API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Sep 23rd, 2017
63
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Diff 16.01 KB | None | 0 0
raw download clone embed print report
  1. From b667b5d57e090e6554094eeaa4fc108dec3ad8dd Mon Sep 17 00:00:00 2001
  2. From: Dan Larkin <danielhlarkin@gmail.com>
  3. Date: Wed, 13 Jul 2011 11:33:48 -0500
  4. Subject: [PATCH 1/2] Preparation for --subme 11, threshold handling will change output
  5.  
  6. ---
  7. encoder/analyse.c |   24 ++++++++++++------------
  8.  1 files changed, 12 insertions(+), 12 deletions(-)
  9.  
  10. diff --git a/encoder/analyse.c b/encoder/analyse.c
  11. index 0dd7235..f27169c 100644
  12. --- a/encoder/analyse.c
  13. +++ b/encoder/analyse.c
  14. @@ -1033,7 +1033,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  15.  
  16.  static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  17.  {
  18. -    if( a->i_satd_i16x16 <= i_satd_thresh )
  19. +    if( a->i_satd_i16x16 < i_satd_thresh )
  20.      {
  21.          h->mb.i_type = I_16x16;
  22.          x264_analyse_update_cache( h, a );
  23. @@ -1042,7 +1042,7 @@ static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  24.      else
  25.          a->i_satd_i16x16 = COST_MAX;
  26.  
  27. -    if( a->i_satd_i4x4 <= i_satd_thresh && a->i_satd_i4x4 < COST_MAX )
  28. +    if( a->i_satd_i4x4 < i_satd_thresh )
  29.      {
  30.          h->mb.i_type = I_4x4;
  31.          x264_analyse_update_cache( h, a );
  32. @@ -1051,7 +1051,7 @@ static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  33.      else
  34.          a->i_satd_i4x4 = COST_MAX;
  35.  
  36. -    if( a->i_satd_i8x8 <= i_satd_thresh && a->i_satd_i8x8 < COST_MAX )
  37. +    if( a->i_satd_i8x8 < i_satd_thresh && a->i_satd_i8x8 < COST_MAX )
  38.      {
  39.          h->mb.i_type = I_8x8;
  40.          x264_analyse_update_cache( h, a );
  41. @@ -2576,7 +2576,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  42.          a->l0.i_rd16x16 = x264_rd_cost_mb( h, a->i_lambda2 );
  43.      }
  44.  
  45. -    if( a->l0.i_cost16x8 <= thresh )
  46. +    if( a->l0.i_cost16x8 < thresh )
  47.      {
  48.          h->mb.i_partition = D_16x8;
  49.          x264_analyse_update_cache( h, a );
  50. @@ -2585,7 +2585,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  51.      else
  52.          a->l0.i_cost16x8 = COST_MAX;
  53.  
  54. -    if( a->l0.i_cost8x16 <= thresh )
  55. +    if( a->l0.i_cost8x16 < thresh )
  56.      {
  57.          h->mb.i_partition = D_8x16;
  58.          x264_analyse_update_cache( h, a );
  59. @@ -2594,7 +2594,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  60.      else
  61.          a->l0.i_cost8x16 = COST_MAX;
  62.  
  63. -    if( a->l0.i_cost8x8 <= thresh )
  64. +    if( a->l0.i_cost8x8 < thresh )
  65.      {
  66.          h->mb.i_type = P_8x8;
  67.          h->mb.i_partition = D_8x8;
  68. @@ -2655,7 +2655,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  69.      //FIXME not all the update_cache calls are needed
  70.      h->mb.i_partition = D_16x16;
  71.      /* L0 */
  72. -    if( a->l0.me16x16.cost <= thresh && a->l0.i_rd16x16 == COST_MAX )
  73. +    if( a->l0.me16x16.cost < thresh && a->l0.i_rd16x16 == COST_MAX )
  74.      {
  75.          h->mb.i_type = B_L0_L0;
  76.          x264_analyse_update_cache( h, a );
  77. @@ -2663,7 +2663,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  78.      }
  79.  
  80.      /* L1 */
  81. -    if( a->l1.me16x16.cost <= thresh && a->l1.i_rd16x16 == COST_MAX )
  82. +    if( a->l1.me16x16.cost < thresh && a->l1.i_rd16x16 == COST_MAX )
  83.      {
  84.          h->mb.i_type = B_L1_L1;
  85.          x264_analyse_update_cache( h, a );
  86. @@ -2671,7 +2671,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  87.      }
  88.  
  89.      /* BI */
  90. -    if( a->i_cost16x16bi <= thresh && a->i_rd16x16bi == COST_MAX )
  91. +    if( a->i_cost16x16bi < thresh && a->i_rd16x16bi == COST_MAX )
  92.      {
  93.          h->mb.i_type = B_BI_BI;
  94.          x264_analyse_update_cache( h, a );
  95. @@ -2679,7 +2679,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  96.      }
  97.  
  98.      /* 8x8 */
  99. -    if( a->i_cost8x8bi <= thresh && a->i_rd8x8bi == COST_MAX )
  100. +    if( a->i_cost8x8bi < thresh && a->i_rd8x8bi == COST_MAX )
  101.      {
  102.          h->mb.i_type = B_8x8;
  103.          h->mb.i_partition = D_8x8;
  104. @@ -2689,7 +2689,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  105.      }
  106.  
  107.      /* 16x8 */
  108. -    if( a->i_cost16x8bi <= thresh && a->i_rd16x8bi == COST_MAX )
  109. +    if( a->i_cost16x8bi < thresh && a->i_rd16x8bi == COST_MAX )
  110.      {
  111.          h->mb.i_type = a->i_mb_type16x8;
  112.          h->mb.i_partition = D_16x8;
  113. @@ -2698,7 +2698,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  114.      }
  115.  
  116.      /* 8x16 */
  117. -    if( a->i_cost8x16bi <= thresh && a->i_rd8x16bi == COST_MAX )
  118. +    if( a->i_cost8x16bi < thresh && a->i_rd8x16bi == COST_MAX )
  119.      {
  120.          h->mb.i_type = a->i_mb_type8x16;
  121.          h->mb.i_partition = D_8x16;
  122. --
  123. 1.7.4.1
  124.  
  125.  
  126. From 0359b664e37fad03d560456870e8eda6a84a2622 Mon Sep 17 00:00:00 2001
  127. From: Dan Larkin <danielhlarkin@gmail.com>
  128. Date: Wed, 13 Jul 2011 12:45:23 -0500
  129. Subject: [PATCH 2/2] Added --subme 11, which disables all early terminations in analysis
  130.  
  131. ---
  132. encoder/analyse.c |   47 +++++++++++++++++++++++++----------------------
  133.  encoder/encoder.c |    4 ++--
  134.  encoder/me.c      |    1 +
  135.  3 files changed, 28 insertions(+), 24 deletions(-)
  136.  
  137. diff --git a/encoder/analyse.c b/encoder/analyse.c
  138. index f27169c..3bf3153 100644
  139. --- a/encoder/analyse.c
  140. +++ b/encoder/analyse.c
  141. @@ -132,6 +132,7 @@ typedef struct
  142.      int i_mb_type8x16;
  143.  
  144.      int b_direct_available;
  145. +    int b_early_terminate;
  146.  
  147.  } x264_mb_analysis_t;
  148.  
  149. @@ -587,6 +588,8 @@ static void x264_mb_analyse_init( x264_t *h, x264_mb_analysis_t *a, int qp )
  150.          else
  151.              a->b_force_intra = 0;
  152.      }
  153. +    
  154. +    a->b_early_terminate = h->param.analyse.i_subpel_refine < 11;
  155.  }
  156.  
  157.  /* Prediction modes allowed for various combinations of neighbors. */
  158. @@ -782,7 +785,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  159.  
  160.      /* Not heavily tuned */
  161.      static const uint8_t i16x16_thresh_lut[11] = { 2, 2, 2, 3, 3, 4, 4, 4, 4, 4, 4 };
  162. -    int i16x16_thresh = a->b_fast_intra ? (i16x16_thresh_lut[h->mb.i_subpel_refine]*i_satd_inter)>>1 : COST_MAX;
  163. +    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;
  164.  
  165.      if( !h->mb.b_lossless && predict_mode[3] >= 0 )
  166.      {
  167. @@ -834,7 +837,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  168.      {
  169.          ALIGNED_ARRAY_16( pixel, edge,[33] );
  170.          x264_pixel_cmp_t sa8d = (h->pixf.mbcmp[0] == h->pixf.satd[0]) ? h->pixf.sa8d[PIXEL_8x8] : h->pixf.mbcmp[PIXEL_8x8];
  171. -        int i_satd_thresh = a->i_mbrd ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
  172. +        int i_satd_thresh = (!a->b_early_terminate || a->i_mbrd) ? COST_MAX : X264_MIN( i_satd_inter, a->i_satd_i16x16 );
  173.  
  174.          // FIXME some bias like in i4x4?
  175.          int i_cost = lambda * 4; /* base predmode costs */
  176. @@ -927,7 +930,7 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  177.          }
  178.          /* Not heavily tuned */
  179.          static const uint8_t i8x8_thresh[11] = { 4, 4, 4, 5, 5, 5, 6, 6, 6, 6, 6 };
  180. -        if( X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  181. +        if( a->b_early_terminate && X264_MIN(i_cost, a->i_satd_i16x16) > (i_satd_inter*i8x8_thresh[h->mb.i_subpel_refine])>>2 )
  182.              return;
  183.      }
  184.  
  185. @@ -935,10 +938,10 @@ static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *a, int i_satd_
  186.      if( flags & X264_ANALYSE_I4x4 )
  187.      {
  188.          int i_cost = lambda * (24+16); /* 24from JVT (SATD0), 16 from base predmode costs */
  189. -        int i_satd_thresh = X264_MIN3( i_satd_inter, a->i_satd_i16x16, a->i_satd_i8x8 );
  190. +        int i_satd_thresh = a->b_early_terminate ? X264_MIN3( i_satd_inter, a->i_satd_i16x16, a->i_satd_i8x8 ) : COST_MAX;
  191.          h->mb.i_cbp_luma = 0;
  192.  
  193. -        if( a->i_mbrd )
  194. +        if( a->b_early_terminate && a->i_mbrd )
  195.              i_satd_thresh = i_satd_thresh * (10-a->b_fast_intra)/8;
  196.  
  197.          if( h->sh.i_type == SLICE_TYPE_B )
  198. @@ -1572,7 +1575,7 @@ static void x264_mb_analyse_inter_p16x8( x264_t *h, x264_mb_analysis_t *a, int i
  199.  
  200.          /* Early termination based on the current SATD score of partition[0]
  201.             plus the estimated SATD score of partition[1] */
  202. -        if( !i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  203. +        if( a->b_early_terminate && (!i && l0m->cost + a->i_cost_est16x8[1] > i_best_satd * (4 + !!a->i_mbrd) / 4) )
  204.          {
  205.              a->l0.i_cost16x8 = COST_MAX;
  206.              return;
  207. @@ -1637,7 +1640,7 @@ static void x264_mb_analyse_inter_p8x16( x264_t *h, x264_mb_analysis_t *a, int i
  208.  
  209.          /* Early termination based on the current SATD score of partition[0]
  210.             plus the estimated SATD score of partition[1] */
  211. -        if( !i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4 )
  212. +        if( a->b_early_terminate && (!i && l0m->cost + a->i_cost_est8x16[1] > i_best_satd * (4 + !!a->i_mbrd) / 4) )
  213.          {
  214.              a->l0.i_cost8x16 = COST_MAX;
  215.              return;
  216. @@ -2454,8 +2457,8 @@ static void x264_mb_analyse_inter_b16x8( x264_t *h, x264_mb_analysis_t *a, int i
  217.  
  218.          /* Early termination based on the current SATD score of partition[0]
  219.             plus the estimated SATD score of partition[1] */
  220. -        if( !i && i_part_cost + a->i_cost_est16x8[1] > i_best_satd
  221. -            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16 )
  222. +        if( a->b_early_terminate && (!i && i_part_cost + a->i_cost_est16x8[1] > i_best_satd
  223. +            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16) )
  224.          {
  225.              a->i_cost16x8bi = COST_MAX;
  226.              return;
  227. @@ -2547,8 +2550,8 @@ static void x264_mb_analyse_inter_b8x16( x264_t *h, x264_mb_analysis_t *a, int i
  228.  
  229.          /* Early termination based on the current SATD score of partition[0]
  230.             plus the estimated SATD score of partition[1] */
  231. -        if( !i && i_part_cost + a->i_cost_est8x16[1] > i_best_satd
  232. -            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16 )
  233. +        if( a->b_early_terminate && (!i && i_part_cost + a->i_cost_est8x16[1] > i_best_satd
  234. +            * (16 + (!!a->i_mbrd + !!h->mb.i_psy_rd))/16) )
  235.          {
  236.              a->i_cost8x16bi = COST_MAX;
  237.              return;
  238. @@ -2566,10 +2569,10 @@ static void x264_mb_analyse_inter_b8x16( x264_t *h, x264_mb_analysis_t *a, int i
  239.  
  240.  static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  241.  {
  242. -    int thresh = i_satd * 5/4;
  243. +    int thresh = a->b_early_terminate ? i_satd * 5/4 : COST_MAX;
  244.  
  245.      h->mb.i_type = P_L0;
  246. -    if( a->l0.i_rd16x16 == COST_MAX && a->l0.me16x16.cost <= i_satd * 3/2 )
  247. +    if( !a->b_early_terminate || (a->l0.i_rd16x16 == COST_MAX && a->l0.me16x16.cost <= i_satd * 3/2) )
  248.      {
  249.          h->mb.i_partition = D_16x16;
  250.          x264_analyse_update_cache( h, a );
  251. @@ -2639,7 +2642,7 @@ static void x264_mb_analyse_p_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd )
  252.  
  253.  static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_inter )
  254.  {
  255. -    int thresh = i_satd_inter * (17 + (!!h->mb.i_psy_rd))/16;
  256. +    int thresh = a->b_early_terminate ? i_satd_inter * (17 + (!!h->mb.i_psy_rd))/16 : COST_MAX;
  257.  
  258.      if( a->b_direct_available && a->i_rd16x16direct == COST_MAX )
  259.      {
  260. @@ -3020,8 +3023,8 @@ intra_analysis:
  261.              i_partition = D_16x16;
  262.              i_cost = analysis.l0.me16x16.cost;
  263.  
  264. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  265. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost )
  266. +            if( ( flags & X264_ANALYSE_PSUB16x16 ) && (!analysis.b_early_terminate ||
  267. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost) )
  268.              {
  269.                  i_type = P_8x8;
  270.                  i_partition = D_8x8;
  271. @@ -3056,8 +3059,8 @@ intra_analysis:
  272.  
  273.              /* Now do 16x8/8x16 */
  274.              i_thresh16x8 = analysis.l0.me8x8[1].cost_mv + analysis.l0.me8x8[2].cost_mv;
  275. -            if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
  276. -                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8 )
  277. +            if( ( flags & X264_ANALYSE_PSUB16x16 ) && (!analysis.b_early_terminate ||
  278. +                analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost + i_thresh16x8) )
  279.              {
  280.                  int i_avg_mv_ref_cost = (analysis.l0.me8x8[2].cost_mv + analysis.l0.me8x8[2].i_ref_cost
  281.                                        + analysis.l0.me8x8[3].cost_mv + analysis.l0.me8x8[3].i_ref_cost + 1) >> 1;
  282. @@ -3177,7 +3180,7 @@ intra_analysis:
  283.                  h->mb.i_partition = i_partition;
  284.                  if( i_cost < COST_MAX )
  285.                      x264_mb_analyse_transform_rd( h, &analysis, &i_satd_inter, &i_cost );
  286. -                x264_intra_rd( h, &analysis, i_satd_inter * 5/4 );
  287. +                x264_intra_rd( h, &analysis, analysis.b_early_terminate ? i_satd_inter * 5/4 : COST_MAX );
  288.              }
  289.  
  290.              COPY2_IF_LT( i_cost, analysis.i_satd_i16x16, i_type, I_16x16 );
  291. @@ -3431,17 +3434,17 @@ intra_analysis:
  292.  
  293.                  /* We can gain a little speed by checking the mode with the lowest estimated cost first */
  294.                  int try_16x8_first = i_cost_est16x8bi_total < i_cost_est8x16bi_total;
  295. -                if( try_16x8_first && i_cost_est16x8bi_total < i_cost )
  296. +                if( !analysis.b_early_terminate || (try_16x8_first && i_cost_est16x8bi_total < i_cost) )
  297.                  {
  298.                      x264_mb_analyse_inter_b16x8( h, &analysis, i_cost );
  299.                      COPY3_IF_LT( i_cost, analysis.i_cost16x8bi, i_type, analysis.i_mb_type16x8, i_partition, D_16x8 );
  300.                  }
  301. -                if( i_cost_est8x16bi_total < i_cost )
  302. +                if( !analysis.b_early_terminate || i_cost_est8x16bi_total < i_cost )
  303.                  {
  304.                      x264_mb_analyse_inter_b8x16( h, &analysis, i_cost );
  305.                      COPY3_IF_LT( i_cost, analysis.i_cost8x16bi, i_type, analysis.i_mb_type8x16, i_partition, D_8x16 );
  306.                  }
  307. -                if( !try_16x8_first && i_cost_est16x8bi_total < i_cost )
  308. +                if( analysis.b_early_terminate && (!try_16x8_first && i_cost_est16x8bi_total < i_cost) )
  309.                  {
  310.                      x264_mb_analyse_inter_b16x8( h, &analysis, i_cost );
  311.                      COPY3_IF_LT( i_cost, analysis.i_cost16x8bi, i_type, analysis.i_mb_type16x8, i_partition, D_16x8 );
  312. diff --git a/encoder/encoder.c b/encoder/encoder.c
  313. index 79880a3..4022819 100644
  314. --- a/encoder/encoder.c
  315. +++ b/encoder/encoder.c
  316. @@ -503,7 +503,7 @@ static int x264_validate_parameters( x264_t *h, int b_open )
  317.      h->param.rc.f_rf_constant = x264_clip3f( h->param.rc.f_rf_constant, -QP_BD_OFFSET, 51 );
  318.      h->param.rc.f_rf_constant_max = x264_clip3f( h->param.rc.f_rf_constant_max, -QP_BD_OFFSET, 51 );
  319.      h->param.rc.i_qp_constant = x264_clip3( h->param.rc.i_qp_constant, 0, QP_MAX );
  320. -    h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 10 );
  321. +    h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 0, 11 );
  322.      h->param.rc.f_ip_factor = X264_MAX( h->param.rc.f_ip_factor, 0.01f );
  323.      h->param.rc.f_pb_factor = X264_MAX( h->param.rc.f_pb_factor, 0.01f );
  324.      if( h->param.rc.i_rc_method == X264_RC_CRF )
  325. @@ -784,7 +784,7 @@ static int x264_validate_parameters( x264_t *h, int b_open )
  326.          h->param.rc.f_aq_strength = 0;
  327.      }
  328.      h->param.analyse.i_noise_reduction = x264_clip3( h->param.analyse.i_noise_reduction, 0, 1<<16 );
  329. -    if( h->param.analyse.i_subpel_refine == 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
  330. +    if( h->param.analyse.i_subpel_refine >= 10 && (h->param.analyse.i_trellis != 2 || !h->param.rc.i_aq_mode) )
  331.          h->param.analyse.i_subpel_refine = 9;
  332.  
  333.      {
  334. diff --git a/encoder/me.c b/encoder/me.c
  335. index ca6c602..10b7ef2 100644
  336. --- a/encoder/me.c
  337. +++ b/encoder/me.c
  338. @@ -46,6 +46,7 @@ static const uint8_t subpel_iterations[][4] =
  339.      {0,0,2,2},
  340.      {0,0,4,10},
  341.      {0,0,4,10},
  342. +    {0,0,4,10},
  343.      {0,0,4,10}};
  344.  
  345.  /* (x-1)%6 */
  346. --
  347. 1.7.4.1
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!