1 /*****************************************************************************
2 * quant.c: quantization and level-run
3 *****************************************************************************
4 * Copyright (C) 2005-2011 x264 project
6 * Authors: Loren Merritt <lorenm@u.washington.edu>
7 * Fiona Glaser <fiona@x264.com>
8 * Christian Heine <sennindemokrit@gmx.net>
9 * Henrik Gramner <hengar-6@student.ltu.se>
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation; either version 2 of the License, or
14 * (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with this program; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
25 * This program is also available under a commercial proprietary license.
26 * For more information, contact us at licensing@x264.com.
27 *****************************************************************************/
32 #include "x86/quant.h"
35 # include "ppc/quant.h"
38 # include "arm/quant.h"
41 #define QUANT_ONE( coef, mf, f ) \
44 (coef) = (f + (coef)) * (mf) >> 16; \
46 (coef) = - ((f - (coef)) * (mf) >> 16); \
50 static int quant_8x8( dctcoef dct[64], udctcoef mf[64], udctcoef bias[64] )
53 for( int i = 0; i < 64; i++ )
54 QUANT_ONE( dct[i], mf[i], bias[i] );
58 static int quant_4x4( dctcoef dct[16], udctcoef mf[16], udctcoef bias[16] )
61 for( int i = 0; i < 16; i++ )
62 QUANT_ONE( dct[i], mf[i], bias[i] );
66 static int quant_4x4_dc( dctcoef dct[16], int mf, int bias )
69 for( int i = 0; i < 16; i++ )
70 QUANT_ONE( dct[i], mf, bias );
74 static int quant_2x2_dc( dctcoef dct[4], int mf, int bias )
77 QUANT_ONE( dct[0], mf, bias );
78 QUANT_ONE( dct[1], mf, bias );
79 QUANT_ONE( dct[2], mf, bias );
80 QUANT_ONE( dct[3], mf, bias );
84 #define DEQUANT_SHL( x ) \
85 dct[x] = ( dct[x] * dequant_mf[i_mf][x] ) << i_qbits
87 #define DEQUANT_SHR( x ) \
88 dct[x] = ( dct[x] * dequant_mf[i_mf][x] + f ) >> (-i_qbits)
90 static void dequant_4x4( dctcoef dct[16], int dequant_mf[6][16], int i_qp )
92 const int i_mf = i_qp%6;
93 const int i_qbits = i_qp/6 - 4;
97 for( int i = 0; i < 16; i++ )
102 const int f = 1 << (-i_qbits-1);
103 for( int i = 0; i < 16; i++ )
108 static void dequant_8x8( dctcoef dct[64], int dequant_mf[6][64], int i_qp )
110 const int i_mf = i_qp%6;
111 const int i_qbits = i_qp/6 - 6;
115 for( int i = 0; i < 64; i++ )
120 const int f = 1 << (-i_qbits-1);
121 for( int i = 0; i < 64; i++ )
126 static void dequant_4x4_dc( dctcoef dct[16], int dequant_mf[6][16], int i_qp )
128 const int i_qbits = i_qp/6 - 6;
132 const int i_dmf = dequant_mf[i_qp%6][0] << i_qbits;
133 for( int i = 0; i < 16; i++ )
138 const int i_dmf = dequant_mf[i_qp%6][0];
139 const int f = 1 << (-i_qbits-1);
140 for( int i = 0; i < 16; i++ )
141 dct[i] = ( dct[i] * i_dmf + f ) >> (-i_qbits);
145 #define IDCT_DEQUANT_2X4_START \
146 int a0 = dct[0] + dct[1]; \
147 int a1 = dct[2] + dct[3]; \
148 int a2 = dct[4] + dct[5]; \
149 int a3 = dct[6] + dct[7]; \
150 int a4 = dct[0] - dct[1]; \
151 int a5 = dct[2] - dct[3]; \
152 int a6 = dct[4] - dct[5]; \
153 int a7 = dct[6] - dct[7]; \
163 static void idct_dequant_2x4_dc( dctcoef dct[8], dctcoef dct4x4[8][16], int dequant_mf[6][16], int i_qp )
165 IDCT_DEQUANT_2X4_START
166 int dmf = dequant_mf[i_qp%6][0] << i_qp/6;
167 dct4x4[0][0] = ((b0 + b1) * dmf + 32) >> 6;
168 dct4x4[1][0] = ((b2 + b3) * dmf + 32) >> 6;
169 dct4x4[2][0] = ((b0 - b1) * dmf + 32) >> 6;
170 dct4x4[3][0] = ((b2 - b3) * dmf + 32) >> 6;
171 dct4x4[4][0] = ((b4 - b5) * dmf + 32) >> 6;
172 dct4x4[5][0] = ((b6 - b7) * dmf + 32) >> 6;
173 dct4x4[6][0] = ((b4 + b5) * dmf + 32) >> 6;
174 dct4x4[7][0] = ((b6 + b7) * dmf + 32) >> 6;
177 static void idct_dequant_2x4_dconly( dctcoef dct[8], int dequant_mf[6][16], int i_qp )
179 IDCT_DEQUANT_2X4_START
180 int dmf = dequant_mf[i_qp%6][0] << i_qp/6;
181 dct[0] = ((b0 + b1) * dmf + 32) >> 6;
182 dct[1] = ((b2 + b3) * dmf + 32) >> 6;
183 dct[2] = ((b0 - b1) * dmf + 32) >> 6;
184 dct[3] = ((b2 - b3) * dmf + 32) >> 6;
185 dct[4] = ((b4 - b5) * dmf + 32) >> 6;
186 dct[5] = ((b6 - b7) * dmf + 32) >> 6;
187 dct[6] = ((b4 + b5) * dmf + 32) >> 6;
188 dct[7] = ((b6 + b7) * dmf + 32) >> 6;
191 static ALWAYS_INLINE void optimize_chroma_idct_dequant_2x4( dctcoef out[8], dctcoef dct[8], int dmf )
193 IDCT_DEQUANT_2X4_START
194 out[0] = ((b0 + b1) * dmf + 2080) >> 6; /* 2080 = 32 + (32<<6) */
195 out[1] = ((b2 + b3) * dmf + 2080) >> 6;
196 out[2] = ((b0 - b1) * dmf + 2080) >> 6;
197 out[3] = ((b2 - b3) * dmf + 2080) >> 6;
198 out[4] = ((b4 - b5) * dmf + 2080) >> 6;
199 out[5] = ((b6 - b7) * dmf + 2080) >> 6;
200 out[6] = ((b4 + b5) * dmf + 2080) >> 6;
201 out[7] = ((b6 + b7) * dmf + 2080) >> 6;
203 #undef IDCT_DEQUANT_2X4_START
205 static ALWAYS_INLINE void optimize_chroma_idct_dequant_2x2( dctcoef out[4], dctcoef dct[4], int dmf )
207 int d0 = dct[0] + dct[1];
208 int d1 = dct[2] + dct[3];
209 int d2 = dct[0] - dct[1];
210 int d3 = dct[2] - dct[3];
211 out[0] = ((d0 + d1) * dmf >> 5) + 32;
212 out[1] = ((d0 - d1) * dmf >> 5) + 32;
213 out[2] = ((d2 + d3) * dmf >> 5) + 32;
214 out[3] = ((d2 - d3) * dmf >> 5) + 32;
217 static ALWAYS_INLINE int optimize_chroma_round( dctcoef *ref, dctcoef *dct, int dequant_mf, int chroma422 )
222 optimize_chroma_idct_dequant_2x4( out, dct, dequant_mf );
224 optimize_chroma_idct_dequant_2x2( out, dct, dequant_mf );
227 for( int i = 0; i < (chroma422?8:4); i++ )
228 sum |= ref[i] ^ out[i];
232 static ALWAYS_INLINE int optimize_chroma_dc_internal( dctcoef *dct, int dequant_mf, int chroma422 )
234 /* dequant_mf = h->dequant4_mf[CQM_4IC + b_inter][i_qp%6][0] << i_qp/6, max 32*64 */
239 optimize_chroma_idct_dequant_2x4( dct_orig, dct, dequant_mf );
241 optimize_chroma_idct_dequant_2x2( dct_orig, dct, dequant_mf );
243 /* If the DC coefficients already round to zero, terminate early. */
245 for( int i = 0; i < (chroma422?8:4); i++ )
250 /* Start with the highest frequency coefficient... is this the best option? */
251 for( nz = 0, coeff = (chroma422?7:3); coeff >= 0; coeff-- )
253 int level = dct[coeff];
254 int sign = level>>31 | 1; /* dct[coeff] < 0 ? -1 : 1 */
258 dct[coeff] = level - sign;
259 if( optimize_chroma_round( dct_orig, dct, dequant_mf, chroma422 ) )
272 static int optimize_chroma_2x2_dc( dctcoef dct[4], int dequant_mf )
274 return optimize_chroma_dc_internal( dct, dequant_mf, 0 );
277 static int optimize_chroma_2x4_dc( dctcoef dct[8], int dequant_mf )
279 return optimize_chroma_dc_internal( dct, dequant_mf, 1 );
282 static void x264_denoise_dct( dctcoef *dct, uint32_t *sum, udctcoef *offset, int size )
284 for( int i = 0; i < size; i++ )
287 int sign = level>>31;
288 level = (level+sign)^sign;
291 dct[i] = level<0 ? 0 : (level^sign)-sign;
296 * x264_mb_decimate_score: given dct coeffs it returns a score to see if we could empty this dct coeffs
297 * to 0 (low score means set it to null)
298 * Used in inter macroblock (luma and chroma)
299 * luma: for a 8x8 block: if score < 4 -> null
300 * for the complete mb: if score < 6 -> null
301 * chroma: for the complete mb: if score < 7 -> null
304 const uint8_t x264_decimate_table4[16] =
306 3,2,2,1,1,1,0,0,0,0,0,0,0,0,0,0
308 const uint8_t x264_decimate_table8[64] =
310 3,3,3,3,2,2,2,2,2,2,2,2,1,1,1,1,
311 1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,
312 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
313 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
316 static int ALWAYS_INLINE x264_decimate_score_internal( dctcoef *dct, int i_max )
318 const uint8_t *ds_table = (i_max == 64) ? x264_decimate_table8 : x264_decimate_table4;
322 while( idx >= 0 && dct[idx] == 0 )
328 if( (unsigned)(dct[idx--] + 1) > 2 )
332 while( idx >= 0 && dct[idx] == 0 )
337 i_score += ds_table[i_run];
343 static int x264_decimate_score15( dctcoef *dct )
345 return x264_decimate_score_internal( dct+1, 15 );
347 static int x264_decimate_score16( dctcoef *dct )
349 return x264_decimate_score_internal( dct, 16 );
351 static int x264_decimate_score64( dctcoef *dct )
353 return x264_decimate_score_internal( dct, 64 );
357 static int x264_coeff_last##num( dctcoef *l )\
360 while( i_last >= 0 && l[i_last] == 0 )\
371 #define level_run(num)\
372 static int x264_coeff_level_run##num( dctcoef *dct, x264_run_level_t *runlevel )\
374 int i_last = runlevel->last = x264_coeff_last##num(dct);\
380 runlevel->level[i_total] = dct[i_last];\
381 mask |= 1 << (i_last);\
382 while( --i_last >= 0 && dct[i_last] == 0 )\
384 runlevel->run[i_total++] = r;\
385 } while( i_last >= 0 );\
386 runlevel->mask = mask;\
396 #define INIT_TRELLIS(cpu)\
397 pf->trellis_cabac_4x4 = x264_trellis_cabac_4x4_##cpu;\
398 pf->trellis_cabac_8x8 = x264_trellis_cabac_8x8_##cpu;\
399 pf->trellis_cabac_4x4_psy = x264_trellis_cabac_4x4_psy_##cpu;\
400 pf->trellis_cabac_8x8_psy = x264_trellis_cabac_8x8_psy_##cpu;\
401 pf->trellis_cabac_dc = x264_trellis_cabac_dc_##cpu;\
402 pf->trellis_cabac_chroma_422_dc = x264_trellis_cabac_chroma_422_dc_##cpu;
404 #define INIT_TRELLIS(...)
407 void x264_quant_init( x264_t *h, int cpu, x264_quant_function_t *pf )
409 pf->quant_8x8 = quant_8x8;
410 pf->quant_4x4 = quant_4x4;
411 pf->quant_4x4_dc = quant_4x4_dc;
412 pf->quant_2x2_dc = quant_2x2_dc;
414 pf->dequant_4x4 = dequant_4x4;
415 pf->dequant_4x4_dc = dequant_4x4_dc;
416 pf->dequant_8x8 = dequant_8x8;
418 pf->idct_dequant_2x4_dc = idct_dequant_2x4_dc;
419 pf->idct_dequant_2x4_dconly = idct_dequant_2x4_dconly;
421 pf->optimize_chroma_2x2_dc = optimize_chroma_2x2_dc;
422 pf->optimize_chroma_2x4_dc = optimize_chroma_2x4_dc;
424 pf->denoise_dct = x264_denoise_dct;
425 pf->decimate_score15 = x264_decimate_score15;
426 pf->decimate_score16 = x264_decimate_score16;
427 pf->decimate_score64 = x264_decimate_score64;
429 pf->coeff_last4 = x264_coeff_last4;
430 pf->coeff_last8 = x264_coeff_last8;
431 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15;
432 pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16;
433 pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64;
434 pf->coeff_level_run4 = x264_coeff_level_run4;
435 pf->coeff_level_run8 = x264_coeff_level_run8;
436 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15;
437 pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16;
441 INIT_TRELLIS( sse2 );
442 if( cpu&X264_CPU_MMX2 )
445 pf->denoise_dct = x264_denoise_dct_mmx;
446 pf->decimate_score15 = x264_decimate_score15_mmx2;
447 pf->decimate_score16 = x264_decimate_score16_mmx2;
448 if( cpu&X264_CPU_SLOW_CTZ )
450 pf->decimate_score15 = x264_decimate_score15_mmx2_slowctz;
451 pf->decimate_score16 = x264_decimate_score16_mmx2_slowctz;
453 pf->decimate_score64 = x264_decimate_score64_mmx2;
454 pf->coeff_last8 = x264_coeff_last8_mmx2;
455 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_mmx2;
456 pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16_mmx2;
457 pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64_mmx2;
458 pf->coeff_level_run8 = x264_coeff_level_run8_mmx2;
459 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_mmx2;
460 pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16_mmx2;
462 pf->coeff_last4 = x264_coeff_last4_mmx2;
463 pf->coeff_level_run4 = x264_coeff_level_run4_mmx2;
464 if( cpu&X264_CPU_LZCNT )
465 pf->coeff_level_run4 = x264_coeff_level_run4_mmx2_lzcnt;
467 if( cpu&X264_CPU_SSE2 )
469 pf->quant_4x4 = x264_quant_4x4_sse2;
470 pf->quant_8x8 = x264_quant_8x8_sse2;
471 pf->quant_2x2_dc = x264_quant_2x2_dc_sse2;
472 pf->quant_4x4_dc = x264_quant_4x4_dc_sse2;
473 pf->dequant_4x4 = x264_dequant_4x4_sse2;
474 pf->dequant_8x8 = x264_dequant_8x8_sse2;
475 pf->dequant_4x4_dc = x264_dequant_4x4dc_sse2;
476 pf->denoise_dct = x264_denoise_dct_sse2;
477 pf->decimate_score15 = x264_decimate_score15_sse2;
478 pf->decimate_score16 = x264_decimate_score16_sse2;
479 pf->decimate_score64 = x264_decimate_score64_sse2;
480 if( cpu&X264_CPU_SLOW_CTZ )
482 pf->decimate_score15 = x264_decimate_score15_sse2_slowctz;
483 pf->decimate_score16 = x264_decimate_score16_sse2_slowctz;
485 pf->coeff_last8 = x264_coeff_last8_sse2;
486 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2;
487 pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2;
488 pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2;
489 pf->coeff_level_run8 = x264_coeff_level_run8_sse2;
490 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2;
491 pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2;
492 if( cpu&X264_CPU_LZCNT )
494 pf->coeff_last4 = x264_coeff_last4_mmx2_lzcnt;
495 pf->coeff_last8 = x264_coeff_last8_sse2_lzcnt;
496 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2_lzcnt;
497 pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2_lzcnt;
498 pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2_lzcnt;
499 pf->coeff_level_run8 = x264_coeff_level_run8_sse2_lzcnt;
500 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2_lzcnt;
501 pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2_lzcnt;
504 if( cpu&X264_CPU_SSSE3 )
506 pf->quant_4x4 = x264_quant_4x4_ssse3;
507 pf->quant_8x8 = x264_quant_8x8_ssse3;
508 pf->quant_2x2_dc = x264_quant_2x2_dc_ssse3;
509 pf->quant_4x4_dc = x264_quant_4x4_dc_ssse3;
510 pf->denoise_dct = x264_denoise_dct_ssse3;
511 pf->decimate_score15 = x264_decimate_score15_ssse3;
512 pf->decimate_score16 = x264_decimate_score16_ssse3;
513 if( cpu&X264_CPU_SLOW_CTZ )
515 pf->decimate_score15 = x264_decimate_score15_ssse3_slowctz;
516 pf->decimate_score16 = x264_decimate_score16_ssse3_slowctz;
518 pf->decimate_score64 = x264_decimate_score64_ssse3;
519 INIT_TRELLIS( ssse3 );
521 if( cpu&X264_CPU_SSE4 )
523 pf->quant_2x2_dc = x264_quant_2x2_dc_sse4;
524 pf->quant_4x4_dc = x264_quant_4x4_dc_sse4;
525 pf->quant_4x4 = x264_quant_4x4_sse4;
526 pf->quant_8x8 = x264_quant_8x8_sse4;
528 if( cpu&X264_CPU_AVX )
530 pf->denoise_dct = x264_denoise_dct_avx;
532 if( cpu&X264_CPU_XOP )
534 pf->dequant_4x4_dc = x264_dequant_4x4dc_xop;
535 if( h->param.i_cqm_preset != X264_CQM_FLAT )
537 pf->dequant_4x4 = x264_dequant_4x4_xop;
538 pf->dequant_8x8 = x264_dequant_8x8_xop;
542 #else // !HIGH_BIT_DEPTH
544 INIT_TRELLIS( sse2 );
545 if( cpu&X264_CPU_MMX )
548 pf->quant_4x4 = x264_quant_4x4_mmx;
549 pf->quant_8x8 = x264_quant_8x8_mmx;
550 pf->dequant_4x4 = x264_dequant_4x4_mmx;
551 pf->dequant_4x4_dc = x264_dequant_4x4dc_mmx2;
552 pf->dequant_8x8 = x264_dequant_8x8_mmx;
553 if( h->param.i_cqm_preset == X264_CQM_FLAT )
555 pf->dequant_4x4 = x264_dequant_4x4_flat16_mmx;
556 pf->dequant_8x8 = x264_dequant_8x8_flat16_mmx;
558 pf->denoise_dct = x264_denoise_dct_mmx;
562 if( cpu&X264_CPU_MMX2 )
564 pf->quant_2x2_dc = x264_quant_2x2_dc_mmx2;
566 pf->quant_4x4_dc = x264_quant_4x4_dc_mmx2;
567 pf->decimate_score15 = x264_decimate_score15_mmx2;
568 pf->decimate_score16 = x264_decimate_score16_mmx2;
569 if( cpu&X264_CPU_SLOW_CTZ )
571 pf->decimate_score15 = x264_decimate_score15_mmx2_slowctz;
572 pf->decimate_score16 = x264_decimate_score16_mmx2_slowctz;
574 pf->decimate_score64 = x264_decimate_score64_mmx2;
575 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_mmx2;
576 pf->coeff_last[ DCT_LUMA_4x4] = x264_coeff_last16_mmx2;
577 pf->coeff_last[ DCT_LUMA_8x8] = x264_coeff_last64_mmx2;
578 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_mmx2;
579 pf->coeff_level_run[ DCT_LUMA_4x4] = x264_coeff_level_run16_mmx2;
581 pf->coeff_last4 = x264_coeff_last4_mmx2;
582 pf->coeff_last8 = x264_coeff_last8_mmx2;
583 pf->coeff_level_run4 = x264_coeff_level_run4_mmx2;
584 pf->coeff_level_run8 = x264_coeff_level_run8_mmx2;
585 if( cpu&X264_CPU_LZCNT )
587 pf->coeff_last4 = x264_coeff_last4_mmx2_lzcnt;
588 pf->coeff_last8 = x264_coeff_last8_mmx2_lzcnt;
589 pf->coeff_level_run4 = x264_coeff_level_run4_mmx2_lzcnt;
590 pf->coeff_level_run8 = x264_coeff_level_run8_mmx2_lzcnt;
594 if( cpu&X264_CPU_SSE2 )
596 pf->quant_4x4_dc = x264_quant_4x4_dc_sse2;
597 pf->quant_4x4 = x264_quant_4x4_sse2;
598 pf->quant_8x8 = x264_quant_8x8_sse2;
599 pf->dequant_4x4 = x264_dequant_4x4_sse2;
600 pf->dequant_4x4_dc = x264_dequant_4x4dc_sse2;
601 pf->dequant_8x8 = x264_dequant_8x8_sse2;
602 if( h->param.i_cqm_preset == X264_CQM_FLAT )
604 pf->dequant_4x4 = x264_dequant_4x4_flat16_sse2;
605 pf->dequant_8x8 = x264_dequant_8x8_flat16_sse2;
607 pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_sse2;
608 pf->denoise_dct = x264_denoise_dct_sse2;
609 pf->decimate_score15 = x264_decimate_score15_sse2;
610 pf->decimate_score16 = x264_decimate_score16_sse2;
611 pf->decimate_score64 = x264_decimate_score64_sse2;
612 if( cpu&X264_CPU_SLOW_CTZ )
614 pf->decimate_score15 = x264_decimate_score15_sse2_slowctz;
615 pf->decimate_score16 = x264_decimate_score16_sse2_slowctz;
617 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2;
618 pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2;
619 pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2;
620 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2;
621 pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2;
622 if( cpu&X264_CPU_LZCNT )
624 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_sse2_lzcnt;
625 pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_sse2_lzcnt;
626 pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_sse2_lzcnt;
627 pf->coeff_level_run[ DCT_LUMA_AC] = x264_coeff_level_run15_sse2_lzcnt;
628 pf->coeff_level_run[DCT_LUMA_4x4] = x264_coeff_level_run16_sse2_lzcnt;
632 if( cpu&X264_CPU_SSSE3 )
634 pf->quant_2x2_dc = x264_quant_2x2_dc_ssse3;
635 pf->quant_4x4_dc = x264_quant_4x4_dc_ssse3;
636 pf->quant_4x4 = x264_quant_4x4_ssse3;
637 pf->quant_8x8 = x264_quant_8x8_ssse3;
638 pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_ssse3;
639 pf->denoise_dct = x264_denoise_dct_ssse3;
640 pf->decimate_score15 = x264_decimate_score15_ssse3;
641 pf->decimate_score16 = x264_decimate_score16_ssse3;
642 if( cpu&X264_CPU_SLOW_CTZ )
644 pf->decimate_score15 = x264_decimate_score15_ssse3_slowctz;
645 pf->decimate_score16 = x264_decimate_score16_ssse3_slowctz;
647 pf->decimate_score64 = x264_decimate_score64_ssse3;
648 INIT_TRELLIS( ssse3 );
651 if( cpu&X264_CPU_SSE4 )
653 pf->quant_4x4_dc = x264_quant_4x4_dc_sse4;
654 pf->quant_4x4 = x264_quant_4x4_sse4;
655 pf->quant_8x8 = x264_quant_8x8_sse4;
656 pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_sse4;
659 if( cpu&X264_CPU_AVX )
661 pf->dequant_4x4_dc = x264_dequant_4x4dc_avx;
662 if( h->param.i_cqm_preset != X264_CQM_FLAT )
664 pf->dequant_4x4 = x264_dequant_4x4_avx;
665 pf->dequant_8x8 = x264_dequant_8x8_avx;
667 pf->optimize_chroma_2x2_dc = x264_optimize_chroma_2x2_dc_avx;
668 pf->denoise_dct = x264_denoise_dct_avx;
671 if( cpu&X264_CPU_XOP )
673 if( h->param.i_cqm_preset != X264_CQM_FLAT )
675 pf->dequant_4x4 = x264_dequant_4x4_xop;
676 pf->dequant_8x8 = x264_dequant_8x8_xop;
682 if( cpu&X264_CPU_ALTIVEC ) {
683 pf->quant_2x2_dc = x264_quant_2x2_dc_altivec;
684 pf->quant_4x4_dc = x264_quant_4x4_dc_altivec;
685 pf->quant_4x4 = x264_quant_4x4_altivec;
686 pf->quant_8x8 = x264_quant_8x8_altivec;
688 pf->dequant_4x4 = x264_dequant_4x4_altivec;
689 pf->dequant_8x8 = x264_dequant_8x8_altivec;
694 if( cpu&X264_CPU_ARMV6 )
695 pf->coeff_last4 = x264_coeff_last4_arm;
697 if( cpu&X264_CPU_NEON )
699 pf->quant_2x2_dc = x264_quant_2x2_dc_neon;
700 pf->quant_4x4 = x264_quant_4x4_neon;
701 pf->quant_4x4_dc = x264_quant_4x4_dc_neon;
702 pf->quant_8x8 = x264_quant_8x8_neon;
703 pf->dequant_4x4 = x264_dequant_4x4_neon;
704 pf->dequant_4x4_dc = x264_dequant_4x4_dc_neon;
705 pf->dequant_8x8 = x264_dequant_8x8_neon;
706 pf->coeff_last[ DCT_LUMA_AC] = x264_coeff_last15_neon;
707 pf->coeff_last[DCT_LUMA_4x4] = x264_coeff_last16_neon;
708 pf->coeff_last[DCT_LUMA_8x8] = x264_coeff_last64_neon;
711 #endif // HIGH_BIT_DEPTH
712 pf->coeff_last[DCT_LUMA_DC] = pf->coeff_last[DCT_CHROMAU_DC] = pf->coeff_last[DCT_CHROMAV_DC] =
713 pf->coeff_last[DCT_CHROMAU_4x4] = pf->coeff_last[DCT_CHROMAV_4x4] = pf->coeff_last[DCT_LUMA_4x4];
714 pf->coeff_last[DCT_CHROMA_AC] = pf->coeff_last[DCT_CHROMAU_AC] =
715 pf->coeff_last[DCT_CHROMAV_AC] = pf->coeff_last[DCT_LUMA_AC];
716 pf->coeff_last[DCT_CHROMAU_8x8] = pf->coeff_last[DCT_CHROMAV_8x8] = pf->coeff_last[DCT_LUMA_8x8];
718 pf->coeff_level_run[DCT_LUMA_DC] = pf->coeff_level_run[DCT_CHROMAU_DC] = pf->coeff_level_run[DCT_CHROMAV_DC] =
719 pf->coeff_level_run[DCT_CHROMAU_4x4] = pf->coeff_level_run[DCT_CHROMAV_4x4] = pf->coeff_level_run[DCT_LUMA_4x4];
720 pf->coeff_level_run[DCT_CHROMA_AC] = pf->coeff_level_run[DCT_CHROMAU_AC] =
721 pf->coeff_level_run[DCT_CHROMAV_AC] = pf->coeff_level_run[DCT_LUMA_AC];