3 * Copyright (c) 2000, 2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>
8 * This file is part of Libav.
10 * Libav is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * Libav is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with Libav; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
30 #include "libavutil/attributes.h"
31 #include "libavutil/imgutils.h"
33 #include "copy_block.h"
36 #include "simple_idct.h"
39 #include "imgconvert.h"
41 #include "mpegvideo.h"
44 uint32_t ff_squareTbl[512] = {0, };
47 #include "dsputil_template.c"
51 #include "dsputil_template.c"
53 // 0x7f7f7f7f or 0x7f7f7f7f7f7f7f7f or whatever, depending on the cpu's native arithmetic size
54 #define pb_7f (~0UL/255 * 0x7f)
55 #define pb_80 (~0UL/255 * 0x80)
57 /* Specific zigzag scan for 248 idct. NOTE that unlike the
58 specification, we interleave the fields */
59 const uint8_t ff_zigzag248_direct[64] = {
60 0, 8, 1, 9, 16, 24, 2, 10,
61 17, 25, 32, 40, 48, 56, 33, 41,
62 18, 26, 3, 11, 4, 12, 19, 27,
63 34, 42, 49, 57, 50, 58, 35, 43,
64 20, 28, 5, 13, 6, 14, 21, 29,
65 36, 44, 51, 59, 52, 60, 37, 45,
66 22, 30, 7, 15, 23, 31, 38, 46,
67 53, 61, 54, 62, 39, 47, 55, 63,
70 const uint8_t ff_alternate_horizontal_scan[64] = {
71 0, 1, 2, 3, 8, 9, 16, 17,
72 10, 11, 4, 5, 6, 7, 15, 14,
73 13, 12, 19, 18, 24, 25, 32, 33,
74 26, 27, 20, 21, 22, 23, 28, 29,
75 30, 31, 34, 35, 40, 41, 48, 49,
76 42, 43, 36, 37, 38, 39, 44, 45,
77 46, 47, 50, 51, 56, 57, 58, 59,
78 52, 53, 54, 55, 60, 61, 62, 63,
81 const uint8_t ff_alternate_vertical_scan[64] = {
82 0, 8, 16, 24, 1, 9, 2, 10,
83 17, 25, 32, 40, 48, 56, 57, 49,
84 41, 33, 26, 18, 3, 11, 4, 12,
85 19, 27, 34, 42, 50, 58, 35, 43,
86 51, 59, 20, 28, 5, 13, 6, 14,
87 21, 29, 36, 44, 52, 60, 37, 45,
88 53, 61, 22, 30, 7, 15, 23, 31,
89 38, 46, 54, 62, 39, 47, 55, 63,
92 /* Input permutation for the simple_idct_mmx */
93 static const uint8_t simple_mmx_permutation[64]={
94 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
95 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
96 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
97 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
98 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
99 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
100 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
101 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
104 static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7};
106 av_cold void ff_init_scantable(uint8_t *permutation, ScanTable *st,
107 const uint8_t *src_scantable)
112 st->scantable= src_scantable;
116 j = src_scantable[i];
117 st->permutated[i] = permutation[j];
123 j = st->permutated[i];
125 st->raster_end[i]= end;
129 av_cold void ff_init_scantable_permutation(uint8_t *idct_permutation,
130 int idct_permutation_type)
134 switch(idct_permutation_type){
135 case FF_NO_IDCT_PERM:
137 idct_permutation[i]= i;
139 case FF_LIBMPEG2_IDCT_PERM:
141 idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
143 case FF_SIMPLE_IDCT_PERM:
145 idct_permutation[i]= simple_mmx_permutation[i];
147 case FF_TRANSPOSE_IDCT_PERM:
149 idct_permutation[i]= ((i&7)<<3) | (i>>3);
151 case FF_PARTTRANS_IDCT_PERM:
153 idct_permutation[i]= (i&0x24) | ((i&3)<<3) | ((i>>3)&3);
155 case FF_SSE2_IDCT_PERM:
157 idct_permutation[i]= (i&0x38) | idct_sse2_row_perm[i&7];
160 av_log(NULL, AV_LOG_ERROR, "Internal error, IDCT permutation not set\n");
164 static int pix_sum_c(uint8_t * pix, int line_size)
169 for (i = 0; i < 16; i++) {
170 for (j = 0; j < 16; j += 8) {
181 pix += line_size - 16;
186 static int pix_norm1_c(uint8_t * pix, int line_size)
189 uint32_t *sq = ff_squareTbl + 256;
192 for (i = 0; i < 16; i++) {
193 for (j = 0; j < 16; j += 8) {
205 register uint64_t x=*(uint64_t*)pix;
207 s += sq[(x>>8)&0xff];
208 s += sq[(x>>16)&0xff];
209 s += sq[(x>>24)&0xff];
210 s += sq[(x>>32)&0xff];
211 s += sq[(x>>40)&0xff];
212 s += sq[(x>>48)&0xff];
213 s += sq[(x>>56)&0xff];
215 register uint32_t x=*(uint32_t*)pix;
217 s += sq[(x>>8)&0xff];
218 s += sq[(x>>16)&0xff];
219 s += sq[(x>>24)&0xff];
220 x=*(uint32_t*)(pix+4);
222 s += sq[(x>>8)&0xff];
223 s += sq[(x>>16)&0xff];
224 s += sq[(x>>24)&0xff];
229 pix += line_size - 16;
234 static void bswap_buf(uint32_t *dst, const uint32_t *src, int w){
237 for(i=0; i+8<=w; i+=8){
238 dst[i+0]= av_bswap32(src[i+0]);
239 dst[i+1]= av_bswap32(src[i+1]);
240 dst[i+2]= av_bswap32(src[i+2]);
241 dst[i+3]= av_bswap32(src[i+3]);
242 dst[i+4]= av_bswap32(src[i+4]);
243 dst[i+5]= av_bswap32(src[i+5]);
244 dst[i+6]= av_bswap32(src[i+6]);
245 dst[i+7]= av_bswap32(src[i+7]);
248 dst[i+0]= av_bswap32(src[i+0]);
252 static void bswap16_buf(uint16_t *dst, const uint16_t *src, int len)
255 *dst++ = av_bswap16(*src++);
258 static int sse4_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)
261 uint32_t *sq = ff_squareTbl + 256;
264 for (i = 0; i < h; i++) {
265 s += sq[pix1[0] - pix2[0]];
266 s += sq[pix1[1] - pix2[1]];
267 s += sq[pix1[2] - pix2[2]];
268 s += sq[pix1[3] - pix2[3]];
275 static int sse8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)
278 uint32_t *sq = ff_squareTbl + 256;
281 for (i = 0; i < h; i++) {
282 s += sq[pix1[0] - pix2[0]];
283 s += sq[pix1[1] - pix2[1]];
284 s += sq[pix1[2] - pix2[2]];
285 s += sq[pix1[3] - pix2[3]];
286 s += sq[pix1[4] - pix2[4]];
287 s += sq[pix1[5] - pix2[5]];
288 s += sq[pix1[6] - pix2[6]];
289 s += sq[pix1[7] - pix2[7]];
296 static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
299 uint32_t *sq = ff_squareTbl + 256;
302 for (i = 0; i < h; i++) {
303 s += sq[pix1[ 0] - pix2[ 0]];
304 s += sq[pix1[ 1] - pix2[ 1]];
305 s += sq[pix1[ 2] - pix2[ 2]];
306 s += sq[pix1[ 3] - pix2[ 3]];
307 s += sq[pix1[ 4] - pix2[ 4]];
308 s += sq[pix1[ 5] - pix2[ 5]];
309 s += sq[pix1[ 6] - pix2[ 6]];
310 s += sq[pix1[ 7] - pix2[ 7]];
311 s += sq[pix1[ 8] - pix2[ 8]];
312 s += sq[pix1[ 9] - pix2[ 9]];
313 s += sq[pix1[10] - pix2[10]];
314 s += sq[pix1[11] - pix2[11]];
315 s += sq[pix1[12] - pix2[12]];
316 s += sq[pix1[13] - pix2[13]];
317 s += sq[pix1[14] - pix2[14]];
318 s += sq[pix1[15] - pix2[15]];
326 static void diff_pixels_c(int16_t *restrict block, const uint8_t *s1,
327 const uint8_t *s2, int stride){
330 /* read the pixels */
332 block[0] = s1[0] - s2[0];
333 block[1] = s1[1] - s2[1];
334 block[2] = s1[2] - s2[2];
335 block[3] = s1[3] - s2[3];
336 block[4] = s1[4] - s2[4];
337 block[5] = s1[5] - s2[5];
338 block[6] = s1[6] - s2[6];
339 block[7] = s1[7] - s2[7];
347 static void put_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels,
352 /* read the pixels */
354 pixels[0] = av_clip_uint8(block[0]);
355 pixels[1] = av_clip_uint8(block[1]);
356 pixels[2] = av_clip_uint8(block[2]);
357 pixels[3] = av_clip_uint8(block[3]);
358 pixels[4] = av_clip_uint8(block[4]);
359 pixels[5] = av_clip_uint8(block[5]);
360 pixels[6] = av_clip_uint8(block[6]);
361 pixels[7] = av_clip_uint8(block[7]);
368 static void put_signed_pixels_clamped_c(const int16_t *block,
369 uint8_t *restrict pixels,
374 for (i = 0; i < 8; i++) {
375 for (j = 0; j < 8; j++) {
378 else if (*block > 127)
381 *pixels = (uint8_t)(*block + 128);
385 pixels += (line_size - 8);
389 static void add_pixels8_c(uint8_t *restrict pixels,
396 pixels[0] += block[0];
397 pixels[1] += block[1];
398 pixels[2] += block[2];
399 pixels[3] += block[3];
400 pixels[4] += block[4];
401 pixels[5] += block[5];
402 pixels[6] += block[6];
403 pixels[7] += block[7];
409 static void add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels,
414 /* read the pixels */
416 pixels[0] = av_clip_uint8(pixels[0] + block[0]);
417 pixels[1] = av_clip_uint8(pixels[1] + block[1]);
418 pixels[2] = av_clip_uint8(pixels[2] + block[2]);
419 pixels[3] = av_clip_uint8(pixels[3] + block[3]);
420 pixels[4] = av_clip_uint8(pixels[4] + block[4]);
421 pixels[5] = av_clip_uint8(pixels[5] + block[5]);
422 pixels[6] = av_clip_uint8(pixels[6] + block[6]);
423 pixels[7] = av_clip_uint8(pixels[7] + block[7]);
429 static int sum_abs_dctelem_c(int16_t *block)
433 sum+= FFABS(block[i]);
437 static void fill_block16_c(uint8_t *block, uint8_t value, int line_size, int h)
441 for (i = 0; i < h; i++) {
442 memset(block, value, 16);
447 static void fill_block8_c(uint8_t *block, uint8_t value, int line_size, int h)
451 for (i = 0; i < h; i++) {
452 memset(block, value, 8);
457 #define avg2(a,b) ((a+b+1)>>1)
458 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
460 static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)
462 const int A=(16-x16)*(16-y16);
463 const int B=( x16)*(16-y16);
464 const int C=(16-x16)*( y16);
465 const int D=( x16)*( y16);
470 dst[0]= (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + rounder)>>8;
471 dst[1]= (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + rounder)>>8;
472 dst[2]= (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + rounder)>>8;
473 dst[3]= (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + rounder)>>8;
474 dst[4]= (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + rounder)>>8;
475 dst[5]= (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + rounder)>>8;
476 dst[6]= (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + rounder)>>8;
477 dst[7]= (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + rounder)>>8;
483 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
484 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
487 const int s= 1<<shift;
497 for(x=0; x<8; x++){ //XXX FIXME optimize
498 int src_x, src_y, frac_x, frac_y, index;
507 if((unsigned)src_x < width){
508 if((unsigned)src_y < height){
509 index= src_x + src_y*stride;
510 dst[y*stride + x]= ( ( src[index ]*(s-frac_x)
511 + src[index +1]* frac_x )*(s-frac_y)
512 + ( src[index+stride ]*(s-frac_x)
513 + src[index+stride+1]* frac_x )* frac_y
516 index= src_x + av_clip(src_y, 0, height)*stride;
517 dst[y*stride + x]= ( ( src[index ]*(s-frac_x)
518 + src[index +1]* frac_x )*s
522 if((unsigned)src_y < height){
523 index= av_clip(src_x, 0, width) + src_y*stride;
524 dst[y*stride + x]= ( ( src[index ]*(s-frac_y)
525 + src[index+stride ]* frac_y )*s
528 index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride;
529 dst[y*stride + x]= src[index ];
541 static inline void put_tpel_pixels_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
543 case 2: put_pixels2_8_c (dst, src, stride, height); break;
544 case 4: put_pixels4_8_c (dst, src, stride, height); break;
545 case 8: put_pixels8_8_c (dst, src, stride, height); break;
546 case 16:put_pixels16_8_c(dst, src, stride, height); break;
550 static inline void put_tpel_pixels_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
552 for (i=0; i < height; i++) {
553 for (j=0; j < width; j++) {
554 dst[j] = (683*(2*src[j] + src[j+1] + 1)) >> 11;
561 static inline void put_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
563 for (i=0; i < height; i++) {
564 for (j=0; j < width; j++) {
565 dst[j] = (683*(src[j] + 2*src[j+1] + 1)) >> 11;
572 static inline void put_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
574 for (i=0; i < height; i++) {
575 for (j=0; j < width; j++) {
576 dst[j] = (683*(2*src[j] + src[j+stride] + 1)) >> 11;
583 static inline void put_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
585 for (i=0; i < height; i++) {
586 for (j=0; j < width; j++) {
587 dst[j] = (2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15;
594 static inline void put_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
596 for (i=0; i < height; i++) {
597 for (j=0; j < width; j++) {
598 dst[j] = (2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15;
605 static inline void put_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
607 for (i=0; i < height; i++) {
608 for (j=0; j < width; j++) {
609 dst[j] = (683*(src[j] + 2*src[j+stride] + 1)) >> 11;
616 static inline void put_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
618 for (i=0; i < height; i++) {
619 for (j=0; j < width; j++) {
620 dst[j] = (2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15;
627 static inline void put_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
629 for (i=0; i < height; i++) {
630 for (j=0; j < width; j++) {
631 dst[j] = (2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15;
638 static inline void avg_tpel_pixels_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
640 case 2: avg_pixels2_8_c (dst, src, stride, height); break;
641 case 4: avg_pixels4_8_c (dst, src, stride, height); break;
642 case 8: avg_pixels8_8_c (dst, src, stride, height); break;
643 case 16:avg_pixels16_8_c(dst, src, stride, height); break;
647 static inline void avg_tpel_pixels_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
649 for (i=0; i < height; i++) {
650 for (j=0; j < width; j++) {
651 dst[j] = (dst[j] + ((683*(2*src[j] + src[j+1] + 1)) >> 11) + 1) >> 1;
658 static inline void avg_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
660 for (i=0; i < height; i++) {
661 for (j=0; j < width; j++) {
662 dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+1] + 1)) >> 11) + 1) >> 1;
669 static inline void avg_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
671 for (i=0; i < height; i++) {
672 for (j=0; j < width; j++) {
673 dst[j] = (dst[j] + ((683*(2*src[j] + src[j+stride] + 1)) >> 11) + 1) >> 1;
680 static inline void avg_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
682 for (i=0; i < height; i++) {
683 for (j=0; j < width; j++) {
684 dst[j] = (dst[j] + ((2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
691 static inline void avg_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
693 for (i=0; i < height; i++) {
694 for (j=0; j < width; j++) {
695 dst[j] = (dst[j] + ((2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
702 static inline void avg_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
704 for (i=0; i < height; i++) {
705 for (j=0; j < width; j++) {
706 dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+stride] + 1)) >> 11) + 1) >> 1;
713 static inline void avg_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
715 for (i=0; i < height; i++) {
716 for (j=0; j < width; j++) {
717 dst[j] = (dst[j] + ((2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
724 static inline void avg_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
726 for (i=0; i < height; i++) {
727 for (j=0; j < width; j++) {
728 dst[j] = (dst[j] + ((2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
735 #define QPEL_MC(r, OPNAME, RND, OP) \
736 static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
737 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
741 OP(dst[0], (src[0]+src[1])*20 - (src[0]+src[2])*6 + (src[1]+src[3])*3 - (src[2]+src[4]));\
742 OP(dst[1], (src[1]+src[2])*20 - (src[0]+src[3])*6 + (src[0]+src[4])*3 - (src[1]+src[5]));\
743 OP(dst[2], (src[2]+src[3])*20 - (src[1]+src[4])*6 + (src[0]+src[5])*3 - (src[0]+src[6]));\
744 OP(dst[3], (src[3]+src[4])*20 - (src[2]+src[5])*6 + (src[1]+src[6])*3 - (src[0]+src[7]));\
745 OP(dst[4], (src[4]+src[5])*20 - (src[3]+src[6])*6 + (src[2]+src[7])*3 - (src[1]+src[8]));\
746 OP(dst[5], (src[5]+src[6])*20 - (src[4]+src[7])*6 + (src[3]+src[8])*3 - (src[2]+src[8]));\
747 OP(dst[6], (src[6]+src[7])*20 - (src[5]+src[8])*6 + (src[4]+src[8])*3 - (src[3]+src[7]));\
748 OP(dst[7], (src[7]+src[8])*20 - (src[6]+src[8])*6 + (src[5]+src[7])*3 - (src[4]+src[6]));\
754 static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
756 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
760 const int src0= src[0*srcStride];\
761 const int src1= src[1*srcStride];\
762 const int src2= src[2*srcStride];\
763 const int src3= src[3*srcStride];\
764 const int src4= src[4*srcStride];\
765 const int src5= src[5*srcStride];\
766 const int src6= src[6*srcStride];\
767 const int src7= src[7*srcStride];\
768 const int src8= src[8*srcStride];\
769 OP(dst[0*dstStride], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
770 OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
771 OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
772 OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
773 OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
774 OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
775 OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
776 OP(dst[7*dstStride], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
782 static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
783 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
788 OP(dst[ 0], (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]));\
789 OP(dst[ 1], (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]));\
790 OP(dst[ 2], (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]));\
791 OP(dst[ 3], (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]));\
792 OP(dst[ 4], (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]));\
793 OP(dst[ 5], (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 9]));\
794 OP(dst[ 6], (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 9])*3 - (src[ 3]+src[10]));\
795 OP(dst[ 7], (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 9])*6 + (src[ 5]+src[10])*3 - (src[ 4]+src[11]));\
796 OP(dst[ 8], (src[ 8]+src[ 9])*20 - (src[ 7]+src[10])*6 + (src[ 6]+src[11])*3 - (src[ 5]+src[12]));\
797 OP(dst[ 9], (src[ 9]+src[10])*20 - (src[ 8]+src[11])*6 + (src[ 7]+src[12])*3 - (src[ 6]+src[13]));\
798 OP(dst[10], (src[10]+src[11])*20 - (src[ 9]+src[12])*6 + (src[ 8]+src[13])*3 - (src[ 7]+src[14]));\
799 OP(dst[11], (src[11]+src[12])*20 - (src[10]+src[13])*6 + (src[ 9]+src[14])*3 - (src[ 8]+src[15]));\
800 OP(dst[12], (src[12]+src[13])*20 - (src[11]+src[14])*6 + (src[10]+src[15])*3 - (src[ 9]+src[16]));\
801 OP(dst[13], (src[13]+src[14])*20 - (src[12]+src[15])*6 + (src[11]+src[16])*3 - (src[10]+src[16]));\
802 OP(dst[14], (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]));\
803 OP(dst[15], (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]));\
809 static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
810 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
815 const int src0= src[0*srcStride];\
816 const int src1= src[1*srcStride];\
817 const int src2= src[2*srcStride];\
818 const int src3= src[3*srcStride];\
819 const int src4= src[4*srcStride];\
820 const int src5= src[5*srcStride];\
821 const int src6= src[6*srcStride];\
822 const int src7= src[7*srcStride];\
823 const int src8= src[8*srcStride];\
824 const int src9= src[9*srcStride];\
825 const int src10= src[10*srcStride];\
826 const int src11= src[11*srcStride];\
827 const int src12= src[12*srcStride];\
828 const int src13= src[13*srcStride];\
829 const int src14= src[14*srcStride];\
830 const int src15= src[15*srcStride];\
831 const int src16= src[16*srcStride];\
832 OP(dst[ 0*dstStride], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
833 OP(dst[ 1*dstStride], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
834 OP(dst[ 2*dstStride], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
835 OP(dst[ 3*dstStride], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
836 OP(dst[ 4*dstStride], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
837 OP(dst[ 5*dstStride], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
838 OP(dst[ 6*dstStride], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
839 OP(dst[ 7*dstStride], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
840 OP(dst[ 8*dstStride], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
841 OP(dst[ 9*dstStride], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
842 OP(dst[10*dstStride], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
843 OP(dst[11*dstStride], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
844 OP(dst[12*dstStride], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
845 OP(dst[13*dstStride], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
846 OP(dst[14*dstStride], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
847 OP(dst[15*dstStride], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
853 static void OPNAME ## qpel8_mc10_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
856 put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
857 OPNAME ## pixels8_l2_8(dst, src, half, stride, stride, 8, 8);\
860 static void OPNAME ## qpel8_mc20_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
862 OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
865 static void OPNAME ## qpel8_mc30_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
868 put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
869 OPNAME ## pixels8_l2_8(dst, src+1, half, stride, stride, 8, 8);\
872 static void OPNAME ## qpel8_mc01_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
876 copy_block9(full, src, 16, stride, 9);\
877 put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
878 OPNAME ## pixels8_l2_8(dst, full, half, stride, 16, 8, 8);\
881 static void OPNAME ## qpel8_mc02_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
884 copy_block9(full, src, 16, stride, 9);\
885 OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
888 static void OPNAME ## qpel8_mc03_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
892 copy_block9(full, src, 16, stride, 9);\
893 put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
894 OPNAME ## pixels8_l2_8(dst, full+16, half, stride, 16, 8, 8);\
896 void ff_ ## OPNAME ## qpel8_mc11_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
902 copy_block9(full, src, 16, stride, 9);\
903 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
904 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
905 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
906 OPNAME ## pixels8_l4_8(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
908 static void OPNAME ## qpel8_mc11_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
913 copy_block9(full, src, 16, stride, 9);\
914 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
915 put ## RND ## pixels8_l2_8(halfH, halfH, full, 8, 8, 16, 9);\
916 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
917 OPNAME ## pixels8_l2_8(dst, halfH, halfHV, stride, 8, 8, 8);\
919 void ff_ ## OPNAME ## qpel8_mc31_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
925 copy_block9(full, src, 16, stride, 9);\
926 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
927 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
928 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
929 OPNAME ## pixels8_l4_8(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
931 static void OPNAME ## qpel8_mc31_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
936 copy_block9(full, src, 16, stride, 9);\
937 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
938 put ## RND ## pixels8_l2_8(halfH, halfH, full+1, 8, 8, 16, 9);\
939 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
940 OPNAME ## pixels8_l2_8(dst, halfH, halfHV, stride, 8, 8, 8);\
942 void ff_ ## OPNAME ## qpel8_mc13_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
948 copy_block9(full, src, 16, stride, 9);\
949 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
950 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
951 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
952 OPNAME ## pixels8_l4_8(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
954 static void OPNAME ## qpel8_mc13_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
959 copy_block9(full, src, 16, stride, 9);\
960 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
961 put ## RND ## pixels8_l2_8(halfH, halfH, full, 8, 8, 16, 9);\
962 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
963 OPNAME ## pixels8_l2_8(dst, halfH+8, halfHV, stride, 8, 8, 8);\
965 void ff_ ## OPNAME ## qpel8_mc33_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
971 copy_block9(full, src, 16, stride, 9);\
972 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full , 8, 16, 9);\
973 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
974 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
975 OPNAME ## pixels8_l4_8(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
977 static void OPNAME ## qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
982 copy_block9(full, src, 16, stride, 9);\
983 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
984 put ## RND ## pixels8_l2_8(halfH, halfH, full+1, 8, 8, 16, 9);\
985 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
986 OPNAME ## pixels8_l2_8(dst, halfH+8, halfHV, stride, 8, 8, 8);\
988 static void OPNAME ## qpel8_mc21_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
992 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
993 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
994 OPNAME ## pixels8_l2_8(dst, halfH, halfHV, stride, 8, 8, 8);\
996 static void OPNAME ## qpel8_mc23_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1000 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
1001 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1002 OPNAME ## pixels8_l2_8(dst, halfH+8, halfHV, stride, 8, 8, 8);\
1004 void ff_ ## OPNAME ## qpel8_mc12_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1006 uint8_t full[16*9];\
1009 uint8_t halfHV[64];\
1010 copy_block9(full, src, 16, stride, 9);\
1011 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1012 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
1013 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1014 OPNAME ## pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);\
1016 static void OPNAME ## qpel8_mc12_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1018 uint8_t full[16*9];\
1020 copy_block9(full, src, 16, stride, 9);\
1021 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1022 put ## RND ## pixels8_l2_8(halfH, halfH, full, 8, 8, 16, 9);\
1023 OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1025 void ff_ ## OPNAME ## qpel8_mc32_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1027 uint8_t full[16*9];\
1030 uint8_t halfHV[64];\
1031 copy_block9(full, src, 16, stride, 9);\
1032 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1033 put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
1034 put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1035 OPNAME ## pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);\
1037 static void OPNAME ## qpel8_mc32_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1039 uint8_t full[16*9];\
1041 copy_block9(full, src, 16, stride, 9);\
1042 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1043 put ## RND ## pixels8_l2_8(halfH, halfH, full+1, 8, 8, 16, 9);\
1044 OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1046 static void OPNAME ## qpel8_mc22_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1049 put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
1050 OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1053 static void OPNAME ## qpel16_mc10_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1056 put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
1057 OPNAME ## pixels16_l2_8(dst, src, half, stride, stride, 16, 16);\
1060 static void OPNAME ## qpel16_mc20_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1062 OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
1065 static void OPNAME ## qpel16_mc30_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1068 put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
1069 OPNAME ## pixels16_l2_8(dst, src+1, half, stride, stride, 16, 16);\
1072 static void OPNAME ## qpel16_mc01_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1074 uint8_t full[24*17];\
1076 copy_block17(full, src, 24, stride, 17);\
1077 put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
1078 OPNAME ## pixels16_l2_8(dst, full, half, stride, 24, 16, 16);\
1081 static void OPNAME ## qpel16_mc02_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1083 uint8_t full[24*17];\
1084 copy_block17(full, src, 24, stride, 17);\
1085 OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
1088 static void OPNAME ## qpel16_mc03_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1090 uint8_t full[24*17];\
1092 copy_block17(full, src, 24, stride, 17);\
1093 put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
1094 OPNAME ## pixels16_l2_8(dst, full+24, half, stride, 24, 16, 16);\
1096 void ff_ ## OPNAME ## qpel16_mc11_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1098 uint8_t full[24*17];\
1099 uint8_t halfH[272];\
1100 uint8_t halfV[256];\
1101 uint8_t halfHV[256];\
1102 copy_block17(full, src, 24, stride, 17);\
1103 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1104 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1105 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1106 OPNAME ## pixels16_l4_8(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1108 static void OPNAME ## qpel16_mc11_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1110 uint8_t full[24*17];\
1111 uint8_t halfH[272];\
1112 uint8_t halfHV[256];\
1113 copy_block17(full, src, 24, stride, 17);\
1114 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1115 put ## RND ## pixels16_l2_8(halfH, halfH, full, 16, 16, 24, 17);\
1116 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1117 OPNAME ## pixels16_l2_8(dst, halfH, halfHV, stride, 16, 16, 16);\
1119 void ff_ ## OPNAME ## qpel16_mc31_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1121 uint8_t full[24*17];\
1122 uint8_t halfH[272];\
1123 uint8_t halfV[256];\
1124 uint8_t halfHV[256];\
1125 copy_block17(full, src, 24, stride, 17);\
1126 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1127 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1128 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1129 OPNAME ## pixels16_l4_8(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1131 static void OPNAME ## qpel16_mc31_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1133 uint8_t full[24*17];\
1134 uint8_t halfH[272];\
1135 uint8_t halfHV[256];\
1136 copy_block17(full, src, 24, stride, 17);\
1137 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1138 put ## RND ## pixels16_l2_8(halfH, halfH, full+1, 16, 16, 24, 17);\
1139 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1140 OPNAME ## pixels16_l2_8(dst, halfH, halfHV, stride, 16, 16, 16);\
1142 void ff_ ## OPNAME ## qpel16_mc13_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1144 uint8_t full[24*17];\
1145 uint8_t halfH[272];\
1146 uint8_t halfV[256];\
1147 uint8_t halfHV[256];\
1148 copy_block17(full, src, 24, stride, 17);\
1149 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1150 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1151 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1152 OPNAME ## pixels16_l4_8(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1154 static void OPNAME ## qpel16_mc13_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1156 uint8_t full[24*17];\
1157 uint8_t halfH[272];\
1158 uint8_t halfHV[256];\
1159 copy_block17(full, src, 24, stride, 17);\
1160 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1161 put ## RND ## pixels16_l2_8(halfH, halfH, full, 16, 16, 24, 17);\
1162 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1163 OPNAME ## pixels16_l2_8(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1165 void ff_ ## OPNAME ## qpel16_mc33_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1167 uint8_t full[24*17];\
1168 uint8_t halfH[272];\
1169 uint8_t halfV[256];\
1170 uint8_t halfHV[256];\
1171 copy_block17(full, src, 24, stride, 17);\
1172 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full , 16, 24, 17);\
1173 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1174 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1175 OPNAME ## pixels16_l4_8(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1177 static void OPNAME ## qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1179 uint8_t full[24*17];\
1180 uint8_t halfH[272];\
1181 uint8_t halfHV[256];\
1182 copy_block17(full, src, 24, stride, 17);\
1183 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1184 put ## RND ## pixels16_l2_8(halfH, halfH, full+1, 16, 16, 24, 17);\
1185 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1186 OPNAME ## pixels16_l2_8(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1188 static void OPNAME ## qpel16_mc21_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1190 uint8_t halfH[272];\
1191 uint8_t halfHV[256];\
1192 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1193 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1194 OPNAME ## pixels16_l2_8(dst, halfH, halfHV, stride, 16, 16, 16);\
1196 static void OPNAME ## qpel16_mc23_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1198 uint8_t halfH[272];\
1199 uint8_t halfHV[256];\
1200 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1201 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1202 OPNAME ## pixels16_l2_8(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1204 void ff_ ## OPNAME ## qpel16_mc12_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1206 uint8_t full[24*17];\
1207 uint8_t halfH[272];\
1208 uint8_t halfV[256];\
1209 uint8_t halfHV[256];\
1210 copy_block17(full, src, 24, stride, 17);\
1211 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1212 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1213 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1214 OPNAME ## pixels16_l2_8(dst, halfV, halfHV, stride, 16, 16, 16);\
1216 static void OPNAME ## qpel16_mc12_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1218 uint8_t full[24*17];\
1219 uint8_t halfH[272];\
1220 copy_block17(full, src, 24, stride, 17);\
1221 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1222 put ## RND ## pixels16_l2_8(halfH, halfH, full, 16, 16, 24, 17);\
1223 OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1225 void ff_ ## OPNAME ## qpel16_mc32_old_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1227 uint8_t full[24*17];\
1228 uint8_t halfH[272];\
1229 uint8_t halfV[256];\
1230 uint8_t halfHV[256];\
1231 copy_block17(full, src, 24, stride, 17);\
1232 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1233 put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1234 put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1235 OPNAME ## pixels16_l2_8(dst, halfV, halfHV, stride, 16, 16, 16);\
1237 static void OPNAME ## qpel16_mc32_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1239 uint8_t full[24*17];\
1240 uint8_t halfH[272];\
1241 copy_block17(full, src, 24, stride, 17);\
1242 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1243 put ## RND ## pixels16_l2_8(halfH, halfH, full+1, 16, 16, 24, 17);\
1244 OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1246 static void OPNAME ## qpel16_mc22_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
1248 uint8_t halfH[272];\
1249 put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1250 OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1253 #define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
1254 #define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
1255 #define op_put(a, b) a = cm[((b) + 16)>>5]
1256 #define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]
1258 QPEL_MC(0, put_ , _ , op_put)
1259 QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
1260 QPEL_MC(0, avg_ , _ , op_avg)
1264 #undef op_put_no_rnd
1266 void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1268 put_pixels8_8_c(dst, src, stride, 8);
1270 void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1272 avg_pixels8_8_c(dst, src, stride, 8);
1274 void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1276 put_pixels16_8_c(dst, src, stride, 16);
1278 void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1280 avg_pixels16_8_c(dst, src, stride, 16);
1283 #define put_qpel8_mc00_c ff_put_pixels8x8_c
1284 #define avg_qpel8_mc00_c ff_avg_pixels8x8_c
1285 #define put_qpel16_mc00_c ff_put_pixels16x16_c
1286 #define avg_qpel16_mc00_c ff_avg_pixels16x16_c
1287 #define put_no_rnd_qpel8_mc00_c ff_put_pixels8x8_c
1288 #define put_no_rnd_qpel16_mc00_c ff_put_pixels16x16_c
1290 static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
1291 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
1295 dst[0]= cm[(9*(src[0] + src[1]) - (src[-1] + src[2]) + 8)>>4];
1296 dst[1]= cm[(9*(src[1] + src[2]) - (src[ 0] + src[3]) + 8)>>4];
1297 dst[2]= cm[(9*(src[2] + src[3]) - (src[ 1] + src[4]) + 8)>>4];
1298 dst[3]= cm[(9*(src[3] + src[4]) - (src[ 2] + src[5]) + 8)>>4];
1299 dst[4]= cm[(9*(src[4] + src[5]) - (src[ 3] + src[6]) + 8)>>4];
1300 dst[5]= cm[(9*(src[5] + src[6]) - (src[ 4] + src[7]) + 8)>>4];
1301 dst[6]= cm[(9*(src[6] + src[7]) - (src[ 5] + src[8]) + 8)>>4];
1302 dst[7]= cm[(9*(src[7] + src[8]) - (src[ 6] + src[9]) + 8)>>4];
1308 #if CONFIG_RV40_DECODER
1309 void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1311 put_pixels16_xy2_8_c(dst, src, stride, 16);
1313 void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1315 avg_pixels16_xy2_8_c(dst, src, stride, 16);
1317 void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1319 put_pixels8_xy2_8_c(dst, src, stride, 8);
1321 void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1323 avg_pixels8_xy2_8_c(dst, src, stride, 8);
1325 #endif /* CONFIG_RV40_DECODER */
1327 static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
1328 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
1332 const int src_1= src[ -srcStride];
1333 const int src0 = src[0 ];
1334 const int src1 = src[ srcStride];
1335 const int src2 = src[2*srcStride];
1336 const int src3 = src[3*srcStride];
1337 const int src4 = src[4*srcStride];
1338 const int src5 = src[5*srcStride];
1339 const int src6 = src[6*srcStride];
1340 const int src7 = src[7*srcStride];
1341 const int src8 = src[8*srcStride];
1342 const int src9 = src[9*srcStride];
1343 dst[0*dstStride]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
1344 dst[1*dstStride]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];
1345 dst[2*dstStride]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];
1346 dst[3*dstStride]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];
1347 dst[4*dstStride]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];
1348 dst[5*dstStride]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];
1349 dst[6*dstStride]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];
1350 dst[7*dstStride]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];
1356 static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1359 wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1360 put_pixels8_l2_8(dst, src, half, stride, stride, 8, 8);
1363 static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1365 wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
1368 static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1371 wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1372 put_pixels8_l2_8(dst, src+1, half, stride, stride, 8, 8);
1375 static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1377 wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
1380 static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1385 wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1386 wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
1387 wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1388 put_pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);
1390 static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1395 wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1396 wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
1397 wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1398 put_pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);
1400 static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
1403 wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1404 wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
1407 static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1413 s += abs(pix1[0] - pix2[0]);
1414 s += abs(pix1[1] - pix2[1]);
1415 s += abs(pix1[2] - pix2[2]);
1416 s += abs(pix1[3] - pix2[3]);
1417 s += abs(pix1[4] - pix2[4]);
1418 s += abs(pix1[5] - pix2[5]);
1419 s += abs(pix1[6] - pix2[6]);
1420 s += abs(pix1[7] - pix2[7]);
1421 s += abs(pix1[8] - pix2[8]);
1422 s += abs(pix1[9] - pix2[9]);
1423 s += abs(pix1[10] - pix2[10]);
1424 s += abs(pix1[11] - pix2[11]);
1425 s += abs(pix1[12] - pix2[12]);
1426 s += abs(pix1[13] - pix2[13]);
1427 s += abs(pix1[14] - pix2[14]);
1428 s += abs(pix1[15] - pix2[15]);
1435 static int pix_abs16_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1441 s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
1442 s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
1443 s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
1444 s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
1445 s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
1446 s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
1447 s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
1448 s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
1449 s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
1450 s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
1451 s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
1452 s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
1453 s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
1454 s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
1455 s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
1456 s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
1463 static int pix_abs16_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1466 uint8_t *pix3 = pix2 + line_size;
1470 s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
1471 s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
1472 s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
1473 s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
1474 s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
1475 s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
1476 s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
1477 s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
1478 s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
1479 s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
1480 s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
1481 s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
1482 s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
1483 s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
1484 s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
1485 s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
1493 static int pix_abs16_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1496 uint8_t *pix3 = pix2 + line_size;
1500 s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
1501 s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
1502 s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
1503 s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
1504 s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
1505 s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
1506 s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
1507 s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
1508 s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
1509 s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
1510 s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
1511 s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
1512 s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
1513 s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
1514 s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
1515 s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
1523 static inline int pix_abs8_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1529 s += abs(pix1[0] - pix2[0]);
1530 s += abs(pix1[1] - pix2[1]);
1531 s += abs(pix1[2] - pix2[2]);
1532 s += abs(pix1[3] - pix2[3]);
1533 s += abs(pix1[4] - pix2[4]);
1534 s += abs(pix1[5] - pix2[5]);
1535 s += abs(pix1[6] - pix2[6]);
1536 s += abs(pix1[7] - pix2[7]);
1543 static int pix_abs8_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1549 s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
1550 s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
1551 s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
1552 s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
1553 s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
1554 s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
1555 s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
1556 s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
1563 static int pix_abs8_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1566 uint8_t *pix3 = pix2 + line_size;
1570 s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
1571 s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
1572 s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
1573 s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
1574 s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
1575 s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
1576 s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
1577 s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
1585 static int pix_abs8_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
1588 uint8_t *pix3 = pix2 + line_size;
1592 s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
1593 s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
1594 s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
1595 s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
1596 s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
1597 s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
1598 s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
1599 s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
1607 static int nsse16_c(void *v, uint8_t *s1, uint8_t *s2, int stride, int h){
1608 MpegEncContext *c = v;
1614 for(x=0; x<16; x++){
1615 score1+= (s1[x ] - s2[x ])*(s1[x ] - s2[x ]);
1618 for(x=0; x<15; x++){
1619 score2+= FFABS( s1[x ] - s1[x +stride]
1620 - s1[x+1] + s1[x+1+stride])
1621 -FFABS( s2[x ] - s2[x +stride]
1622 - s2[x+1] + s2[x+1+stride]);
1629 if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
1630 else return score1 + FFABS(score2)*8;
1633 static int nsse8_c(void *v, uint8_t *s1, uint8_t *s2, int stride, int h){
1634 MpegEncContext *c = v;
1641 score1+= (s1[x ] - s2[x ])*(s1[x ] - s2[x ]);
1645 score2+= FFABS( s1[x ] - s1[x +stride]
1646 - s1[x+1] + s1[x+1+stride])
1647 -FFABS( s2[x ] - s2[x +stride]
1648 - s2[x+1] + s2[x+1+stride]);
1655 if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
1656 else return score1 + FFABS(score2)*8;
1659 static int try_8x8basis_c(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){
1663 for(i=0; i<8*8; i++){
1664 int b= rem[i] + ((basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT));
1667 assert(-512<b && b<512);
1669 sum += (w*b)*(w*b)>>4;
1674 static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale){
1677 for(i=0; i<8*8; i++){
1678 rem[i] += (basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT);
1682 static int zero_cmp(void *s, uint8_t *a, uint8_t *b, int stride, int h){
1686 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type){
1689 memset(cmp, 0, sizeof(void*)*6);
1697 cmp[i]= c->hadamard8_diff[i];
1703 cmp[i]= c->dct_sad[i];
1706 cmp[i]= c->dct264_sad[i];
1709 cmp[i]= c->dct_max[i];
1712 cmp[i]= c->quant_psnr[i];
1733 av_log(NULL, AV_LOG_ERROR,"internal error in cmp function selection\n");
1738 static void add_bytes_c(uint8_t *dst, uint8_t *src, int w){
1740 for (i = 0; i <= w - (int) sizeof(long); i += sizeof(long)) {
1741 long a = *(long*)(src+i);
1742 long b = *(long*)(dst+i);
1743 *(long*)(dst+i) = ((a&pb_7f) + (b&pb_7f)) ^ ((a^b)&pb_80);
1746 dst[i+0] += src[i+0];
1749 static void diff_bytes_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
1751 #if !HAVE_FAST_UNALIGNED
1752 if((long)src2 & (sizeof(long)-1)){
1753 for(i=0; i+7<w; i+=8){
1754 dst[i+0] = src1[i+0]-src2[i+0];
1755 dst[i+1] = src1[i+1]-src2[i+1];
1756 dst[i+2] = src1[i+2]-src2[i+2];
1757 dst[i+3] = src1[i+3]-src2[i+3];
1758 dst[i+4] = src1[i+4]-src2[i+4];
1759 dst[i+5] = src1[i+5]-src2[i+5];
1760 dst[i+6] = src1[i+6]-src2[i+6];
1761 dst[i+7] = src1[i+7]-src2[i+7];
1765 for (i = 0; i <= w - (int) sizeof(long); i += sizeof(long)) {
1766 long a = *(long*)(src1+i);
1767 long b = *(long*)(src2+i);
1768 *(long*)(dst+i) = ((a|pb_80) - (b&pb_7f)) ^ ((a^b^pb_80)&pb_80);
1771 dst[i+0] = src1[i+0]-src2[i+0];
1774 static void add_hfyu_median_prediction_c(uint8_t *dst, const uint8_t *src1, const uint8_t *diff, int w, int *left, int *left_top){
1782 l= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF) + diff[i];
1791 static void sub_hfyu_median_prediction_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top){
1799 const int pred= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF);
1809 static int add_hfyu_left_prediction_c(uint8_t *dst, const uint8_t *src, int w, int acc){
1812 for(i=0; i<w-1; i++){
1839 static void add_hfyu_left_prediction_bgr32_c(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha){
1869 #define BUTTERFLY2(o1,o2,i1,i2) \
1873 #define BUTTERFLY1(x,y) \
1882 #define BUTTERFLYA(x,y) (FFABS((x)+(y)) + FFABS((x)-(y)))
1884 static int hadamard8_diff8x8_c(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
1892 //FIXME try pointer walks
1893 BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-dst[stride*i+0],src[stride*i+1]-dst[stride*i+1]);
1894 BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-dst[stride*i+2],src[stride*i+3]-dst[stride*i+3]);
1895 BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4]-dst[stride*i+4],src[stride*i+5]-dst[stride*i+5]);
1896 BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6]-dst[stride*i+6],src[stride*i+7]-dst[stride*i+7]);
1898 BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
1899 BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
1900 BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
1901 BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
1903 BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
1904 BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
1905 BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
1906 BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
1910 BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
1911 BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
1912 BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
1913 BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
1915 BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
1916 BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
1917 BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
1918 BUTTERFLY1(temp[8*5+i], temp[8*7+i]);
1921 BUTTERFLYA(temp[8*0+i], temp[8*4+i])
1922 +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
1923 +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
1924 +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
1929 static int hadamard8_intra8x8_c(/*MpegEncContext*/ void *s, uint8_t *src, uint8_t *dummy, int stride, int h){
1937 //FIXME try pointer walks
1938 BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0],src[stride*i+1]);
1939 BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2],src[stride*i+3]);
1940 BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4],src[stride*i+5]);
1941 BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6],src[stride*i+7]);
1943 BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
1944 BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
1945 BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
1946 BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
1948 BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
1949 BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
1950 BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
1951 BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
1955 BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
1956 BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
1957 BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
1958 BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
1960 BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
1961 BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
1962 BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
1963 BUTTERFLY1(temp[8*5+i], temp[8*7+i]);
1966 BUTTERFLYA(temp[8*0+i], temp[8*4+i])
1967 +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
1968 +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
1969 +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
1972 sum -= FFABS(temp[8*0] + temp[8*4]); // -mean
1977 static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
1978 MpegEncContext * const s= (MpegEncContext *)c;
1979 LOCAL_ALIGNED_16(int16_t, temp, [64]);
1983 s->dsp.diff_pixels(temp, src1, src2, stride);
1985 return s->dsp.sum_abs_dctelem(temp);
1990 const int s07 = SRC(0) + SRC(7);\
1991 const int s16 = SRC(1) + SRC(6);\
1992 const int s25 = SRC(2) + SRC(5);\
1993 const int s34 = SRC(3) + SRC(4);\
1994 const int a0 = s07 + s34;\
1995 const int a1 = s16 + s25;\
1996 const int a2 = s07 - s34;\
1997 const int a3 = s16 - s25;\
1998 const int d07 = SRC(0) - SRC(7);\
1999 const int d16 = SRC(1) - SRC(6);\
2000 const int d25 = SRC(2) - SRC(5);\
2001 const int d34 = SRC(3) - SRC(4);\
2002 const int a4 = d16 + d25 + (d07 + (d07>>1));\
2003 const int a5 = d07 - d34 - (d25 + (d25>>1));\
2004 const int a6 = d07 + d34 - (d16 + (d16>>1));\
2005 const int a7 = d16 - d25 + (d34 + (d34>>1));\
2007 DST(1, a4 + (a7>>2)) ;\
2008 DST(2, a2 + (a3>>1)) ;\
2009 DST(3, a5 + (a6>>2)) ;\
2011 DST(5, a6 - (a5>>2)) ;\
2012 DST(6, (a2>>1) - a3 ) ;\
2013 DST(7, (a4>>2) - a7 ) ;\
2016 static int dct264_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
2017 MpegEncContext * const s= (MpegEncContext *)c;
2022 s->dsp.diff_pixels(dct[0], src1, src2, stride);
2024 #define SRC(x) dct[i][x]
2025 #define DST(x,v) dct[i][x]= v
2026 for( i = 0; i < 8; i++ )
2031 #define SRC(x) dct[x][i]
2032 #define DST(x,v) sum += FFABS(v)
2033 for( i = 0; i < 8; i++ )
2041 static int dct_max8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
2042 MpegEncContext * const s= (MpegEncContext *)c;
2043 LOCAL_ALIGNED_16(int16_t, temp, [64]);
2048 s->dsp.diff_pixels(temp, src1, src2, stride);
2052 sum= FFMAX(sum, FFABS(temp[i]));
2057 static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
2058 MpegEncContext * const s= (MpegEncContext *)c;
2059 LOCAL_ALIGNED_16(int16_t, temp, [64*2]);
2060 int16_t * const bak = temp+64;
2066 s->dsp.diff_pixels(temp, src1, src2, stride);
2068 memcpy(bak, temp, 64*sizeof(int16_t));
2070 s->block_last_index[0/*FIXME*/]= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
2071 s->dct_unquantize_inter(s, temp, 0, s->qscale);
2072 ff_simple_idct_8(temp); //FIXME
2075 sum+= (temp[i]-bak[i])*(temp[i]-bak[i]);
2080 static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
2081 MpegEncContext * const s= (MpegEncContext *)c;
2082 const uint8_t *scantable= s->intra_scantable.permutated;
2083 LOCAL_ALIGNED_16(int16_t, temp, [64]);
2084 LOCAL_ALIGNED_16(uint8_t, lsrc1, [64]);
2085 LOCAL_ALIGNED_16(uint8_t, lsrc2, [64]);
2086 int i, last, run, bits, level, distortion, start_i;
2087 const int esc_length= s->ac_esc_length;
2089 uint8_t * last_length;
2093 copy_block8(lsrc1, src1, 8, stride, 8);
2094 copy_block8(lsrc2, src2, 8, stride, 8);
2096 s->dsp.diff_pixels(temp, lsrc1, lsrc2, 8);
2098 s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
2104 length = s->intra_ac_vlc_length;
2105 last_length= s->intra_ac_vlc_last_length;
2106 bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
2109 length = s->inter_ac_vlc_length;
2110 last_length= s->inter_ac_vlc_last_length;
2115 for(i=start_i; i<last; i++){
2116 int j= scantable[i];
2121 if((level&(~127)) == 0){
2122 bits+= length[UNI_AC_ENC_INDEX(run, level)];
2131 level= temp[i] + 64;
2135 if((level&(~127)) == 0){
2136 bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
2144 s->dct_unquantize_intra(s, temp, 0, s->qscale);
2146 s->dct_unquantize_inter(s, temp, 0, s->qscale);
2149 s->dsp.idct_add(lsrc2, 8, temp);
2151 distortion= s->dsp.sse[1](NULL, lsrc2, lsrc1, 8, 8);
2153 return distortion + ((bits*s->qscale*s->qscale*109 + 64)>>7);
2156 static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
2157 MpegEncContext * const s= (MpegEncContext *)c;
2158 const uint8_t *scantable= s->intra_scantable.permutated;
2159 LOCAL_ALIGNED_16(int16_t, temp, [64]);
2160 int i, last, run, bits, level, start_i;
2161 const int esc_length= s->ac_esc_length;
2163 uint8_t * last_length;
2167 s->dsp.diff_pixels(temp, src1, src2, stride);
2169 s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
2175 length = s->intra_ac_vlc_length;
2176 last_length= s->intra_ac_vlc_last_length;
2177 bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
2180 length = s->inter_ac_vlc_length;
2181 last_length= s->inter_ac_vlc_last_length;
2186 for(i=start_i; i<last; i++){
2187 int j= scantable[i];
2192 if((level&(~127)) == 0){
2193 bits+= length[UNI_AC_ENC_INDEX(run, level)];
2202 level= temp[i] + 64;
2206 if((level&(~127)) == 0){
2207 bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
2215 #define VSAD_INTRA(size) \
2216 static int vsad_intra##size##_c(/*MpegEncContext*/ void *c, uint8_t *s, uint8_t *dummy, int stride, int h){ \
2220 for(y=1; y<h; y++){ \
2221 for(x=0; x<size; x+=4){ \
2222 score+= FFABS(s[x ] - s[x +stride]) + FFABS(s[x+1] - s[x+1+stride]) \
2223 +FFABS(s[x+2] - s[x+2+stride]) + FFABS(s[x+3] - s[x+3+stride]); \
2233 static int vsad16_c(/*MpegEncContext*/ void *c, uint8_t *s1, uint8_t *s2, int stride, int h){
2238 for(x=0; x<16; x++){
2239 score+= FFABS(s1[x ] - s2[x ] - s1[x +stride] + s2[x +stride]);
2248 #define SQ(a) ((a)*(a))
2249 #define VSSE_INTRA(size) \
2250 static int vsse_intra##size##_c(/*MpegEncContext*/ void *c, uint8_t *s, uint8_t *dummy, int stride, int h){ \
2254 for(y=1; y<h; y++){ \
2255 for(x=0; x<size; x+=4){ \
2256 score+= SQ(s[x ] - s[x +stride]) + SQ(s[x+1] - s[x+1+stride]) \
2257 +SQ(s[x+2] - s[x+2+stride]) + SQ(s[x+3] - s[x+3+stride]); \
2267 static int vsse16_c(/*MpegEncContext*/ void *c, uint8_t *s1, uint8_t *s2, int stride, int h){
2272 for(x=0; x<16; x++){
2273 score+= SQ(s1[x ] - s2[x ] - s1[x +stride] + s2[x +stride]);
2282 static int ssd_int8_vs_int16_c(const int8_t *pix1, const int16_t *pix2,
2286 for(i=0; i<size; i++)
2287 score += (pix1[i]-pix2[i])*(pix1[i]-pix2[i]);
2291 #define WRAPPER8_16_SQ(name8, name16)\
2292 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
2294 score +=name8(s, dst , src , stride, 8);\
2295 score +=name8(s, dst+8 , src+8 , stride, 8);\
2299 score +=name8(s, dst , src , stride, 8);\
2300 score +=name8(s, dst+8 , src+8 , stride, 8);\
2305 WRAPPER8_16_SQ(hadamard8_diff8x8_c, hadamard8_diff16_c)
2306 WRAPPER8_16_SQ(hadamard8_intra8x8_c, hadamard8_intra16_c)
2307 WRAPPER8_16_SQ(dct_sad8x8_c, dct_sad16_c)
2309 WRAPPER8_16_SQ(dct264_sad8x8_c, dct264_sad16_c)
2311 WRAPPER8_16_SQ(dct_max8x8_c, dct_max16_c)
2312 WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
2313 WRAPPER8_16_SQ(rd8x8_c, rd16_c)
2314 WRAPPER8_16_SQ(bit8x8_c, bit16_c)
2316 static inline uint32_t clipf_c_one(uint32_t a, uint32_t mini,
2317 uint32_t maxi, uint32_t maxisign)
2320 if(a > mini) return mini;
2321 else if((a^(1U<<31)) > maxisign) return maxi;
2325 static void vector_clipf_c_opposite_sign(float *dst, const float *src, float *min, float *max, int len){
2327 uint32_t mini = *(uint32_t*)min;
2328 uint32_t maxi = *(uint32_t*)max;
2329 uint32_t maxisign = maxi ^ (1U<<31);
2330 uint32_t *dsti = (uint32_t*)dst;
2331 const uint32_t *srci = (const uint32_t*)src;
2332 for(i=0; i<len; i+=8) {
2333 dsti[i + 0] = clipf_c_one(srci[i + 0], mini, maxi, maxisign);
2334 dsti[i + 1] = clipf_c_one(srci[i + 1], mini, maxi, maxisign);
2335 dsti[i + 2] = clipf_c_one(srci[i + 2], mini, maxi, maxisign);
2336 dsti[i + 3] = clipf_c_one(srci[i + 3], mini, maxi, maxisign);
2337 dsti[i + 4] = clipf_c_one(srci[i + 4], mini, maxi, maxisign);
2338 dsti[i + 5] = clipf_c_one(srci[i + 5], mini, maxi, maxisign);
2339 dsti[i + 6] = clipf_c_one(srci[i + 6], mini, maxi, maxisign);
2340 dsti[i + 7] = clipf_c_one(srci[i + 7], mini, maxi, maxisign);
2343 static void vector_clipf_c(float *dst, const float *src, float min, float max, int len){
2345 if(min < 0 && max > 0) {
2346 vector_clipf_c_opposite_sign(dst, src, &min, &max, len);
2348 for(i=0; i < len; i+=8) {
2349 dst[i ] = av_clipf(src[i ], min, max);
2350 dst[i + 1] = av_clipf(src[i + 1], min, max);
2351 dst[i + 2] = av_clipf(src[i + 2], min, max);
2352 dst[i + 3] = av_clipf(src[i + 3], min, max);
2353 dst[i + 4] = av_clipf(src[i + 4], min, max);
2354 dst[i + 5] = av_clipf(src[i + 5], min, max);
2355 dst[i + 6] = av_clipf(src[i + 6], min, max);
2356 dst[i + 7] = av_clipf(src[i + 7], min, max);
2361 static int32_t scalarproduct_int16_c(const int16_t * v1, const int16_t * v2, int order)
2366 res += *v1++ * *v2++;
2371 static int32_t scalarproduct_and_madd_int16_c(int16_t *v1, const int16_t *v2, const int16_t *v3, int order, int mul)
2376 *v1++ += mul * *v3++;
2381 static void vector_clip_int32_c(int32_t *dst, const int32_t *src, int32_t min,
2382 int32_t max, unsigned int len)
2385 *dst++ = av_clip(*src++, min, max);
2386 *dst++ = av_clip(*src++, min, max);
2387 *dst++ = av_clip(*src++, min, max);
2388 *dst++ = av_clip(*src++, min, max);
2389 *dst++ = av_clip(*src++, min, max);
2390 *dst++ = av_clip(*src++, min, max);
2391 *dst++ = av_clip(*src++, min, max);
2392 *dst++ = av_clip(*src++, min, max);
2397 static void jref_idct_put(uint8_t *dest, int line_size, int16_t *block)
2399 ff_j_rev_dct (block);
2400 put_pixels_clamped_c(block, dest, line_size);
2402 static void jref_idct_add(uint8_t *dest, int line_size, int16_t *block)
2404 ff_j_rev_dct (block);
2405 add_pixels_clamped_c(block, dest, line_size);
2408 /* init static data */
2409 av_cold void ff_dsputil_static_init(void)
2413 for(i=0;i<512;i++) {
2414 ff_squareTbl[i] = (i - 256) * (i - 256);
2418 av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
2421 if (avctx->bits_per_raw_sample == 10) {
2422 c->fdct = ff_jpeg_fdct_islow_10;
2423 c->fdct248 = ff_fdct248_islow_10;
2425 if(avctx->dct_algo==FF_DCT_FASTINT) {
2426 c->fdct = ff_fdct_ifast;
2427 c->fdct248 = ff_fdct_ifast248;
2429 else if(avctx->dct_algo==FF_DCT_FAAN) {
2430 c->fdct = ff_faandct;
2431 c->fdct248 = ff_faandct248;
2434 c->fdct = ff_jpeg_fdct_islow_8; //slow/accurate/default
2435 c->fdct248 = ff_fdct248_islow_8;
2438 #endif //CONFIG_ENCODERS
2440 if (avctx->bits_per_raw_sample == 10) {
2441 c->idct_put = ff_simple_idct_put_10;
2442 c->idct_add = ff_simple_idct_add_10;
2443 c->idct = ff_simple_idct_10;
2444 c->idct_permutation_type = FF_NO_IDCT_PERM;
2446 if(avctx->idct_algo==FF_IDCT_INT){
2447 c->idct_put= jref_idct_put;
2448 c->idct_add= jref_idct_add;
2449 c->idct = ff_j_rev_dct;
2450 c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
2451 }else if(avctx->idct_algo==FF_IDCT_FAAN){
2452 c->idct_put= ff_faanidct_put;
2453 c->idct_add= ff_faanidct_add;
2454 c->idct = ff_faanidct;
2455 c->idct_permutation_type= FF_NO_IDCT_PERM;
2456 }else{ //accurate/default
2457 c->idct_put = ff_simple_idct_put_8;
2458 c->idct_add = ff_simple_idct_add_8;
2459 c->idct = ff_simple_idct_8;
2460 c->idct_permutation_type= FF_NO_IDCT_PERM;
2464 c->diff_pixels = diff_pixels_c;
2465 c->put_pixels_clamped = put_pixels_clamped_c;
2466 c->put_signed_pixels_clamped = put_signed_pixels_clamped_c;
2467 c->add_pixels_clamped = add_pixels_clamped_c;
2468 c->sum_abs_dctelem = sum_abs_dctelem_c;
2471 c->pix_sum = pix_sum_c;
2472 c->pix_norm1 = pix_norm1_c;
2474 c->fill_block_tab[0] = fill_block16_c;
2475 c->fill_block_tab[1] = fill_block8_c;
2477 /* TODO [0] 16 [1] 8 */
2478 c->pix_abs[0][0] = pix_abs16_c;
2479 c->pix_abs[0][1] = pix_abs16_x2_c;
2480 c->pix_abs[0][2] = pix_abs16_y2_c;
2481 c->pix_abs[0][3] = pix_abs16_xy2_c;
2482 c->pix_abs[1][0] = pix_abs8_c;
2483 c->pix_abs[1][1] = pix_abs8_x2_c;
2484 c->pix_abs[1][2] = pix_abs8_y2_c;
2485 c->pix_abs[1][3] = pix_abs8_xy2_c;
2487 c->put_tpel_pixels_tab[ 0] = put_tpel_pixels_mc00_c;
2488 c->put_tpel_pixels_tab[ 1] = put_tpel_pixels_mc10_c;
2489 c->put_tpel_pixels_tab[ 2] = put_tpel_pixels_mc20_c;
2490 c->put_tpel_pixels_tab[ 4] = put_tpel_pixels_mc01_c;
2491 c->put_tpel_pixels_tab[ 5] = put_tpel_pixels_mc11_c;
2492 c->put_tpel_pixels_tab[ 6] = put_tpel_pixels_mc21_c;
2493 c->put_tpel_pixels_tab[ 8] = put_tpel_pixels_mc02_c;
2494 c->put_tpel_pixels_tab[ 9] = put_tpel_pixels_mc12_c;
2495 c->put_tpel_pixels_tab[10] = put_tpel_pixels_mc22_c;
2497 c->avg_tpel_pixels_tab[ 0] = avg_tpel_pixels_mc00_c;
2498 c->avg_tpel_pixels_tab[ 1] = avg_tpel_pixels_mc10_c;
2499 c->avg_tpel_pixels_tab[ 2] = avg_tpel_pixels_mc20_c;
2500 c->avg_tpel_pixels_tab[ 4] = avg_tpel_pixels_mc01_c;
2501 c->avg_tpel_pixels_tab[ 5] = avg_tpel_pixels_mc11_c;
2502 c->avg_tpel_pixels_tab[ 6] = avg_tpel_pixels_mc21_c;
2503 c->avg_tpel_pixels_tab[ 8] = avg_tpel_pixels_mc02_c;
2504 c->avg_tpel_pixels_tab[ 9] = avg_tpel_pixels_mc12_c;
2505 c->avg_tpel_pixels_tab[10] = avg_tpel_pixels_mc22_c;
2507 #define dspfunc(PFX, IDX, NUM) \
2508 c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
2509 c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
2510 c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
2511 c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
2512 c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
2513 c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
2514 c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
2515 c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
2516 c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
2517 c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
2518 c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
2519 c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
2520 c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
2521 c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
2522 c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
2523 c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c
2525 dspfunc(put_qpel, 0, 16);
2526 dspfunc(put_no_rnd_qpel, 0, 16);
2528 dspfunc(avg_qpel, 0, 16);
2530 dspfunc(put_qpel, 1, 8);
2531 dspfunc(put_no_rnd_qpel, 1, 8);
2533 dspfunc(avg_qpel, 1, 8);
2537 c->put_mspel_pixels_tab[0]= ff_put_pixels8x8_c;
2538 c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c;
2539 c->put_mspel_pixels_tab[2]= put_mspel8_mc20_c;
2540 c->put_mspel_pixels_tab[3]= put_mspel8_mc30_c;
2541 c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
2542 c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
2543 c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
2544 c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;
2546 #define SET_CMP_FUNC(name) \
2547 c->name[0]= name ## 16_c;\
2548 c->name[1]= name ## 8x8_c;
2550 SET_CMP_FUNC(hadamard8_diff)
2551 c->hadamard8_diff[4]= hadamard8_intra16_c;
2552 c->hadamard8_diff[5]= hadamard8_intra8x8_c;
2553 SET_CMP_FUNC(dct_sad)
2554 SET_CMP_FUNC(dct_max)
2556 SET_CMP_FUNC(dct264_sad)
2558 c->sad[0]= pix_abs16_c;
2559 c->sad[1]= pix_abs8_c;
2563 SET_CMP_FUNC(quant_psnr)
2566 c->vsad[0]= vsad16_c;
2567 c->vsad[4]= vsad_intra16_c;
2568 c->vsad[5]= vsad_intra8_c;
2569 c->vsse[0]= vsse16_c;
2570 c->vsse[4]= vsse_intra16_c;
2571 c->vsse[5]= vsse_intra8_c;
2572 c->nsse[0]= nsse16_c;
2573 c->nsse[1]= nsse8_c;
2575 c->ssd_int8_vs_int16 = ssd_int8_vs_int16_c;
2577 c->add_bytes= add_bytes_c;
2578 c->diff_bytes= diff_bytes_c;
2579 c->add_hfyu_median_prediction= add_hfyu_median_prediction_c;
2580 c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
2581 c->add_hfyu_left_prediction = add_hfyu_left_prediction_c;
2582 c->add_hfyu_left_prediction_bgr32 = add_hfyu_left_prediction_bgr32_c;
2583 c->bswap_buf= bswap_buf;
2584 c->bswap16_buf = bswap16_buf;
2586 c->try_8x8basis= try_8x8basis_c;
2587 c->add_8x8basis= add_8x8basis_c;
2589 c->vector_clipf = vector_clipf_c;
2590 c->scalarproduct_int16 = scalarproduct_int16_c;
2591 c->scalarproduct_and_madd_int16 = scalarproduct_and_madd_int16_c;
2592 c->vector_clip_int32 = vector_clip_int32_c;
2594 c->shrink[0]= av_image_copy_plane;
2595 c->shrink[1]= ff_shrink22;
2596 c->shrink[2]= ff_shrink44;
2597 c->shrink[3]= ff_shrink88;
2599 c->add_pixels8 = add_pixels8_c;
2603 #define FUNC(f, depth) f ## _ ## depth
2604 #define FUNCC(f, depth) f ## _ ## depth ## _c
2606 c->draw_edges = FUNCC(draw_edges, 8);
2607 c->clear_block = FUNCC(clear_block, 8);
2608 c->clear_blocks = FUNCC(clear_blocks, 8);
2610 #define BIT_DEPTH_FUNCS(depth) \
2611 c->get_pixels = FUNCC(get_pixels, depth);
2613 switch (avctx->bits_per_raw_sample) {
2616 BIT_DEPTH_FUNCS(16);
2625 ff_dsputil_init_arm(c, avctx);
2627 ff_dsputil_init_bfin(c, avctx);
2629 ff_dsputil_init_ppc(c, avctx);
2631 ff_dsputil_init_x86(c, avctx);
2633 ff_init_scantable_permutation(c->idct_permutation,
2634 c->idct_permutation_type);