API tools faq
paste
Guest User

techouse

a guest
Jun 4th, 2008
330
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 9.57 KB | None | 0 0
raw download clone embed print report
  1. --- common/common.c Sun Jan 27 13:39:08 2008
  2. +++ common/common.c Sun Jan 27 13:39:08 2008
  3. @@ -483,6 +483,8 @@
  4. p->analyse.i_mv_range_thread = atoi(value);
  5. OPT2("subme", "subq")
  6. p->analyse.i_subpel_refine = atoi(value);
  7. + OPT2("me-prepass", "meprepass")
  8. + p->analyse.i_me_prepass = atobool(value);
  9. OPT("rdcmp")
  10. b_error |= parse_enum( value, x264_rdcmp_names, &p->analyse.i_rdcmp );
  11. OPT("bime")
  12. @@ -881,6 +883,7 @@
  13. s += sprintf( s, " analyse=%#x:%#x", p->analyse.intra, p->analyse.inter );
  14. s += sprintf( s, " me=%s", x264_motion_est_names[ p->analyse.i_me_method ] );
  15. s += sprintf( s, " subme=%d", p->analyse.i_subpel_refine );
  16. + s += sprintf( s, " me-prepass=%d", p->analyse.i_me_prepass );
  17. s += sprintf( s, " rdcmp=%s", x264_rdcmp_names[ p->analyse.i_rdcmp ] );
  18. s += sprintf( s, " brdo=%d", p->analyse.b_bframe_rdo );
  19. s += sprintf( s, " mixed_ref=%d", p->analyse.b_mixed_references );
  20. --- x264.c Sun Jan 27 13:39:08 2008
  21. +++ x264.c Sun Jan 27 13:39:08 2008
  22. @@ -248,7 +248,8 @@
  23. H1( " --mvrange-thread <int> Minimum buffer between threads [-1 (auto)]\n" );
  24. H0( " -m, --subme <integer> Subpixel motion estimation and partition\n"
  25. " decision quality: 1=fast, 7=best. [%d]\n", defaults->analyse.i_subpel_refine );
  26. - H0( " --b-rdo RD based mode decision for B-frames. Requires subme 6.\n" );
  27. + H0( " --me-prepass Run an ME prepass on predictors. Requires subme 3 or higher.\n");
  28. + H0( " --b-rdo RD based mode decision for B-frames. Requires subme 6 or higher.\n" );
  29. H0( " --rdcmp Metric used for RD mode decision [\"%s\"]\n",
  30. strtable_lookup( x264_rdcmp_names, defaults->analyse.i_rdcmp ) );
  31. H0( " - ssd: normal (maximum PSNR)\n"
  32. @@ -427,6 +428,7 @@
  33. { "mvrange", required_argument, NULL, 0 },
  34. { "mvrange-thread", required_argument, NULL, 0 },
  35. { "subme", required_argument, NULL, 'm' },
  36. + { "me-prepass", no_argument, NULL, 0 },
  37. { "rdcmp", required_argument, NULL, 0 },
  38. { "b-rdo", no_argument, NULL, 0 },
  39. { "mixed-refs", no_argument, NULL, 0 },
  40. --- x264.h Sun Jan 27 13:39:08 2008
  41. +++ x264.h Sun Jan 27 13:39:08 2008
  42. @@ -233,6 +233,7 @@
  43. int i_mv_range; /* maximum length of a mv (in pixels). -1 = auto, based on level */
  44. int i_mv_range_thread; /* minimum space between threads. -1 = auto, based on number of threads. */
  45. int i_subpel_refine; /* subpixel motion estimation quality */
  46. + int i_me_prepass; /* run an ME prepass on predictors */
  47. int b_bidir_me; /* jointly optimize both MVs in B-frames */
  48. int b_chroma_me; /* chroma ME for subpel and mode decision in P-frames */
  49. int b_bframe_rdo; /* RD based mode decision for B-frames */
  50. --- encoder/me.c Sun Jan 27 13:39:08 2008
  51. +++ encoder/me.c Sun Jan 27 20:25:50 2008
  52. @@ -69,6 +69,23 @@
  53. COPY3_IF_LT( bpred_cost, cost, bpred_mx, mx, bpred_my, my ); \
  54. }
  55.  
  56. +#define COST_MV_HPEL2( mx, my, cost ) \
  57. +{ \
  58. + int stride = 16; \
  59. + uint8_t *src = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh ); \
  60. + cost = h->pixf.fpelcmp[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \
  61. + + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \
  62. +}
  63. +
  64. +#define COST_MV_HPEL3( mx, my) \
  65. +{ \
  66. + int stride = 16; \
  67. + uint8_t *src = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh ); \
  68. + int cost = h->pixf.fpelcmp[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \
  69. + + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \
  70. + COPY3_IF_LT( bestcost, cost, bestx, mx, besty, my ); \
  71. +}
  72. +
  73. #define COST_MV_X3_DIR( m0x, m0y, m1x, m1y, m2x, m2y, costs )\
  74. {\
  75. uint8_t *pix_base = p_fref + bmx + bmy*m->i_stride[0];\
  76. @@ -171,8 +188,13 @@
  77. int mv_y_min = h->mb.mv_min_fpel[1];
  78. int mv_x_max = h->mb.mv_max_fpel[0];
  79. int mv_y_max = h->mb.mv_max_fpel[1];
  80. + int mv_x_min4 = h->mb.mv_min_fpel[0]<<2;
  81. + int mv_y_min4 = h->mb.mv_min_fpel[1]<<2;
  82. + int mv_x_max4 = h->mb.mv_max_fpel[0]<<2;
  83. + int mv_y_max4 = h->mb.mv_max_fpel[1]<<2;
  84.  
  85. #define CHECK_MVRANGE(mx,my) ( mx >= mv_x_min && mx <= mv_x_max && my >= mv_y_min && my <= mv_y_max )
  86. +#define CHECK_MVRANGE4(mx,my) ( mx >= mv_x_min4 && mx <= mv_x_max4 && my >= mv_y_min4 && my <= mv_y_max4 )
  87.  
  88. const int16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0];
  89. const int16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1];
  90. @@ -183,19 +205,88 @@
  91. pmy = ( bmy + 2 ) >> 2;
  92. bcost = COST_MAX;
  93.  
  94. - /* try extra predictors if provided */
  95. - if( h->mb.i_subpel_refine >= 3 )
  96. - {
  97. - COST_MV_HPEL( bmx, bmy );
  98. - for( i = 0; i < i_mvc; i++ )
  99. + /* try extra predictors if provided */
  100. + if( h->mb.i_subpel_refine >= 3 )
  101. + {
  102. + COST_MV_HPEL( bmx, bmy );
  103. + if(!h->param.analyse.i_me_prepass)
  104. + {
  105. + for( i = 0; i < i_mvc; i++ )
  106. + {
  107. + const int mx = x264_clip3( mvc[i][0], mv_x_min*4, mv_x_max*4 );
  108. + const int my = x264_clip3( mvc[i][1], mv_y_min*4, mv_y_max*4 );
  109. + if( mx != bpred_mx || my != bpred_my )
  110. + COST_MV_HPEL( mx, my );
  111. + }
  112. + }
  113. + else
  114. {
  115. - int mx = mvc[i][0];
  116. - int my = mvc[i][1];
  117. - if( (mx | my) && ((mx-bmx) | (my-bmy)) )
  118. + for( i = 0; i < i_mvc; i++ )
  119. {
  120. - mx = x264_clip3( mx, mv_x_min*4, mv_x_max*4 );
  121. - my = x264_clip3( my, mv_y_min*4, mv_y_max*4 );
  122. - COST_MV_HPEL( mx, my );
  123. + const int mx = x264_clip3( mvc[i][0], mv_x_min*4, mv_x_max*4 );
  124. + const int my = x264_clip3( mvc[i][1], mv_y_min*4, mv_y_max*4 );
  125. + int doSearch = 1;
  126. + int j;
  127. + for(j = 0; j < i; j++)
  128. + {
  129. + if(mvc[i][0] == mvc[j][0] && mvc[i][1] == mvc[j][1]) doSearch = 0;
  130. + }
  131. + if( ( mx != bpred_mx || my != bpred_my ) && doSearch)
  132. + {
  133. + int bestcost;
  134. + int bestx = mx;
  135. + int besty = my;
  136. + COST_MV_HPEL2( mx, my, bestcost );
  137. + COPY3_IF_LT( bpred_cost, bestcost, bpred_mx, bestx, bpred_my, besty );
  138. + if(bestcost < 2*bpred_cost)
  139. + {
  140. + int n;
  141. + int dir = -2;
  142. + COST_MV_HPEL2(bestx-4,besty,costs[0]);
  143. + COST_MV_HPEL2(bestx-2,besty+4,costs[1]);
  144. + COST_MV_HPEL2(bestx+2,besty+4,costs[2]);
  145. + COST_MV_HPEL2(bestx+4,besty,costs[3]);
  146. + COST_MV_HPEL2(bestx+2,besty-4,costs[4]);
  147. + COST_MV_HPEL2(bestx-2,besty-4,costs[5]);
  148. + COPY2_IF_LT( bestcost, costs[0], dir, 0 );
  149. + COPY2_IF_LT( bestcost, costs[1], dir, 1 );
  150. + COPY2_IF_LT( bestcost, costs[2], dir, 2 );
  151. + COPY2_IF_LT( bestcost, costs[3], dir, 3 );
  152. + COPY2_IF_LT( bestcost, costs[4], dir, 4 );
  153. + COPY2_IF_LT( bestcost, costs[5], dir, 5 );
  154. + if( dir != -2 )
  155. + {
  156. + static const int hex2[8][2] = {{-2,-4}, {-4,0}, {-2,4}, {2,4}, {4,0}, {2,-4}, {-2,-4}, {-4,0}};
  157. + bestx += hex2[dir+1][0];
  158. + besty += hex2[dir+1][1];
  159. + for( n = 1; n < i_me_range && CHECK_MVRANGE4(bestx, besty); n++ )
  160. + {
  161. + static const int mod6[8] = {5,0,1,2,3,4,5,0};
  162. + const int odir = mod6[dir+1];
  163. + COST_MV_HPEL2(hex2[odir+0][0]+bestx,hex2[odir+0][1]+besty,costs[0]);
  164. + COST_MV_HPEL2(hex2[odir+1][0]+bestx,hex2[odir+1][1]+besty,costs[1]);
  165. + COST_MV_HPEL2(hex2[odir+2][0]+bestx,hex2[odir+2][1]+besty,costs[2]);
  166. + dir = -2;
  167. + COPY2_IF_LT( bestcost, costs[0], dir, odir-1 );
  168. + COPY2_IF_LT( bestcost, costs[1], dir, odir );
  169. + COPY2_IF_LT( bestcost, costs[2], dir, odir+1 );
  170. + if( dir == -2 )
  171. + break;
  172. + bestx += hex2[dir+1][0];
  173. + besty += hex2[dir+1][1];
  174. + }
  175. + }
  176. + COST_MV_HPEL3(bestx+2,besty-2);
  177. + COST_MV_HPEL3(bestx+2,besty);
  178. + COST_MV_HPEL3(bestx+2,besty+2);
  179. + COST_MV_HPEL3(bestx,besty-2);
  180. + COST_MV_HPEL3(bestx,besty+2);
  181. + COST_MV_HPEL3(bestx-2,besty-2);
  182. + COST_MV_HPEL3(bestx-2,besty);
  183. + COST_MV_HPEL3(bestx-2,besty+2);
  184. + COPY3_IF_LT(bpred_cost,bestcost,bpred_mx,bestx,bpred_my,besty);
  185. + }
  186. + }
  187. }
  188. }
  189. bmx = ( bpred_mx + 2 ) >> 2;
  190.  
Advertisement
Add Comment
Please, Sign In to add comment
Public Pastes
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!