {
03273 for (x=0; x<4; x++) {
03274 *(dest_cb + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
03275 }
03276 }
03277 }
03278 for(i=20; i<20+4; i++) {
03279 for (y=0; y<4; y++) {
03280 for (x=0; x<4; x++) {
03281 *(dest_cr + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
03282 }
03283 }
03284 }
03285 } else {
03286 if(IS_INTRA(mb_type)){
03287 if(h->deblocking_filter) {
03288 if (h->mb_aff_frame) {
03289 if (!bottom)
03290 xchg_pair_border(h, dest_y, dest_cb, dest_cr, s->linesize, s->uvlinesize, 1);
03291 } else {
03292 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1);
03293 }
03294 }
03295
03296 if(!(s->flags&CODEC_FLAG_GRAY)){
03297 h->pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize);
03298 h->pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize);
03299 }
03300
03301 if(IS_INTRA4x4(mb_type)){
03302 if(!s->encoding){
03303 if(IS_8x8DCT(mb_type)){
03304 for(i=0; i<16; i+=4){
03305 uint8_t * const ptr= dest_y + block_offset[i];
03306 const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
03307 h->pred8x8l[ dir ](ptr, (h->topleft_samples_available<<i)&0x8000,
03308 (h->topright_samples_available<<(i+1))&0x8000, linesize);
03309 if(h->non_zero_count_cache[ scan8[i] ])
03310 idct_add(ptr, h->mb + i*16, linesize);
03311 }
03312 }else
03313 for(i=0; i<16; i++){
03314 uint8_t * const ptr= dest_y + block_offset[i];
03315 uint8_t *topright;
03316 const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
03317 int tr;
03318
03319 if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
03320 const int topright_avail= (h->topright_samples_available<<i)&0x8000;
03321 assert(mb_y || linesize <= block_offset[i]);
03322 if(!topright_avail){
03323 tr= ptr[3 - linesize]*0x01010101;
03324 topright= (uint8_t*) &tr;
03325 }else
03326 topright= ptr + 4 - linesize;
03327 }else
03328 topright= NULL;
03329
03330 h->pred4x4[ dir ](ptr, topright, linesize);
03331 if(h->non_zero_count_cache[ scan8[i] ]){
03332 if(s->codec_id == CODEC_ID_H264)
03333 idct_add(ptr, h->mb + i*16, linesize);
03334 else
03335 svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
03336 }
03337 }
03338 }
03339 }else{
03340 h->pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize);
03341 if(s->codec_id == CODEC_ID_H264){
03342 if(!transform_bypass)
03343 h264_luma_dc_dequant_idct_c(h->mb, s->qscale);
03344 }else
03345 svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
03346 }
03347 if(h->deblocking_filter) {
03348 if (h->mb_aff_frame) {
03349 if (bottom) {
03350 uint8_t *pair_dest_y = s->current_picture.data[0] + ((mb_y-1) * 16* s->linesize ) + mb_x * 16;
03351 uint8_t *pair_dest_cb = s->current_picture.data[1] + ((mb_y-1) * 8 * s->uvlinesize) + mb_x * 8;
03352 uint8_t *pair_dest_cr = s->current_picture.data[2] + ((mb_y-1) * 8 * s->uvlinesize) + mb_x * 8;
03353 s->mb_y--;
03354 xchg_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize, 0);
03355 s->mb_y++;
03356 }
03357 } else {
03358 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
03359 }
03360 }
03361 }else if(s->codec_id == CODEC_ID_H264){
03362 hl_motion(h, dest_y, dest_cb, dest_cr,
03363 s->dsp.put_h264_qpel_pixels_tab, s->dsp.put_h264_chroma_pixels_tab,
03364 s->dsp.avg_h264_qpel_pixels_tab, s->dsp.avg_h264_chroma_pixels_tab,
03365 s->dsp.weight_h264_pixels_tab, s->dsp.biweight_h264_pixels_tab);
03366 }
03367
03368
03369 if(!IS_INTRA4x4(mb_type)){
03370 if(s->codec_id == CODEC_ID_H264){
03371 const int di = IS_8x8DCT(mb_type) ? 4 : 1;
03372 for(i=0; i<16; i+=di){
03373 if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
03374 uint8_t * const ptr= dest_y + block_offset[i];
03375 idct_add(ptr, h->mb + i*16, linesize);
03376 }
03377 }
03378 }else{
03379 for(i=0; i<16; i++){
03380 if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
03381 uint8_t * const ptr= dest_y + block_offset[i];
03382 svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
03383 }
03384 }
<