API tools faq
paste
Advertisement
Guest User

techouse

a guest
Jun 8th, 2008
395
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 21.20 KB | None | 0 0
raw download clone embed print report
  1. diff --git a/common/common.c b/common/common.c
  2. index ff8ce77..8a3a593 100644
  3. --- a/common/common.c
  4. +++ b/common/common.c
  5. @@ -95,6 +95,7 @@ void x264_param_default( x264_param_t *param )
  6. param->rc.f_pb_factor = 1.3;
  7. param->rc.i_aq_mode = X264_AQ_GLOBAL;
  8. param->rc.f_aq_strength = 1.0;
  9. + param->analyse.i_fgo = 0;
  10.  
  11. param->rc.b_stat_write = 0;
  12. param->rc.psz_stat_out = "x264_2pass.log";
  13. @@ -519,6 +520,8 @@ int x264_param_parse( x264_param_t *p, const char *name, const char *value )
  14. p->rc.i_aq_mode = atoi(value);
  15. OPT("aq-strength")
  16. p->rc.f_aq_strength = atof(value);
  17. + OPT("fgo")
  18. + p->analyse.i_fgo = atoi(value);
  19. OPT("pass")
  20. {
  21. int i = x264_clip3( atoi(value), 0, 3 );
  22. @@ -867,6 +870,7 @@ char *x264_param2string( x264_param_t *p, int b_res )
  23. s += sprintf( s, " nr=%d", p->analyse.i_noise_reduction );
  24. s += sprintf( s, " decimate=%d", p->analyse.b_dct_decimate );
  25. s += sprintf( s, " mbaff=%d", p->b_interlaced );
  26. + s += sprintf( s, " fgo=%d", p->analyse.i_fgo );
  27.  
  28. s += sprintf( s, " bframes=%d", p->i_bframe );
  29. if( p->i_bframe )
  30. diff --git a/common/pixel.c b/common/pixel.c
  31. index 1d5567b..71fc811 100644
  32. --- a/common/pixel.c
  33. +++ b/common/pixel.c
  34. @@ -70,7 +70,7 @@ PIXEL_SAD_C( x264_pixel_sad_4x4, 4, 4 )
  35. ****************************************************************************/
  36. #define PIXEL_SSD_C( name, lx, ly ) \
  37. static int name( uint8_t *pix1, int i_stride_pix1, \
  38. - uint8_t *pix2, int i_stride_pix2 ) \
  39. + uint8_t *pix2, int i_stride_pix2, int weight ) \
  40. { \
  41. int i_sum = 0; \
  42. int x, y; \
  43. @@ -95,6 +95,61 @@ PIXEL_SSD_C( x264_pixel_ssd_8x4, 8, 4 )
  44. PIXEL_SSD_C( x264_pixel_ssd_4x8, 4, 8 )
  45. PIXEL_SSD_C( x264_pixel_ssd_4x4, 4, 4 )
  46.  
  47. +#define PIXEL_NOISE_C( lx, ly) \
  48. +static int x264_pixel_noise_##lx##x##ly( uint8_t *pix, int i_stride ) \
  49. +{\
  50. + int score = 0;\
  51. + int x,y;\
  52. + for(y=0; y<ly; y++){\
  53. + if(y+1<ly){\
  54. + for(x=0; x<lx-1; x++)\
  55. + score += abs( pix[x] - pix[x+i_stride]\
  56. + - pix[x+1] + pix[x+1+i_stride]);\
  57. + }\
  58. + pix += i_stride;\
  59. + }\
  60. + return score;\
  61. +}
  62. +
  63. +PIXEL_NOISE_C( 16, 16 )
  64. +PIXEL_NOISE_C( 16, 8 )
  65. +PIXEL_NOISE_C( 8, 16 )
  66. +PIXEL_NOISE_C( 8, 8 )
  67. +PIXEL_NOISE_C( 8, 4 )
  68. +PIXEL_NOISE_C( 4, 8 )
  69. +PIXEL_NOISE_C( 4, 4 )
  70. +
  71. +#define PIXEL_NSSD( nssdname, lx, ly, ssdname, noisename)\
  72. +static int x264_pixel_nssd_##lx##x##ly##nssdname( uint8_t *pix1,\
  73. +int i_stride_pix1, uint8_t *pix2, int i_stride_pix2, int weight ) \
  74. +{\
  75. + int ssd = x264_pixel_ssd_##lx##x##ly##ssdname( pix1, i_stride_pix1, \
  76. + pix2, i_stride_pix2, weight );\
  77. + int noise1 = x264_pixel_noise_##lx##x##ly##noisename( pix1, i_stride_pix1 );\
  78. + int noise2 = x264_pixel_noise_##lx##x##ly##noisename( pix2, i_stride_pix2 );\
  79. + return ssd + abs(noise1 - noise2) * weight;\
  80. +}
  81. +
  82. +PIXEL_NSSD( , 16, 16, , )
  83. +PIXEL_NSSD( , 16, 8, , )
  84. +PIXEL_NSSD( , 8, 16, , )
  85. +PIXEL_NSSD( , 8, 8, , )
  86. +PIXEL_NSSD( , 8, 4, , )
  87. +PIXEL_NSSD( , 4, 8, , )
  88. +PIXEL_NSSD( , 4, 4, , )
  89. +PIXEL_NSSD( _mmxext, 16, 16, _mmx, _mmxext)
  90. +PIXEL_NSSD( _mmxext, 16, 8, _mmx, _mmxext)
  91. +PIXEL_NSSD( _mmxext, 8, 16, _mmx, _mmxext)
  92. +PIXEL_NSSD( _mmxext, 8, 8, _mmx, _mmxext)
  93. +PIXEL_NSSD( _mmxext, 8, 4, _mmx, _mmxext)
  94. +PIXEL_NSSD( _mmxext, 4, 8, _mmx, )
  95. +PIXEL_NSSD( _mmxext, 4, 4, _mmx, )
  96. +PIXEL_NSSD(_sse2, 16, 16, _sse2, _mmxext)
  97. +PIXEL_NSSD(_sse2, 16, 8, _sse2, _mmxext)
  98. +PIXEL_NSSD(_sse2, 8, 16, _sse2, _mmxext)
  99. +PIXEL_NSSD(_sse2, 8, 8, _sse2, _mmxext)
  100. +PIXEL_NSSD(_sse2, 8, 4, _sse2, _mmxext)
  101. +
  102. int64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2, int i_width, int i_height )
  103. {
  104. int64_t i_ssd = 0;
  105. @@ -102,7 +157,7 @@ int64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, uint8_t *pix1, int i_pix1
  106. int align = !(((long)pix1 | (long)pix2 | i_pix1 | i_pix2) & 15);
  107.  
  108. #define SSD(size) i_ssd += pf->ssd[size]( pix1 + y*i_pix1 + x, i_pix1, \
  109. - pix2 + y*i_pix2 + x, i_pix2 );
  110. + pix2 + y*i_pix2 + x, i_pix2, 0 );
  111. for( y = 0; y < i_height-15; y += 16 )
  112. {
  113. x = 0;
  114. @@ -527,6 +582,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
  115. INIT7( sad_x3, );
  116. INIT7( sad_x4, );
  117. INIT7( ssd, );
  118. + INIT7( nssd, );
  119. INIT7( satd, );
  120. INIT7( satd_x3, );
  121. INIT7( satd_x4, );
  122. @@ -550,6 +606,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
  123. INIT7( satd, _mmxext );
  124. INIT7( satd_x3, _mmxext );
  125. INIT7( satd_x4, _mmxext );
  126. + INIT7( nssd, _mmxext );
  127. INIT_ADS( _mmxext );
  128.  
  129. #ifdef ARCH_X86
  130. @@ -605,6 +662,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
  131. if( cpu&X264_CPU_SSE2 )
  132. {
  133. INIT5( ssd, _sse2 );
  134. + INIT5( nssd, _sse2);
  135. INIT5( satd, _sse2 );
  136. INIT5( satd_x3, _sse2 );
  137. INIT5( satd_x4, _sse2 );
  138. diff --git a/common/pixel.h b/common/pixel.h
  139. index c95a304..c22de07 100644
  140. --- a/common/pixel.h
  141. +++ b/common/pixel.h
  142. @@ -27,6 +27,7 @@
  143. // SSD assumes all args aligned
  144. // other cmp functions assume first arg aligned
  145. typedef int (*x264_pixel_cmp_t) ( uint8_t *, int, uint8_t *, int );
  146. +typedef int (*x264_pixel_cmp_weight_t) ( uint8_t *, int, uint8_t *, int, int );
  147. typedef void (*x264_pixel_cmp_x3_t) ( uint8_t *, uint8_t *, uint8_t *, uint8_t *, int, int[3] );
  148. typedef void (*x264_pixel_cmp_x4_t) ( uint8_t *, uint8_t *, uint8_t *, uint8_t *, uint8_t *, int, int[4] );
  149.  
  150. @@ -66,10 +67,12 @@ static const uint8_t x264_size2pixel[5][5] = {
  151. typedef struct
  152. {
  153. x264_pixel_cmp_t sad[7];
  154. - x264_pixel_cmp_t ssd[7];
  155. + x264_pixel_cmp_weight_t ssd[7];
  156. + x264_pixel_cmp_weight_t nssd[7];
  157. x264_pixel_cmp_t satd[7];
  158. x264_pixel_cmp_t ssim[7];
  159. x264_pixel_cmp_t sa8d[4];
  160. + x264_pixel_cmp_weight_t rdcmp[7]; /* either ssd or nsse for mode decision */
  161. x264_pixel_cmp_t mbcmp[7]; /* either satd or sad for subpel refine and mode decision */
  162. x264_pixel_cmp_t fpelcmp[7]; /* either satd or sad for fullpel motion search */
  163. x264_pixel_cmp_x3_t fpelcmp_x3[7];
  164. diff --git a/common/x86/pixel-a.asm b/common/x86/pixel-a.asm
  165. index b4d0656..4efcd04 100644
  166. --- a/common/x86/pixel-a.asm
  167. +++ b/common/x86/pixel-a.asm
  168. @@ -1285,6 +1285,92 @@ SATD_W4 ssse3 ; mmx, but uses pabsw from ssse3.
  169.  
  170.  
  171. ;=============================================================================
  172. +; NSSD
  173. +;=============================================================================
  174. +
  175. +%macro NOISE_CORE_LOAD_FIRST 3
  176. + mova %1, [r0+%3]
  177. + movu %2, [r0+%3+1]
  178. +%endmacro
  179. +
  180. +%macro NOISE_CORE_LOAD_LAST 3
  181. + mova %1, [r0+%3]
  182. + mova %2, %1
  183. + psllq %1, 8
  184. + psrlq %2, 8
  185. + psrlq %1, 8
  186. +%endmacro
  187. +
  188. +%macro NOISE_CORE_START 5
  189. + NOISE_CORE_LOAD %1, %2, %5
  190. + mova %3, %1
  191. + mova %4, %2
  192. + punpcklbw %1, m7
  193. + punpcklbw %2, m7
  194. + punpckhbw %3, m7
  195. + punpckhbw %4, m7
  196. + psubw %1, %2
  197. + psubw %3, %4
  198. +%endmacro
  199. +
  200. +%macro NOISE_CORE 7
  201. + NOISE_CORE_START %1, %2, %3, %4, %7
  202. + psubw %5, %1
  203. + psubw %6, %3
  204. + ABS2 %5, %6, %4, %2
  205. + paddw %6, %5
  206. + paddw m6, %6
  207. +%endmacro
  208. +
  209. +;arguments: src, stride
  210. +;macro arguments: width, height, name
  211. +%macro NOISE 3
  212. +%if %1 == 16
  213. +cglobal x264_pixel_noise_%1x%2_%3, 2,3
  214. + mov r2, r0
  215. +%else
  216. +cglobal x264_pixel_noise_%1x%2_%3, 2,2
  217. +x264_pixel_noise_%1x%2_%3 %+ .skip_prologue
  218. +%endif
  219. + pxor m7, m7
  220. + pxor m6, m6
  221. + NOISE_CORE_START m0, m1, m2, m3, 0
  222. + NOISE_CORE m4, m1, m5, m3, m0, m2, r1
  223. + lea r0, [r0+r1*2]
  224. +%rep (%2 - 2) / 2
  225. + NOISE_CORE m0, m1, m2, m3, m4, m5, 0
  226. + NOISE_CORE m4, m1, m5, m3, m0, m2, r1
  227. + lea r0, [r0+r1*2]
  228. +%endrep
  229. + mova m0, m6
  230. + punpcklwd m0, m7
  231. + punpckhwd m6, m7
  232. + paddd m6, m0
  233. + mova m0, m6
  234. + psrlq m6, 32
  235. + paddd m0, m6
  236. +%if %1 == 16
  237. + lea r0, [r2+8]
  238. + movd r2d, m0
  239. + call x264_pixel_noise_8x%2_%3 %+ .skip_prologue
  240. + add eax, r2d
  241. +%else
  242. + movd eax, m0
  243. +%endif
  244. + RET
  245. +%endmacro
  246. +
  247. +INIT_MMX
  248. +%define ABS2 ABS2_MMX
  249. +%define NOISE_CORE_LOAD NOISE_CORE_LOAD_LAST
  250. +NOISE 8, 16, mmxext
  251. +NOISE 8, 8, mmxext
  252. +NOISE 8, 4, mmxext
  253. +%define NOISE_CORE_LOAD NOISE_CORE_LOAD_FIRST
  254. +NOISE 16, 16, mmxext
  255. +NOISE 16, 8, mmxext
  256. +
  257. +;=============================================================================
  258. ; SSIM
  259. ;=============================================================================
  260.  
  261. diff --git a/common/x86/pixel.h b/common/x86/pixel.h
  262. index fcacaf2..2f80e82 100644
  263. --- a/common/x86/pixel.h
  264. +++ b/common/x86/pixel.h
  265. @@ -46,8 +46,9 @@ DECL_X1( sad, sse3 )
  266. DECL_X4( sad, mmxext )
  267. DECL_X4( sad, sse2 )
  268. DECL_X4( sad, sse3 )
  269. -DECL_X1( ssd, mmx )
  270. -DECL_X1( ssd, sse2 )
  271. +DECL_PIXELS( int, ssd, mmx, ( uint8_t *, int, uint8_t *, int, int ) )
  272. +DECL_PIXELS( int, ssd, sse2, ( uint8_t *, int, uint8_t *, int, int ) )
  273. +DECL_PIXELS( int, noise, mmxext, ( uint8_t *, int ) )
  274. DECL_X1( satd, mmxext )
  275. DECL_X1( satd, sse2 )
  276. DECL_X1( satd, ssse3 )
  277. diff --git a/encoder/analyse.c b/encoder/analyse.c
  278. index de3cf57..fd7e478 100644
  279. --- a/encoder/analyse.c
  280. +++ b/encoder/analyse.c
  281. @@ -765,7 +765,8 @@ static void x264_intra_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_thresh )
  282. else
  283. a->i_satd_i4x4 = COST_MAX;
  284.  
  285. - if( a->i_satd_i8x8 <= i_satd_thresh && a->i_satd_i8x8 < COST_MAX )
  286. + if( (a->i_satd_i8x8 <= i_satd_thresh || h->param.analyse.i_fgo)
  287. + && a->i_satd_i8x8 < COST_MAX )
  288. {
  289. h->mb.i_type = I_8x8;
  290. x264_analyse_update_cache( h, a );
  291. @@ -1928,7 +1929,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  292. //FIXME not all the update_cache calls are needed
  293. h->mb.i_partition = D_16x16;
  294. /* L0 */
  295. - if( a->l0.me16x16.cost <= thresh && a->l0.i_rd16x16 == COST_MAX )
  296. + if( (a->l0.me16x16.cost <= thresh || h->param.analyse.i_fgo) && a->l0.i_rd16x16 == COST_MAX )
  297. {
  298. h->mb.i_type = B_L0_L0;
  299. x264_analyse_update_cache( h, a );
  300. @@ -1936,7 +1937,7 @@ static void x264_mb_analyse_b_rd( x264_t *h, x264_mb_analysis_t *a, int i_satd_i
  301. }
  302.  
  303. /* L1 */
  304. - if( a->l1.me16x16.cost <= thresh && a->l1.i_rd16x16 == COST_MAX )
  305. + if( (a->l1.me16x16.cost <= thresh || h->param.analyse.i_fgo) && a->l1.i_rd16x16 == COST_MAX )
  306. {
  307. h->mb.i_type = B_L1_L1;
  308. x264_analyse_update_cache( h, a );
  309. @@ -2277,7 +2278,8 @@ void x264_macroblock_analyse( x264_t *h )
  310.  
  311. if( analysis.b_mbrd )
  312. {
  313. - x264_mb_analyse_p_rd( h, &analysis, X264_MIN(i_satd_inter, i_satd_intra) );
  314. + x264_mb_analyse_p_rd( h, &analysis, h->param.analyse.i_fgo ?
  315. + i_satd_inter : X264_MIN(i_satd_inter, i_satd_intra) );
  316. i_type = P_L0;
  317. i_partition = D_16x16;
  318. i_cost = analysis.l0.me16x16.cost;
  319. diff --git a/encoder/encoder.c b/encoder/encoder.c
  320. index 2b81e64..83f0141 100644
  321. --- a/encoder/encoder.c
  322. +++ b/encoder/encoder.c
  323. @@ -402,6 +402,7 @@ static int x264_validate_parameters( x264_t *h )
  324. h->param.analyse.i_noise_reduction = 0;
  325. h->param.analyse.i_subpel_refine = x264_clip3( h->param.analyse.i_subpel_refine, 1, 6 );
  326. h->param.rc.i_aq_mode = 0;
  327. + h->param.analyse.i_fgo = 0;
  328. }
  329. if( h->param.rc.i_rc_method == X264_RC_CQP )
  330. {
  331. @@ -500,6 +501,29 @@ static int x264_validate_parameters( x264_t *h )
  332. if( h->param.analyse.i_direct_8x8_inference < 0 )
  333. h->param.analyse.i_direct_8x8_inference = l->direct8x8;
  334. }
  335. +
  336. + if( h->param.analyse.i_fgo )
  337. + {
  338. + if(h->param.analyse.i_subpel_refine < 6 ||
  339. + (!h->param.analyse.b_bframe_rdo && h->param.i_bframe) )
  340. + {
  341. + if(h->param.i_bframe)
  342. + x264_log( h, X264_LOG_WARNING, "fgo requires b-rdo and subme >= 6\n" );
  343. + else
  344. + x264_log( h, X264_LOG_WARNING, "fgo requires subme >= 6\n" );
  345. + h->param.analyse.i_fgo = 0;
  346. + }
  347. + else
  348. + {
  349. + /* Arbitrary clipping. */
  350. + h->param.analyse.i_fgo = x264_clip3(h->param.analyse.i_fgo, 0, 50);
  351. + /* P-skip's threshold isn't necessarily accurate when using NSSD/FGO */
  352. + h->param.analyse.b_fast_pskip = 0;
  353. + /* B-frame QPs need to be lower to retain grain */
  354. + /* Arbitrary formula to scale pbratio based on fgo strength. */
  355. + h->param.rc.f_pb_factor = 1 + (h->param.rc.f_pb_factor - 1) / pow(h->param.analyse.i_fgo,0.3);
  356. + }
  357. + }
  358.  
  359. if( h->param.i_threads > 1 )
  360. {
  361. @@ -562,6 +586,7 @@ static void mbcmp_init( x264_t *h )
  362. memcpy( h->pixf.fpelcmp, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.fpelcmp) );
  363. memcpy( h->pixf.fpelcmp_x3, satd ? h->pixf.satd_x3 : h->pixf.sad_x3, sizeof(h->pixf.fpelcmp_x3) );
  364. memcpy( h->pixf.fpelcmp_x4, satd ? h->pixf.satd_x4 : h->pixf.sad_x4, sizeof(h->pixf.fpelcmp_x4) );
  365. + memcpy( h->pixf.rdcmp, h->param.analyse.i_fgo ? h->pixf.nssd : h->pixf.ssd, sizeof(h->pixf.rdcmp) );
  366. }
  367.  
  368. /****************************************************************************
  369. diff --git a/encoder/ratecontrol.c b/encoder/ratecontrol.c
  370. index 11790ea..52b436f 100644
  371. --- a/encoder/ratecontrol.c
  372. +++ b/encoder/ratecontrol.c
  373. @@ -194,7 +194,7 @@ static NOINLINE int ac_energy_mb( x264_t *h, int mb_x, int mb_y, int *satd )
  374. int pix = i ? PIXEL_8x8 : PIXEL_16x16;
  375. stride <<= h->mb.b_interlaced;
  376. sad = h->pixf.sad[pix]( flat, 0, h->fenc->plane[i]+offset, stride );
  377. - ssd = h->pixf.ssd[pix]( flat, 0, h->fenc->plane[i]+offset, stride );
  378. + ssd = h->pixf.ssd[pix]( flat, 0, h->fenc->plane[i]+offset, stride, 0 );
  379. var += ssd - (sad * sad >> (i?6:8));
  380. // SATD to represent the block's overall complexity (bit cost) for intra encoding.
  381. // exclude the DC coef, because nothing short of an actual intra prediction will estimate DC cost.
  382. @@ -269,7 +269,7 @@ int x264_ratecontrol_new( x264_t *h )
  383.  
  384. rc->b_abr = h->param.rc.i_rc_method != X264_RC_CQP && !h->param.rc.b_stat_read;
  385. rc->b_2pass = h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.b_stat_read;
  386. -
  387. +
  388. /* FIXME: use integers */
  389. if(h->param.i_fps_num > 0 && h->param.i_fps_den > 0)
  390. rc->fps = (float) h->param.i_fps_num / h->param.i_fps_den;
  391. @@ -679,7 +679,7 @@ void x264_ratecontrol_summary( x264_t *h )
  392. if( rc->b_abr && h->param.rc.i_rc_method == X264_RC_ABR && rc->cbr_decay > .9999 )
  393. {
  394. double base_cplx = h->mb.i_mb_count * (h->param.i_bframe ? 120 : 80);
  395. - x264_log( h, X264_LOG_INFO, "final ratefactor: %.2f\n",
  396. + x264_log( h, X264_LOG_INFO, "final ratefactor: %.2f\n",
  397. qscale2qp( pow( base_cplx, 1 - h->param.rc.f_qcompress )
  398. * rc->cplxr_sum / rc->wanted_bits_window ) );
  399. }
  400. @@ -838,7 +838,7 @@ double predict_row_size( x264_t *h, int y, int qp )
  401. x264_ratecontrol_t *rc = h->rc;
  402. double pred_s = predict_size( rc->row_pred, qp2qscale(qp), h->fdec->i_row_satd[y] );
  403. double pred_t = 0;
  404. - if( h->sh.i_type != SLICE_TYPE_I
  405. + if( h->sh.i_type != SLICE_TYPE_I
  406. && h->fref0[0]->i_type == h->fdec->i_type
  407. && h->fref0[0]->i_row_satd[y] > 0 )
  408. {
  409. @@ -1007,7 +1007,7 @@ void x264_ratecontrol_end( x264_t *h, int bits )
  410. int dir_frame = h->stat.frame.i_direct_score[1] - h->stat.frame.i_direct_score[0];
  411. int dir_avg = h->stat.i_direct_score[1] - h->stat.i_direct_score[0];
  412. char c_direct = h->mb.b_direct_auto_write ?
  413. - ( dir_frame>0 ? 's' : dir_frame<0 ? 't' :
  414. + ( dir_frame>0 ? 's' : dir_frame<0 ? 't' :
  415. dir_avg>0 ? 's' : dir_avg<0 ? 't' : '-' )
  416. : '-';
  417. fprintf( rc->p_stat_file_out,
  418. diff --git a/encoder/rdo.c b/encoder/rdo.c
  419. index 8607e07..ec956fe 100644
  420. --- a/encoder/rdo.c
  421. +++ b/encoder/rdo.c
  422. @@ -55,18 +55,22 @@ static uint16_t cabac_prefix_size[15][128];
  423.  
  424. static int ssd_mb( x264_t *h )
  425. {
  426. - return h->pixf.ssd[PIXEL_16x16]( h->mb.pic.p_fenc[0], FENC_STRIDE,
  427. - h->mb.pic.p_fdec[0], FDEC_STRIDE )
  428. - + h->pixf.ssd[PIXEL_8x8]( h->mb.pic.p_fenc[1], FENC_STRIDE,
  429. - h->mb.pic.p_fdec[1], FDEC_STRIDE )
  430. - + h->pixf.ssd[PIXEL_8x8]( h->mb.pic.p_fenc[2], FENC_STRIDE,
  431. - h->mb.pic.p_fdec[2], FDEC_STRIDE );
  432. + return h->pixf.rdcmp[PIXEL_16x16]( h->mb.pic.p_fenc[0], FENC_STRIDE,
  433. + h->mb.pic.p_fdec[0], FDEC_STRIDE,
  434. + h->param.analyse.i_fgo )
  435. + + h->pixf.rdcmp[PIXEL_8x8] ( h->mb.pic.p_fenc[1], FENC_STRIDE,
  436. + h->mb.pic.p_fdec[1], FDEC_STRIDE,
  437. + h->param.analyse.i_fgo )
  438. + + h->pixf.rdcmp[PIXEL_8x8] ( h->mb.pic.p_fenc[2], FENC_STRIDE,
  439. + h->mb.pic.p_fdec[2], FDEC_STRIDE,
  440. + h->param.analyse.i_fgo );
  441. }
  442.  
  443. static int ssd_plane( x264_t *h, int size, int p, int x, int y )
  444. {
  445. - return h->pixf.ssd[size]( h->mb.pic.p_fenc[p] + x+y*FENC_STRIDE, FENC_STRIDE,
  446. - h->mb.pic.p_fdec[p] + x+y*FDEC_STRIDE, FDEC_STRIDE );
  447. + return h->pixf.rdcmp[size]( h->mb.pic.p_fenc[p] + x+y*FENC_STRIDE, FENC_STRIDE,
  448. + h->mb.pic.p_fdec[p] + x+y*FDEC_STRIDE, FDEC_STRIDE,
  449. + h->param.analyse.i_fgo );
  450. }
  451.  
  452. static int x264_rd_cost_mb( x264_t *h, int i_lambda2 )
  453. diff --git a/tools/checkasm.c b/tools/checkasm.c
  454. index 73faf12..ecfb540 100644
  455. --- a/tools/checkasm.c
  456. +++ b/tools/checkasm.c
  457. @@ -69,9 +69,32 @@ static int check_pixel( int cpu_ref, int cpu_new )
  458. } \
  459. } \
  460. report( "pixel " #name " :" );
  461. +
  462. +#define TEST_PIXEL_WEIGHT( name, align ) \
  463. + for( i = 0, ok = 1, used_asm = 0; i < 7; i++ ) \
  464. + { \
  465. + int res_c, res_asm; \
  466. + if( pixel_asm.name[i] != pixel_ref.name[i] ) \
  467. + { \
  468. + for( j=0; j<64; j++ ) \
  469. + { \
  470. + used_asm = 1; \
  471. + res_c = call_c( pixel_c.name[i], buf1, 32, buf2+j*!align, 16, j ); \
  472. + res_asm = call_a( pixel_asm.name[i], buf1, 32, buf2+j*!align, 16, j ); \
  473. + if( res_c != res_asm ) \
  474. + { \
  475. + ok = 0; \
  476. + fprintf( stderr, #name "[%d]: %d != %d [FAILED]\n", i, res_c, res_asm ); \
  477. + break; \
  478. + } \
  479. + } \
  480. + } \
  481. + } \
  482. + report( "pixel " #name " :" );
  483.  
  484. TEST_PIXEL( sad, 0 );
  485. - TEST_PIXEL( ssd, 1 );
  486. + TEST_PIXEL_WEIGHT( ssd, 1 );
  487. + TEST_PIXEL_WEIGHT( nssd, 1 );
  488. TEST_PIXEL( satd, 0 );
  489. TEST_PIXEL( sa8d, 0 );
  490.  
  491. diff --git a/x264.c b/x264.c
  492. index 70adb71..88c8c78 100644
  493. --- a/x264.c
  494. +++ b/x264.c
  495. @@ -196,6 +196,9 @@ static void Help( x264_param_t *defaults, int b_longhelp )
  496. " textured areas. [%.1f]\n"
  497. " - 0.5: weak AQ\n"
  498. " - 1.5: strong AQ\n", defaults->rc.f_aq_strength );
  499. + H0( " --fgo <int> Activates Film Grain Optimization.[%d]\n"
  500. + " - 5: weak FGO\n"
  501. + " - 15: strong FGO\n", defaults->analyse.i_fgo);
  502. H0( "\n" );
  503. H0( " -p, --pass <1|2|3> Enable multipass ratecontrol\n"
  504. " - 1: First pass, creates stats file\n"
  505. @@ -420,6 +423,7 @@ static int Parse( int argc, char **argv,
  506. { "no-dct-decimate", no_argument, NULL, 0 },
  507. { "aq-strength", required_argument, NULL, 0 },
  508. { "aq-mode", required_argument, NULL, 0 },
  509. + { "fgo", required_argument, NULL, 0 },
  510. { "deadzone-inter", required_argument, NULL, '0' },
  511. { "deadzone-intra", required_argument, NULL, '0' },
  512. { "level", required_argument, NULL, 0 },
  513. diff --git a/x264.h b/x264.h
  514. index d2c6510..c0156ea 100644
  515. --- a/x264.h
  516. +++ b/x264.h
  517. @@ -236,6 +236,7 @@ typedef struct x264_param_t
  518. int b_fast_pskip; /* early SKIP detection on P-frames */
  519. int b_dct_decimate; /* transform coefficient thresholding on P-frames */
  520. int i_noise_reduction; /* adaptive pseudo-deadzone */
  521. + int i_fgo; /* psy film grain optimization */
  522.  
  523. /* the deadzone size that will be used in luma quantization */
  524. int i_luma_deadzone[2]; /* {inter, intra} */
  525.  
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!