interlaced motion estimation
[libav.git] / libavcodec / dsputil.h
CommitLineData
ff4ec49e
FB
1/*
2 * DSP utils
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
4 *
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
9 *
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 */
24641185
MN
19
20/**
21 * @file dsputil.h
983e3246 22 * DSP utils.
5755c27f
MN
23 * note, many functions in here may use MMX which trashes the FPU state, it is
24 * absolutely necessary to call emms_c() between dsp & float/double code
24641185
MN
25 */
26
de6d9b64
FB
27#ifndef DSPUTIL_H
28#define DSPUTIL_H
29
30#include "common.h"
43f1708f 31#include "avcodec.h"
de6d9b64 32
24641185 33
44eb4951 34//#define DEBUG
de6d9b64
FB
35/* dct code */
36typedef short DCTELEM;
37
03c94ede 38void fdct_ifast (DCTELEM *data);
48b1f800 39void fdct_ifast248 (DCTELEM *data);
28db7fce 40void ff_jpeg_fdct_islow (DCTELEM *data);
10acc479 41void ff_fdct248_islow (DCTELEM *data);
de6d9b64
FB
42
43void j_rev_dct (DCTELEM *data);
44
3f09f52a 45void ff_fdct_mmx(DCTELEM *block);
cf3bf5bb 46void ff_fdct_mmx2(DCTELEM *block);
de6d9b64 47
e0eac44e 48/* encoding scans */
0c1a9eda
ZK
49extern const uint8_t ff_alternate_horizontal_scan[64];
50extern const uint8_t ff_alternate_vertical_scan[64];
51extern const uint8_t ff_zigzag_direct[64];
10acc479 52extern const uint8_t ff_zigzag248_direct[64];
5a240838 53
de6d9b64
FB
54/* pixel operations */
55#define MAX_NEG_CROP 384
56
57/* temporary */
0c1a9eda
ZK
58extern uint32_t squareTbl[512];
59extern uint8_t cropTbl[256 + 2 * MAX_NEG_CROP];
de6d9b64 60
de6d9b64 61
b7c27ee6 62/* minimum alignment rules ;)
eb4b3dd3 63if u notice errors in the align stuff, need more alignment for some asm code for some cpu
b7c27ee6
MN
64or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ...
65
66!warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
67i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
de6d9b64 68
b7c27ee6
MN
69!future video codecs might need functions with less strict alignment
70*/
71
eb4b3dd3 72/*
0c1a9eda
ZK
73void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
74void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
75void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
76void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
649c00c9 77void clear_blocks_c(DCTELEM *blocks);
eb4b3dd3 78*/
de6d9b64
FB
79
80/* add and put pixel (decoding) */
b7c27ee6 81// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
0c1a9eda 82typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
669ac79c 83typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
0c1a9eda 84typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
0da71265 85typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
b3184779 86
db794953 87#define DEF_OLD_QPEL(name)\
0c1a9eda
ZK
88void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
89void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
90void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
db794953
MN
91
92DEF_OLD_QPEL(qpel16_mc11_old_c)
93DEF_OLD_QPEL(qpel16_mc31_old_c)
94DEF_OLD_QPEL(qpel16_mc12_old_c)
95DEF_OLD_QPEL(qpel16_mc32_old_c)
96DEF_OLD_QPEL(qpel16_mc13_old_c)
97DEF_OLD_QPEL(qpel16_mc33_old_c)
98DEF_OLD_QPEL(qpel8_mc11_old_c)
99DEF_OLD_QPEL(qpel8_mc31_old_c)
100DEF_OLD_QPEL(qpel8_mc12_old_c)
101DEF_OLD_QPEL(qpel8_mc32_old_c)
102DEF_OLD_QPEL(qpel8_mc13_old_c)
103DEF_OLD_QPEL(qpel8_mc33_old_c)
b3184779
MN
104
105#define CALL_2X_PIXELS(a, b, n)\
106static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
107 b(block , pixels , line_size, h);\
108 b(block+n, pixels+n, line_size, h);\
109}
44eb4951 110
de6d9b64
FB
111/* motion estimation */
112
bb198e19 113typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
1457ab52 114
0da71265 115
24641185
MN
116/**
117 * DSPContext.
118 */
eb4b3dd3
ZK
119typedef struct DSPContext {
120 /* pixel ops : interface with DCT */
0c1a9eda
ZK
121 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
122 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
123 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
124 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
d518aebd
MN
125 /**
126 * translational global motion compensation.
127 */
0c1a9eda 128 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
d518aebd
MN
129 /**
130 * global motion compensation.
131 */
0c1a9eda 132 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
eb4b3dd3
ZK
133 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
134 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
0c1a9eda
ZK
135 int (*pix_sum)(uint8_t * pix, int line_size);
136 int (*pix_norm1)(uint8_t * pix, int line_size);
bb198e19
MN
137// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
138
139 me_cmp_func sad[4]; /* identical to pix_absAxA except additional void * */
140 me_cmp_func sse[4];
141 me_cmp_func hadamard8_diff[4];
142 me_cmp_func dct_sad[4];
143 me_cmp_func quant_psnr[4];
144 me_cmp_func bit[4];
145 me_cmp_func rd[4];
1457ab52
MN
146 int (*hadamard8_abs )(uint8_t *src, int stride, int mean);
147
bb198e19
MN
148 me_cmp_func me_pre_cmp[5];
149 me_cmp_func me_cmp[5];
150 me_cmp_func me_sub_cmp[5];
151 me_cmp_func mb_cmp[5];
eb4b3dd3 152
669ac79c 153 /* maybe create an array for 16/8/4/2 functions */
d518aebd
MN
154 /**
155 * Halfpel motion compensation with rounding (a+b+1)>>1.
669ac79c
MN
156 * this is an array[4][4] of motion compensation funcions for 4
157 * horizontal blocksizes (2,4,8,16) and the 4 halfpel positions<br>
5755c27f 158 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
159 * @param block destination where the result is stored
160 * @param pixels source
161 * @param line_size number of bytes in a horizontal line of block
162 * @param h height
163 */
669ac79c 164 op_pixels_func put_pixels_tab[4][4];
d518aebd
MN
165
166 /**
167 * Halfpel motion compensation with rounding (a+b+1)>>1.
4e8eed2f 168 * This is an array[4][4] of motion compensation functions for 4
da3b9756 169 * horizontal blocksizes (2,4,8,16) and the 4 halfpel positions<br>
5755c27f 170 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
171 * @param block destination into which the result is averaged (a+b+1)>>1
172 * @param pixels source
173 * @param line_size number of bytes in a horizontal line of block
174 * @param h height
175 */
da3b9756 176 op_pixels_func avg_pixels_tab[4][4];
d518aebd
MN
177
178 /**
179 * Halfpel motion compensation with no rounding (a+b)>>1.
eb14c713
MN
180 * this is an array[2][4] of motion compensation funcions for 2
181 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 182 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
183 * @param block destination where the result is stored
184 * @param pixels source
185 * @param line_size number of bytes in a horizontal line of block
186 * @param h height
187 */
eb4b3dd3 188 op_pixels_func put_no_rnd_pixels_tab[2][4];
d518aebd
MN
189
190 /**
191 * Halfpel motion compensation with no rounding (a+b)>>1.
eb14c713
MN
192 * this is an array[2][4] of motion compensation funcions for 2
193 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 194 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
195 * @param block destination into which the result is averaged (a+b)>>1
196 * @param pixels source
197 * @param line_size number of bytes in a horizontal line of block
198 * @param h height
199 */
eb4b3dd3 200 op_pixels_func avg_no_rnd_pixels_tab[2][4];
669ac79c
MN
201
202 /**
203 * Thirdpel motion compensation with rounding (a+b+1)>>1.
204 * this is an array[12] of motion compensation funcions for the 9 thirdpel positions<br>
205 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
206 * @param block destination where the result is stored
207 * @param pixels source
208 * @param line_size number of bytes in a horizontal line of block
209 * @param h height
210 */
211 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
da3b9756
MM
212 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
213
eb4b3dd3
ZK
214 qpel_mc_func put_qpel_pixels_tab[2][16];
215 qpel_mc_func avg_qpel_pixels_tab[2][16];
216 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
217 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
1457ab52 218 qpel_mc_func put_mspel_pixels_tab[8];
0da71265
MN
219
220 /**
221 * h264 Chram MC
222 */
223 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
224 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
eb4b3dd3 225
0da71265
MN
226 qpel_mc_func put_h264_qpel_pixels_tab[3][16];
227 qpel_mc_func avg_h264_qpel_pixels_tab[3][16];
228
bb198e19 229 me_cmp_func pix_abs[2][4];
11f18faf
MN
230
231 /* huffyuv specific */
11f18faf 232 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
1457ab52 233 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
84705403
MN
234 /**
235 * subtract huffyuv's variant of median prediction
236 * note, this might read from src1[-1], src2[-1]
237 */
238 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
3d2e8cce 239 void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
b0368839 240
332f9ac4
MN
241 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
242 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
243
b0368839
MN
244 /* (I)DCT */
245 void (*fdct)(DCTELEM *block/* align 16*/);
10acc479 246 void (*fdct248)(DCTELEM *block/* align 16*/);
4fb518c3
MN
247
248 /* IDCT really*/
249 void (*idct)(DCTELEM *block/* align 16*/);
24641185
MN
250
251 /**
77c92c2d 252 * block -> idct -> clip to unsigned 8 bit -> dest.
24641185 253 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
a1adf436 254 * @param line_size size in bytes of a horizotal line of dest
24641185 255 */
b0368839 256 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
24641185
MN
257
258 /**
259 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
a1adf436 260 * @param line_size size in bytes of a horizotal line of dest
24641185 261 */
b0368839 262 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
24641185
MN
263
264 /**
77c92c2d 265 * idct input permutation.
05493021
MN
266 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
267 * IDCT)
268 * this permutation must be performed before the idct_put/add, note, normally this can be merged
269 * with the zigzag/alternate scan<br>
24641185
MN
270 * an example to avoid confusion:
271 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
272 * - (x -> referece dct -> reference idct -> x)
273 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
274 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
275 */
b0368839
MN
276 uint8_t idct_permutation[64];
277 int idct_permutation_type;
278#define FF_NO_IDCT_PERM 1
279#define FF_LIBMPEG2_IDCT_PERM 2
280#define FF_SIMPLE_IDCT_PERM 3
281#define FF_TRANSPOSE_IDCT_PERM 4
282
eb4b3dd3
ZK
283} DSPContext;
284
59cf08ce 285void dsputil_static_init(void);
b0368839 286void dsputil_init(DSPContext* p, AVCodecContext *avctx);
de6d9b64 287
7801d21d
MN
288/**
289 * permute block according to permuatation.
290 * @param last last non zero element in scantable order
291 */
0c1a9eda 292void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
e0eac44e 293
d8085ea7
MN
294#define BYTE_VEC32(c) ((c)*0x01010101UL)
295
296static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
297{
298 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
299}
300
301static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
302{
303 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
304}
305
24641185 306/**
77c92c2d 307 * Empty mmx state.
24641185
MN
308 * this must be called between any dsp function and float/double code.
309 * for example sin(); dsp->idct_put(); emms_c(); cos()
310 */
eb4b3dd3
ZK
311#define emms_c()
312
e629ab68
RD
313/* should be defined by architectures supporting
314 one or more MultiMedia extension */
315int mm_support(void);
316
3d03c0a2 317#if defined(HAVE_MMX)
de6d9b64 318
18f77016 319#undef emms_c
eb4b3dd3 320
de6d9b64
FB
321#define MM_MMX 0x0001 /* standard MMX */
322#define MM_3DNOW 0x0004 /* AMD 3DNOW */
323#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
324#define MM_SSE 0x0008 /* SSE functions */
325#define MM_SSE2 0x0010 /* PIV SSE2 functions */
326
327extern int mm_flags;
328
0c1a9eda
ZK
329void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
330void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
de6d9b64
FB
331
332static inline void emms(void)
333{
fb16b7e7
FB
334 __asm __volatile ("emms;":::"memory");
335}
336
1457ab52 337
fb16b7e7
FB
338#define emms_c() \
339{\
340 if (mm_flags & MM_MMX)\
341 emms();\
de6d9b64
FB
342}
343
344#define __align8 __attribute__ ((aligned (8)))
345
b0368839
MN
346void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
347void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
2720569a 348
3d03c0a2
FB
349#elif defined(ARCH_ARMV4L)
350
3d03c0a2
FB
351/* This is to use 4 bytes read to the IDCT pointers for some 'zero'
352 line ptimizations */
353#define __align8 __attribute__ ((aligned (4)))
354
b0368839 355void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx);
3d03c0a2 356
c34270f5 357#elif defined(HAVE_MLIB)
c34270f5
FB
358
359/* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
360#define __align8 __attribute__ ((aligned (8)))
361
b0368839 362void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
c34270f5 363
1e98dffb
NK
364#elif defined(ARCH_ALPHA)
365
1e98dffb
NK
366#define __align8 __attribute__ ((aligned (8)))
367
b0368839 368void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
1e98dffb 369
59925ef2
BF
370#elif defined(ARCH_POWERPC)
371
404d2241
BF
372#define MM_ALTIVEC 0x0001 /* standard AltiVec */
373
374extern int mm_flags;
375
3b991c54 376#if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
9a197a24 377#define pixel altivec_pixel
3b991c54 378#include <altivec.h>
9a197a24 379#undef pixel
3b991c54
RD
380#endif
381
59925ef2
BF
382#define __align8 __attribute__ ((aligned (16)))
383
b0368839 384void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
59925ef2 385
d46aba26
LS
386#elif defined(HAVE_MMI)
387
d46aba26
LS
388#define __align8 __attribute__ ((aligned (16)))
389
b0368839 390void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
d46aba26 391
0c6bd2ea
B
392#elif defined(ARCH_SH4)
393
394#define __align8 __attribute__ ((aligned (8)))
395
396void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
397
de6d9b64
FB
398#else
399
400#define __align8
401
402#endif
403
6d4985bb
FB
404#ifdef __GNUC__
405
406struct unaligned_64 { uint64_t l; } __attribute__((packed));
407struct unaligned_32 { uint32_t l; } __attribute__((packed));
669ac79c 408struct unaligned_16 { uint16_t l; } __attribute__((packed));
6d4985bb 409
669ac79c 410#define LD16(a) (((const struct unaligned_16 *) (a))->l)
6d4985bb
FB
411#define LD32(a) (((const struct unaligned_32 *) (a))->l)
412#define LD64(a) (((const struct unaligned_64 *) (a))->l)
413
414#define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
415
416#else /* __GNUC__ */
417
669ac79c 418#define LD16(a) (*((uint16_t*)(a)))
6d4985bb
FB
419#define LD32(a) (*((uint32_t*)(a)))
420#define LD64(a) (*((uint64_t*)(a)))
421
422#define ST32(a, b) *((uint32_t*)(a)) = (b)
423
424#endif /* !__GNUC__ */
425
43f1708f 426/* PSNR */
0c1a9eda 427void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
43f1708f
J
428 int orig_linesize[3], int coded_linesize,
429 AVCodecContext *avctx);
bb6f5690
FB
430
431/* FFT computation */
432
433/* NOTE: soon integer code will be added, so you must use the
434 FFTSample type */
435typedef float FFTSample;
436
437typedef struct FFTComplex {
438 FFTSample re, im;
439} FFTComplex;
440
441typedef struct FFTContext {
442 int nbits;
443 int inverse;
444 uint16_t *revtab;
445 FFTComplex *exptab;
446 FFTComplex *exptab1; /* only used by SSE code */
447 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
448} FFTContext;
449
450int fft_init(FFTContext *s, int nbits, int inverse);
451void fft_permute(FFTContext *s, FFTComplex *z);
452void fft_calc_c(FFTContext *s, FFTComplex *z);
453void fft_calc_sse(FFTContext *s, FFTComplex *z);
8d268a7d
FB
454void fft_calc_altivec(FFTContext *s, FFTComplex *z);
455
bb6f5690
FB
456static inline void fft_calc(FFTContext *s, FFTComplex *z)
457{
458 s->fft_calc(s, z);
459}
460void fft_end(FFTContext *s);
461
462/* MDCT computation */
463
464typedef struct MDCTContext {
465 int n; /* size of MDCT (i.e. number of input data * 2) */
466 int nbits; /* n = 2^nbits */
467 /* pre/post rotation tables */
468 FFTSample *tcos;
469 FFTSample *tsin;
470 FFTContext fft;
471} MDCTContext;
472
82696bee 473int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
eb4b3dd3 474void ff_imdct_calc(MDCTContext *s, FFTSample *output,
bb6f5690 475 const FFTSample *input, FFTSample *tmp);
eb4b3dd3 476void ff_mdct_calc(MDCTContext *s, FFTSample *out,
bb6f5690 477 const FFTSample *input, FFTSample *tmp);
82696bee 478void ff_mdct_end(MDCTContext *s);
bb6f5690 479
bb198e19
MN
480#define WARPER8_16(name8, name16)\
481static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
482 return name8(s, dst , src , stride, h)\
483 +name8(s, dst+8 , src+8 , stride, h);\
484}
485
486#define WARPER8_16_SQ(name8, name16)\
487static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
488 int score=0;\
489 score +=name8(s, dst , src , stride, 8);\
490 score +=name8(s, dst+8 , src+8 , stride, 8);\
491 if(h==16){\
492 dst += 8*stride;\
493 src += 8*stride;\
494 score +=name8(s, dst , src , stride, 8);\
495 score +=name8(s, dst+8 , src+8 , stride, 8);\
496 }\
497 return score;\
1457ab52
MN
498}
499
95e2ce4a 500#ifndef HAVE_LRINTF
9d85cbd9
FB
501/* XXX: add ISOC specific test to avoid specific BSD testing. */
502/* better than nothing implementation. */
6234d753 503/* btw, rintf() is existing on fbsd too -- alex */
9d85cbd9
FB
504static inline long int lrintf(float x)
505{
ea937d01
FB
506#ifdef CONFIG_WIN32
507 /* XXX: incorrect, but make it compile */
508 return (int)(x);
509#else
9d85cbd9 510 return (int)(rint(x));
ea937d01 511#endif
9d85cbd9
FB
512}
513#endif
514
de6d9b64 515#endif