1 ;*****************************************************************************
2 ;* sad16-a.asm: x86 high depth sad functions
3 ;*****************************************************************************
4 ;* Copyright (C) 2010-2012 x264 project
6 ;* Authors: Oskar Arvidsson <oskar@irock.se>
8 ;* This program is free software; you can redistribute it and/or modify
9 ;* it under the terms of the GNU General Public License as published by
10 ;* the Free Software Foundation; either version 2 of the License, or
11 ;* (at your option) any later version.
13 ;* This program is distributed in the hope that it will be useful,
14 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
15 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 ;* GNU General Public License for more details.
18 ;* You should have received a copy of the GNU General Public License
19 ;* along with this program; if not, write to the Free Software
20 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
22 ;* This program is also available under a commercial proprietary license.
23 ;* For more information, contact us at licensing@x264.com.
24 ;*****************************************************************************
27 %include "x86util.asm"
35 ;=============================================================================
37 ;=============================================================================
39 %macro SAD_INC_1x16P_MMX 0
48 ABSW2 m1, m2, m1, m2, m5, m6
49 ABSW2 m3, m4, m3, m4, m7, m5
58 %macro SAD_INC_2x8P_MMX 0
67 ABSW2 m1, m2, m1, m2, m5, m6
68 ABSW2 m3, m4, m3, m4, m7, m5
77 %macro SAD_INC_2x4P_MMX 0
82 ABSW2 m1, m2, m1, m2, m3, m4
89 ;-----------------------------------------------------------------------------
90 ; int pixel_sad_NxM( uint16_t *, intptr_t, uint16_t *, intptr_t )
91 ;-----------------------------------------------------------------------------
93 cglobal pixel_sad_%1x%2, 4,4
119 ;=============================================================================
121 ;=============================================================================
123 %macro SAD_INC_2x16P_XMM 0
126 movu m3, [r2+2*r3+ 0]
127 movu m4, [r2+2*r3+16]
130 psubw m3, [r0+2*r1+ 0]
131 psubw m4, [r0+2*r1+16]
132 ABSW2 m1, m2, m1, m2, m5, m6
135 ABSW2 m3, m4, m3, m4, m7, m5
142 %macro SAD_INC_2x8P_XMM 0
147 ABSW2 m1, m2, m1, m2, m3, m4
154 ;-----------------------------------------------------------------------------
155 ; int pixel_sad_NxM( uint16_t *, intptr_t, uint16_t *, intptr_t )
156 ;-----------------------------------------------------------------------------
158 cglobal pixel_sad_%1x%2, 4,4,8
174 INIT_XMM sse2, aligned
185 INIT_XMM ssse3, aligned
191 ;=============================================================================
193 ;=============================================================================
195 %macro SAD_X3_INC_P 0
196 add r0, 4*FENC_STRIDE
202 %macro SAD_X3_ONE_START 0
210 ABSW2 m0, m1, m0, m1, m4, m5
222 ABSW2 m3, m4, m3, m4, m7, m6
230 %if mmsize == 8 && %1*%2 == 256
252 %macro SAD_X4_INC_P 0
253 add r0, 4*FENC_STRIDE
260 %macro SAD_X4_ONE_START 0
270 ABSW2 m0, m1, m0, m1, m5, m6
271 ABSW2 m2, m3, m2, m3, m4, m7
285 ABSW2 m5, m6, m5, m6, m9, m10
286 ABSW2 m7, m8, m7, m8, m9, m10
306 %else ; num_mmregs == 8 && !ssse3
317 ABSW2 m5, m6, m5, m6, m7, m4
324 %if mmsize == 8 && %1*%2 == 256
346 SAD_X%1_ONE x*mmsize, x*mmsize
347 SAD_X%1_ONE 2*FENC_STRIDE+x*mmsize, 2*%2+x*mmsize
353 cglobal pixel_vsad, 3,3,8
357 mova m3, [r0+2*r1+16]
361 ABSW2 m0, m1, m0, m1, m4, m5
369 mova m7, [r0+2*r1+16]
389 HADDW m0, m1 ; max sum: 62(pixel diffs)*511(pixel_max)=31682
391 HADDUW m0, m1 ; max sum: 62(pixel diffs)*1023(pixel_max)=63426
403 ;-----------------------------------------------------------------------------
404 ; void pixel_sad_xK_MxN( uint16_t *fenc, uint16_t *pix0, uint16_t *pix1,
405 ; uint16_t *pix2, intptr_t i_stride, int scores[3] )
406 ;-----------------------------------------------------------------------------
408 cglobal pixel_sad_x%1_%2x%3, 6,7,XMM_REGS
410 %xdefine STRIDE r %+ regnum
413 SAD_X%1_ONE 2*FENC_STRIDE, 2*STRIDE
414 SAD_X_2xNP %1, STRIDE, 1, %2/(mmsize/2)-1
417 SAD_X_2xNP %1, STRIDE, 0, %2/(mmsize/2)
472 ;-----------------------------------------------------------------------------
473 ; void intra_sad_x3_4x4( uint16_t *fenc, uint16_t *fdec, int res[3] );
474 ;-----------------------------------------------------------------------------
476 %macro INTRA_SAD_X3_4x4 0
477 cglobal intra_sad_x3_4x4, 3,3,7
478 movq m0, [r1-1*FDEC_STRIDEB]
479 movq m1, [r0+0*FENC_STRIDEB]
480 movq m2, [r0+2*FENC_STRIDEB]
481 pshuflw m6, m0, q1032
483 pshuflw m5, m6, q2301
485 punpcklqdq m6, m6 ;A+B+C+D 8 times
487 movhps m1, [r0+1*FENC_STRIDEB]
488 movhps m2, [r0+3*FENC_STRIDEB]
495 movd [r2], m0 ;V prediction cost
496 movd m3, [r1+0*FDEC_STRIDEB-4]
497 movhps m3, [r1+1*FDEC_STRIDEB-8]
498 movd m4, [r1+2*FDEC_STRIDEB-4]
499 movhps m4, [r1+3*FDEC_STRIDEB-8]
500 pshufhw m3, m3, q3333
501 pshufhw m4, m4, q3333
502 pshuflw m3, m3, q1111 ; FF FF EE EE
503 pshuflw m4, m4, q1111 ; HH HH GG GG
522 movd [r2+8], m5 ;DC prediction cost
523 movd [r2+4], m1 ;H prediction cost
534 ;-----------------------------------------------------------------------------
535 ; void intra_sad_x3_8x8( pixel *fenc, pixel edge[36], int res[3] );
536 ;-----------------------------------------------------------------------------
547 %macro INTRA_SAD_HVDC_ITER 2
548 mova m4, [r0+(%1-4)*FENC_STRIDEB]
551 ACCUM paddw, 1, 4, %1
552 mova m4, [r0+(%1-4)*FENC_STRIDEB]
555 ACCUM paddw, 2, 4, %1
557 psubw m5, [r0+(%1-4)*FENC_STRIDEB]
559 ACCUM paddw, 3, 5, %1
562 %macro INTRA_SAD_X3_8x8 0
563 cglobal intra_sad_x3_8x8, 3,3,8
564 add r0, 4*FENC_STRIDEB
565 movu m0, [r1+7*SIZEOF_PIXEL]
566 mova m6, [r1+16*SIZEOF_PIXEL] ;V prediction
574 INTRA_SAD_HVDC_ITER 0, q3333
575 INTRA_SAD_HVDC_ITER 1, q2222
576 INTRA_SAD_HVDC_ITER 2, q1111
577 INTRA_SAD_HVDC_ITER 3, q0000
578 movq m7, [r1+7*SIZEOF_PIXEL]
580 INTRA_SAD_HVDC_ITER 4, q3333
581 INTRA_SAD_HVDC_ITER 5, q2222
582 INTRA_SAD_HVDC_ITER 6, q1111
583 INTRA_SAD_HVDC_ITER 7, q0000