API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Sep 23rd, 2017
82
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Diff 15.45 KB | None | 0 0
raw download clone embed print report
  1. diff --git a/encoder/analyse.c b/encoder/analyse.c
  2. index 0dd7235..e7ed1cc 100644
  3. --- a/encoder/analyse.c
  4. +++ b/encoder/analyse.c
  5. @@ -132,6 +132,7 @@ typedef struct
  6.      int i_mb_type8x16;
  7.  
  8.      int b_direct_available;
  9. +    int b_early_terminate;
  10.  
  11.  } x264_mb_analysis_t;
  12.  
  13. @@ -587,6 +588,8 @@ static void x264_mb_analyse_init( x264_t *h, x264_mb_analysis_t *a, int qp )
  14.          else
  15.              a->b_force_intra = 0;
  16.      }
  17. +    
  18. +    a->b_early_terminate = h->param.analyse.i_subpel_refine < 11;
  19.  }
  20.  
  21.  /* Prediction modes allowed for various combinations of neighbors. */
  22. @@ -782,7 +785,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  23.  
  24.      /* Not heavily tuned */
  25.      static const uint8_t i16x16_thresh_lut[11] = { 2, 2, 2, 3, 3, 4, 4, 4, 4, 4, 4 };
  26. -    int i16x16_thresh = a->b_fast_intra ? (i16x16_thresh_lut[h->mb.i_subpel_refine]*i_satd_inter)>>1 : COST_MAX;
  27. +    int i16x16_thresh = (a->b_early_terminate && a->b_fast_intra) ? (i16x16_thresh_lut[h->mb.i_subpel_refine]*i_satd_inter)>>1 : COST_MAX;
  28.  
  29.      if( !h->mb.b_lossless && predict_mode[3] >= 0 )
  30.      {
  31. @@ -834,7 +837,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  32.      {
  33.          ALIGNED_ARRAY_16( pixel, edge,[33] );
  34.          x264_pixel_cmp_t sa8d = (h->pixf.mbcmp[0] == h->pixf.satd[0]) ? h->pixf.sa8d[PIXEL_8x8] : h->pixf.mbcmp[PIXEL_8x8];
  35. -        int i_satd_thresh = a->i_mbrd ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
  36. +        int i_satd_thresh = (!a->b_early_terminate || a->i_mbrd) ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
  37.  
  38.          // FIXME some bias like in i4x4?
  39.          int i_cost = lambda * 4; /* base predmode costs */
  40. @@ -927,7 +930,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  41.          }
  42.          /* Not heavily tuned */
  43.          static const uint8_t i8x8_thresh[11] = { 4, 4, 4, 5, 5, 5, 6, 6, 6, 6, 6 };
  44. -        if( X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  45. +        if( a->b_early_terminate && X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  46.              return;
  47.      }
  48.  
  49. @@ -935,10 +938,10 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  50.      if( flags & X264_ANALYSE_I4x4 )
  51.      {
  52.          int i_cost = lambda * (24+16); /* 24from JVT (SATD0), 16 from base predmode costs */
  53. -        int i_satd_thresh = X264_MIN3( i_satd_inter, a->i_satd_i16x16, a->i_satd_i8x8 );
  54. +        int i_satd_thresh = a->b_early_terminate ? X264_MIN3( i_satd_inter, a->i_satd_i16x16, a->i_satd_i8x8 ) : COST_MAX;
  55.          h->mb.i_cbp_luma = 0;
  56.  
  57. -        if( a->i_mbrd )
  58. +        if( a->b_early_terminate && a->i_mbrd )
  59.              i_satd_thresh = i_satd_thresh * (10-a->b_fast_intra)/8;
  60.  
  61.          if( h->sh.i_type == SLICE_TYPE_B )
  62. @@ -1033,7 +1036,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  63.  
  64.  static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  65.  {
  66. -    if( a->i_satd_i16x16 <= i_satd_thresh )
  67. +    if( a->i_satd_i16x16 < i_satd_thresh )
  68.      {
  69.          h->mb.i_type = I_16x16;
  70.          x264_analyse_update_cache( h, a );
  71. @@ -1042,7 +1045,7 @@ static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  72.      else
  73.          a->i_satd_i16x16 = COST_MAX;
  74.  
  75. -    if( a->i_satd_i4x4 <= i_satd_thresh && a->i_satd_i4x4 < COST_MAX )
  76. +    if( a->i_satd_i4x4 < i_satd_thresh && a->i_satd_i4x4 < COST_MAX )
  77.      {
  78.          h->mb.i_type = I_4x4;
  79.          x264_analyse_update_cache( h, a );
  80. @@ -1051,7 +1054,7 @@ static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  81.      else
  82.          a->i_satd_i4x4 = COST_MAX;
  83.  
  84. -    if( a->i_satd_i8x8 <= i_satd_thresh && a->i_satd_i8x8 < COST_MAX )
  85. +    if( a->i_satd_i8x8 < i_satd_thresh && a->i_satd_i8x8 < COST_MAX )
  86.      {
  87.          h->mb.i_type = I_8x8;
  88.          x264_analyse_update_cache( h, a );
  89. @@ -1572,7 +1575,7 @@ static void x264_mb_analyse_inter_p16x8( x264_t *h, x264_mb_analysis_t *a, int i
  90.  
  91.          /* Early termination based on the current SATD score of partition[0]
  92.             plus the estimated SATD score of partition[1] */
  93. -        if( !i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  94. +        if( a->b_early_terminate && (!i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4) )
  95.          {
  96.              a->l0.i_cost16x8 = COST_MAX;
  97.              return;
  98. @@ -1637,7 +1640,7 @@ static void x264_mb_analyse_inter_p8x16( x264_t *h, x264_mb_analysis_t *a, int i
  99.  
  100.          /* Early termination based on the current SATD score of partition[0]
  101.             plus the estimated SATD score of partition[1] */
  102. -        if( !i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  103. +        if( a->b_early_terminate && (!i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4) )
  104.          {
  105.              a->l0.i_cost8x16 = COST_MAX;
  106.              return;
  107. @@ -2454,8 +2457,8 @@ static void x264_mb_analyse_inter_b16x8( x264_t *h, x264_mb_analysis_t *a, int i
  108.  
  109.          /* Early termination based on the current SATD score of partition[0]
  110.             plus the estimated SATD score of partition[1] */
  111. -        if( !i && i_part_cost + a->i_cost_est16x8[1] > i_best_satd
  112. -            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16 )
  113. +        if( a->b_early_terminate && (!i && i_part_cost + a->i_cost_est16x8[1] > i_best_satd
  114. +            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16) )
  115.          {
  116.              a->i_cost16x8bi = COST_MAX;
  117.              return;
  118. @@ -2547,8 +2550,8 @@ static void x264_mb_analyse_inter_b8x16( x264_t *h, x264_mb_analysis_t *a, int i
  119.  
  120.          /* Early termination based on the current SATD score of partition[0]
  121.             plus the estimated SATD score of partition[1] */
  122. -        if( !i && i_part_cost + a->i_cost_est8x16[1] > i_best_satd
  123. -            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16 )
  124. +        if( a->b_early_terminate && (!i && i_part_cost + a->i_cost_est8x16[1] > i_best_satd
  125. +            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16) )
  126.          {
  127.              a->i_cost8x16bi = COST_MAX;
  128.              return;
  129. @@ -2566,17 +2569,17 @@ static void x264_mb_analyse_inter_b8x16( x264_t *h, x264_mb_analysis_t *a, int i
  130.  
  131.  static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  132.  {
  133. -    int thresh = i_satd * 5/4;
  134. +    int thresh = a->b_early_terminate ? COST_MAX : i_satd * 5/4;
  135.  
  136.      h->mb.i_type = P_L0;
  137. -    if( a->l0.i_rd16x16 == COST_MAX && a->l0.me16x16.cost <= i_satd * 3/2 )
  138. +    if( !a->b_early_terminate || (a->l0.i_rd16x16 == COST_MAX && a->l0.me16x16.cost <= i_satd * 3/2) )
  139.      {
  140.          h->mb.i_partition = D_16x16;
  141.          x264_analyse_update_cache( h, a );
  142.          a->l0.i_rd16x16 = x264_rd_cost_mb( h, a->i_lambda2 );
  143.      }
  144.  
  145. -    if( a->l0.i_cost16x8 <= thresh )
  146. +    if( a->l0.i_cost16x8 < thresh )
  147.      {
  148.          h->mb.i_partition = D_16x8;
  149.          x264_analyse_update_cache( h, a );
  150. @@ -2585,7 +2588,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  151.      else
  152.          a->l0.i_cost16x8 = COST_MAX;
  153.  
  154. -    if( a->l0.i_cost8x16 <= thresh )
  155. +    if( a->l0.i_cost8x16 < thresh )
  156.      {
  157.          h->mb.i_partition = D_8x16;
  158.          x264_analyse_update_cache( h, a );
  159. @@ -2594,7 +2597,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  160.      else
  161.          a->l0.i_cost8x16 = COST_MAX;
  162.  
  163. -    if( a->l0.i_cost8x8 <= thresh )
  164. +    if( a->l0.i_cost8x8 < thresh )
  165.      {
  166.          h->mb.i_type = P_8x8;
  167.          h->mb.i_partition = D_8x8;
  168. @@ -2639,7 +2642,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  169.  
  170.  static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_inter )
  171.  {
  172. -    int thresh = i_satd_inter * (17 + (!!h->mb.i_psy_rd))/16;
  173. +    int thresh = !a->b_early_terminate ? COST_MAX : i_satd_inter * (17 + (!!h->mb.i_psy_rd))/16;
  174.  
  175.      if( a->b_direct_available && a->i_rd16x16direct == COST_MAX )
  176.      {
  177. @@ -2655,7 +2658,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  178.      //FIXME not all the update_cache calls are needed
  179.      h->mb.i_partition = D_16x16;
  180.      /* L0 */
  181. -    if( a->l0.me16x16.cost <= thresh && a->l0.i_rd16x16 == COST_MAX )
  182. +    if( a->l0.me16x16.cost < thresh && a->l0.i_rd16x16 == COST_MAX )
  183.      {
  184.          h->mb.i_type = B_L0_L0;
  185.          x264_analyse_update_cache( h, a );
  186. @@ -2663,7 +2666,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  187.      }
  188.  
  189.      /* L1 */
  190. -    if( a->l1.me16x16.cost <= thresh && a->l1.i_rd16x16 == COST_MAX )
  191. +    if( a->l1.me16x16.cost < thresh && a->l1.i_rd16x16 == COST_MAX )
  192.      {
  193.          h->mb.i_type = B_L1_L1;
  194.          x264_analyse_update_cache( h, a );
  195. @@ -2671,7 +2674,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  196.      }
  197.  
  198.      /* BI */
  199. -    if( a->i_cost16x16bi <= thresh && a->i_rd16x16bi == COST_MAX )
  200. +    if( a->i_cost16x16bi < thresh && a->i_rd16x16bi == COST_MAX )
  201.      {
  202.          h->mb.i_type = B_BI_BI;
  203.          x264_analyse_update_cache( h, a );
  204. @@ -2679,7 +2682,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  205.      }
  206.  
  207.      /* 8x8 */
  208. -    if( a->i_cost8x8bi <= thresh && a->i_rd8x8bi == COST_MAX )
  209. +    if( a->i_cost8x8bi < thresh && a->i_rd8x8bi == COST_MAX )
  210.      {
  211.          h->mb.i_type = B_8x8;
  212.          h->mb.i_partition = D_8x8;
  213. @@ -2689,7 +2692,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  214.      }
  215.  
  216.      /* 16x8 */
  217. -    if( a->i_cost16x8bi <= thresh && a->i_rd16x8bi == COST_MAX )
  218. +    if( a->i_cost16x8bi < thresh && a->i_rd16x8bi == COST_MAX )
  219.      {
  220.          h->mb.i_type = a->i_mb_type16x8;
  221.          h->mb.i_partition = D_16x8;
  222. @@ -2698,7 +2701,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  223.      }
  224.  
  225.      /* 8x16 */
  226. -    if( a->i_cost8x16bi <= thresh && a->i_rd8x16bi == COST_MAX )
  227. +    if( a->i_cost8x16bi < thresh && a->i_rd8x16bi == COST_MAX )
  228.      {
  229.          h->mb.i_type = a->i_mb_type8x16;
  230.          h->mb.i_partition = D_8x16;
  231. @@ -3020,8 +3023,8 @@ intra_analysis:
  232.              i_partition = D_16x16;
  233.              i_cost = analysis.l0.me16x16.cost;
  234.  
  235. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  236. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost )
  237. +            if( ( flags & X264_ANALYSE_PSUB16x16 ) && (!analysis.b_early_terminate ||
  238. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost) )
  239.              {
  240.                  i_type = P_8x8;
  241.                  i_partition = D_8x8;
  242. @@ -3056,8 +3059,8 @@ intra_analysis:
  243.  
  244.              /* Now do 16x8/8x16 */
  245.              i_thresh16x8 = analysis.l0.me8x8[1].cost_mv + analysis.l0.me8x8[2].cost_mv;
  246. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  247. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8 )
  248. +            if( ( flags & X264_ANALYSE_PSUB16x16 ) && (!analysis.b_early_terminate ||
  249. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8) )
  250.              {
  251.                  int i_avg_mv_ref_cost = (analysis.l0.me8x8[2].cost_mv + analysis.l0.me8x8[2].i_ref_cost
  252.                                        + analysis.l0.me8x8[3].cost_mv + analysis.l0.me8x8[3].i_ref_cost + 1) >> 1;
  253. @@ -3177,7 +3180,7 @@ intra_analysis:
  254.                  h->mb.i_partition = i_partition;
  255.                  if( i_cost < COST_MAX )
  256.                      x264_mb_analyse_transform_rd( h, &analysis, &i_satd_inter, &i_cost );
  257. -                x264_intra_rd( h, &analysis, i_satd_inter * 5/4 );
  258. +                x264_intra_rd( h, &analysis, analysis.b_early_terminate ? i_satd_inter * 5/4 : COST_MAX );
  259.              }
  260.  
  261.              COPY2_IF_LT( i_cost, analysis.i_satd_i16x16, i_type, I_16x16 );
  262. @@ -3431,17 +3434,17 @@ intra_analysis:
  263.  
  264.                  /* We can gain a little speed by checking the mode with the lowest estimated cost first */
  265.                  int try_16x8_first = i_cost_est16x8bi_total < i_cost_est8x16bi_total;
  266. -                if( try_16x8_first && i_cost_est16x8bi_total < i_cost )
  267. +                if( !analysis.b_early_terminate || (try_16x8_first && i_cost_est16x8bi_total < i_cost) )
  268.                  {
  269.                      x264_mb_analyse_inter_b16x8( h, &analysis, i_cost );
  270.                      COPY3_IF_LT( i_cost, analysis.i_cost16x8bi, i_type, analysis.i_mb_type16x8, i_partition, D_16x8 );
  271.                  }
  272. -                if( i_cost_est8x16bi_total < i_cost )
  273. +                if( !analysis.b_early_terminate || (i_cost_est8x16bi_total < i_cost) )
  274.                  {
  275.                      x264_mb_analyse_inter_b8x16( h, &analysis, i_cost );
  276.                      COPY3_IF_LT( i_cost, analysis.i_cost8x16bi, i_type, analysis.i_mb_type8x16, i_partition, D_8x16 );
  277.                  }
  278. -                if( !try_16x8_first && i_cost_est16x8bi_total < i_cost )
  279. +                if( analysis.b_early_terminate && (!try_16x8_first && i_cost_est16x8bi_total < i_cost) )
  280.                  {
  281.                      x264_mb_analyse_inter_b16x8( h, &analysis, i_cost );
  282.                      COPY3_IF_LT( i_cost, analysis.i_cost16x8bi, i_type, analysis.i_mb_type16x8, i_partition, D_16x8 );
  283. @@ -3573,7 +3576,7 @@ intra_analysis:
  284.              if( analysis.i_mbrd )
  285.              {
  286.                  x264_mb_analyse_transform_rd( h, &analysis, &i_satd_inter, &i_cost );
  287. -                x264_intra_rd( h, &analysis, i_satd_inter * 17/16 );
  288. +                x264_intra_rd( h, &analysis, analysis.b_early_terminate ? i_satd_inter * 17/16 : COST_MAX );
  289.              }
  290.  
  291.              COPY2_IF_LT( i_cost, analysis.i_satd_i16x16, i_type, I_16x16 );
  292. diff --git a/encoder/encoder.c b/encoder/encoder.c
  293. index 85ad046..9c63859 100644
  294. --- a/encoder/encoder.c
  295. +++ b/encoder/encoder.c
  296. @@ -503,7 +503,7 @@ static int x264_validate_parameters( x264_t *h, int b_open )
  297.      h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
  298.      h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
  299.      h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, QP_MAX );
  300. -    h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 10 );
  301. +    h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
  302.      h->param.rc.f_ip_factor = X264_MAX( h->param.rc.f_ip_factor, 0.01f );
  303.      h->param.rc.f_pb_factor = X264_MAX( h->param.rc.f_pb_factor, 0.01f );
  304.      if( h->param.rc.i_rc_method == X264_RC_CRF )
  305. @@ -784,7 +784,7 @@ static int x264_validate_parameters( x264_t *h, int b_open )
  306.          h->param.rc.f_aq_strength = 0;
  307.      }
  308.      h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
  309. -    if( h->param.analyse.i_subpel_refine == 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
  310. +    if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
  311.          h->param.analyse.i_subpel_refine = 9;
  312.  
  313.      {
  314. diff --git a/encoder/me.c b/encoder/me.c
  315. index ca6c602..10b7ef2 100644
  316. --- a/encoder/me.c
  317. +++ b/encoder/me.c
  318. @@ -46,6 +46,7 @@ static const uint8_t subpel_iterations[][4] =
  319.      {0,0,2,2},
  320.      {0,0,4,10},
  321.      {0,0,4,10},
  322. +    {0,0,4,10},
  323.      {0,0,4,10}};
  324.  
  325.  /* (x-1)%6 */
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!