simplify
[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
7d67aa9b 82//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
0c1a9eda 83typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
669ac79c 84typedef 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 85typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
0da71265 86typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
b3184779 87
db794953 88#define DEF_OLD_QPEL(name)\
0c1a9eda
ZK
89void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
90void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
91void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
db794953
MN
92
93DEF_OLD_QPEL(qpel16_mc11_old_c)
94DEF_OLD_QPEL(qpel16_mc31_old_c)
95DEF_OLD_QPEL(qpel16_mc12_old_c)
96DEF_OLD_QPEL(qpel16_mc32_old_c)
97DEF_OLD_QPEL(qpel16_mc13_old_c)
98DEF_OLD_QPEL(qpel16_mc33_old_c)
99DEF_OLD_QPEL(qpel8_mc11_old_c)
100DEF_OLD_QPEL(qpel8_mc31_old_c)
101DEF_OLD_QPEL(qpel8_mc12_old_c)
102DEF_OLD_QPEL(qpel8_mc32_old_c)
103DEF_OLD_QPEL(qpel8_mc13_old_c)
104DEF_OLD_QPEL(qpel8_mc33_old_c)
b3184779
MN
105
106#define CALL_2X_PIXELS(a, b, n)\
107static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
108 b(block , pixels , line_size, h);\
109 b(block+n, pixels+n, line_size, h);\
110}
44eb4951 111
de6d9b64 112/* motion estimation */
7d67aa9b
MN
113// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
114// allthough currently h<4 is not used as functions with width <8 are not used and neither implemented
bb198e19 115typedef 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 116
0da71265 117
24641185
MN
118/**
119 * DSPContext.
120 */
eb4b3dd3
ZK
121typedef struct DSPContext {
122 /* pixel ops : interface with DCT */
0c1a9eda
ZK
123 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
124 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
125 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
126 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
d518aebd
MN
127 /**
128 * translational global motion compensation.
129 */
0c1a9eda 130 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
d518aebd
MN
131 /**
132 * global motion compensation.
133 */
0c1a9eda 134 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
eb4b3dd3
ZK
135 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
136 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
0c1a9eda
ZK
137 int (*pix_sum)(uint8_t * pix, int line_size);
138 int (*pix_norm1)(uint8_t * pix, int line_size);
bb198e19
MN
139// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
140
141 me_cmp_func sad[4]; /* identical to pix_absAxA except additional void * */
142 me_cmp_func sse[4];
143 me_cmp_func hadamard8_diff[4];
144 me_cmp_func dct_sad[4];
145 me_cmp_func quant_psnr[4];
146 me_cmp_func bit[4];
147 me_cmp_func rd[4];
1457ab52
MN
148 int (*hadamard8_abs )(uint8_t *src, int stride, int mean);
149
bb198e19
MN
150 me_cmp_func me_pre_cmp[5];
151 me_cmp_func me_cmp[5];
152 me_cmp_func me_sub_cmp[5];
153 me_cmp_func mb_cmp[5];
eb4b3dd3 154
669ac79c 155 /* maybe create an array for 16/8/4/2 functions */
d518aebd
MN
156 /**
157 * Halfpel motion compensation with rounding (a+b+1)>>1.
669ac79c 158 * this is an array[4][4] of motion compensation funcions for 4
e5771f4f 159 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 160 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
161 * @param block destination where the result is stored
162 * @param pixels source
163 * @param line_size number of bytes in a horizontal line of block
164 * @param h height
165 */
669ac79c 166 op_pixels_func put_pixels_tab[4][4];
d518aebd
MN
167
168 /**
169 * Halfpel motion compensation with rounding (a+b+1)>>1.
4e8eed2f 170 * This is an array[4][4] of motion compensation functions for 4
e5771f4f 171 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 172 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
173 * @param block destination into which the result is averaged (a+b+1)>>1
174 * @param pixels source
175 * @param line_size number of bytes in a horizontal line of block
176 * @param h height
177 */
da3b9756 178 op_pixels_func avg_pixels_tab[4][4];
d518aebd
MN
179
180 /**
181 * Halfpel motion compensation with no rounding (a+b)>>1.
eb14c713
MN
182 * this is an array[2][4] of motion compensation funcions for 2
183 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 184 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
185 * @param block destination where the result is stored
186 * @param pixels source
187 * @param line_size number of bytes in a horizontal line of block
188 * @param h height
189 */
eb4b3dd3 190 op_pixels_func put_no_rnd_pixels_tab[2][4];
d518aebd
MN
191
192 /**
193 * Halfpel motion compensation with no rounding (a+b)>>1.
eb14c713
MN
194 * this is an array[2][4] of motion compensation funcions for 2
195 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 196 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
197 * @param block destination into which the result is averaged (a+b)>>1
198 * @param pixels source
199 * @param line_size number of bytes in a horizontal line of block
200 * @param h height
201 */
eb4b3dd3 202 op_pixels_func avg_no_rnd_pixels_tab[2][4];
669ac79c
MN
203
204 /**
205 * Thirdpel motion compensation with rounding (a+b+1)>>1.
206 * this is an array[12] of motion compensation funcions for the 9 thirdpel positions<br>
207 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
208 * @param block destination where the result is stored
209 * @param pixels source
210 * @param line_size number of bytes in a horizontal line of block
211 * @param h height
212 */
213 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
da3b9756
MM
214 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
215
eb4b3dd3
ZK
216 qpel_mc_func put_qpel_pixels_tab[2][16];
217 qpel_mc_func avg_qpel_pixels_tab[2][16];
218 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
219 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
1457ab52 220 qpel_mc_func put_mspel_pixels_tab[8];
0da71265
MN
221
222 /**
223 * h264 Chram MC
224 */
225 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
226 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
eb4b3dd3 227
0da71265
MN
228 qpel_mc_func put_h264_qpel_pixels_tab[3][16];
229 qpel_mc_func avg_h264_qpel_pixels_tab[3][16];
230
bb198e19 231 me_cmp_func pix_abs[2][4];
11f18faf
MN
232
233 /* huffyuv specific */
11f18faf 234 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
1457ab52 235 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
84705403
MN
236 /**
237 * subtract huffyuv's variant of median prediction
238 * note, this might read from src1[-1], src2[-1]
239 */
240 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
3d2e8cce 241 void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
b0368839 242
332f9ac4
MN
243 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
244 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
245
b0368839
MN
246 /* (I)DCT */
247 void (*fdct)(DCTELEM *block/* align 16*/);
10acc479 248 void (*fdct248)(DCTELEM *block/* align 16*/);
4fb518c3
MN
249
250 /* IDCT really*/
251 void (*idct)(DCTELEM *block/* align 16*/);
24641185
MN
252
253 /**
77c92c2d 254 * block -> idct -> clip to unsigned 8 bit -> dest.
24641185 255 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
a1adf436 256 * @param line_size size in bytes of a horizotal line of dest
24641185 257 */
b0368839 258 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
24641185
MN
259
260 /**
261 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
a1adf436 262 * @param line_size size in bytes of a horizotal line of dest
24641185 263 */
b0368839 264 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
24641185
MN
265
266 /**
77c92c2d 267 * idct input permutation.
05493021
MN
268 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
269 * IDCT)
270 * this permutation must be performed before the idct_put/add, note, normally this can be merged
271 * with the zigzag/alternate scan<br>
24641185
MN
272 * an example to avoid confusion:
273 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
274 * - (x -> referece dct -> reference idct -> x)
275 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
276 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
277 */
b0368839
MN
278 uint8_t idct_permutation[64];
279 int idct_permutation_type;
280#define FF_NO_IDCT_PERM 1
281#define FF_LIBMPEG2_IDCT_PERM 2
282#define FF_SIMPLE_IDCT_PERM 3
283#define FF_TRANSPOSE_IDCT_PERM 4
284
eb4b3dd3
ZK
285} DSPContext;
286
59cf08ce 287void dsputil_static_init(void);
b0368839 288void dsputil_init(DSPContext* p, AVCodecContext *avctx);
de6d9b64 289
7801d21d
MN
290/**
291 * permute block according to permuatation.
292 * @param last last non zero element in scantable order
293 */
0c1a9eda 294void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
e0eac44e 295
d8085ea7
MN
296#define BYTE_VEC32(c) ((c)*0x01010101UL)
297
298static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
299{
300 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
301}
302
303static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
304{
305 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
306}
307
24641185 308/**
77c92c2d 309 * Empty mmx state.
24641185
MN
310 * this must be called between any dsp function and float/double code.
311 * for example sin(); dsp->idct_put(); emms_c(); cos()
312 */
eb4b3dd3
ZK
313#define emms_c()
314
e629ab68
RD
315/* should be defined by architectures supporting
316 one or more MultiMedia extension */
317int mm_support(void);
318
3d03c0a2 319#if defined(HAVE_MMX)
de6d9b64 320
18f77016 321#undef emms_c
eb4b3dd3 322
de6d9b64
FB
323#define MM_MMX 0x0001 /* standard MMX */
324#define MM_3DNOW 0x0004 /* AMD 3DNOW */
325#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
326#define MM_SSE 0x0008 /* SSE functions */
327#define MM_SSE2 0x0010 /* PIV SSE2 functions */
328
329extern int mm_flags;
330
0c1a9eda
ZK
331void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
332void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
de6d9b64
FB
333
334static inline void emms(void)
335{
fb16b7e7
FB
336 __asm __volatile ("emms;":::"memory");
337}
338
1457ab52 339
fb16b7e7
FB
340#define emms_c() \
341{\
342 if (mm_flags & MM_MMX)\
343 emms();\
de6d9b64
FB
344}
345
346#define __align8 __attribute__ ((aligned (8)))
347
b0368839
MN
348void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
349void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
2720569a 350
3d03c0a2
FB
351#elif defined(ARCH_ARMV4L)
352
3d03c0a2
FB
353/* This is to use 4 bytes read to the IDCT pointers for some 'zero'
354 line ptimizations */
355#define __align8 __attribute__ ((aligned (4)))
356
b0368839 357void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx);
3d03c0a2 358
c34270f5 359#elif defined(HAVE_MLIB)
c34270f5
FB
360
361/* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
362#define __align8 __attribute__ ((aligned (8)))
363
b0368839 364void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
c34270f5 365
1e98dffb
NK
366#elif defined(ARCH_ALPHA)
367
1e98dffb
NK
368#define __align8 __attribute__ ((aligned (8)))
369
b0368839 370void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
1e98dffb 371
59925ef2
BF
372#elif defined(ARCH_POWERPC)
373
404d2241
BF
374#define MM_ALTIVEC 0x0001 /* standard AltiVec */
375
376extern int mm_flags;
377
3b991c54 378#if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
9a197a24 379#define pixel altivec_pixel
3b991c54 380#include <altivec.h>
9a197a24 381#undef pixel
3b991c54
RD
382#endif
383
59925ef2
BF
384#define __align8 __attribute__ ((aligned (16)))
385
b0368839 386void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
59925ef2 387
d46aba26
LS
388#elif defined(HAVE_MMI)
389
d46aba26
LS
390#define __align8 __attribute__ ((aligned (16)))
391
b0368839 392void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
d46aba26 393
0c6bd2ea
B
394#elif defined(ARCH_SH4)
395
396#define __align8 __attribute__ ((aligned (8)))
397
398void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
399
de6d9b64
FB
400#else
401
402#define __align8
403
404#endif
405
6d4985bb
FB
406#ifdef __GNUC__
407
408struct unaligned_64 { uint64_t l; } __attribute__((packed));
409struct unaligned_32 { uint32_t l; } __attribute__((packed));
669ac79c 410struct unaligned_16 { uint16_t l; } __attribute__((packed));
6d4985bb 411
669ac79c 412#define LD16(a) (((const struct unaligned_16 *) (a))->l)
6d4985bb
FB
413#define LD32(a) (((const struct unaligned_32 *) (a))->l)
414#define LD64(a) (((const struct unaligned_64 *) (a))->l)
415
416#define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
417
418#else /* __GNUC__ */
419
669ac79c 420#define LD16(a) (*((uint16_t*)(a)))
6d4985bb
FB
421#define LD32(a) (*((uint32_t*)(a)))
422#define LD64(a) (*((uint64_t*)(a)))
423
424#define ST32(a, b) *((uint32_t*)(a)) = (b)
425
426#endif /* !__GNUC__ */
427
43f1708f 428/* PSNR */
0c1a9eda 429void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
43f1708f
J
430 int orig_linesize[3], int coded_linesize,
431 AVCodecContext *avctx);
bb6f5690
FB
432
433/* FFT computation */
434
435/* NOTE: soon integer code will be added, so you must use the
436 FFTSample type */
437typedef float FFTSample;
438
439typedef struct FFTComplex {
440 FFTSample re, im;
441} FFTComplex;
442
443typedef struct FFTContext {
444 int nbits;
445 int inverse;
446 uint16_t *revtab;
447 FFTComplex *exptab;
448 FFTComplex *exptab1; /* only used by SSE code */
449 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
450} FFTContext;
451
452int fft_init(FFTContext *s, int nbits, int inverse);
453void fft_permute(FFTContext *s, FFTComplex *z);
454void fft_calc_c(FFTContext *s, FFTComplex *z);
455void fft_calc_sse(FFTContext *s, FFTComplex *z);
8d268a7d
FB
456void fft_calc_altivec(FFTContext *s, FFTComplex *z);
457
bb6f5690
FB
458static inline void fft_calc(FFTContext *s, FFTComplex *z)
459{
460 s->fft_calc(s, z);
461}
462void fft_end(FFTContext *s);
463
464/* MDCT computation */
465
466typedef struct MDCTContext {
467 int n; /* size of MDCT (i.e. number of input data * 2) */
468 int nbits; /* n = 2^nbits */
469 /* pre/post rotation tables */
470 FFTSample *tcos;
471 FFTSample *tsin;
472 FFTContext fft;
473} MDCTContext;
474
82696bee 475int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
eb4b3dd3 476void ff_imdct_calc(MDCTContext *s, FFTSample *output,
bb6f5690 477 const FFTSample *input, FFTSample *tmp);
eb4b3dd3 478void ff_mdct_calc(MDCTContext *s, FFTSample *out,
bb6f5690 479 const FFTSample *input, FFTSample *tmp);
82696bee 480void ff_mdct_end(MDCTContext *s);
bb6f5690 481
bb198e19
MN
482#define WARPER8_16(name8, name16)\
483static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
484 return name8(s, dst , src , stride, h)\
485 +name8(s, dst+8 , src+8 , stride, h);\
486}
487
488#define WARPER8_16_SQ(name8, name16)\
489static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
490 int score=0;\
491 score +=name8(s, dst , src , stride, 8);\
492 score +=name8(s, dst+8 , src+8 , stride, 8);\
493 if(h==16){\
494 dst += 8*stride;\
495 src += 8*stride;\
496 score +=name8(s, dst , src , stride, 8);\
497 score +=name8(s, dst+8 , src+8 , stride, 8);\
498 }\
499 return score;\
1457ab52
MN
500}
501
95e2ce4a 502#ifndef HAVE_LRINTF
9d85cbd9
FB
503/* XXX: add ISOC specific test to avoid specific BSD testing. */
504/* better than nothing implementation. */
6234d753 505/* btw, rintf() is existing on fbsd too -- alex */
9d85cbd9
FB
506static inline long int lrintf(float x)
507{
ea937d01
FB
508#ifdef CONFIG_WIN32
509 /* XXX: incorrect, but make it compile */
510 return (int)(x);
511#else
9d85cbd9 512 return (int)(rint(x));
ea937d01 513#endif
9d85cbd9
FB
514}
515#endif
516
de6d9b64 517#endif