X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Frv34.c;h=7e83a99959ccaaf06a1a92672db8ef97fab89329;hb=9afb7061f938831248942050cfdb449e014ed427;hp=478c93ada0dd6d1caa3eb4cb869b61c79cea7336;hpb=d78062386e425deafe9a08d109cff70b7a2de22c;p=ffmpeg diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c index 478c93ada0d..7e83a99959c 100644 --- a/libavcodec/rv34.c +++ b/libavcodec/rv34.c @@ -131,10 +131,10 @@ static void rv34_gen_vlc(const uint8_t *bits, int size, VLC *vlc, const uint8_t vlc->table = &table_data[table_offs[num]]; vlc->table_allocated = table_offs[num + 1] - table_offs[num]; - init_vlc_sparse(vlc, FFMIN(maxbits, 9), realsize, - bits2, 1, 1, - cw, 2, 2, - syms, 2, 2, INIT_VLC_USE_NEW_STATIC); + ff_init_vlc_sparse(vlc, FFMIN(maxbits, 9), realsize, + bits2, 1, 1, + cw, 2, 2, + syms, 2, 2, INIT_VLC_USE_NEW_STATIC); } /** @@ -240,15 +240,15 @@ static inline void decode_subblock(DCTELEM *dst, int code, const int is_block2, { int flags = modulo_three_table[code]; - decode_coeff( dst+0, (flags >> 6) , 3, gb, vlc, q); + decode_coeff( dst+0*4+0, (flags >> 6) , 3, gb, vlc, q); if(is_block2){ - decode_coeff(dst+8, (flags >> 4) & 3, 2, gb, vlc, q); - decode_coeff(dst+1, (flags >> 2) & 3, 2, gb, vlc, q); + decode_coeff(dst+1*4+0, (flags >> 4) & 3, 2, gb, vlc, q); + decode_coeff(dst+0*4+1, (flags >> 2) & 3, 2, gb, vlc, q); }else{ - decode_coeff(dst+1, (flags >> 4) & 3, 2, gb, vlc, q); - decode_coeff(dst+8, (flags >> 2) & 3, 2, gb, vlc, q); + decode_coeff(dst+0*4+1, (flags >> 4) & 3, 2, gb, vlc, q); + decode_coeff(dst+1*4+0, (flags >> 2) & 3, 2, gb, vlc, q); } - decode_coeff( dst+9, (flags >> 0) & 3, 2, gb, vlc, q); + decode_coeff( dst+1*4+1, (flags >> 0) & 3, 2, gb, vlc, q); } /** @@ -260,20 +260,15 @@ static inline void decode_subblock1(DCTELEM *dst, int code, GetBitContext *gb, V decode_coeff(dst, coeff, 3, gb, vlc, q); } -static inline void decode_subblock3(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc, +static inline void decode_subblock3(DCTELEM *dst, int code, GetBitContext *gb, VLC *vlc, int q_dc, int q_ac1, int q_ac2) { int flags = modulo_three_table[code]; - decode_coeff( dst+0, (flags >> 6) , 3, gb, vlc, q_dc); - if(is_block2){ - decode_coeff(dst+8, (flags >> 4) & 3, 2, gb, vlc, q_ac1); - decode_coeff(dst+1, (flags >> 2) & 3, 2, gb, vlc, q_ac1); - }else{ - decode_coeff(dst+1, (flags >> 4) & 3, 2, gb, vlc, q_ac1); - decode_coeff(dst+8, (flags >> 2) & 3, 2, gb, vlc, q_ac1); - } - decode_coeff( dst+9, (flags >> 0) & 3, 2, gb, vlc, q_ac2); + decode_coeff(dst+0*4+0, (flags >> 6) , 3, gb, vlc, q_dc); + decode_coeff(dst+0*4+1, (flags >> 4) & 3, 2, gb, vlc, q_ac1); + decode_coeff(dst+1*4+0, (flags >> 2) & 3, 2, gb, vlc, q_ac1); + decode_coeff(dst+1*4+1, (flags >> 0) & 3, 2, gb, vlc, q_ac2); } /** @@ -287,7 +282,7 @@ static inline void decode_subblock3(DCTELEM *dst, int code, const int is_block2, * o--o */ -static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc, int q_dc, int q_ac1, int q_ac2) +static int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc, int q_dc, int q_ac1, int q_ac2) { int code, pattern, has_ac = 1; @@ -298,7 +293,7 @@ static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rv code >>= 3; if (modulo_three_table[code] & 0x3F) { - decode_subblock3(dst, code, 0, gb, &rvlc->coefficient, q_dc, q_ac1, q_ac2); + decode_subblock3(dst, code, gb, &rvlc->coefficient, q_dc, q_ac1, q_ac2); } else { decode_subblock1(dst, code, gb, &rvlc->coefficient, q_dc); if (!pattern) @@ -308,17 +303,17 @@ static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rv if(pattern & 4){ code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2); - decode_subblock(dst + 2, code, 0, gb, &rvlc->coefficient, q_ac2); + decode_subblock(dst + 4*0+2, code, 0, gb, &rvlc->coefficient, q_ac2); } if(pattern & 2){ // Looks like coefficients 1 and 2 are swapped for this block code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2); - decode_subblock(dst + 8*2, code, 1, gb, &rvlc->coefficient, q_ac2); + decode_subblock(dst + 4*2+0, code, 1, gb, &rvlc->coefficient, q_ac2); } if(pattern & 1){ code = get_vlc2(gb, rvlc->third_pattern[sc].table, 9, 2); - decode_subblock(dst + 8*2+2, code, 0, gb, &rvlc->coefficient, q_ac2); + decode_subblock(dst + 4*2+2, code, 0, gb, &rvlc->coefficient, q_ac2); } - return has_ac || pattern; + return has_ac | pattern; } /** @@ -472,26 +467,26 @@ static void rv34_pred_mv(RV34DecContext *r, int block_type, int subblock_no, int int A[2] = {0}, B[2], C[2]; int i, j; int mx, my; - int avail_index = avail_indexes[subblock_no]; + int* avail = r->avail_cache + avail_indexes[subblock_no]; int c_off = part_sizes_w[block_type]; mv_pos += (subblock_no & 1) + (subblock_no >> 1)*s->b8_stride; if(subblock_no == 3) c_off = -1; - if(r->avail_cache[avail_index - 1]){ + if(avail[-1]){ A[0] = s->current_picture_ptr->f.motion_val[0][mv_pos-1][0]; A[1] = s->current_picture_ptr->f.motion_val[0][mv_pos-1][1]; } - if(r->avail_cache[avail_index - 4]){ + if(avail[-4]){ B[0] = s->current_picture_ptr->f.motion_val[0][mv_pos-s->b8_stride][0]; B[1] = s->current_picture_ptr->f.motion_val[0][mv_pos-s->b8_stride][1]; }else{ B[0] = A[0]; B[1] = A[1]; } - if(!r->avail_cache[avail_index - 4 + c_off]){ - if(r->avail_cache[avail_index - 4] && (r->avail_cache[avail_index - 1] || r->rv30)){ + if(!avail[c_off-4]){ + if(avail[-4] && (avail[-1] || r->rv30)){ C[0] = s->current_picture_ptr->f.motion_val[0][mv_pos-s->b8_stride-1][0]; C[1] = s->current_picture_ptr->f.motion_val[0][mv_pos-s->b8_stride-1][1]; }else{ @@ -521,7 +516,7 @@ static void rv34_pred_mv(RV34DecContext *r, int block_type, int subblock_no, int */ static int calc_add_mv(RV34DecContext *r, int dir, int val) { - int mul = dir ? -r->weight2 : r->weight1; + int mul = dir ? -r->mv_weight2 : r->mv_weight1; return (val * mul + 0x2000) >> 14; } @@ -554,7 +549,7 @@ static void rv34_pred_mv_b(RV34DecContext *r, int block_type, int dir) MpegEncContext *s = &r->s; int mb_pos = s->mb_x + s->mb_y * s->mb_stride; int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride; - int A[2], B[2], C[2]; + int A[2] = { 0 }, B[2] = { 0 }, C[2] = { 0 }; int has_A = 0, has_B = 0, has_C = 0; int mx, my; int i, j; @@ -562,9 +557,6 @@ static void rv34_pred_mv_b(RV34DecContext *r, int block_type, int dir) const int mask = dir ? MB_TYPE_L1 : MB_TYPE_L0; int type = cur_pic->f.mb_type[mb_pos]; - memset(A, 0, sizeof(A)); - memset(B, 0, sizeof(B)); - memset(C, 0, sizeof(C)); if((r->avail_cache[6-1] & type) & mask){ A[0] = cur_pic->f.motion_val[dir][mv_pos - 1][0]; A[1] = cur_pic->f.motion_val[dir][mv_pos - 1][1]; @@ -611,21 +603,21 @@ static void rv34_pred_mv_rv3(RV34DecContext *r, int block_type, int dir) int A[2] = {0}, B[2], C[2]; int i, j, k; int mx, my; - int avail_index = avail_indexes[0]; + int* avail = r->avail_cache + avail_indexes[0]; - if(r->avail_cache[avail_index - 1]){ + if(avail[-1]){ A[0] = s->current_picture_ptr->f.motion_val[0][mv_pos - 1][0]; A[1] = s->current_picture_ptr->f.motion_val[0][mv_pos - 1][1]; } - if(r->avail_cache[avail_index - 4]){ + if(avail[-4]){ B[0] = s->current_picture_ptr->f.motion_val[0][mv_pos - s->b8_stride][0]; B[1] = s->current_picture_ptr->f.motion_val[0][mv_pos - s->b8_stride][1]; }else{ B[0] = A[0]; B[1] = A[1]; } - if(!r->avail_cache[avail_index - 4 + 2]){ - if(r->avail_cache[avail_index - 4] && (r->avail_cache[avail_index - 1])){ + if(!avail[-4 + 2]){ + if(avail[-4] && (avail[-1])){ C[0] = s->current_picture_ptr->f.motion_val[0][mv_pos - s->b8_stride - 1][0]; C[1] = s->current_picture_ptr->f.motion_val[0][mv_pos - s->b8_stride - 1][1]; }else{ @@ -711,8 +703,7 @@ static inline void rv34_mc(RV34DecContext *r, const int block_type, if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) { /* wait for the referenced mb row to be finished */ - int mb_row = FFMIN(s->mb_height - 1, - s->mb_y + ((yoff + my + 5 + 8 * height) >> 4)); + int mb_row = s->mb_y + ((yoff + my + 5 + 8 * height) >> 4); AVFrame *f = dir ? &s->next_picture_ptr->f : &s->last_picture_ptr->f; ff_thread_await_progress(f, mb_row, 0); } @@ -780,24 +771,24 @@ static void rv34_mc_1mv(RV34DecContext *r, const int block_type, static void rv4_weight(RV34DecContext *r) { - r->rdsp.rv40_weight_pixels_tab[0](r->s.dest[0], - r->tmp_b_block_y[0], - r->tmp_b_block_y[1], - r->weight1, - r->weight2, - r->s.linesize); - r->rdsp.rv40_weight_pixels_tab[1](r->s.dest[1], - r->tmp_b_block_uv[0], - r->tmp_b_block_uv[2], - r->weight1, - r->weight2, - r->s.uvlinesize); - r->rdsp.rv40_weight_pixels_tab[1](r->s.dest[2], - r->tmp_b_block_uv[1], - r->tmp_b_block_uv[3], - r->weight1, - r->weight2, - r->s.uvlinesize); + r->rdsp.rv40_weight_pixels_tab[r->scaled_weight][0](r->s.dest[0], + r->tmp_b_block_y[0], + r->tmp_b_block_y[1], + r->weight1, + r->weight2, + r->s.linesize); + r->rdsp.rv40_weight_pixels_tab[r->scaled_weight][1](r->s.dest[1], + r->tmp_b_block_uv[0], + r->tmp_b_block_uv[2], + r->weight1, + r->weight2, + r->s.uvlinesize); + r->rdsp.rv40_weight_pixels_tab[r->scaled_weight][1](r->s.dest[2], + r->tmp_b_block_uv[1], + r->tmp_b_block_uv[3], + r->weight1, + r->weight2, + r->s.uvlinesize); } static void rv34_mc_2mv(RV34DecContext *r, const int block_type) @@ -998,49 +989,48 @@ static inline int adjust_pred16(int itype, int up, int left) return itype; } +static inline void rv34_process_block(RV34DecContext *r, + uint8_t *pdst, int stride, + int fc, int sc, int q_dc, int q_ac) +{ + MpegEncContext *s = &r->s; + DCTELEM *ptr = s->block[0]; + int has_ac = rv34_decode_block(ptr, &s->gb, r->cur_vlcs, + fc, sc, q_dc, q_ac, q_ac); + if(has_ac){ + r->rdsp.rv34_idct_add(pdst, stride, ptr); + }else{ + r->rdsp.rv34_idct_dc_add(pdst, stride, ptr[0]); + ptr[0] = 0; + } +} + static void rv34_output_i16x16(RV34DecContext *r, int8_t *intra_types, int cbp) { - LOCAL_ALIGNED_16(DCTELEM, block16, [64]); + LOCAL_ALIGNED_16(DCTELEM, block16, [16]); MpegEncContext *s = &r->s; - DSPContext *dsp = &s->dsp; GetBitContext *gb = &s->gb; int q_dc = rv34_qscale_tab[ r->luma_dc_quant_i[s->qscale] ], q_ac = rv34_qscale_tab[s->qscale]; uint8_t *dst = s->dest[0]; DCTELEM *ptr = s->block[0]; - int avail[6*8] = {0}; int i, j, itype, has_ac; - memset(block16, 0, 64 * sizeof(*block16)); - - // Set neighbour information. - if(r->avail_cache[1]) - avail[0] = 1; - if(r->avail_cache[2]) - avail[1] = avail[2] = 1; - if(r->avail_cache[3]) - avail[3] = avail[4] = 1; - if(r->avail_cache[4]) - avail[5] = 1; - if(r->avail_cache[5]) - avail[8] = avail[16] = 1; - if(r->avail_cache[9]) - avail[24] = avail[32] = 1; + memset(block16, 0, 16 * sizeof(*block16)); has_ac = rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0, q_dc, q_dc, q_ac); if(has_ac) - r->rdsp.rv34_inv_transform_tab[1](block16); + r->rdsp.rv34_inv_transform(block16); else - r->rdsp.rv34_inv_transform_dc_tab[1](block16); + r->rdsp.rv34_inv_transform_dc(block16); itype = ittrans16[intra_types[0]]; itype = adjust_pred16(itype, r->avail_cache[6-4], r->avail_cache[6-1]); r->h.pred16x16[itype](dst, s->linesize); - dsp->clear_block(ptr); for(j = 0; j < 4; j++){ for(i = 0; i < 4; i++, cbp >>= 1){ - int dc = block16[i + j*8]; + int dc = block16[i + j*4]; if(cbp & 1){ has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac); @@ -1050,7 +1040,6 @@ static void rv34_output_i16x16(RV34DecContext *r, int8_t *intra_types, int cbp) if(has_ac){ ptr[0] = dc; r->rdsp.rv34_idct_add(dst+4*i, s->linesize, ptr); - dsp->clear_block(ptr); }else r->rdsp.rv34_idct_dc_add(dst+4*i, s->linesize, dc); } @@ -1073,14 +1062,8 @@ static void rv34_output_i16x16(RV34DecContext *r, int8_t *intra_types, int cbp) if(!(cbp & 1)) continue; pdst = dst + (i&1)*4 + (i&2)*2*s->uvlinesize; - has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->chroma_vlc, 1, q_dc, q_ac, q_ac); - if(has_ac){ - r->rdsp.rv34_idct_add(pdst, s->uvlinesize, ptr); - dsp->clear_block(ptr); - }else{ - r->rdsp.rv34_idct_dc_add(pdst, s->uvlinesize, ptr[0]); - ptr[0] = 0; - } + rv34_process_block(r, pdst, s->uvlinesize, + r->chroma_vlc, 1, q_dc, q_ac); } } } @@ -1088,14 +1071,10 @@ static void rv34_output_i16x16(RV34DecContext *r, int8_t *intra_types, int cbp) static void rv34_output_intra(RV34DecContext *r, int8_t *intra_types, int cbp) { MpegEncContext *s = &r->s; - DSPContext *dsp = &s->dsp; - GetBitContext *gb = &s->gb; - DCTELEM *ptr = s->block[0]; uint8_t *dst = s->dest[0]; int avail[6*8] = {0}; int i, j, k; - int idx, has_ac; - int q_ac, q_dc; + int idx, q_ac, q_dc; // Set neighbour information. if(r->avail_cache[1]) @@ -1119,14 +1098,8 @@ static void rv34_output_intra(RV34DecContext *r, int8_t *intra_types, int cbp) avail[idx] = 1; if(!(cbp & 1)) continue; - has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac); - if(has_ac){ - r->rdsp.rv34_idct_add(dst, s->linesize, ptr); - dsp->clear_block(ptr); - }else{ - r->rdsp.rv34_idct_dc_add(dst, s->linesize, ptr[0]); - ptr[0] = 0; - } + rv34_process_block(r, dst, s->linesize, + r->luma_vlc, 0, q_ac, q_ac); } dst += s->linesize * 4 - 4*4; intra_types += r->intra_types_stride; @@ -1150,15 +1123,8 @@ static void rv34_output_intra(RV34DecContext *r, int8_t *intra_types, int cbp) if(!(cbp&1)) continue; - has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->chroma_vlc, 1, q_dc, q_ac, q_ac); - if(has_ac){ - r->rdsp.rv34_idct_add(dst + 4*i, s->uvlinesize, ptr); - dsp->clear_block(ptr); - } - else { - r->rdsp.rv34_idct_dc_add(dst + 4*i, s->uvlinesize, ptr[0]); - ptr[0] = 0; - } + rv34_process_block(r, dst + 4*i, s->uvlinesize, + r->chroma_vlc, 1, q_dc, q_ac); } dst += 4*s->uvlinesize; @@ -1166,33 +1132,6 @@ static void rv34_output_intra(RV34DecContext *r, int8_t *intra_types, int cbp) } } -/** - * mask for retrieving all bits in coded block pattern - * corresponding to one 8x8 block - */ -#define LUMA_CBP_BLOCK_MASK 0x33 - -#define U_CBP_MASK 0x0F0000 -#define V_CBP_MASK 0xF00000 - -/** @} */ // recons group - - -static void rv34_apply_differences(RV34DecContext *r, int cbp) -{ - static const int shifts[4] = { 0, 2, 8, 10 }; - MpegEncContext *s = &r->s; - int i; - - for(i = 0; i < 4; i++) - if((cbp & (LUMA_CBP_BLOCK_MASK << shifts[i])) || r->block_type == RV34_MB_P_MIX16x16) - s->dsp.add_pixels_clamped(s->block[i], s->dest[0] + (i & 1)*8 + (i&2)*4*s->linesize, s->linesize); - if(cbp & U_CBP_MASK) - s->dsp.add_pixels_clamped(s->block[4], s->dest[1], s->uvlinesize); - if(cbp & V_CBP_MASK) - s->dsp.add_pixels_clamped(s->block[5], s->dest[2], s->uvlinesize); -} - static int is_mv_diff_gt_3(int16_t (*motion_val)[2], int step) { int d; @@ -1237,14 +1176,15 @@ static int rv34_set_deblock_coef(RV34DecContext *r) static int rv34_decode_inter_macroblock(RV34DecContext *r, int8_t *intra_types) { - MpegEncContext *s = &r->s; - GetBitContext *gb = &s->gb; + MpegEncContext *s = &r->s; + GetBitContext *gb = &s->gb; + uint8_t *dst = s->dest[0]; + DCTELEM *ptr = s->block[0]; + int mb_pos = s->mb_x + s->mb_y * s->mb_stride; int cbp, cbp2; int q_dc, q_ac, has_ac; - int i, blknum, blkoff; - LOCAL_ALIGNED_16(DCTELEM, block16, [64]); + int i, j; int dist; - int mb_pos = s->mb_x + s->mb_y * s->mb_stride; // Calculate which neighbours are available. Maybe it's worth optimizing too. memset(r->avail_cache, 0, sizeof(r->avail_cache)); @@ -1278,64 +1218,66 @@ static int rv34_decode_inter_macroblock(RV34DecContext *r, int8_t *intra_types) } if(r->is16){ - int luma_dc_quant = r->block_type == RV34_MB_P_MIX16x16 - ? r->luma_dc_quant_p[s->qscale] - : r->luma_dc_quant_i[s->qscale]; - q_dc = rv34_qscale_tab[luma_dc_quant]; + // Only for RV34_MB_P_MIX16x16 + LOCAL_ALIGNED_16(DCTELEM, block16, [16]); + memset(block16, 0, 16 * sizeof(*block16)); + q_dc = rv34_qscale_tab[ r->luma_dc_quant_p[s->qscale] ]; q_ac = rv34_qscale_tab[s->qscale]; - s->dsp.clear_block(block16); if (rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0, q_dc, q_dc, q_ac)) - r->rdsp.rv34_inv_transform_tab[1](block16); + r->rdsp.rv34_inv_transform(block16); else - r->rdsp.rv34_inv_transform_dc_tab[1](block16); + r->rdsp.rv34_inv_transform_dc(block16); q_ac = rv34_qscale_tab[s->qscale]; - for(i = 0; i < 16; i++, cbp >>= 1){ - DCTELEM *ptr; - blknum = ((i & 2) >> 1) + ((i & 8) >> 2); - blkoff = ((i & 1) << 2) + ((i & 4) << 3); - ptr = s->block[blknum] + blkoff; - if(cbp & 1) - has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac); - else - has_ac = 0; - ptr[0] = block16[(i & 3) | ((i & 0xC) << 1)]; - if(has_ac) - r->rdsp.rv34_inv_transform_tab[0](ptr); - else - r->rdsp.rv34_inv_transform_dc_tab[0](ptr); + + for(j = 0; j < 4; j++){ + for(i = 0; i < 4; i++, cbp >>= 1){ + int dc = block16[i + j*4]; + + if(cbp & 1){ + has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac); + }else + has_ac = 0; + + if(has_ac){ + ptr[0] = dc; + r->rdsp.rv34_idct_add(dst+4*i, s->linesize, ptr); + }else + r->rdsp.rv34_idct_dc_add(dst+4*i, s->linesize, dc); + } + + dst += 4*s->linesize; } + + r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1); }else{ q_ac = rv34_qscale_tab[s->qscale]; - for(i = 0; i < 16; i++, cbp >>= 1){ - DCTELEM *ptr; - if(!(cbp & 1)) continue; - blknum = ((i & 2) >> 1) + ((i & 8) >> 2); - blkoff = ((i & 1) << 2) + ((i & 4) << 3); - ptr = s->block[blknum] + blkoff; - has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac); - if(has_ac) - r->rdsp.rv34_inv_transform_tab[0](ptr); - else - r->rdsp.rv34_inv_transform_dc_tab[0](ptr); + + for(j = 0; j < 4; j++){ + for(i = 0; i < 4; i++, cbp >>= 1){ + if(!(cbp & 1)) continue; + + rv34_process_block(r, dst + 4*i, s->linesize, + r->luma_vlc, 0, q_ac, q_ac); + } + dst += 4*s->linesize; } } - if(r->block_type == RV34_MB_P_MIX16x16) - r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1); + q_dc = rv34_qscale_tab[rv34_chroma_quant[1][s->qscale]]; q_ac = rv34_qscale_tab[rv34_chroma_quant[0][s->qscale]]; - for(; i < 24; i++, cbp >>= 1){ - DCTELEM *ptr; - if(!(cbp & 1)) continue; - blknum = ((i & 4) >> 2) + 4; - blkoff = ((i & 1) << 2) + ((i & 2) << 4); - ptr = s->block[blknum] + blkoff; - if (rv34_decode_block(ptr, gb, r->cur_vlcs, r->chroma_vlc, 1, q_dc, q_ac, q_ac)) - r->rdsp.rv34_inv_transform_tab[0](ptr); - else - r->rdsp.rv34_inv_transform_dc_tab[0](ptr); + + for(j = 1; j < 3; j++){ + dst = s->dest[j]; + for(i = 0; i < 4; i++, cbp >>= 1){ + uint8_t *pdst; + if(!(cbp & 1)) continue; + pdst = dst + (i&1)*4 + (i&2)*2*s->uvlinesize; + + rv34_process_block(r, pdst, s->uvlinesize, + r->chroma_vlc, 1, q_dc, q_ac); + } } - rv34_apply_differences(r, cbp2); return 0; } @@ -1395,11 +1337,58 @@ static int check_slice_end(RV34DecContext *r, MpegEncContext *s) return 0; } + +static void rv34_decoder_free(RV34DecContext *r) +{ + av_freep(&r->intra_types_hist); + r->intra_types = NULL; + av_freep(&r->tmp_b_block_base); + av_freep(&r->mb_type); + av_freep(&r->cbp_luma); + av_freep(&r->cbp_chroma); + av_freep(&r->deblock_coefs); +} + + +static int rv34_decoder_alloc(RV34DecContext *r) +{ + r->intra_types_stride = r->s.mb_width * 4 + 4; + + r->cbp_chroma = av_malloc(r->s.mb_stride * r->s.mb_height * + sizeof(*r->cbp_chroma)); + r->cbp_luma = av_malloc(r->s.mb_stride * r->s.mb_height * + sizeof(*r->cbp_luma)); + r->deblock_coefs = av_malloc(r->s.mb_stride * r->s.mb_height * + sizeof(*r->deblock_coefs)); + r->intra_types_hist = av_malloc(r->intra_types_stride * 4 * 2 * + sizeof(*r->intra_types_hist)); + r->mb_type = av_mallocz(r->s.mb_stride * r->s.mb_height * + sizeof(*r->mb_type)); + + if (!(r->cbp_chroma && r->cbp_luma && r->deblock_coefs && + r->intra_types_hist && r->mb_type)) { + rv34_decoder_free(r); + return AVERROR(ENOMEM); + } + + r->intra_types = r->intra_types_hist + r->intra_types_stride * 4; + + return 0; +} + + +static int rv34_decoder_realloc(RV34DecContext *r) +{ + rv34_decoder_free(r); + return rv34_decoder_alloc(r); +} + + static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int buf_size) { MpegEncContext *s = &r->s; GetBitContext *gb = &s->gb; - int mb_pos; + int mb_pos, slice_type; int res; init_get_bits(&r->s.gb, buf, buf_size*8); @@ -1409,63 +1398,10 @@ static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int return -1; } - if ((s->mb_x == 0 && s->mb_y == 0) || s->current_picture_ptr==NULL) { - if(s->width != r->si.width || s->height != r->si.height){ - av_log(s->avctx, AV_LOG_DEBUG, "Changing dimensions to %dx%d\n", r->si.width,r->si.height); - MPV_common_end(s); - s->width = r->si.width; - s->height = r->si.height; - avcodec_set_dimensions(s->avctx, s->width, s->height); - if(MPV_common_init(s) < 0) - return -1; - r->intra_types_stride = s->mb_width*4 + 4; - r->intra_types_hist = av_realloc(r->intra_types_hist, r->intra_types_stride * 4 * 2 * sizeof(*r->intra_types_hist)); - r->intra_types = r->intra_types_hist + r->intra_types_stride * 4; - r->mb_type = av_realloc(r->mb_type, r->s.mb_stride * r->s.mb_height * sizeof(*r->mb_type)); - r->cbp_luma = av_realloc(r->cbp_luma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_luma)); - r->cbp_chroma = av_realloc(r->cbp_chroma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_chroma)); - r->deblock_coefs = av_realloc(r->deblock_coefs, r->s.mb_stride * r->s.mb_height * sizeof(*r->deblock_coefs)); - av_freep(&r->tmp_b_block_base); - } - s->pict_type = r->si.type ? r->si.type : AV_PICTURE_TYPE_I; - if(MPV_frame_start(s, s->avctx) < 0) - return -1; - ff_er_frame_start(s); - if (!r->tmp_b_block_base) { - int i; - - r->tmp_b_block_base = av_malloc(s->linesize * 48); - for (i = 0; i < 2; i++) - r->tmp_b_block_y[i] = r->tmp_b_block_base + i * 16 * s->linesize; - for (i = 0; i < 4; i++) - r->tmp_b_block_uv[i] = r->tmp_b_block_base + 32 * s->linesize - + (i >> 1) * 8 * s->uvlinesize + (i & 1) * 16; - } - r->cur_pts = r->si.pts; - if(s->pict_type != AV_PICTURE_TYPE_B){ - r->last_pts = r->next_pts; - r->next_pts = r->cur_pts; - }else{ - int refdist = GET_PTS_DIFF(r->next_pts, r->last_pts); - int dist0 = GET_PTS_DIFF(r->cur_pts, r->last_pts); - int dist1 = GET_PTS_DIFF(r->next_pts, r->cur_pts); - - if(!refdist){ - r->weight1 = r->weight2 = 8192; - }else{ - r->weight1 = (dist0 << 14) / refdist; - r->weight2 = (dist1 << 14) / refdist; - } - } - s->mb_x = s->mb_y = 0; - ff_thread_finish_setup(s->avctx); - } else { - int slice_type = r->si.type ? r->si.type : AV_PICTURE_TYPE_I; - - if (slice_type != s->pict_type) { - av_log(s->avctx, AV_LOG_ERROR, "Slice type mismatch\n"); - return AVERROR_INVALIDDATA; - } + slice_type = r->si.type ? r->si.type : AV_PICTURE_TYPE_I; + if (slice_type != s->pict_type) { + av_log(s->avctx, AV_LOG_ERROR, "Slice type mismatch\n"); + return AVERROR_INVALIDDATA; } r->si.end = end; @@ -1487,7 +1423,6 @@ static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int ff_init_block_index(s); while(!check_slice_end(r, s)) { ff_update_block_index(s); - s->dsp.clear_blocks(s->block[0]); if(r->si.type) res = rv34_decode_inter_macroblock(r, r->intra_types + s->mb_x * 4 + 4); @@ -1531,8 +1466,9 @@ av_cold int ff_rv34_decode_init(AVCodecContext *avctx) { RV34DecContext *r = avctx->priv_data; MpegEncContext *s = &r->s; + int ret; - MPV_decode_defaults(s); + ff_MPV_decode_defaults(s); s->avctx = avctx; s->out_format = FMT_H263; s->codec_id = avctx->codec_id; @@ -1547,29 +1483,22 @@ av_cold int ff_rv34_decode_init(AVCodecContext *avctx) avctx->has_b_frames = 1; s->low_delay = 0; - if (MPV_common_init(s) < 0) - return -1; + if ((ret = ff_MPV_common_init(s)) < 0) + return ret; - ff_h264_pred_init(&r->h, CODEC_ID_RV40, 8, 1); + ff_h264_pred_init(&r->h, AV_CODEC_ID_RV40, 8, 1); #if CONFIG_RV30_DECODER - if (avctx->codec_id == CODEC_ID_RV30) + if (avctx->codec_id == AV_CODEC_ID_RV30) ff_rv30dsp_init(&r->rdsp, &r->s.dsp); #endif #if CONFIG_RV40_DECODER - if (avctx->codec_id == CODEC_ID_RV40) + if (avctx->codec_id == AV_CODEC_ID_RV40) ff_rv40dsp_init(&r->rdsp, &r->s.dsp); #endif - r->intra_types_stride = 4*s->mb_stride + 4; - r->intra_types_hist = av_malloc(r->intra_types_stride * 4 * 2 * sizeof(*r->intra_types_hist)); - r->intra_types = r->intra_types_hist + r->intra_types_stride * 4; - - r->mb_type = av_mallocz(r->s.mb_stride * r->s.mb_height * sizeof(*r->mb_type)); - - r->cbp_luma = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_luma)); - r->cbp_chroma = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_chroma)); - r->deblock_coefs = av_malloc(r->s.mb_stride * r->s.mb_height * sizeof(*r->deblock_coefs)); + if ((ret = rv34_decoder_alloc(r)) < 0) + return ret; if(!intra_vlcs[0].cbppattern[0].bits) rv34_init_tables(); @@ -1579,40 +1508,17 @@ av_cold int ff_rv34_decode_init(AVCodecContext *avctx) int ff_rv34_decode_init_thread_copy(AVCodecContext *avctx) { + int err; RV34DecContext *r = avctx->priv_data; r->s.avctx = avctx; if (avctx->internal->is_copy) { - r->cbp_chroma = av_malloc(r->s.mb_stride * r->s.mb_height * - sizeof(*r->cbp_chroma)); - r->cbp_luma = av_malloc(r->s.mb_stride * r->s.mb_height * - sizeof(*r->cbp_luma)); - r->deblock_coefs = av_malloc(r->s.mb_stride * r->s.mb_height * - sizeof(*r->deblock_coefs)); - r->intra_types_hist = av_malloc(r->intra_types_stride * 4 * 2 * - sizeof(*r->intra_types_hist)); - r->mb_type = av_malloc(r->s.mb_stride * r->s.mb_height * - sizeof(*r->mb_type)); - - if (!(r->cbp_chroma && r->cbp_luma && r->deblock_coefs && - r->intra_types_hist && r->mb_type)) { - av_freep(&r->cbp_chroma); - av_freep(&r->cbp_luma); - av_freep(&r->deblock_coefs); - av_freep(&r->intra_types_hist); - av_freep(&r->mb_type); - r->intra_types = NULL; - return AVERROR(ENOMEM); - } - - r->intra_types = r->intra_types_hist + r->intra_types_stride * 4; r->tmp_b_block_base = NULL; - - memset(r->mb_type, 0, r->s.mb_stride * r->s.mb_height * - sizeof(*r->mb_type)); - - MPV_common_init(&r->s); + if ((err = ff_MPV_common_init(&r->s)) < 0) + return err; + if ((err = rv34_decoder_alloc(r)) < 0) + return err; } return 0; } @@ -1626,6 +1532,16 @@ int ff_rv34_decode_update_thread_context(AVCodecContext *dst, const AVCodecConte if (dst == src || !s1->context_initialized) return 0; + if (s->height != s1->height || s->width != s1->width) { + ff_MPV_common_end(s); + s->height = s1->height; + s->width = s1->width; + if ((err = ff_MPV_common_init(s)) < 0) + return err; + if ((err = rv34_decoder_realloc(r)) < 0) + return err; + } + if ((err = ff_mpeg_update_thread_context(dst, src))) return err; @@ -1635,10 +1551,6 @@ int ff_rv34_decode_update_thread_context(AVCodecContext *dst, const AVCodecConte memset(&r->si, 0, sizeof(r->si)); - /* necessary since it is it the condition checked for in decode_slice - * to call MPV_frame_start. cmp. comment at the end of decode_frame */ - s->current_picture_ptr = NULL; - return 0; } @@ -1648,8 +1560,34 @@ static int get_slice_offset(AVCodecContext *avctx, const uint8_t *buf, int n) else return AV_RL32(buf + n*8 - 4) == 1 ? AV_RL32(buf + n*8) : AV_RB32(buf + n*8); } +static int finish_frame(AVCodecContext *avctx, AVFrame *pict) +{ + RV34DecContext *r = avctx->priv_data; + MpegEncContext *s = &r->s; + int got_picture = 0; + + ff_er_frame_end(s); + ff_MPV_frame_end(s); + s->mb_num_left = 0; + + if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) + ff_thread_report_progress(&s->current_picture_ptr->f, INT_MAX, 0); + + if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) { + *pict = s->current_picture_ptr->f; + got_picture = 1; + } else if (s->last_picture_ptr != NULL) { + *pict = s->last_picture_ptr->f; + got_picture = 1; + } + if (got_picture) + ff_print_debug_info(s, pict); + + return got_picture; +} + int ff_rv34_decode_frame(AVCodecContext *avctx, - void *data, int *data_size, + void *data, int *got_picture_ptr, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; @@ -1667,10 +1605,10 @@ int ff_rv34_decode_frame(AVCodecContext *avctx, if (buf_size == 0) { /* special case for last picture */ if (s->low_delay==0 && s->next_picture_ptr) { - *pict = *(AVFrame*)s->next_picture_ptr; + *pict = s->next_picture_ptr->f; s->next_picture_ptr = NULL; - *data_size = sizeof(AVFrame); + *got_picture_ptr = 1; } return 0; } @@ -1687,20 +1625,105 @@ int ff_rv34_decode_frame(AVCodecContext *avctx, if(get_slice_offset(avctx, slices_hdr, 0) < 0 || get_slice_offset(avctx, slices_hdr, 0) > buf_size){ av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n"); - return -1; + return AVERROR_INVALIDDATA; } init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, 0), (buf_size-get_slice_offset(avctx, slices_hdr, 0))*8); if(r->parse_slice_header(r, &r->s.gb, &si) < 0 || si.start){ av_log(avctx, AV_LOG_ERROR, "First slice header is incorrect\n"); - return -1; + return AVERROR_INVALIDDATA; + } + if ((!s->last_picture_ptr || !s->last_picture_ptr->f.data[0]) && + si.type == AV_PICTURE_TYPE_B) { + av_log(avctx, AV_LOG_ERROR, "Invalid decoder state: B-frame without " + "reference data.\n"); + return AVERROR_INVALIDDATA; } - if ((!s->last_picture_ptr || !s->last_picture_ptr->f.data[0]) && si.type == AV_PICTURE_TYPE_B) - return -1; if( (avctx->skip_frame >= AVDISCARD_NONREF && si.type==AV_PICTURE_TYPE_B) || (avctx->skip_frame >= AVDISCARD_NONKEY && si.type!=AV_PICTURE_TYPE_I) || avctx->skip_frame >= AVDISCARD_ALL) return avpkt->size; + /* first slice */ + if (si.start == 0) { + if (s->mb_num_left > 0) { + av_log(avctx, AV_LOG_ERROR, "New frame but still %d MB left.", + s->mb_num_left); + ff_er_frame_end(s); + ff_MPV_frame_end(s); + } + + if (s->width != si.width || s->height != si.height) { + int err; + + if (HAVE_THREADS && + (s->avctx->active_thread_type & FF_THREAD_FRAME)) { + av_log_missing_feature(s->avctx, "Width/height changing with " + "frame threading is", 0); + return AVERROR_PATCHWELCOME; + } + + av_log(s->avctx, AV_LOG_WARNING, "Changing dimensions to %dx%d\n", + si.width, si.height); + ff_MPV_common_end(s); + s->width = si.width; + s->height = si.height; + avcodec_set_dimensions(s->avctx, s->width, s->height); + if ((err = ff_MPV_common_init(s)) < 0) + return err; + if ((err = rv34_decoder_realloc(r)) < 0) + return err; + } + s->pict_type = si.type ? si.type : AV_PICTURE_TYPE_I; + if (ff_MPV_frame_start(s, s->avctx) < 0) + return -1; + ff_er_frame_start(s); + if (!r->tmp_b_block_base) { + int i; + + r->tmp_b_block_base = av_malloc(s->linesize * 48); + for (i = 0; i < 2; i++) + r->tmp_b_block_y[i] = r->tmp_b_block_base + + i * 16 * s->linesize; + for (i = 0; i < 4; i++) + r->tmp_b_block_uv[i] = r->tmp_b_block_base + 32 * s->linesize + + (i >> 1) * 8 * s->uvlinesize + + (i & 1) * 16; + } + r->cur_pts = si.pts; + if (s->pict_type != AV_PICTURE_TYPE_B) { + r->last_pts = r->next_pts; + r->next_pts = r->cur_pts; + } else { + int refdist = GET_PTS_DIFF(r->next_pts, r->last_pts); + int dist0 = GET_PTS_DIFF(r->cur_pts, r->last_pts); + int dist1 = GET_PTS_DIFF(r->next_pts, r->cur_pts); + + if(!refdist){ + r->mv_weight1 = r->mv_weight2 = r->weight1 = r->weight2 = 8192; + r->scaled_weight = 0; + }else{ + r->mv_weight1 = (dist0 << 14) / refdist; + r->mv_weight2 = (dist1 << 14) / refdist; + if((r->mv_weight1|r->mv_weight2) & 511){ + r->weight1 = r->mv_weight1; + r->weight2 = r->mv_weight2; + r->scaled_weight = 0; + }else{ + r->weight1 = r->mv_weight1 >> 9; + r->weight2 = r->mv_weight2 >> 9; + r->scaled_weight = 1; + } + } + } + s->mb_x = s->mb_y = 0; + ff_thread_finish_setup(s->avctx); + } else if (HAVE_THREADS && + (s->avctx->active_thread_type & FF_THREAD_FRAME)) { + av_log(s->avctx, AV_LOG_ERROR, "Decoder needs full frames in frame " + "multithreading mode (start MB is %d).\n", si.start); + return AVERROR_INVALIDDATA; + } + for(i = 0; i < slice_count; i++){ int offset = get_slice_offset(avctx, slices_hdr, i); int size; @@ -1715,6 +1738,8 @@ int ff_rv34_decode_frame(AVCodecContext *avctx, } r->si.end = s->mb_width * s->mb_height; + s->mb_num_left = r->s.mb_x + r->s.mb_y*r->s.mb_width - r->si.start; + if(i+1 < slice_count){ if (get_slice_offset(avctx, slices_hdr, i+1) < 0 || get_slice_offset(avctx, slices_hdr, i+1) > buf_size) { @@ -1735,31 +1760,29 @@ int ff_rv34_decode_frame(AVCodecContext *avctx, break; } last = rv34_decode_slice(r, r->si.end, buf + offset, size); - s->mb_num_left = r->s.mb_x + r->s.mb_y*r->s.mb_width - r->si.start; if(last) break; } - if(last && s->current_picture_ptr){ - if(r->loop_filter) - r->loop_filter(r, s->mb_height - 1); - if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) - ff_thread_report_progress(&s->current_picture_ptr->f, - s->mb_height - 1, 0); - ff_er_frame_end(s); - MPV_frame_end(s); - if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) { - *pict = *(AVFrame*)s->current_picture_ptr; - } else if (s->last_picture_ptr != NULL) { - *pict = *(AVFrame*)s->last_picture_ptr; - } - - if(s->last_picture_ptr || s->low_delay){ - *data_size = sizeof(AVFrame); - ff_print_debug_info(s, pict); + if (s->current_picture_ptr) { + if (last) { + if(r->loop_filter) + r->loop_filter(r, s->mb_height - 1); + + *got_picture_ptr = finish_frame(avctx, pict); + } else if (HAVE_THREADS && + (s->avctx->active_thread_type & FF_THREAD_FRAME)) { + av_log(avctx, AV_LOG_INFO, "marking unfished frame as finished\n"); + /* always mark the current frame as finished, frame-mt supports + * only complete frames */ + ff_er_frame_end(s); + ff_MPV_frame_end(s); + s->mb_num_left = 0; + ff_thread_report_progress(&s->current_picture_ptr->f, INT_MAX, 0); + return AVERROR_INVALIDDATA; } - s->current_picture_ptr = NULL; //so we can detect if frame_end wasnt called (find some nicer solution...) } + return avpkt->size; } @@ -1767,15 +1790,8 @@ av_cold int ff_rv34_decode_end(AVCodecContext *avctx) { RV34DecContext *r = avctx->priv_data; - MPV_common_end(&r->s); - - av_freep(&r->intra_types_hist); - r->intra_types = NULL; - av_freep(&r->tmp_b_block_base); - av_freep(&r->mb_type); - av_freep(&r->cbp_luma); - av_freep(&r->cbp_chroma); - av_freep(&r->deblock_coefs); + ff_MPV_common_end(&r->s); + rv34_decoder_free(r); return 0; }