00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00027 #include "avcodec.h"
00028 #include "dsputil.h"
00029 #include "mpegvideo.h"
00030 #include "golomb.h"
00031 #include "mathops.h"
00032 #include "rectangle.h"
00033
00034 #include "rv34vlc.h"
00035 #include "rv34data.h"
00036 #include "rv34.h"
00037
00038
00039
00041 static const int rv34_mb_type_to_lavc[12] = {
00042 MB_TYPE_INTRA,
00043 MB_TYPE_INTRA16x16 | MB_TYPE_SEPARATE_DC,
00044 MB_TYPE_16x16 | MB_TYPE_L0,
00045 MB_TYPE_8x8 | MB_TYPE_L0,
00046 MB_TYPE_16x16 | MB_TYPE_L0,
00047 MB_TYPE_16x16 | MB_TYPE_L1,
00048 MB_TYPE_SKIP,
00049 MB_TYPE_DIRECT2 | MB_TYPE_16x16,
00050 MB_TYPE_16x8 | MB_TYPE_L0,
00051 MB_TYPE_8x16 | MB_TYPE_L0,
00052 MB_TYPE_16x16 | MB_TYPE_L0L1,
00053 MB_TYPE_16x16 | MB_TYPE_L0 | MB_TYPE_SEPARATE_DC
00054 };
00055
00056
00057 static RV34VLC intra_vlcs[NUM_INTRA_TABLES], inter_vlcs[NUM_INTER_TABLES];
00058
00071 static void rv34_gen_vlc(const uint8_t *bits, int size, VLC *vlc, const uint8_t *insyms)
00072 {
00073 int i;
00074 int counts[17] = {0}, codes[17];
00075 uint16_t cw[size], syms[size];
00076 uint8_t bits2[size];
00077 int maxbits = 0, realsize = 0;
00078
00079 for(i = 0; i < size; i++){
00080 if(bits[i]){
00081 bits2[realsize] = bits[i];
00082 syms[realsize] = insyms ? insyms[i] : i;
00083 realsize++;
00084 maxbits = FFMAX(maxbits, bits[i]);
00085 counts[bits[i]]++;
00086 }
00087 }
00088
00089 codes[0] = 0;
00090 for(i = 0; i < 16; i++)
00091 codes[i+1] = (codes[i] + counts[i]) << 1;
00092 for(i = 0; i < realsize; i++)
00093 cw[i] = codes[bits2[i]]++;
00094
00095 init_vlc_sparse(vlc, FFMIN(maxbits, 9), realsize,
00096 bits2, 1, 1,
00097 cw, 2, 2,
00098 syms, 2, 2, INIT_VLC_USE_STATIC);
00099 }
00100
00104 static av_cold void rv34_init_tables(void)
00105 {
00106 int i, j, k;
00107
00108 for(i = 0; i < NUM_INTRA_TABLES; i++){
00109 for(j = 0; j < 2; j++){
00110 rv34_gen_vlc(rv34_table_intra_cbppat [i][j], CBPPAT_VLC_SIZE, &intra_vlcs[i].cbppattern[j], NULL);
00111 rv34_gen_vlc(rv34_table_intra_secondpat[i][j], OTHERBLK_VLC_SIZE, &intra_vlcs[i].second_pattern[j], NULL);
00112 rv34_gen_vlc(rv34_table_intra_thirdpat [i][j], OTHERBLK_VLC_SIZE, &intra_vlcs[i].third_pattern[j], NULL);
00113 for(k = 0; k < 4; k++)
00114 rv34_gen_vlc(rv34_table_intra_cbp[i][j+k*2], CBP_VLC_SIZE, &intra_vlcs[i].cbp[j][k], rv34_cbp_code);
00115 }
00116 for(j = 0; j < 4; j++)
00117 rv34_gen_vlc(rv34_table_intra_firstpat[i][j], FIRSTBLK_VLC_SIZE, &intra_vlcs[i].first_pattern[j], NULL);
00118 rv34_gen_vlc(rv34_intra_coeff[i], COEFF_VLC_SIZE, &intra_vlcs[i].coefficient, NULL);
00119 }
00120
00121 for(i = 0; i < NUM_INTER_TABLES; i++){
00122 rv34_gen_vlc(rv34_inter_cbppat[i], CBPPAT_VLC_SIZE, &inter_vlcs[i].cbppattern[0], NULL);
00123 for(j = 0; j < 4; j++)
00124 rv34_gen_vlc(rv34_inter_cbp[i][j], CBP_VLC_SIZE, &inter_vlcs[i].cbp[0][j], rv34_cbp_code);
00125 for(j = 0; j < 2; j++){
00126 rv34_gen_vlc(rv34_table_inter_firstpat [i][j], FIRSTBLK_VLC_SIZE, &inter_vlcs[i].first_pattern[j], NULL);
00127 rv34_gen_vlc(rv34_table_inter_secondpat[i][j], OTHERBLK_VLC_SIZE, &inter_vlcs[i].second_pattern[j], NULL);
00128 rv34_gen_vlc(rv34_table_inter_thirdpat [i][j], OTHERBLK_VLC_SIZE, &inter_vlcs[i].third_pattern[j], NULL);
00129 }
00130 rv34_gen_vlc(rv34_inter_coeff[i], COEFF_VLC_SIZE, &inter_vlcs[i].coefficient, NULL);
00131 }
00132 }
00133
00135
00136
00142 static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
00143 {
00144 int i;
00145
00146 for(i=0; i<4; i++){
00147 const int z0= 13*(block[i+8*0] + block[i+8*2]);
00148 const int z1= 13*(block[i+8*0] - block[i+8*2]);
00149 const int z2= 7* block[i+8*1] - 17*block[i+8*3];
00150 const int z3= 17* block[i+8*1] + 7*block[i+8*3];
00151
00152 temp[4*i+0]= z0+z3;
00153 temp[4*i+1]= z1+z2;
00154 temp[4*i+2]= z1-z2;
00155 temp[4*i+3]= z0-z3;
00156 }
00157 }
00158
00163 static void rv34_inv_transform(DCTELEM *block){
00164 int temp[16];
00165 int i;
00166
00167 rv34_row_transform(temp, block);
00168
00169 for(i=0; i<4; i++){
00170 const int z0= 13*(temp[4*0+i] + temp[4*2+i]) + 0x200;
00171 const int z1= 13*(temp[4*0+i] - temp[4*2+i]) + 0x200;
00172 const int z2= 7* temp[4*1+i] - 17*temp[4*3+i];
00173 const int z3= 17* temp[4*1+i] + 7*temp[4*3+i];
00174
00175 block[i*8+0]= (z0 + z3)>>10;
00176 block[i*8+1]= (z1 + z2)>>10;
00177 block[i*8+2]= (z1 - z2)>>10;
00178 block[i*8+3]= (z0 - z3)>>10;
00179 }
00180
00181 }
00182
00189 static void rv34_inv_transform_noround(DCTELEM *block){
00190 int temp[16];
00191 int i;
00192
00193 rv34_row_transform(temp, block);
00194
00195 for(i=0; i<4; i++){
00196 const int z0= 13*(temp[4*0+i] + temp[4*2+i]);
00197 const int z1= 13*(temp[4*0+i] - temp[4*2+i]);
00198 const int z2= 7* temp[4*1+i] - 17*temp[4*3+i];
00199 const int z3= 17* temp[4*1+i] + 7*temp[4*3+i];
00200
00201 block[i*8+0]= ((z0 + z3)*3)>>11;
00202 block[i*8+1]= ((z1 + z2)*3)>>11;
00203 block[i*8+2]= ((z1 - z2)*3)>>11;
00204 block[i*8+3]= ((z0 - z3)*3)>>11;
00205 }
00206
00207 }
00208
00210
00211
00220 static int rv34_decode_cbp(GetBitContext *gb, RV34VLC *vlc, int table)
00221 {
00222 int pattern, code, cbp=0;
00223 int ones;
00224 static const int cbp_masks[3] = {0x100000, 0x010000, 0x110000};
00225 static const int shifts[4] = { 0, 2, 8, 10 };
00226 const int *curshift = shifts;
00227 int i, t, mask;
00228
00229 code = get_vlc2(gb, vlc->cbppattern[table].table, 9, 2);
00230 pattern = code & 0xF;
00231 code >>= 4;
00232
00233 ones = rv34_count_ones[pattern];
00234
00235 for(mask = 8; mask; mask >>= 1, curshift++){
00236 if(pattern & mask)
00237 cbp |= get_vlc2(gb, vlc->cbp[table][ones].table, vlc->cbp[table][ones].bits, 1) << curshift[0];
00238 }
00239
00240 for(i = 0; i < 4; i++){
00241 t = modulo_three_table[code][i];
00242 if(t == 1)
00243 cbp |= cbp_masks[get_bits1(gb)] << i;
00244 if(t == 2)
00245 cbp |= cbp_masks[2] << i;
00246 }
00247 return cbp;
00248 }
00249
00253 static inline void decode_coeff(DCTELEM *dst, int coef, int esc, GetBitContext *gb, VLC* vlc)
00254 {
00255 if(coef){
00256 if(coef == esc){
00257 coef = get_vlc2(gb, vlc->table, 9, 2);
00258 if(coef > 23){
00259 coef -= 23;
00260 coef = 22 + ((1 << coef) | get_bits(gb, coef));
00261 }
00262 coef += esc;
00263 }
00264 if(get_bits1(gb))
00265 coef = -coef;
00266 *dst = coef;
00267 }
00268 }
00269
00273 static inline void decode_subblock(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc)
00274 {
00275 int coeffs[4];
00276
00277 coeffs[0] = modulo_three_table[code][0];
00278 coeffs[1] = modulo_three_table[code][1];
00279 coeffs[2] = modulo_three_table[code][2];
00280 coeffs[3] = modulo_three_table[code][3];
00281 decode_coeff(dst , coeffs[0], 3, gb, vlc);
00282 if(is_block2){
00283 decode_coeff(dst+8, coeffs[1], 2, gb, vlc);
00284 decode_coeff(dst+1, coeffs[2], 2, gb, vlc);
00285 }else{
00286 decode_coeff(dst+1, coeffs[1], 2, gb, vlc);
00287 decode_coeff(dst+8, coeffs[2], 2, gb, vlc);
00288 }
00289 decode_coeff(dst+9, coeffs[3], 2, gb, vlc);
00290 }
00291
00303 static inline void rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc)
00304 {
00305 int code, pattern;
00306
00307 code = get_vlc2(gb, rvlc->first_pattern[fc].table, 9, 2);
00308
00309 pattern = code & 0x7;
00310
00311 code >>= 3;
00312 decode_subblock(dst, code, 0, gb, &rvlc->coefficient);
00313
00314 if(pattern & 4){
00315 code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
00316 decode_subblock(dst + 2, code, 0, gb, &rvlc->coefficient);
00317 }
00318 if(pattern & 2){
00319 code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
00320 decode_subblock(dst + 8*2, code, 1, gb, &rvlc->coefficient);
00321 }
00322 if(pattern & 1){
00323 code = get_vlc2(gb, rvlc->third_pattern[sc].table, 9, 2);
00324 decode_subblock(dst + 8*2+2, code, 0, gb, &rvlc->coefficient);
00325 }
00326
00327 }
00328
00333 static inline void rv34_dequant4x4(DCTELEM *block, int Qdc, int Q)
00334 {
00335 int i, j;
00336
00337 block[0] = (block[0] * Qdc + 8) >> 4;
00338 for(i = 0; i < 4; i++)
00339 for(j = !i; j < 4; j++)
00340 block[j + i*8] = (block[j + i*8] * Q + 8) >> 4;
00341 }
00342
00347 static inline void rv34_dequant4x4_16x16(DCTELEM *block, int Qdc, int Q)
00348 {
00349 int i;
00350
00351 for(i = 0; i < 3; i++)
00352 block[rv34_dezigzag[i]] = (block[rv34_dezigzag[i]] * Qdc + 8) >> 4;
00353 for(; i < 16; i++)
00354 block[rv34_dezigzag[i]] = (block[rv34_dezigzag[i]] * Q + 8) >> 4;
00355 }
00357
00358
00368 int ff_rv34_get_start_offset(GetBitContext *gb, int mb_size)
00369 {
00370 int i;
00371 for(i = 0; i < 5; i++)
00372 if(rv34_mb_max_sizes[i] >= mb_size - 1)
00373 break;
00374 return rv34_mb_bits_sizes[i];
00375 }
00376
00380 static inline RV34VLC* choose_vlc_set(int quant, int mod, int type)
00381 {
00382 if(mod == 2 && quant < 19) quant += 10;
00383 else if(mod && quant < 26) quant += 5;
00384 return type ? &inter_vlcs[rv34_quant_to_vlc_set[1][av_clip(quant, 0, 30)]]
00385 : &intra_vlcs[rv34_quant_to_vlc_set[0][av_clip(quant, 0, 30)]];
00386 }
00387
00391 static inline int rv34_decode_dquant(GetBitContext *gb, int quant)
00392 {
00393 if(get_bits1(gb))
00394 return rv34_dquant_tab[get_bits1(gb)][quant];
00395 else
00396 return get_bits(gb, 5);
00397 }
00398
00400
00407 static const uint8_t part_sizes_w[RV34_MB_TYPES] = { 2, 2, 2, 1, 2, 2, 2, 2, 2, 1, 2, 2 };
00408
00410 static const uint8_t part_sizes_h[RV34_MB_TYPES] = { 2, 2, 2, 1, 2, 2, 2, 2, 1, 2, 2, 2 };
00411
00413 static const uint8_t avail_indexes[4] = { 5, 6, 9, 10 };
00414
00422 static void rv34_pred_mv(RV34DecContext *r, int block_type, int subblock_no, int dmv_no)
00423 {
00424 MpegEncContext *s = &r->s;
00425 int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
00426 int A[2] = {0}, B[2], C[2];
00427 int i, j;
00428 int mx, my;
00429 int avail_index = avail_indexes[subblock_no];
00430 int c_off = part_sizes_w[block_type];
00431
00432 mv_pos += (subblock_no & 1) + (subblock_no >> 1)*s->b8_stride;
00433 if(subblock_no == 3)
00434 c_off = -1;
00435
00436 if(r->avail_cache[avail_index - 1]){
00437 A[0] = s->current_picture_ptr->motion_val[0][mv_pos-1][0];
00438 A[1] = s->current_picture_ptr->motion_val[0][mv_pos-1][1];
00439 }
00440 if(r->avail_cache[avail_index - 4]){
00441 B[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride][0];
00442 B[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride][1];
00443 }else{
00444 B[0] = A[0];
00445 B[1] = A[1];
00446 }
00447 if(!r->avail_cache[avail_index - 4 + c_off]){
00448 if(r->avail_cache[avail_index - 4] && (r->avail_cache[avail_index - 1] || r->rv30)){
00449 C[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride-1][0];
00450 C[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride-1][1];
00451 }else{
00452 C[0] = A[0];
00453 C[1] = A[1];
00454 }
00455 }else{
00456 C[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride+c_off][0];
00457 C[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride+c_off][1];
00458 }
00459 mx = mid_pred(A[0], B[0], C[0]);
00460 my = mid_pred(A[1], B[1], C[1]);
00461 mx += r->dmv[dmv_no][0];
00462 my += r->dmv[dmv_no][1];
00463 for(j = 0; j < part_sizes_h[block_type]; j++){
00464 for(i = 0; i < part_sizes_w[block_type]; i++){
00465 s->current_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][0] = mx;
00466 s->current_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][1] = my;
00467 }
00468 }
00469 }
00470
00471 #define GET_PTS_DIFF(a, b) ((a - b + 8192) & 0x1FFF)
00472
00476 static int calc_add_mv(RV34DecContext *r, int dir, int val)
00477 {
00478 int refdist = GET_PTS_DIFF(r->next_pts, r->last_pts);
00479 int dist = dir ? -GET_PTS_DIFF(r->next_pts, r->cur_pts) : GET_PTS_DIFF(r->cur_pts, r->last_pts);
00480 int mul;
00481
00482 if(!refdist) return 0;
00483 mul = (dist << 14) / refdist;
00484 return (val * mul + 0x2000) >> 14;
00485 }
00486
00490 static inline void rv34_pred_b_vector(int A[2], int B[2], int C[2],
00491 int A_avail, int B_avail, int C_avail,
00492 int *mx, int *my)
00493 {
00494 if(A_avail + B_avail + C_avail != 3){
00495 *mx = A[0] + B[0] + C[0];
00496 *my = A[1] + B[1] + C[1];
00497 if(A_avail + B_avail + C_avail == 2){
00498 *mx /= 2;
00499 *my /= 2;
00500 }
00501 }else{
00502 *mx = mid_pred(A[0], B[0], C[0]);
00503 *my = mid_pred(A[1], B[1], C[1]);
00504 }
00505 }
00506
00510 static void rv34_pred_mv_b(RV34DecContext *r, int block_type, int dir)
00511 {
00512 MpegEncContext *s = &r->s;
00513 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
00514 int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
00515 int A[2], B[2], C[2];
00516 int has_A = 0, has_B = 0, has_C = 0;
00517 int mx, my;
00518 int i, j;
00519 Picture *cur_pic = s->current_picture_ptr;
00520 const int mask = dir ? MB_TYPE_L1 : MB_TYPE_L0;
00521 int type = cur_pic->mb_type[mb_pos];
00522
00523 memset(A, 0, sizeof(A));
00524 memset(B, 0, sizeof(B));
00525 memset(C, 0, sizeof(C));
00526 if((r->avail_cache[5-1] & type) & mask){
00527 A[0] = cur_pic->motion_val[dir][mv_pos - 1][0];
00528 A[1] = cur_pic->motion_val[dir][mv_pos - 1][1];
00529 has_A = 1;
00530 }
00531 if((r->avail_cache[5-4] & type) & mask){
00532 B[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][0];
00533 B[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][1];
00534 has_B = 1;
00535 }
00536 if((r->avail_cache[5-2] & type) & mask){
00537 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][0];
00538 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][1];
00539 has_C = 1;
00540 }else if((s->mb_x+1) == s->mb_width && (r->avail_cache[5-5] & type) & mask){
00541 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][0];
00542 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][1];
00543 has_C = 1;
00544 }
00545
00546 rv34_pred_b_vector(A, B, C, has_A, has_B, has_C, &mx, &my);
00547
00548 mx += r->dmv[dir][0];
00549 my += r->dmv[dir][1];
00550
00551 for(j = 0; j < 2; j++){
00552 for(i = 0; i < 2; i++){
00553 cur_pic->motion_val[dir][mv_pos + i + j*s->b8_stride][0] = mx;
00554 cur_pic->motion_val[dir][mv_pos + i + j*s->b8_stride][1] = my;
00555 }
00556 }
00557 if(block_type == RV34_MB_B_BACKWARD || block_type == RV34_MB_B_FORWARD)
00558 fill_rectangle(cur_pic->motion_val[!dir][mv_pos], 2, 2, s->b8_stride, 0, 4);
00559 }
00560
00564 static void rv34_pred_mv_rv3(RV34DecContext *r, int block_type, int dir)
00565 {
00566 MpegEncContext *s = &r->s;
00567 int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
00568 int A[2] = {0}, B[2], C[2];
00569 int i, j, k;
00570 int mx, my;
00571 int avail_index = avail_indexes[0];
00572
00573 if(r->avail_cache[avail_index - 1]){
00574 A[0] = s->current_picture_ptr->motion_val[0][mv_pos-1][0];
00575 A[1] = s->current_picture_ptr->motion_val[0][mv_pos-1][1];
00576 }
00577 if(r->avail_cache[avail_index - 4]){
00578 B[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride][0];
00579 B[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride][1];
00580 }else{
00581 B[0] = A[0];
00582 B[1] = A[1];
00583 }
00584 if(!r->avail_cache[avail_index - 4 + 2]){
00585 if(r->avail_cache[avail_index - 4] && (r->avail_cache[avail_index - 1])){
00586 C[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride-1][0];
00587 C[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride-1][1];
00588 }else{
00589 C[0] = A[0];
00590 C[1] = A[1];
00591 }
00592 }else{
00593 C[0] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride+2][0];
00594 C[1] = s->current_picture_ptr->motion_val[0][mv_pos-s->b8_stride+2][1];
00595 }
00596 mx = mid_pred(A[0], B[0], C[0]);
00597 my = mid_pred(A[1], B[1], C[1]);
00598 mx += r->dmv[0][0];
00599 my += r->dmv[0][1];
00600 for(j = 0; j < 2; j++){
00601 for(i = 0; i < 2; i++){
00602 for(k = 0; k < 2; k++){
00603 s->current_picture_ptr->motion_val[k][mv_pos + i + j*s->b8_stride][0] = mx;
00604 s->current_picture_ptr->motion_val[k][mv_pos + i + j*s->b8_stride][1] = my;
00605 }
00606 }
00607 }
00608 }
00609
00610 static const int chroma_coeffs[3] = { 0, 3, 5 };
00611
00627 static inline void rv34_mc(RV34DecContext *r, const int block_type,
00628 const int xoff, const int yoff, int mv_off,
00629 const int width, const int height, int dir,
00630 const int thirdpel,
00631 qpel_mc_func (*qpel_mc)[16],
00632 h264_chroma_mc_func (*chroma_mc))
00633 {
00634 MpegEncContext *s = &r->s;
00635 uint8_t *Y, *U, *V, *srcY, *srcU, *srcV;
00636 int dxy, mx, my, umx, umy, lx, ly, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;
00637 int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride + mv_off;
00638 int is16x16 = 1;
00639
00640 if(thirdpel){
00641 int chroma_mx, chroma_my;
00642 mx = (s->current_picture_ptr->motion_val[dir][mv_pos][0] + (3 << 24)) / 3 - (1 << 24);
00643 my = (s->current_picture_ptr->motion_val[dir][mv_pos][1] + (3 << 24)) / 3 - (1 << 24);
00644 lx = (s->current_picture_ptr->motion_val[dir][mv_pos][0] + (3 << 24)) % 3;
00645 ly = (s->current_picture_ptr->motion_val[dir][mv_pos][1] + (3 << 24)) % 3;
00646 chroma_mx = (s->current_picture_ptr->motion_val[dir][mv_pos][0] + 1) >> 1;
00647 chroma_my = (s->current_picture_ptr->motion_val[dir][mv_pos][1] + 1) >> 1;
00648 umx = (chroma_mx + (3 << 24)) / 3 - (1 << 24);
00649 umy = (chroma_my + (3 << 24)) / 3 - (1 << 24);
00650 uvmx = chroma_coeffs[(chroma_mx + (3 << 24)) % 3];
00651 uvmy = chroma_coeffs[(chroma_my + (3 << 24)) % 3];
00652 }else{
00653 int cx, cy;
00654 mx = s->current_picture_ptr->motion_val[dir][mv_pos][0] >> 2;
00655 my = s->current_picture_ptr->motion_val[dir][mv_pos][1] >> 2;
00656 lx = s->current_picture_ptr->motion_val[dir][mv_pos][0] & 3;
00657 ly = s->current_picture_ptr->motion_val[dir][mv_pos][1] & 3;
00658 cx = s->current_picture_ptr->motion_val[dir][mv_pos][0] / 2;
00659 cy = s->current_picture_ptr->motion_val[dir][mv_pos][1] / 2;
00660 umx = cx >> 2;
00661 umy = cy >> 2;
00662 uvmx = (cx & 3) << 1;
00663 uvmy = (cy & 3) << 1;
00664
00665 if(uvmx == 6 && uvmy == 6)
00666 uvmx = uvmy = 4;
00667 }
00668 dxy = ly*4 + lx;
00669 srcY = dir ? s->next_picture_ptr->data[0] : s->last_picture_ptr->data[0];
00670 srcU = dir ? s->next_picture_ptr->data[1] : s->last_picture_ptr->data[1];
00671 srcV = dir ? s->next_picture_ptr->data[2] : s->last_picture_ptr->data[2];
00672 src_x = s->mb_x * 16 + xoff + mx;
00673 src_y = s->mb_y * 16 + yoff + my;
00674 uvsrc_x = s->mb_x * 8 + (xoff >> 1) + umx;
00675 uvsrc_y = s->mb_y * 8 + (yoff >> 1) + umy;
00676 srcY += src_y * s->linesize + src_x;
00677 srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
00678 srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
00679 if( (unsigned)(src_x - !!lx*2) > s->h_edge_pos - !!lx*2 - (width <<3) - 4
00680 || (unsigned)(src_y - !!ly*2) > s->v_edge_pos - !!ly*2 - (height<<3) - 4){
00681 uint8_t *uvbuf= s->edge_emu_buffer + 22 * s->linesize;
00682
00683 srcY -= 2 + 2*s->linesize;
00684 ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, (width<<3)+6, (height<<3)+6,
00685 src_x - 2, src_y - 2, s->h_edge_pos, s->v_edge_pos);
00686 srcY = s->edge_emu_buffer + 2 + 2*s->linesize;
00687 ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, (width<<2)+1, (height<<2)+1,
00688 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
00689 ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, (width<<2)+1, (height<<2)+1,
00690 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
00691 srcU = uvbuf;
00692 srcV = uvbuf + 16;
00693 }
00694 Y = s->dest[0] + xoff + yoff *s->linesize;
00695 U = s->dest[1] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
00696 V = s->dest[2] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
00697
00698 if(block_type == RV34_MB_P_16x8){
00699 qpel_mc[1][dxy](Y, srcY, s->linesize);
00700 Y += 8;
00701 srcY += 8;
00702 }else if(block_type == RV34_MB_P_8x16){
00703 qpel_mc[1][dxy](Y, srcY, s->linesize);
00704 Y += 8 * s->linesize;
00705 srcY += 8 * s->linesize;
00706 }
00707 is16x16 = (block_type != RV34_MB_P_8x8) && (block_type != RV34_MB_P_16x8) && (block_type != RV34_MB_P_8x16);
00708 qpel_mc[!is16x16][dxy](Y, srcY, s->linesize);
00709 chroma_mc[2-width] (U, srcU, s->uvlinesize, height*4, uvmx, uvmy);
00710 chroma_mc[2-width] (V, srcV, s->uvlinesize, height*4, uvmx, uvmy);
00711 }
00712
00713 static void rv34_mc_1mv(RV34DecContext *r, const int block_type,
00714 const int xoff, const int yoff, int mv_off,
00715 const int width, const int height, int dir)
00716 {
00717 rv34_mc(r, block_type, xoff, yoff, mv_off, width, height, dir, r->rv30,
00718 r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
00719 : r->s.dsp.put_rv40_qpel_pixels_tab,
00720 r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
00721 : r->s.dsp.put_rv40_chroma_pixels_tab);
00722 }
00723
00724 static void rv34_mc_2mv(RV34DecContext *r, const int block_type)
00725 {
00726 rv34_mc(r, block_type, 0, 0, 0, 2, 2, 0, r->rv30,
00727 r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
00728 : r->s.dsp.put_rv40_qpel_pixels_tab,
00729 r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
00730 : r->s.dsp.put_rv40_chroma_pixels_tab);
00731 rv34_mc(r, block_type, 0, 0, 0, 2, 2, 1, r->rv30,
00732 r->rv30 ? r->s.dsp.avg_rv30_tpel_pixels_tab
00733 : r->s.dsp.avg_rv40_qpel_pixels_tab,
00734 r->rv30 ? r->s.dsp.avg_h264_chroma_pixels_tab
00735 : r->s.dsp.avg_rv40_chroma_pixels_tab);
00736 }
00737
00738 static void rv34_mc_2mv_skip(RV34DecContext *r)
00739 {
00740 int i, j;
00741 for(j = 0; j < 2; j++)
00742 for(i = 0; i < 2; i++){
00743 rv34_mc(r, RV34_MB_P_8x8, i*8, j*8, i+j*r->s.b8_stride, 1, 1, 0, r->rv30,
00744 r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
00745 : r->s.dsp.put_rv40_qpel_pixels_tab,
00746 r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
00747 : r->s.dsp.put_rv40_chroma_pixels_tab);
00748 rv34_mc(r, RV34_MB_P_8x8, i*8, j*8, i+j*r->s.b8_stride, 1, 1, 1, r->rv30,
00749 r->rv30 ? r->s.dsp.avg_rv30_tpel_pixels_tab
00750 : r->s.dsp.avg_rv40_qpel_pixels_tab,
00751 r->rv30 ? r->s.dsp.avg_h264_chroma_pixels_tab
00752 : r->s.dsp.avg_rv40_chroma_pixels_tab);
00753 }
00754 }
00755
00757 static const int num_mvs[RV34_MB_TYPES] = { 0, 0, 1, 4, 1, 1, 0, 0, 2, 2, 2, 1 };
00758
00763 static int rv34_decode_mv(RV34DecContext *r, int block_type)
00764 {
00765 MpegEncContext *s = &r->s;
00766 GetBitContext *gb = &s->gb;
00767 int i, j, k, l;
00768 int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
00769 int next_bt;
00770
00771 memset(r->dmv, 0, sizeof(r->dmv));
00772 for(i = 0; i < num_mvs[block_type]; i++){
00773 r->dmv[i][0] = svq3_get_se_golomb(gb);
00774 r->dmv[i][1] = svq3_get_se_golomb(gb);
00775 }
00776 switch(block_type){
00777 case RV34_MB_TYPE_INTRA:
00778 case RV34_MB_TYPE_INTRA16x16:
00779 fill_rectangle(s->current_picture_ptr->motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
00780 return 0;
00781 case RV34_MB_SKIP:
00782 if(s->pict_type == FF_P_TYPE){
00783 fill_rectangle(s->current_picture_ptr->motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
00784 rv34_mc_1mv (r, block_type, 0, 0, 0, 2, 2, 0);
00785 break;
00786 }
00787 case RV34_MB_B_DIRECT:
00788
00789 next_bt = s->next_picture_ptr->mb_type[s->mb_x + s->mb_y * s->mb_stride];
00790 if(IS_INTRA(next_bt))
00791 fill_rectangle(s->current_picture_ptr->motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
00792 else
00793 for(j = 0; j < 2; j++)
00794 for(i = 0; i < 2; i++)
00795 for(k = 0; k < 2; k++)
00796 for(l = 0; l < 2; l++)
00797 s->current_picture_ptr->motion_val[l][mv_pos + i + j*s->b8_stride][k] = calc_add_mv(r, l, s->next_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][k]);
00798 if(IS_16X16(next_bt))
00799 rv34_mc_2mv(r, block_type);
00800 else
00801 rv34_mc_2mv_skip(r);
00802 fill_rectangle(s->current_picture_ptr->motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
00803 break;
00804 case RV34_MB_P_16x16:
00805 case RV34_MB_P_MIX16x16:
00806 rv34_pred_mv(r, block_type, 0, 0);
00807 rv34_mc_1mv (r, block_type, 0, 0, 0, 2, 2, 0);
00808 break;
00809 case RV34_MB_B_FORWARD:
00810 case RV34_MB_B_BACKWARD:
00811 r->dmv[1][0] = r->dmv[0][0];
00812 r->dmv[1][1] = r->dmv[0][1];
00813 if(r->rv30)
00814 rv34_pred_mv_rv3(r, block_type, block_type == RV34_MB_B_BACKWARD);
00815 else
00816 rv34_pred_mv_b (r, block_type, block_type == RV34_MB_B_BACKWARD);
00817 rv34_mc_1mv (r, block_type, 0, 0, 0, 2, 2, block_type == RV34_MB_B_BACKWARD);
00818 break;
00819 case RV34_MB_P_16x8:
00820 case RV34_MB_P_8x16:
00821 rv34_pred_mv(r, block_type, 0, 0);
00822 rv34_pred_mv(r, block_type, 1 + (block_type == RV34_MB_P_16x8), 1);
00823 if(block_type == RV34_MB_P_16x8){
00824 rv34_mc_1mv(r, block_type, 0, 0, 0, 2, 1, 0);
00825 rv34_mc_1mv(r, block_type, 0, 8, s->b8_stride, 2, 1, 0);
00826 }
00827 if(block_type == RV34_MB_P_8x16){
00828 rv34_mc_1mv(r, block_type, 0, 0, 0, 1, 2, 0);
00829 rv34_mc_1mv(r, block_type, 8, 0, 1, 1, 2, 0);
00830 }
00831 break;
00832 case RV34_MB_B_BIDIR:
00833 rv34_pred_mv_b (r, block_type, 0);
00834 rv34_pred_mv_b (r, block_type, 1);
00835 rv34_mc_2mv (r, block_type);
00836 break;
00837 case RV34_MB_P_8x8:
00838 for(i=0;i< 4;i++){
00839 rv34_pred_mv(r, block_type, i, i);
00840 rv34_mc_1mv (r, block_type, (i&1)<<3, (i&2)<<2, (i&1)+(i>>1)*s->b8_stride, 1, 1, 0);
00841 }
00842 break;
00843 }
00844
00845 return 0;
00846 }
00848
00854 static const int ittrans[9] = {
00855 DC_PRED, VERT_PRED, HOR_PRED, DIAG_DOWN_RIGHT_PRED, DIAG_DOWN_LEFT_PRED,
00856 VERT_RIGHT_PRED, VERT_LEFT_PRED, HOR_UP_PRED, HOR_DOWN_PRED,
00857 };
00858
00860 static const int ittrans16[4] = {
00861 DC_PRED8x8, VERT_PRED8x8, HOR_PRED8x8, PLANE_PRED8x8,
00862 };
00863
00867 static void rv34_pred_4x4_block(RV34DecContext *r, uint8_t *dst, int stride, int itype, int up, int left, int down, int right)
00868 {
00869 uint8_t *prev = dst - stride + 4;
00870 uint32_t topleft;
00871
00872 if(!up && !left)
00873 itype = DC_128_PRED;
00874 else if(!up){
00875 if(itype == VERT_PRED) itype = HOR_PRED;
00876 if(itype == DC_PRED) itype = LEFT_DC_PRED;
00877 }else if(!left){
00878 if(itype == HOR_PRED) itype = VERT_PRED;
00879 if(itype == DC_PRED) itype = TOP_DC_PRED;
00880 if(itype == DIAG_DOWN_LEFT_PRED) itype = DIAG_DOWN_LEFT_PRED_RV40_NODOWN;
00881 }
00882 if(!down){
00883 if(itype == DIAG_DOWN_LEFT_PRED) itype = DIAG_DOWN_LEFT_PRED_RV40_NODOWN;
00884 if(itype == HOR_UP_PRED) itype = HOR_UP_PRED_RV40_NODOWN;
00885 if(itype == VERT_LEFT_PRED) itype = VERT_LEFT_PRED_RV40_NODOWN;
00886 }
00887 if(!right && up){
00888 topleft = dst[-stride + 3] * 0x01010101;
00889 prev = (uint8_t*)&topleft;
00890 }
00891 r->h.pred4x4[itype](dst, prev, stride);
00892 }
00893
00895 static void rv34_add_4x4_block(uint8_t *dst, int stride, DCTELEM block[64], int off)
00896 {
00897 int x, y;
00898 for(y = 0; y < 4; y++)
00899 for(x = 0; x < 4; x++)
00900 dst[x + y*stride] = av_clip_uint8(dst[x + y*stride] + block[off + x+y*8]);
00901 }
00902
00903 static inline int adjust_pred16(int itype, int up, int left)
00904 {
00905 if(!up && !left)
00906 itype = DC_128_PRED8x8;
00907 else if(!up){
00908 if(itype == PLANE_PRED8x8)itype = HOR_PRED8x8;
00909 if(itype == VERT_PRED8x8) itype = HOR_PRED8x8;
00910 if(itype == DC_PRED8x8) itype = LEFT_DC_PRED8x8;
00911 }else if(!left){
00912 if(itype == PLANE_PRED8x8)itype = VERT_PRED8x8;
00913 if(itype == HOR_PRED8x8) itype = VERT_PRED8x8;
00914 if(itype == DC_PRED8x8) itype = TOP_DC_PRED8x8;
00915 }
00916 return itype;
00917 }
00918
00919 static void rv34_output_macroblock(RV34DecContext *r, int8_t *intra_types, int cbp, int is16)
00920 {
00921 MpegEncContext *s = &r->s;
00922 DSPContext *dsp = &s->dsp;
00923 int i, j;
00924 uint8_t *Y, *U, *V;
00925 int itype;
00926 int avail[6*8] = {0};
00927 int idx;
00928
00929
00930 if(r->avail_cache[0])
00931 avail[0] = 1;
00932 if(r->avail_cache[1])
00933 avail[1] = avail[2] = 1;
00934 if(r->avail_cache[2])
00935 avail[3] = avail[4] = 1;
00936 if(r->avail_cache[3])
00937 avail[5] = 1;
00938 if(r->avail_cache[4])
00939 avail[8] = avail[16] = 1;
00940 if(r->avail_cache[8])
00941 avail[24] = avail[32] = 1;
00942
00943 Y = s->dest[0];
00944 U = s->dest[1];
00945 V = s->dest[2];
00946 if(!is16){
00947 for(j = 0; j < 4; j++){
00948 idx = 9 + j*8;
00949 for(i = 0; i < 4; i++, cbp >>= 1, Y += 4, idx++){
00950 rv34_pred_4x4_block(r, Y, s->linesize, ittrans[intra_types[i]], avail[idx-8], avail[idx-1], avail[idx+7], avail[idx-7]);
00951 avail[idx] = 1;
00952 if(cbp & 1)
00953 rv34_add_4x4_block(Y, s->linesize, s->block[(i>>1)+(j&2)], (i&1)*4+(j&1)*32);
00954 }
00955 Y += s->linesize * 4 - 4*4;
00956 intra_types += s->b4_stride;
00957 }
00958 intra_types -= s->b4_stride * 4;
00959 fill_rectangle(r->avail_cache + 5, 2, 2, 4, 0, 4);
00960 for(j = 0; j < 2; j++){
00961 idx = 5 + j*4;
00962 for(i = 0; i < 2; i++, cbp >>= 1, idx++){
00963 rv34_pred_4x4_block(r, U + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*s->b4_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]);
00964 rv34_pred_4x4_block(r, V + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*s->b4_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]);
00965 r->avail_cache[idx] = 1;
00966 if(cbp & 0x01)
00967 rv34_add_4x4_block(U + i*4 + j*4*s->uvlinesize, s->uvlinesize, s->block[4], i*4+j*32);
00968 if(cbp & 0x10)
00969 rv34_add_4x4_block(V + i*4 + j*4*s->uvlinesize, s->uvlinesize, s->block[5], i*4+j*32);
00970 }
00971 }
00972 }else{
00973 itype = ittrans16[intra_types[0]];
00974 itype = adjust_pred16(itype, r->avail_cache[5-4], r->avail_cache[5-1]);
00975 r->h.pred16x16[itype](Y, s->linesize);
00976 dsp->add_pixels_clamped(s->block[0], Y, s->linesize);
00977 dsp->add_pixels_clamped(s->block[1], Y + 8, s->linesize);
00978 Y += s->linesize * 8;
00979 dsp->add_pixels_clamped(s->block[2], Y, s->linesize);
00980 dsp->add_pixels_clamped(s->block[3], Y + 8, s->linesize);
00981
00982 itype = ittrans16[intra_types[0]];
00983 if(itype == PLANE_PRED8x8) itype = DC_PRED8x8;
00984 itype = adjust_pred16(itype, r->avail_cache[5-4], r->avail_cache[5-1]);
00985 r->h.pred8x8[itype](U, s->uvlinesize);
00986 dsp->add_pixels_clamped(s->block[4], U, s->uvlinesize);
00987 r->h.pred8x8[itype](V, s->uvlinesize);
00988 dsp->add_pixels_clamped(s->block[5], V, s->uvlinesize);
00989 }
00990 }
00991
00993
00998 static int rv34_decode_mb_header(RV34DecContext *r, int8_t *intra_types)
00999 {
01000 MpegEncContext *s = &r->s;
01001 GetBitContext *gb = &s->gb;
01002 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
01003 int i, t;
01004
01005 if(!r->si.type){
01006 r->is16 = get_bits1(gb);
01007 if(!r->is16 && !r->rv30){
01008 if(!get_bits1(gb))
01009 av_log(s->avctx, AV_LOG_ERROR, "Need DQUANT\n");
01010 }
01011 s->current_picture_ptr->mb_type[mb_pos] = r->is16 ? MB_TYPE_INTRA16x16 : MB_TYPE_INTRA;
01012 r->block_type = r->is16 ? RV34_MB_TYPE_INTRA16x16 : RV34_MB_TYPE_INTRA;
01013 }else{
01014 r->block_type = r->decode_mb_info(r);
01015 if(r->block_type == -1)
01016 return -1;
01017 s->current_picture_ptr->mb_type[mb_pos] = rv34_mb_type_to_lavc[r->block_type];
01018 r->mb_type[mb_pos] = r->block_type;
01019 if(r->block_type == RV34_MB_SKIP){
01020 if(s->pict_type == FF_P_TYPE)
01021 r->mb_type[mb_pos] = RV34_MB_P_16x16;
01022 if(s->pict_type == FF_B_TYPE)
01023 r->mb_type[mb_pos] = RV34_MB_B_DIRECT;
01024 }
01025 r->is16 = !!IS_INTRA16x16(s->current_picture_ptr->mb_type[mb_pos]);
01026 rv34_decode_mv(r, r->block_type);
01027 if(r->block_type == RV34_MB_SKIP){
01028 fill_rectangle(intra_types, 4, 4, s->b4_stride, 0, sizeof(intra_types[0]));
01029 return 0;
01030 }
01031 r->chroma_vlc = 1;
01032 r->luma_vlc = 0;
01033 }
01034 if(IS_INTRA(s->current_picture_ptr->mb_type[mb_pos])){
01035 if(r->is16){
01036 t = get_bits(gb, 2);
01037 fill_rectangle(intra_types, 4, 4, s->b4_stride, t, sizeof(intra_types[0]));
01038 r->luma_vlc = 2;
01039 }else{
01040 if(r->decode_intra_types(r, gb, intra_types) < 0)
01041 return -1;
01042 r->luma_vlc = 1;
01043 }
01044 r->chroma_vlc = 0;
01045 r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 0);
01046 }else{
01047 for(i = 0; i < 16; i++)
01048 intra_types[(i & 3) + (i>>2) * s->b4_stride] = 0;
01049 r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1);
01050 if(r->mb_type[mb_pos] == RV34_MB_P_MIX16x16){
01051 r->is16 = 1;
01052 r->chroma_vlc = 1;
01053 r->luma_vlc = 2;
01054 r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 0);
01055 }
01056 }
01057
01058 return rv34_decode_cbp(gb, r->cur_vlcs, r->is16);
01059 }
01060
01069 #define LUMA_CBP_BLOCK_MASK 0x33
01070
01071 #define U_CBP_MASK 0x0F0000
01072 #define V_CBP_MASK 0xF00000
01073
01074
01075 static void rv34_apply_differences(RV34DecContext *r, int cbp)
01076 {
01077 static const int shifts[4] = { 0, 2, 8, 10 };
01078 MpegEncContext *s = &r->s;
01079 int i;
01080
01081 for(i = 0; i < 4; i++)
01082 if((cbp & (LUMA_CBP_BLOCK_MASK << shifts[i])) || r->block_type == RV34_MB_P_MIX16x16)
01083 s->dsp.add_pixels_clamped(s->block[i], s->dest[0] + (i & 1)*8 + (i&2)*4*s->linesize, s->linesize);
01084 if(cbp & U_CBP_MASK)
01085 s->dsp.add_pixels_clamped(s->block[4], s->dest[1], s->uvlinesize);
01086 if(cbp & V_CBP_MASK)
01087 s->dsp.add_pixels_clamped(s->block[5], s->dest[2], s->uvlinesize);
01088 }
01089
01090 static int is_mv_diff_gt_3(int16_t (*motion_val)[2], int step)
01091 {
01092 int d;
01093 d = motion_val[0][0] - motion_val[-step][0];
01094 if(d < -3 || d > 3)
01095 return 1;
01096 d = motion_val[0][1] - motion_val[-step][1];
01097 if(d < -3 || d > 3)
01098 return 1;
01099 return 0;
01100 }
01101
01102 static int rv34_set_deblock_coef(RV34DecContext *r)
01103 {
01104 MpegEncContext *s = &r->s;
01105 int hmvmask = 0, vmvmask = 0, i, j;
01106 int midx = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
01107 int16_t (*motion_val)[2] = s->current_picture_ptr->motion_val[0][midx];
01108 for(j = 0; j < 16; j += 8){
01109 for(i = 0; i < 2; i++){
01110 if(is_mv_diff_gt_3(motion_val + i, 1))
01111 vmvmask |= 0x11 << (j + i*2);
01112 if((j || s->mb_y) && is_mv_diff_gt_3(motion_val + i, s->b8_stride))
01113 hmvmask |= 0x03 << (j + i*2);
01114 }
01115 motion_val += s->b8_stride;
01116 }
01117 if(s->first_slice_line)
01118 hmvmask &= ~0x000F;
01119 if(!s->mb_x)
01120 vmvmask &= ~0x1111;
01121 if(r->rv30){
01122 vmvmask |= (vmvmask & 0x4444) >> 1;
01123 hmvmask |= (hmvmask & 0x0F00) >> 4;
01124 if(s->mb_x)
01125 r->deblock_coefs[s->mb_x - 1 + s->mb_y*s->mb_stride] |= (vmvmask & 0x1111) << 3;
01126 if(!s->first_slice_line)
01127 r->deblock_coefs[s->mb_x + (s->mb_y - 1)*s->mb_stride] |= (hmvmask & 0xF) << 12;
01128 }
01129 return hmvmask | vmvmask;
01130 }
01131
01132 static int rv34_decode_macroblock(RV34DecContext *r, int8_t *intra_types)
01133 {
01134 MpegEncContext *s = &r->s;
01135 GetBitContext *gb = &s->gb;
01136 int cbp, cbp2;
01137 int i, blknum, blkoff;
01138 DCTELEM block16[64];
01139 int luma_dc_quant;
01140 int dist;
01141 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
01142
01143
01144 memset(r->avail_cache, 0, sizeof(r->avail_cache));
01145 fill_rectangle(r->avail_cache + 5, 2, 2, 4, 1, 4);
01146 dist = (s->mb_x - s->resync_mb_x) + (s->mb_y - s->resync_mb_y) * s->mb_width;
01147 if(s->mb_x && dist)
01148 r->avail_cache[4] =
01149 r->avail_cache[8] = s->current_picture_ptr->mb_type[mb_pos - 1];
01150 if(dist >= s->mb_width)
01151 r->avail_cache[1] =
01152 r->avail_cache[2] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride];
01153 if(((s->mb_x+1) < s->mb_width) && dist >= s->mb_width - 1)
01154 r->avail_cache[3] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride + 1];
01155 if(s->mb_x && dist > s->mb_width)
01156 r->avail_cache[0] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride - 1];
01157
01158 s->qscale = r->si.quant;
01159 cbp = cbp2 = rv34_decode_mb_header(r, intra_types);
01160 r->cbp_luma [mb_pos] = cbp;
01161 r->cbp_chroma[mb_pos] = cbp >> 16;
01162 if(s->pict_type == FF_I_TYPE)
01163 r->deblock_coefs[mb_pos] = 0xFFFF;
01164 else
01165 r->deblock_coefs[mb_pos] = rv34_set_deblock_coef(r) | r->cbp_luma[mb_pos];
01166 s->current_picture_ptr->qscale_table[mb_pos] = s->qscale;
01167
01168 if(cbp == -1)
01169 return -1;
01170
01171 luma_dc_quant = r->block_type == RV34_MB_P_MIX16x16 ? r->luma_dc_quant_p[s->qscale] : r->luma_dc_quant_i[s->qscale];
01172 if(r->is16){
01173 memset(block16, 0, sizeof(block16));
01174 rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0);
01175 rv34_dequant4x4_16x16(block16, rv34_qscale_tab[luma_dc_quant],rv34_qscale_tab[s->qscale]);
01176 rv34_inv_transform_noround(block16);
01177 }
01178
01179 for(i = 0; i < 16; i++, cbp >>= 1){
01180 if(!r->is16 && !(cbp & 1)) continue;
01181 blknum = ((i & 2) >> 1) + ((i & 8) >> 2);
01182 blkoff = ((i & 1) << 2) + ((i & 4) << 3);
01183 if(cbp & 1)
01184 rv34_decode_block(s->block[blknum] + blkoff, gb, r->cur_vlcs, r->luma_vlc, 0);
01185 rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[s->qscale],rv34_qscale_tab[s->qscale]);
01186 if(r->is16)
01187 s->block[blknum][blkoff] = block16[(i & 3) | ((i & 0xC) << 1)];
01188 rv34_inv_transform(s->block[blknum] + blkoff);
01189 }
01190 if(r->block_type == RV34_MB_P_MIX16x16)
01191 r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1);
01192 for(; i < 24; i++, cbp >>= 1){
01193 if(!(cbp & 1)) continue;
01194 blknum = ((i & 4) >> 2) + 4;
01195 blkoff = ((i & 1) << 2) + ((i & 2) << 4);
01196 rv34_decode_block(s->block[blknum] + blkoff, gb, r->cur_vlcs, r->chroma_vlc, 1);
01197 rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[rv34_chroma_quant[1][s->qscale]],rv34_qscale_tab[rv34_chroma_quant[0][s->qscale]]);
01198 rv34_inv_transform(s->block[blknum] + blkoff);
01199 }
01200 if(IS_INTRA(s->current_picture_ptr->mb_type[mb_pos]))
01201 rv34_output_macroblock(r, intra_types, cbp2, r->is16);
01202 else
01203 rv34_apply_differences(r, cbp2);
01204
01205 return 0;
01206 }
01207
01208 static int check_slice_end(RV34DecContext *r, MpegEncContext *s)
01209 {
01210 int bits;
01211 if(s->mb_y >= s->mb_height)
01212 return 1;
01213 if(!s->mb_num_left)
01214 return 1;
01215 if(r->s.mb_skip_run > 1)
01216 return 0;
01217 bits = r->bits - get_bits_count(&s->gb);
01218 if(bits < 0 || (bits < 8 && !show_bits(&s->gb, bits)))
01219 return 1;
01220 return 0;
01221 }
01222
01223 static inline int slice_compare(SliceInfo *si1, SliceInfo *si2)
01224 {
01225 return si1->type != si2->type ||
01226 si1->start >= si2->start ||
01227 si1->width != si2->width ||
01228 si1->height != si2->height||
01229 si1->pts != si2->pts;
01230 }
01231
01232 static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int buf_size)
01233 {
01234 MpegEncContext *s = &r->s;
01235 GetBitContext *gb = &s->gb;
01236 int mb_pos;
01237 int res;
01238
01239 init_get_bits(&r->s.gb, buf, buf_size*8);
01240 res = r->parse_slice_header(r, gb, &r->si);
01241 if(res < 0){
01242 av_log(s->avctx, AV_LOG_ERROR, "Incorrect or unknown slice header\n");
01243 return -1;
01244 }
01245
01246 if ((s->mb_x == 0 && s->mb_y == 0) || s->current_picture_ptr==NULL) {
01247 if(s->width != r->si.width || s->height != r->si.height){
01248 av_log(s->avctx, AV_LOG_DEBUG, "Changing dimensions to %dx%d\n", r->si.width,r->si.height);
01249 MPV_common_end(s);
01250 s->width = r->si.width;
01251 s->height = r->si.height;
01252 avcodec_set_dimensions(s->avctx, s->width, s->height);
01253 if(MPV_common_init(s) < 0)
01254 return -1;
01255 r->intra_types_hist = av_realloc(r->intra_types_hist, s->b4_stride * 4 * 2 * sizeof(*r->intra_types_hist));
01256 r->intra_types = r->intra_types_hist + s->b4_stride * 4;
01257 r->mb_type = av_realloc(r->mb_type, r->s.mb_stride * r->s.mb_height * sizeof(*r->mb_type));
01258 r->cbp_luma = av_realloc(r->cbp_luma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_luma));
01259 r->cbp_chroma = av_realloc(r->cbp_chroma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_chroma));
01260 r->deblock_coefs = av_realloc(r->deblock_coefs, r->s.mb_stride * r->s.mb_height * sizeof(*r->deblock_coefs));
01261 }
01262 s->pict_type = r->si.type ? r->si.type : FF_I_TYPE;
01263 if(MPV_frame_start(s, s->avctx) < 0)
01264 return -1;
01265 ff_er_frame_start(s);
01266 r->cur_pts = r->si.pts;
01267 if(s->pict_type != FF_B_TYPE){
01268 r->last_pts = r->next_pts;
01269 r->next_pts = r->cur_pts;
01270 }
01271 s->mb_x = s->mb_y = 0;
01272 } else {
01273 int slice_type = r->si.type ? r->si.type : FF_I_TYPE;
01274
01275 if (slice_type != s->pict_type) {
01276 av_log(s->avctx, AV_LOG_ERROR, "Slice type mismatch\n");
01277 return AVERROR_INVALIDDATA;
01278 }
01279 if (s->width != r->si.width || s->height != r->si.height) {
01280 av_log(s->avctx, AV_LOG_ERROR, "Size mismatch\n");
01281 return AVERROR_INVALIDDATA;
01282 }
01283 }
01284
01285 r->si.end = end;
01286 s->qscale = r->si.quant;
01287 r->bits = buf_size*8;
01288 s->mb_num_left = r->si.end - r->si.start;
01289 r->s.mb_skip_run = 0;
01290
01291 mb_pos = s->mb_x + s->mb_y * s->mb_width;
01292 if(r->si.start != mb_pos){
01293 av_log(s->avctx, AV_LOG_ERROR, "Slice indicates MB offset %d, got %d\n", r->si.start, mb_pos);
01294 s->mb_x = r->si.start % s->mb_width;
01295 s->mb_y = r->si.start / s->mb_width;
01296 }
01297 memset(r->intra_types_hist, -1, s->b4_stride * 4 * 2 * sizeof(*r->intra_types_hist));
01298 s->first_slice_line = 1;
01299 s->resync_mb_x= s->mb_x;
01300 s->resync_mb_y= s->mb_y;
01301
01302 ff_init_block_index(s);
01303 while(!check_slice_end(r, s)) {
01304 ff_update_block_index(s);
01305 s->dsp.clear_blocks(s->block[0]);
01306
01307 if(rv34_decode_macroblock(r, r->intra_types + s->mb_x * 4 + 1) < 0){
01308 ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, AC_ERROR|DC_ERROR|MV_ERROR);
01309 return -1;
01310 }
01311 if (++s->mb_x == s->mb_width) {
01312 s->mb_x = 0;
01313 s->mb_y++;
01314 ff_init_block_index(s);
01315
01316 memmove(r->intra_types_hist, r->intra_types, s->b4_stride * 4 * sizeof(*r->intra_types_hist));
01317 memset(r->intra_types, -1, s->b4_stride * 4 * sizeof(*r->intra_types_hist));
01318
01319 if(r->loop_filter && s->mb_y >= 2)
01320 r->loop_filter(r, s->mb_y - 2);
01321 }
01322 if(s->mb_x == s->resync_mb_x)
01323 s->first_slice_line=0;
01324 s->mb_num_left--;
01325 }
01326 ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, AC_END|DC_END|MV_END);
01327
01328 return s->mb_y == s->mb_height;
01329 }
01330
01332
01336 av_cold int ff_rv34_decode_init(AVCodecContext *avctx)
01337 {
01338 RV34DecContext *r = avctx->priv_data;
01339 MpegEncContext *s = &r->s;
01340
01341 MPV_decode_defaults(s);
01342 s->avctx= avctx;
01343 s->out_format = FMT_H263;
01344 s->codec_id= avctx->codec_id;
01345
01346 s->width = avctx->width;
01347 s->height = avctx->height;
01348
01349 r->s.avctx = avctx;
01350 avctx->flags |= CODEC_FLAG_EMU_EDGE;
01351 r->s.flags |= CODEC_FLAG_EMU_EDGE;
01352 avctx->pix_fmt = PIX_FMT_YUV420P;
01353 avctx->has_b_frames = 1;
01354 s->low_delay = 0;
01355
01356 if (MPV_common_init(s) < 0)
01357 return -1;
01358
01359 ff_h264_pred_init(&r->h, CODEC_ID_RV40);
01360
01361 r->intra_types_hist = av_malloc(s->b4_stride * 4 * 2 * sizeof(*r->intra_types_hist));
01362 r->intra_types = r->intra_types_hist + s->b4_stride * 4;
01363
01364 r->mb_type = av_mallocz(r->s.mb_stride * r->s.mb_height * sizeof(*r->mb_type));
01365
01366 r->cbp_luma = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_luma));
01367 r->cbp_chroma = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_chroma));
01368 r->deblock_coefs = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->deblock_coefs));
01369
01370 if(!intra_vlcs[0].cbppattern[0].bits)
01371 rv34_init_tables();
01372
01373 return 0;
01374 }
01375
01376 static int get_slice_offset(AVCodecContext *avctx, const uint8_t *buf, int n)
01377 {
01378 if(avctx->slice_count) return avctx->slice_offset[n];
01379 else return AV_RL32(buf + n*8 - 4) == 1 ? AV_RL32(buf + n*8) : AV_RB32(buf + n*8);
01380 }
01381
01382 int ff_rv34_decode_frame(AVCodecContext *avctx,
01383 void *data, int *data_size,
01384 const uint8_t *buf, int buf_size)
01385 {
01386 RV34DecContext *r = avctx->priv_data;
01387 MpegEncContext *s = &r->s;
01388 AVFrame *pict = data;
01389 SliceInfo si;
01390 int i;
01391 int slice_count;
01392 const uint8_t *slices_hdr = NULL;
01393 int last = 0;
01394
01395
01396 if (buf_size == 0) {
01397
01398 if (s->low_delay==0 && s->next_picture_ptr) {
01399 *pict= *(AVFrame*)s->next_picture_ptr;
01400 s->next_picture_ptr= NULL;
01401
01402 *data_size = sizeof(AVFrame);
01403 }
01404 return 0;
01405 }
01406
01407 if(!avctx->slice_count){
01408 slice_count = (*buf++) + 1;
01409 slices_hdr = buf + 4;
01410 buf += 8 * slice_count;
01411 }else
01412 slice_count = avctx->slice_count;
01413
01414
01415 if(get_slice_offset(avctx, slices_hdr, 0) < 0 ||
01416 get_slice_offset(avctx, slices_hdr, 0) > buf_size){
01417 av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n");
01418 return -1;
01419 }
01420 init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, 0), buf_size-get_slice_offset(avctx, slices_hdr, 0));
01421 if(r->parse_slice_header(r, &r->s.gb, &si) < 0 || si.start){
01422 av_log(avctx, AV_LOG_ERROR, "First slice header is incorrect\n");
01423 return -1;
01424 }
01425 if((!s->last_picture_ptr || !s->last_picture_ptr->data[0]) && si.type == FF_B_TYPE)
01426 return -1;
01427
01428 if(avctx->hurry_up && si.type==FF_B_TYPE) return buf_size;
01429 if( (avctx->skip_frame >= AVDISCARD_NONREF && si.type==FF_B_TYPE)
01430 || (avctx->skip_frame >= AVDISCARD_NONKEY && si.type!=FF_I_TYPE)
01431 || avctx->skip_frame >= AVDISCARD_ALL)
01432 return buf_size;
01433
01434 if(avctx->hurry_up>=5)
01435 return buf_size;
01436
01437 for(i=0; i<slice_count; i++){
01438 int offset= get_slice_offset(avctx, slices_hdr, i);
01439 int size;
01440 if(i+1 == slice_count)
01441 size= buf_size - offset;
01442 else
01443 size= get_slice_offset(avctx, slices_hdr, i+1) - offset;
01444
01445 if(offset < 0 || offset > buf_size || size < 0){
01446 av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n");
01447 break;
01448 }
01449
01450 r->si.end = s->mb_width * s->mb_height;
01451 if(i+1 < slice_count){
01452 init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, i+1), (buf_size-get_slice_offset(avctx, slices_hdr, i+1))*8);
01453 if(r->parse_slice_header(r, &r->s.gb, &si) < 0){
01454 if(i+2 < slice_count)
01455 size = get_slice_offset(avctx, slices_hdr, i+2) - offset;
01456 else
01457 size = buf_size - offset;
01458 }else
01459 r->si.end = si.start;
01460 }
01461 last = rv34_decode_slice(r, r->si.end, buf + offset, size);
01462 s->mb_num_left = r->s.mb_x + r->s.mb_y*r->s.mb_width - r->si.start;
01463 if(last)
01464 break;
01465 }
01466
01467 if(last && s->current_picture_ptr){
01468 if(r->loop_filter)
01469 r->loop_filter(r, s->mb_height - 1);
01470 ff_er_frame_end(s);
01471 MPV_frame_end(s);
01472 if (s->pict_type == FF_B_TYPE || s->low_delay) {
01473 *pict= *(AVFrame*)s->current_picture_ptr;
01474 } else if (s->last_picture_ptr != NULL) {
01475 *pict= *(AVFrame*)s->last_picture_ptr;
01476 }
01477
01478 if(s->last_picture_ptr || s->low_delay){
01479 *data_size = sizeof(AVFrame);
01480 ff_print_debug_info(s, pict);
01481 }
01482 s->current_picture_ptr= NULL;
01483 }
01484 return buf_size;
01485 }
01486
01487 av_cold int ff_rv34_decode_end(AVCodecContext *avctx)
01488 {
01489 RV34DecContext *r = avctx->priv_data;
01490
01491 MPV_common_end(&r->s);
01492
01493 av_freep(&r->intra_types_hist);
01494 r->intra_types = NULL;
01495 av_freep(&r->mb_type);
01496 av_freep(&r->cbp_luma);
01497 av_freep(&r->cbp_chroma);
01498 av_freep(&r->deblock_coefs);
01499
01500 return 0;
01501 }