Subversion Repositories Kolibri OS

Rev

Go to most recent revision | Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * HEVC video Decoder
  3.  *
  4.  * Copyright (C) 2012 - 2013 Guillaume Martres
  5.  * Copyright (C) 2013 Seppo Tomperi
  6.  * Copyright (C) 2013 Wassim Hamidouche
  7.  *
  8.  * This file is part of FFmpeg.
  9.  *
  10.  * FFmpeg is free software; you can redistribute it and/or
  11.  * modify it under the terms of the GNU Lesser General Public
  12.  * License as published by the Free Software Foundation; either
  13.  * version 2.1 of the License, or (at your option) any later version.
  14.  *
  15.  * FFmpeg is distributed in the hope that it will be useful,
  16.  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18.  * Lesser General Public License for more details.
  19.  *
  20.  * You should have received a copy of the GNU Lesser General Public
  21.  * License along with FFmpeg; if not, write to the Free Software
  22.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  23.  */
  24.  
  25. #include "libavutil/common.h"
  26. #include "libavutil/internal.h"
  27.  
  28. #include "cabac_functions.h"
  29. #include "golomb.h"
  30. #include "hevc.h"
  31. #include "bit_depth_template.c"
  32.  
  33. #define LUMA 0
  34. #define CB 1
  35. #define CR 2
  36.  
  37. static const uint8_t tctable[54] = {
  38.      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, // QP  0...18
  39.      1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, // QP 19...37
  40.      5, 5, 6, 6, 7, 8, 9,10,11,13,14,16,18,20,22,24           // QP 38...53
  41. };
  42.  
  43. static const uint8_t betatable[52] = {
  44.      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, 7, 8, // QP 0...18
  45.      9,10,11,12,13,14,15,16,17,18,20,22,24,26,28,30,32,34,36, // QP 19...37
  46.     38,40,42,44,46,48,50,52,54,56,58,60,62,64                 // QP 38...51
  47. };
  48.  
  49. static int chroma_tc(HEVCContext *s, int qp_y, int c_idx, int tc_offset)
  50. {
  51.     static const int qp_c[] = { 29, 30, 31, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37 };
  52.     int qp_i, offset;
  53.     int qp;
  54.     int idxt;
  55.  
  56.     // slice qp offset is not used for deblocking
  57.     if (c_idx == 1)
  58.         offset = s->pps->cb_qp_offset;
  59.     else
  60.         offset = s->pps->cr_qp_offset;
  61.  
  62.     qp_i = av_clip_c(qp_y + offset, 0, 57);
  63.     if (qp_i < 30)
  64.         qp = qp_i;
  65.     else if (qp_i > 43)
  66.         qp = qp_i - 6;
  67.     else
  68.         qp = qp_c[qp_i - 30];
  69.  
  70.     idxt = av_clip_c(qp + DEFAULT_INTRA_TC_OFFSET + tc_offset, 0, 53);
  71.     return tctable[idxt];
  72. }
  73.  
  74. static int get_qPy_pred(HEVCContext *s, int xC, int yC, int xBase, int yBase, int log2_cb_size)
  75. {
  76.     HEVCLocalContext *lc     = s->HEVClc;
  77.     int ctb_size_mask        = (1 << s->sps->log2_ctb_size) - 1;
  78.     int MinCuQpDeltaSizeMask = (1 << (s->sps->log2_ctb_size - s->pps->diff_cu_qp_delta_depth)) - 1;
  79.     int xQgBase              = xBase - ( xBase & MinCuQpDeltaSizeMask );
  80.     int yQgBase              = yBase - ( yBase & MinCuQpDeltaSizeMask );
  81.     int min_cb_width         = s->sps->min_cb_width;
  82.     int min_cb_height        = s->sps->min_cb_height;
  83.     int x_cb                 = xQgBase >> s->sps->log2_min_cb_size;
  84.     int y_cb                 = yQgBase >> s->sps->log2_min_cb_size;
  85.     int availableA           = (xBase & ctb_size_mask) && (xQgBase & ctb_size_mask);
  86.     int availableB           = (yBase & ctb_size_mask) && (yQgBase & ctb_size_mask);
  87.     int qPy_pred;
  88.     int qPy_a;
  89.     int qPy_b;
  90.  
  91.     // qPy_pred
  92.     if (lc->first_qp_group) {
  93.         lc->first_qp_group = !lc->tu.is_cu_qp_delta_coded;
  94.         qPy_pred = s->sh.slice_qp;
  95.     } else {
  96.         qPy_pred = lc->qp_y;
  97.         if (log2_cb_size < s->sps->log2_ctb_size - s->pps->diff_cu_qp_delta_depth) {
  98.             static const int offsetX[8][8] = {
  99.                     {-1, 1, 3, 1, 7, 1, 3, 1},
  100.                     { 0, 0, 0, 0, 0, 0, 0, 0},
  101.                     { 1, 3, 1, 3, 1, 3, 1, 3},
  102.                     { 2, 2, 2, 2, 2, 2, 2, 2},
  103.                     { 3, 5, 7, 5, 3, 5, 7, 5},
  104.                     { 4, 4, 4, 4, 4, 4, 4, 4},
  105.                     { 5, 7, 5, 7, 5, 7, 5, 7},
  106.                     { 6, 6, 6, 6, 6, 6, 6, 6}
  107.             };
  108.             static const int offsetY[8][8] = {
  109.                     { 7, 0, 1, 2, 3, 4, 5, 6},
  110.                     { 0, 1, 2, 3, 4, 5, 6, 7},
  111.                     { 1, 0, 3, 2, 5, 4, 7, 6},
  112.                     { 0, 1, 2, 3, 4, 5, 6, 7},
  113.                     { 3, 0, 1, 2, 7, 4, 5, 6},
  114.                     { 0, 1, 2, 3, 4, 5, 6, 7},
  115.                     { 1, 0, 3, 2, 5, 4, 7, 6},
  116.                     { 0, 1, 2, 3, 4, 5, 6, 7}
  117.             };
  118.             int xC0b = (xC - (xC & ctb_size_mask)) >> s->sps->log2_min_cb_size;
  119.             int yC0b = (yC - (yC & ctb_size_mask)) >> s->sps->log2_min_cb_size;
  120.             int idxX = (xQgBase & ctb_size_mask)   >> s->sps->log2_min_cb_size;
  121.             int idxY = (yQgBase & ctb_size_mask)   >> s->sps->log2_min_cb_size;
  122.             int idx_mask = ctb_size_mask >> s->sps->log2_min_cb_size;
  123.             int x, y;
  124.  
  125.             x = FFMIN(xC0b +  offsetX[idxX][idxY],             min_cb_width  - 1);
  126.             y = FFMIN(yC0b + (offsetY[idxX][idxY] & idx_mask), min_cb_height - 1);
  127.  
  128.             if (xC0b == (lc->start_of_tiles_x >> s->sps->log2_min_cb_size) &&
  129.                 offsetX[idxX][idxY] == -1) {
  130.                 x = (lc->end_of_tiles_x >> s->sps->log2_min_cb_size) - 1;
  131.                 y = yC0b - 1;
  132.             }
  133.             qPy_pred = s->qp_y_tab[y * min_cb_width + x];
  134.         }
  135.     }
  136.  
  137.     // qPy_a
  138.     if (availableA == 0)
  139.         qPy_a = qPy_pred;
  140.     else
  141.         qPy_a = s->qp_y_tab[(x_cb - 1) + y_cb * min_cb_width];
  142.  
  143.     // qPy_b
  144.     if (availableB == 0)
  145.         qPy_b = qPy_pred;
  146.     else
  147.         qPy_b = s->qp_y_tab[x_cb + (y_cb - 1) * min_cb_width];
  148.  
  149.     return (qPy_a + qPy_b + 1) >> 1;
  150. }
  151.  
  152. void ff_hevc_set_qPy(HEVCContext *s, int xC, int yC, int xBase, int yBase, int log2_cb_size)
  153. {
  154.     int qp_y = get_qPy_pred(s, xC, yC, xBase, yBase, log2_cb_size);
  155.  
  156.     if (s->HEVClc->tu.cu_qp_delta != 0) {
  157.         int off = s->sps->qp_bd_offset;
  158.         s->HEVClc->qp_y = ((qp_y + s->HEVClc->tu.cu_qp_delta + 52 + 2 * off) % (52 + off)) - off;
  159.     } else
  160.         s->HEVClc->qp_y = qp_y;
  161. }
  162.  
  163. static int get_qPy(HEVCContext *s, int xC, int yC)
  164. {
  165.     int log2_min_cb_size  = s->sps->log2_min_cb_size;
  166.     int x                 = xC >> log2_min_cb_size;
  167.     int y                 = yC >> log2_min_cb_size;
  168.     return s->qp_y_tab[x + y * s->sps->min_cb_width];
  169. }
  170.  
  171. static void copy_CTB(uint8_t *dst, uint8_t *src, int width, int height, int stride)
  172. {
  173.     int i;
  174.  
  175.     for(i=0; i< height; i++){
  176.         memcpy(dst, src, width);
  177.         dst += stride;
  178.         src += stride;
  179.     }
  180. }
  181.  
  182. #define CTB(tab, x, y) ((tab)[(y) * s->sps->ctb_width + (x)])
  183.  
  184. static void sao_filter_CTB(HEVCContext *s, int x, int y)
  185. {
  186.     //  TODO: This should be easily parallelizable
  187.     //  TODO: skip CBs when (cu_transquant_bypass_flag || (pcm_loop_filter_disable_flag && pcm_flag))
  188.     int c_idx = 0;
  189.     int class = 1, class_index;
  190.     int  edges[4]; // 0 left 1 top 2 right 3 bottom
  191.     SAOParams *sao[4];
  192.     int classes[4];
  193.     int x_shift = 0, y_shift = 0;
  194.     int x_ctb = x>>s->sps->log2_ctb_size;
  195.     int y_ctb = y>>s->sps->log2_ctb_size;
  196.     int ctb_addr_rs = y_ctb * s->sps->ctb_width + x_ctb;
  197.     int ctb_addr_ts = s->pps->ctb_addr_rs_to_ts[ctb_addr_rs];
  198.  
  199.     // flags indicating unfilterable edges
  200.     uint8_t vert_edge[] = {0,0,0,0};
  201.     uint8_t horiz_edge[] = {0,0,0,0};
  202.     uint8_t diag_edge[] = {0,0,0,0};
  203.     uint8_t lfase[3]; // current, above, left
  204.     uint8_t no_tile_filter = s->pps->tiles_enabled_flag && !s->pps->loop_filter_across_tiles_enabled_flag;
  205.     uint8_t left_tile_edge = 0;
  206.     uint8_t up_tile_edge = 0;
  207.  
  208.     sao[0]     = &CTB(s->sao, x_ctb, y_ctb);
  209.     edges[0]   = x_ctb == 0;
  210.     edges[1]   = y_ctb == 0;
  211.     edges[2]   = x_ctb == (s->sps->ctb_width - 1);
  212.     edges[3]   = y_ctb == (s->sps->ctb_height - 1);
  213.     lfase[0]   = CTB(s->filter_slice_edges, x_ctb, y_ctb);
  214.     classes[0] = 0;
  215.  
  216.     if (!edges[0]) {
  217.         left_tile_edge = no_tile_filter && s->pps->tile_id[ctb_addr_ts] != s->pps->tile_id[s->pps->ctb_addr_rs_to_ts[ctb_addr_rs-1]];
  218.         sao[class] = &CTB(s->sao, x_ctb - 1, y_ctb);
  219.         vert_edge[0] = (!lfase[0] && CTB(s->tab_slice_address, x_ctb, y_ctb) != CTB(s->tab_slice_address, x_ctb - 1, y_ctb)) || left_tile_edge;
  220.         vert_edge[2] = vert_edge[0];
  221.         lfase[2]     = CTB(s->filter_slice_edges, x_ctb - 1, y_ctb);
  222.         classes[class] = 2;
  223.         class++;
  224.         x_shift = 8;
  225.     }
  226.  
  227.     if (!edges[1]) {
  228.         up_tile_edge = no_tile_filter && s->pps->tile_id[ctb_addr_ts] != s->pps->tile_id[s->pps->ctb_addr_rs_to_ts[ctb_addr_rs - s->sps->ctb_width]];
  229.         sao[class] = &CTB(s->sao, x_ctb, y_ctb - 1);
  230.         horiz_edge[0] = (!lfase[0] && CTB(s->tab_slice_address, x_ctb, y_ctb) != CTB(s->tab_slice_address, x_ctb, y_ctb - 1)) || up_tile_edge;
  231.         horiz_edge[1] = horiz_edge[0];
  232.         lfase[1] = CTB(s->filter_slice_edges, x_ctb, y_ctb - 1);
  233.         classes[class] = 1;
  234.         class++;
  235.         y_shift = 4;
  236.  
  237.         if (!edges[0]) {
  238.             classes[class] = 3;
  239.             sao[class] = &CTB(s->sao, x_ctb - 1, y_ctb - 1);
  240.             class++;
  241.  
  242.             // Tile check here is done current CTB row/col, not above/left like you'd expect,
  243.             //but that is because the tile boundary always extends through the whole pic
  244.             vert_edge[1] = (!lfase[1] && CTB(s->tab_slice_address, x_ctb, y_ctb - 1) != CTB(s->tab_slice_address, x_ctb - 1, y_ctb - 1)) || left_tile_edge;
  245.             vert_edge[3] = vert_edge[1];
  246.             horiz_edge[2] = (!lfase[2] && CTB(s->tab_slice_address, x_ctb - 1, y_ctb) != CTB(s->tab_slice_address, x_ctb - 1, y_ctb - 1)) || up_tile_edge;
  247.             horiz_edge[3] = horiz_edge[2];
  248.             diag_edge[0] = (!lfase[0] && CTB(s->tab_slice_address, x_ctb, y_ctb) != CTB(s->tab_slice_address, x_ctb - 1, y_ctb - 1)) || left_tile_edge || up_tile_edge;
  249.             diag_edge[3] = diag_edge[0];
  250.  
  251.             // Does left CTB comes after above CTB?
  252.             if(CTB(s->tab_slice_address, x_ctb - 1, y_ctb) > CTB(s->tab_slice_address, x_ctb, y_ctb - 1)) {
  253.                 diag_edge[2] = !lfase[2] || left_tile_edge || up_tile_edge;
  254.                 diag_edge[1] = diag_edge[2];
  255.             } else if(CTB(s->tab_slice_address, x_ctb - 1, y_ctb) < CTB(s->tab_slice_address, x_ctb, y_ctb - 1)) {
  256.                 diag_edge[1] = !lfase[1] || left_tile_edge || up_tile_edge;
  257.                 diag_edge[2] = diag_edge[1];
  258.             } else {
  259.                 // Same slice, only consider tiles
  260.                 diag_edge[2] = left_tile_edge || up_tile_edge;
  261.                 diag_edge[1] = diag_edge[2];
  262.             }
  263.         }
  264.     }
  265.  
  266.     for (c_idx = 0; c_idx < 3; c_idx++) {
  267.         int chroma = c_idx ? 1 : 0;
  268.         int x0 = x >> chroma;
  269.         int y0 = y >> chroma;
  270.         int stride = s->frame->linesize[c_idx];
  271.         int ctb_size = (1 << (s->sps->log2_ctb_size)) >> s->sps->hshift[c_idx];
  272.         int width = FFMIN(ctb_size,
  273.                           (s->sps->width >> s->sps->hshift[c_idx]) - x0);
  274.         int height = FFMIN(ctb_size,
  275.                            (s->sps->height >> s->sps->vshift[c_idx]) - y0);
  276.  
  277.         uint8_t *src = &s->frame->data[c_idx][y0 * stride + (x0 << s->sps->pixel_shift)];
  278.         uint8_t *dst = &s->sao_frame->data[c_idx][y0 * stride + (x0 << s->sps->pixel_shift)];
  279.         int offset = (y_shift >> chroma) * stride + ((x_shift >> chroma) << s->sps->pixel_shift);
  280.  
  281.         copy_CTB(dst - offset, src - offset,
  282.                  (edges[2] ? width  + (x_shift >> chroma) : width)  << s->sps->pixel_shift,
  283.                  (edges[3] ? height + (y_shift >> chroma) : height), stride);
  284.  
  285.         for (class_index = 0; class_index < class; class_index++) {
  286.  
  287.             switch (sao[class_index]->type_idx[c_idx]) {
  288.             case SAO_BAND:
  289.                 s->hevcdsp.sao_band_filter[classes[class_index]](dst, src, stride, sao[class_index], edges, width, height, c_idx);
  290.                     break;
  291.             case SAO_EDGE:
  292.                 s->hevcdsp.sao_edge_filter[classes[class_index]](dst, src, stride, sao[class_index],  edges, width, height, c_idx, vert_edge[classes[class_index]], horiz_edge[classes[class_index]], diag_edge[classes[class_index]]);
  293.                 break;
  294.             }
  295.         }
  296.     }
  297. }
  298.  
  299. static int get_pcm(HEVCContext *s, int x, int y)
  300. {
  301.     int log2_min_pu_size = s->sps->log2_min_pu_size;
  302.     int x_pu             = x >> log2_min_pu_size;
  303.     int y_pu             = y >> log2_min_pu_size;
  304.  
  305.     if (x < 0 || x_pu >= s->sps->min_pu_width ||
  306.         y < 0 || y_pu >= s->sps->min_pu_height)
  307.         return 2;
  308.     return s->is_pcm[y_pu * s->sps->min_pu_width + x_pu];
  309. }
  310.  
  311. #define TC_CALC(qp, bs) tctable[av_clip((qp) + DEFAULT_INTRA_TC_OFFSET * ((bs) - 1) + ((tc_offset >> 1) << 1), 0, MAX_QP + DEFAULT_INTRA_TC_OFFSET)]
  312.  
  313. static void deblocking_filter_CTB(HEVCContext *s, int x0, int y0)
  314. {
  315.     uint8_t *src;
  316.     int x, y;
  317.     int chroma;
  318.     int c_tc[2];
  319.     int beta[2];
  320.     int tc[2];
  321.     uint8_t no_p[2] = {0};
  322.     uint8_t no_q[2] = {0};
  323.  
  324.     int log2_ctb_size =  s->sps->log2_ctb_size;
  325.     int x_end, y_end;
  326.     int ctb_size    = 1<<log2_ctb_size;
  327.     int ctb         = (x0 >> log2_ctb_size) + (y0 >> log2_ctb_size) * s->sps->ctb_width;
  328.     int cur_tc_offset   = s->deblock[ctb].tc_offset;
  329.     int cur_beta_offset = s->deblock[ctb].beta_offset;
  330.     int left_tc_offset, left_beta_offset;
  331.     int tc_offset, beta_offset;
  332.     int pcmf        = (s->sps->pcm_enabled_flag && s->sps->pcm.loop_filter_disable_flag) ||
  333.                       s->pps->transquant_bypass_enable_flag;
  334.  
  335.     if (x0) {
  336.         left_tc_offset   = s->deblock[ctb-1].tc_offset;
  337.         left_beta_offset = s->deblock[ctb-1].beta_offset;
  338.     }
  339.  
  340.     x_end = x0+ctb_size;
  341.     if (x_end > s->sps->width)
  342.         x_end = s->sps->width;
  343.     y_end = y0+ctb_size;
  344.     if (y_end > s->sps->height)
  345.         y_end = s->sps->height;
  346.  
  347.     tc_offset = cur_tc_offset;
  348.     beta_offset = cur_beta_offset;
  349.  
  350.     // vertical filtering luma
  351.     for (y = y0; y < y_end; y += 8) {
  352.         for (x = x0 ? x0 : 8; x < x_end; x += 8) {
  353.             const int bs0 = s->vertical_bs[(x >> 3) + (y       >> 2) * s->bs_width];
  354.             const int bs1 = s->vertical_bs[(x >> 3) + ((y + 4) >> 2) * s->bs_width];
  355.             if (bs0 || bs1) {
  356.                 const int qp0 = (get_qPy(s, x - 1, y) + get_qPy(s, x, y) + 1) >> 1;
  357.                 const int qp1 = (get_qPy(s, x - 1, y + 4) + get_qPy(s, x, y + 4) + 1) >> 1;
  358.  
  359.                 beta[0] = betatable[av_clip(qp0 + ((beta_offset >> 1) << 1), 0, MAX_QP)];
  360.                 beta[1] = betatable[av_clip(qp1 + ((beta_offset >> 1) << 1), 0, MAX_QP)];
  361.                 tc[0] = bs0 ? TC_CALC(qp0, bs0) : 0;
  362.                 tc[1] = bs1 ? TC_CALC(qp1, bs1) : 0;
  363.                 src = &s->frame->data[LUMA][y * s->frame->linesize[LUMA] + (x << s->sps->pixel_shift)];
  364.                 if (pcmf) {
  365.                     no_p[0] = get_pcm(s, x - 1, y);
  366.                     no_p[1] = get_pcm(s, x - 1, y + 4);
  367.                     no_q[0] = get_pcm(s, x, y);
  368.                     no_q[1] = get_pcm(s, x, y + 4);
  369.                     s->hevcdsp.hevc_v_loop_filter_luma_c(src, s->frame->linesize[LUMA], beta, tc, no_p, no_q);
  370.                 } else
  371.                     s->hevcdsp.hevc_v_loop_filter_luma(src, s->frame->linesize[LUMA], beta, tc, no_p, no_q);
  372.             }
  373.         }
  374.     }
  375.  
  376.     // vertical filtering chroma
  377.     for (chroma = 1; chroma <= 2; chroma++) {
  378.         for (y = y0; y < y_end; y += 16) {
  379.             for (x = x0 ? x0:16; x < x_end; x += 16) {
  380.                 const int bs0 = s->vertical_bs[(x >> 3) + (y >> 2) * s->bs_width];
  381.                 const int bs1 = s->vertical_bs[(x >> 3) + ((y + 8) >> 2) * s->bs_width];
  382.                 if ((bs0 == 2) || (bs1 == 2)) {
  383.                     const int qp0 = (get_qPy(s, x - 1, y) + get_qPy(s, x, y) + 1) >> 1;
  384.                     const int qp1 = (get_qPy(s, x - 1, y + 8) + get_qPy(s, x, y + 8) + 1) >> 1;
  385.  
  386.                     c_tc[0] = (bs0 == 2) ? chroma_tc(s, qp0, chroma, tc_offset) : 0;
  387.                     c_tc[1] = (bs1 == 2) ? chroma_tc(s, qp1, chroma, tc_offset) : 0;
  388.                     src = &s->frame->data[chroma][(y / 2) * s->frame->linesize[chroma] + ((x / 2) << s->sps->pixel_shift)];
  389.                     if (pcmf) {
  390.                         no_p[0] = get_pcm(s, x - 1, y);
  391.                         no_p[1] = get_pcm(s, x - 1, y + 8);
  392.                         no_q[0] = get_pcm(s, x, y);
  393.                         no_q[1] = get_pcm(s, x, y + 8);
  394.                         s->hevcdsp.hevc_v_loop_filter_chroma_c(src, s->frame->linesize[chroma], c_tc, no_p, no_q);
  395.                     } else
  396.                         s->hevcdsp.hevc_v_loop_filter_chroma(src, s->frame->linesize[chroma], c_tc, no_p, no_q);
  397.                 }
  398.             }
  399.         }
  400.     }
  401.  
  402.     // horizontal filtering luma
  403.     if (x_end != s->sps->width)
  404.         x_end -= 8;
  405.     for (y = y0 ? y0 : 8; y < y_end; y += 8) {
  406.         for (x = x0 ? x0 - 8 : 0; x < x_end; x += 8) {
  407.             const int bs0 = s->horizontal_bs[(x +     y * s->bs_width) >> 2];
  408.             const int bs1 = s->horizontal_bs[(x + 4 + y * s->bs_width) >> 2];
  409.             if (bs0 || bs1) {
  410.                 const int qp0 = (get_qPy(s, x, y - 1)     + get_qPy(s, x, y)     + 1) >> 1;
  411.                 const int qp1 = (get_qPy(s, x + 4, y - 1) + get_qPy(s, x + 4, y) + 1) >> 1;
  412.  
  413.                 tc_offset = x >= x0 ? cur_tc_offset : left_tc_offset;
  414.                 beta_offset = x >= x0 ? cur_beta_offset : left_beta_offset;
  415.  
  416.                 beta[0]  = betatable[av_clip(qp0 + ((beta_offset >> 1) << 1), 0, MAX_QP)];
  417.                 beta[1]  = betatable[av_clip(qp1 + ((beta_offset >> 1) << 1), 0, MAX_QP)];
  418.                 tc[0] = bs0 ? TC_CALC(qp0, bs0) : 0;
  419.                 tc[1] = bs1 ? TC_CALC(qp1, bs1) : 0;
  420.                 src = &s->frame->data[LUMA][y * s->frame->linesize[LUMA] + (x << s->sps->pixel_shift)];
  421.                 if (pcmf) {
  422.                     no_p[0] = get_pcm(s, x, y - 1);
  423.                     no_p[1] = get_pcm(s, x + 4, y - 1);
  424.                     no_q[0] = get_pcm(s, x, y);
  425.                     no_q[1] = get_pcm(s, x + 4, y);
  426.                     s->hevcdsp.hevc_h_loop_filter_luma_c(src, s->frame->linesize[LUMA], beta, tc, no_p, no_q);
  427.                 } else
  428.                     s->hevcdsp.hevc_h_loop_filter_luma(src, s->frame->linesize[LUMA], beta, tc, no_p, no_q);
  429.             }
  430.         }
  431.     }
  432.  
  433.     // horizontal filtering chroma
  434.     for (chroma = 1; chroma <= 2; chroma++) {
  435.         for (y = y0 ? y0 : 16; y < y_end; y += 16) {
  436.             for (x = x0 - 8; x < x_end; x += 16) {
  437.                 int bs0, bs1;
  438.                 // to make sure no memory access over boundary when x = -8
  439.                 // TODO: simplify with row based deblocking
  440.                 if (x < 0) {
  441.                     bs0 = 0;
  442.                     bs1 = s->horizontal_bs[(x + 8 + y * s->bs_width) >> 2];
  443.                 } else if (x >= x_end - 8) {
  444.                     bs0 = s->horizontal_bs[(x + y * s->bs_width) >> 2];
  445.                     bs1 = 0;
  446.                 } else {
  447.                     bs0 = s->horizontal_bs[(x + y * s->bs_width) >> 2];
  448.                     bs1 = s->horizontal_bs[(x + 8 + y * s->bs_width) >> 2];
  449.                 }
  450.  
  451.                 if ((bs0 == 2) || (bs1 == 2)) {
  452.                     const int qp0 = (bs0 == 2) ? ((get_qPy(s, x, y - 1)     + get_qPy(s, x, y)     + 1) >> 1) : 0;
  453.                     const int qp1 = (bs1 == 2) ? ((get_qPy(s, x + 8, y - 1) + get_qPy(s, x + 8, y) + 1) >> 1) : 0;
  454.  
  455.                     tc_offset = x >= x0 ? cur_tc_offset : left_tc_offset;
  456.                     c_tc[0] = (bs0 == 2) ? chroma_tc(s, qp0, chroma, tc_offset) : 0;
  457.                     c_tc[1] = (bs1 == 2) ? chroma_tc(s, qp1, chroma, cur_tc_offset) : 0;
  458.                     src = &s->frame->data[chroma][(y / 2) * s->frame->linesize[chroma] + ((x / 2) << s->sps->pixel_shift)];
  459.                     if (pcmf) {
  460.                         no_p[0] = get_pcm(s, x, y - 1);
  461.                         no_p[1] = get_pcm(s, x + 8, y - 1);
  462.                         no_q[0] = get_pcm(s, x, y);
  463.                         no_q[1] = get_pcm(s, x + 8, y);
  464.                         s->hevcdsp.hevc_h_loop_filter_chroma_c(src, s->frame->linesize[chroma], c_tc, no_p, no_q);
  465.                     } else
  466.                         s->hevcdsp.hevc_h_loop_filter_chroma(src, s->frame->linesize[chroma], c_tc, no_p, no_q);
  467.                 }
  468.             }
  469.         }
  470.     }
  471. }
  472.  
  473. static int boundary_strength(HEVCContext *s, MvField *curr,
  474.                              uint8_t curr_cbf_luma, MvField *neigh,
  475.                              uint8_t neigh_cbf_luma, RefPicList *neigh_refPicList,
  476.                              int tu_border)
  477. {
  478.     int mvs = curr->pred_flag[0] + curr->pred_flag[1];
  479.  
  480.     if (tu_border) {
  481.         if (curr->is_intra || neigh->is_intra)
  482.             return 2;
  483.         if (curr_cbf_luma || neigh_cbf_luma)
  484.             return 1;
  485.     }
  486.  
  487.     if (mvs == neigh->pred_flag[0] + neigh->pred_flag[1]) {
  488.         if (mvs == 2) {
  489.             // same L0 and L1
  490.             if (s->ref->refPicList[0].list[curr->ref_idx[0]] == neigh_refPicList[0].list[neigh->ref_idx[0]]   &&
  491.                 s->ref->refPicList[0].list[curr->ref_idx[0]] == s->ref->refPicList[1].list[curr->ref_idx[1]] &&
  492.                 neigh_refPicList[0].list[neigh->ref_idx[0]] == neigh_refPicList[1].list[neigh->ref_idx[1]]) {
  493.                 if ((abs(neigh->mv[0].x - curr->mv[0].x) >= 4 || abs(neigh->mv[0].y - curr->mv[0].y) >= 4 ||
  494.                      abs(neigh->mv[1].x - curr->mv[1].x) >= 4 || abs(neigh->mv[1].y - curr->mv[1].y) >= 4) &&
  495.                     (abs(neigh->mv[1].x - curr->mv[0].x) >= 4 || abs(neigh->mv[1].y - curr->mv[0].y) >= 4 ||
  496.                      abs(neigh->mv[0].x - curr->mv[1].x) >= 4 || abs(neigh->mv[0].y - curr->mv[1].y) >= 4))
  497.                     return 1;
  498.                 else
  499.                     return 0;
  500.             } else if (neigh_refPicList[0].list[neigh->ref_idx[0]] == s->ref->refPicList[0].list[curr->ref_idx[0]] &&
  501.                        neigh_refPicList[1].list[neigh->ref_idx[1]] == s->ref->refPicList[1].list[curr->ref_idx[1]]) {
  502.                 if (abs(neigh->mv[0].x - curr->mv[0].x) >= 4 || abs(neigh->mv[0].y - curr->mv[0].y) >= 4 ||
  503.                     abs(neigh->mv[1].x - curr->mv[1].x) >= 4 || abs(neigh->mv[1].y - curr->mv[1].y) >= 4)
  504.                     return 1;
  505.                 else
  506.                     return 0;
  507.             } else if (neigh_refPicList[1].list[neigh->ref_idx[1]] == s->ref->refPicList[0].list[curr->ref_idx[0]] &&
  508.                        neigh_refPicList[0].list[neigh->ref_idx[0]] == s->ref->refPicList[1].list[curr->ref_idx[1]]) {
  509.                 if (abs(neigh->mv[1].x - curr->mv[0].x) >= 4 || abs(neigh->mv[1].y - curr->mv[0].y) >= 4 ||
  510.                     abs(neigh->mv[0].x - curr->mv[1].x) >= 4 || abs(neigh->mv[0].y - curr->mv[1].y) >= 4)
  511.                     return 1;
  512.                 else
  513.                     return 0;
  514.             } else {
  515.                 return 1;
  516.             }
  517.         } else { // 1 MV
  518.             Mv A, B;
  519.             int ref_A;
  520.             int ref_B;
  521.  
  522.             if (curr->pred_flag[0]) {
  523.                 A = curr->mv[0];
  524.                 ref_A = s->ref->refPicList[0].list[curr->ref_idx[0]];
  525.             } else {
  526.                 A = curr->mv[1];
  527.                 ref_A = s->ref->refPicList[1].list[curr->ref_idx[1]];
  528.             }
  529.  
  530.             if (neigh->pred_flag[0]) {
  531.                 B = neigh->mv[0];
  532.                 ref_B = neigh_refPicList[0].list[neigh->ref_idx[0]];
  533.             } else {
  534.                 B = neigh->mv[1];
  535.                 ref_B = neigh_refPicList[1].list[neigh->ref_idx[1]];
  536.             }
  537.  
  538.             if (ref_A == ref_B) {
  539.                 if (abs(A.x - B.x) >= 4 || abs(A.y - B.y) >= 4)
  540.                     return 1;
  541.                 else
  542.                     return 0;
  543.             } else
  544.                 return 1;
  545.         }
  546.     }
  547.  
  548.     return 1;
  549. }
  550.  
  551. void ff_hevc_deblocking_boundary_strengths(HEVCContext *s, int x0, int y0, int log2_trafo_size,
  552.                                            int slice_or_tiles_up_boundary, int slice_or_tiles_left_boundary)
  553. {
  554.     MvField *tab_mvf     = s->ref->tab_mvf;
  555.     int log2_min_pu_size = s->sps->log2_min_pu_size;
  556.     int log2_min_tu_size = s->sps->log2_min_tb_size;
  557.     int min_pu_width     = s->sps->min_pu_width;
  558.     int min_tu_width     = s->sps->min_tb_width;
  559.     int is_intra         = tab_mvf[(y0 >> log2_min_pu_size) * min_pu_width + (x0 >> log2_min_pu_size)].is_intra;
  560.  
  561.     int i, j;
  562.     int bs;
  563.  
  564.     if (y0 > 0 && (y0 & 7) == 0) {
  565.         int yp_pu = (y0 - 1) >> log2_min_pu_size;
  566.         int yq_pu = y0 >> log2_min_pu_size;
  567.         int yp_tu = (y0 - 1) >> log2_min_tu_size;
  568.         int yq_tu = y0 >> log2_min_tu_size;
  569.  
  570.         for (i = 0; i < (1 << log2_trafo_size); i += 4) {
  571.             int x_pu = (x0 + i) >> log2_min_pu_size;
  572.             int x_tu = (x0 + i) >> log2_min_tu_size;
  573.             MvField *top  = &tab_mvf[yp_pu * min_pu_width + x_pu];
  574.             MvField *curr = &tab_mvf[yq_pu * min_pu_width + x_pu];
  575.             uint8_t top_cbf_luma  = s->cbf_luma[yp_tu * min_tu_width + x_tu];
  576.             uint8_t curr_cbf_luma = s->cbf_luma[yq_tu * min_tu_width + x_tu];
  577.             RefPicList* top_refPicList = ff_hevc_get_ref_list(s, s->ref, x0 + i, y0 - 1);
  578.  
  579.             bs = boundary_strength(s, curr, curr_cbf_luma, top, top_cbf_luma, top_refPicList, 1);
  580.             if (!s->sh.slice_loop_filter_across_slices_enabled_flag && (slice_or_tiles_up_boundary & 1) && (y0 % (1 << s->sps->log2_ctb_size)) == 0)
  581.                 bs = 0;
  582.             else if (!s->pps->loop_filter_across_tiles_enabled_flag && (slice_or_tiles_up_boundary & 2)  && (y0 % (1 << s->sps->log2_ctb_size)) == 0)
  583.                 bs = 0;
  584.             if (y0 == 0 || s->sh.disable_deblocking_filter_flag == 1)
  585.                 bs = 0;
  586.             if (bs)
  587.                 s->horizontal_bs[((x0 + i) + y0 * s->bs_width) >> 2] = bs;
  588.         }
  589.     }
  590.  
  591.     // bs for TU internal horizontal PU boundaries
  592.     if (log2_trafo_size > s->sps->log2_min_pu_size && !is_intra)
  593.         for (j = 8; j < (1 << log2_trafo_size); j += 8) {
  594.             int yp_pu = (y0 + j - 1) >> log2_min_pu_size;
  595.             int yq_pu = (y0 + j)     >> log2_min_pu_size;
  596.             int yp_tu = (y0 + j - 1) >> log2_min_tu_size;
  597.             int yq_tu = (y0 + j)     >> log2_min_tu_size;
  598.  
  599.  
  600.             for (i = 0; i < (1<<log2_trafo_size); i += 4) {
  601.                 int x_pu = (x0 + i) >> log2_min_pu_size;
  602.                 int x_tu = (x0 + i) >> log2_min_tu_size;
  603.                 MvField *top  = &tab_mvf[yp_pu * min_pu_width + x_pu];
  604.                 MvField *curr = &tab_mvf[yq_pu * min_pu_width + x_pu];
  605.                 uint8_t top_cbf_luma  = s->cbf_luma[yp_tu * min_tu_width + x_tu];
  606.                 uint8_t curr_cbf_luma = s->cbf_luma[yq_tu * min_tu_width + x_tu];
  607.                 RefPicList* top_refPicList = ff_hevc_get_ref_list(s, s->ref, x0 + i, y0 + j - 1);
  608.  
  609.                 bs = boundary_strength(s, curr, curr_cbf_luma, top, top_cbf_luma, top_refPicList, 0);
  610.                 if (s->sh.disable_deblocking_filter_flag == 1)
  611.                     bs = 0;
  612.                 if (bs)
  613.                     s->horizontal_bs[((x0 + i) + (y0 + j) * s->bs_width) >> 2] = bs;
  614.             }
  615.         }
  616.  
  617.     // bs for vertical TU boundaries
  618.     if (x0 > 0 && (x0 & 7) == 0) {
  619.         int xp_pu = (x0 - 1) >> log2_min_pu_size;
  620.         int xq_pu =  x0      >> log2_min_pu_size;
  621.         int xp_tu = (x0 - 1) >> log2_min_tu_size;
  622.         int xq_tu =  x0      >> log2_min_tu_size;
  623.  
  624.         for (i = 0; i < (1 << log2_trafo_size); i += 4) {
  625.             int y_pu = (y0 + i) >> log2_min_pu_size;
  626.             int y_tu = (y0 + i) >> log2_min_tu_size;
  627.             MvField *left = &tab_mvf[y_pu * min_pu_width + xp_pu];
  628.             MvField *curr = &tab_mvf[y_pu * min_pu_width + xq_pu];
  629.  
  630.             uint8_t left_cbf_luma = s->cbf_luma[y_tu * min_tu_width + xp_tu];
  631.             uint8_t curr_cbf_luma = s->cbf_luma[y_tu * min_tu_width + xq_tu];
  632.             RefPicList* left_refPicList = ff_hevc_get_ref_list(s, s->ref, x0 - 1, y0 + i);
  633.  
  634.             bs = boundary_strength(s, curr, curr_cbf_luma, left, left_cbf_luma, left_refPicList, 1);
  635.             if (!s->sh.slice_loop_filter_across_slices_enabled_flag && (slice_or_tiles_left_boundary & 1) && (x0 % (1 << s->sps->log2_ctb_size)) == 0)
  636.                 bs = 0;
  637.             else if (!s->pps->loop_filter_across_tiles_enabled_flag && (slice_or_tiles_left_boundary & 2) && (x0 % (1 << s->sps->log2_ctb_size)) == 0)
  638.                 bs = 0;
  639.             if (x0 == 0 || s->sh.disable_deblocking_filter_flag == 1)
  640.                 bs = 0;
  641.             if (bs)
  642.                 s->vertical_bs[(x0 >> 3) + ((y0 + i) >> 2) * s->bs_width] = bs;
  643.         }
  644.     }
  645.  
  646.     // bs for TU internal vertical PU boundaries
  647.     if (log2_trafo_size > log2_min_pu_size && !is_intra)
  648.         for (j = 0; j < (1 << log2_trafo_size); j += 4) {
  649.             int y_pu = (y0 + j) >> log2_min_pu_size;
  650.             int y_tu = (y0 + j) >> log2_min_tu_size;
  651.  
  652.             for (i = 8; i < (1 << log2_trafo_size); i += 8) {
  653.                 int xp_pu = (x0 + i - 1) >> log2_min_pu_size;
  654.                 int xq_pu = (x0 + i) >> log2_min_pu_size;
  655.                 int xp_tu = (x0 + i - 1) >> log2_min_tu_size;
  656.                 int xq_tu = (x0 + i) >> log2_min_tu_size;
  657.                 MvField *left = &tab_mvf[y_pu * min_pu_width + xp_pu];
  658.                 MvField *curr = &tab_mvf[y_pu * min_pu_width + xq_pu];
  659.                 uint8_t left_cbf_luma = s->cbf_luma[y_tu * min_tu_width + xp_tu];
  660.                 uint8_t curr_cbf_luma = s->cbf_luma[y_tu * min_tu_width + xq_tu];
  661.                 RefPicList* left_refPicList = ff_hevc_get_ref_list(s, s->ref, x0 + i - 1, y0 + j);
  662.  
  663.                 bs = boundary_strength(s, curr, curr_cbf_luma, left, left_cbf_luma, left_refPicList, 0);
  664.                 if (s->sh.disable_deblocking_filter_flag == 1)
  665.                     bs = 0;
  666.                 if (bs)
  667.                     s->vertical_bs[((x0 + i) >> 3) + ((y0 + j) >> 2) * s->bs_width] = bs;
  668.             }
  669.         }
  670. }
  671. #undef LUMA
  672. #undef CB
  673. #undef CR
  674.  
  675. void ff_hevc_hls_filter(HEVCContext *s, int x, int y)
  676. {
  677.     deblocking_filter_CTB(s, x, y);
  678.     if (s->sps->sao_enabled)
  679.         sao_filter_CTB(s, x, y);
  680. }
  681.  
  682. void ff_hevc_hls_filters(HEVCContext *s, int x_ctb, int y_ctb, int ctb_size)
  683. {
  684.     if (y_ctb && x_ctb)
  685.         ff_hevc_hls_filter(s, x_ctb - ctb_size, y_ctb - ctb_size);
  686.     if (y_ctb && x_ctb >= s->sps->width - ctb_size) {
  687.         ff_hevc_hls_filter(s, x_ctb, y_ctb - ctb_size);
  688.         if (s->threads_type == FF_THREAD_FRAME )
  689.             ff_thread_report_progress(&s->ref->tf, y_ctb - ctb_size, 0);
  690.     }
  691.     if (x_ctb && y_ctb >= s->sps->height - ctb_size)
  692.         ff_hevc_hls_filter(s, x_ctb - ctb_size, y_ctb);
  693. }
  694.