3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * note, many functions in here may use MMX which trashes the FPU state, it is
25 * absolutely necessary to call emms_c() between dsp & float/double code
37 typedef short DCTELEM;
39 void fdct_ifast (DCTELEM *data);
40 void fdct_ifast248 (DCTELEM *data);
41 void ff_jpeg_fdct_islow (DCTELEM *data);
42 void ff_fdct248_islow (DCTELEM *data);
44 void j_rev_dct (DCTELEM *data);
45 void j_rev_dct4 (DCTELEM *data);
46 void j_rev_dct2 (DCTELEM *data);
47 void j_rev_dct1 (DCTELEM *data);
49 void ff_fdct_mmx(DCTELEM *block);
50 void ff_fdct_mmx2(DCTELEM *block);
51 void ff_fdct_sse2(DCTELEM *block);
53 void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
54 void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride);
55 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
56 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
59 extern const uint8_t ff_alternate_horizontal_scan[64];
60 extern const uint8_t ff_alternate_vertical_scan[64];
61 extern const uint8_t ff_zigzag_direct[64];
62 extern const uint8_t ff_zigzag248_direct[64];
64 /* pixel operations */
65 #define MAX_NEG_CROP 1024
68 extern uint32_t squareTbl[512];
69 extern uint8_t cropTbl[256 + 2 * MAX_NEG_CROP];
71 /* VP3 DSP functions */
72 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
73 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
74 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
76 /* minimum alignment rules ;)
77 if u notice errors in the align stuff, need more alignment for some asm code for some cpu
78 or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ...
80 !warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
81 i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
83 !future video codecs might need functions with less strict alignment
87 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
88 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
89 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
90 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
91 void clear_blocks_c(DCTELEM *blocks);
94 /* add and put pixel (decoding) */
95 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
96 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
97 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
98 typedef 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);
99 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
100 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
101 typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
102 typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
104 #define DEF_OLD_QPEL(name)\
105 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
106 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
107 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
109 DEF_OLD_QPEL(qpel16_mc11_old_c)
110 DEF_OLD_QPEL(qpel16_mc31_old_c)
111 DEF_OLD_QPEL(qpel16_mc12_old_c)
112 DEF_OLD_QPEL(qpel16_mc32_old_c)
113 DEF_OLD_QPEL(qpel16_mc13_old_c)
114 DEF_OLD_QPEL(qpel16_mc33_old_c)
115 DEF_OLD_QPEL(qpel8_mc11_old_c)
116 DEF_OLD_QPEL(qpel8_mc31_old_c)
117 DEF_OLD_QPEL(qpel8_mc12_old_c)
118 DEF_OLD_QPEL(qpel8_mc32_old_c)
119 DEF_OLD_QPEL(qpel8_mc13_old_c)
120 DEF_OLD_QPEL(qpel8_mc33_old_c)
122 #define CALL_2X_PIXELS(a, b, n)\
123 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
124 b(block , pixels , line_size, h);\
125 b(block+n, pixels+n, line_size, h);\
128 /* motion estimation */
129 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
130 // allthough currently h<4 is not used as functions with width <8 are not used and neither implemented
131 typedef 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))*/;
137 typedef struct DSPContext {
138 /* pixel ops : interface with DCT */
139 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
140 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
141 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
142 void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
143 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
144 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
145 void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
147 * translational global motion compensation.
149 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
151 * global motion compensation.
153 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
154 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
155 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
156 int (*pix_sum)(uint8_t * pix, int line_size);
157 int (*pix_norm1)(uint8_t * pix, int line_size);
158 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
160 me_cmp_func sad[5]; /* identical to pix_absAxA except additional void * */
162 me_cmp_func hadamard8_diff[5];
163 me_cmp_func dct_sad[5];
164 me_cmp_func quant_psnr[5];
172 me_cmp_func dct_max[5];
173 me_cmp_func dct264_sad[5];
175 me_cmp_func me_pre_cmp[5];
176 me_cmp_func me_cmp[5];
177 me_cmp_func me_sub_cmp[5];
178 me_cmp_func mb_cmp[5];
179 me_cmp_func ildct_cmp[5]; //only width 16 used
180 me_cmp_func frame_skip_cmp[5]; //only width 8 used
183 * Halfpel motion compensation with rounding (a+b+1)>>1.
184 * this is an array[4][4] of motion compensation funcions for 4
185 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
186 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
187 * @param block destination where the result is stored
188 * @param pixels source
189 * @param line_size number of bytes in a horizontal line of block
192 op_pixels_func put_pixels_tab[4][4];
195 * Halfpel motion compensation with rounding (a+b+1)>>1.
196 * This is an array[4][4] of motion compensation functions for 4
197 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
198 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
199 * @param block destination into which the result is averaged (a+b+1)>>1
200 * @param pixels source
201 * @param line_size number of bytes in a horizontal line of block
204 op_pixels_func avg_pixels_tab[4][4];
207 * Halfpel motion compensation with no rounding (a+b)>>1.
208 * this is an array[2][4] of motion compensation funcions for 2
209 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
210 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
211 * @param block destination where the result is stored
212 * @param pixels source
213 * @param line_size number of bytes in a horizontal line of block
216 op_pixels_func put_no_rnd_pixels_tab[4][4];
219 * Halfpel motion compensation with no rounding (a+b)>>1.
220 * this is an array[2][4] of motion compensation funcions for 2
221 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
222 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
223 * @param block destination into which the result is averaged (a+b)>>1
224 * @param pixels source
225 * @param line_size number of bytes in a horizontal line of block
228 op_pixels_func avg_no_rnd_pixels_tab[4][4];
230 void (*put_no_rnd_pixels_l2[2])(uint8_t *block/*align width (8 or 16)*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
233 * Thirdpel motion compensation with rounding (a+b+1)>>1.
234 * this is an array[12] of motion compensation funcions for the 9 thirdpel positions<br>
235 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
236 * @param block destination where the result is stored
237 * @param pixels source
238 * @param line_size number of bytes in a horizontal line of block
241 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
242 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
244 qpel_mc_func put_qpel_pixels_tab[2][16];
245 qpel_mc_func avg_qpel_pixels_tab[2][16];
246 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
247 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
248 qpel_mc_func put_mspel_pixels_tab[8];
253 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
254 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
256 qpel_mc_func put_h264_qpel_pixels_tab[4][16];
257 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
259 h264_weight_func weight_h264_pixels_tab[10];
260 h264_biweight_func biweight_h264_pixels_tab[10];
262 me_cmp_func pix_abs[2][4];
264 /* huffyuv specific */
265 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
266 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
268 * subtract huffyuv's variant of median prediction
269 * note, this might read from src1[-1], src2[-1]
271 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
272 void (*bswap_buf)(uint32_t *dst, uint32_t *src, int w);
274 void (*h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
275 void (*h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
276 void (*h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
277 void (*h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
278 void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
279 void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
281 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
282 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
284 void (*h261_loop_filter)(uint8_t *src, int stride);
287 void (*fdct)(DCTELEM *block/* align 16*/);
288 void (*fdct248)(DCTELEM *block/* align 16*/);
291 void (*idct)(DCTELEM *block/* align 16*/);
294 * block -> idct -> clip to unsigned 8 bit -> dest.
295 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
296 * @param line_size size in bytes of a horizotal line of dest
298 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
301 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
302 * @param line_size size in bytes of a horizotal line of dest
304 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
307 * idct input permutation.
308 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
310 * this permutation must be performed before the idct_put/add, note, normally this can be merged
311 * with the zigzag/alternate scan<br>
312 * an example to avoid confusion:
313 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
314 * - (x -> referece dct -> reference idct -> x)
315 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
316 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
318 uint8_t idct_permutation[64];
319 int idct_permutation_type;
320 #define FF_NO_IDCT_PERM 1
321 #define FF_LIBMPEG2_IDCT_PERM 2
322 #define FF_SIMPLE_IDCT_PERM 3
323 #define FF_TRANSPOSE_IDCT_PERM 4
324 #define FF_PARTTRANS_IDCT_PERM 5
326 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
327 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
328 #define BASIS_SHIFT 16
329 #define RECON_SHIFT 6
331 void (*h264_idct_add)(uint8_t *dst, DCTELEM *block, int stride);
332 void (*h264_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
335 void dsputil_static_init(void);
336 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
339 * permute block according to permuatation.
340 * @param last last non zero element in scantable order
342 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
344 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
346 #define BYTE_VEC32(c) ((c)*0x01010101UL)
348 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
350 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
353 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
355 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
358 static inline int get_penalty_factor(int lambda, int lambda2, int type){
362 return lambda>>FF_LAMBDA_SHIFT;
364 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
366 return (4*lambda)>>(FF_LAMBDA_SHIFT);
368 return (2*lambda)>>(FF_LAMBDA_SHIFT);
371 return (2*lambda)>>FF_LAMBDA_SHIFT;
376 return lambda2>>FF_LAMBDA_SHIFT;
384 * this must be called between any dsp function and float/double code.
385 * for example sin(); dsp->idct_put(); emms_c(); cos()
389 /* should be defined by architectures supporting
390 one or more MultiMedia extension */
391 int mm_support(void);
394 #define DECLARE_ALIGNED_16(t,v) t v __attribute__ ((aligned (16)))
396 #define DECLARE_ALIGNED_16(t,v) __declspec(align(16)) t v
399 #if defined(HAVE_MMX)
403 #define MM_MMX 0x0001 /* standard MMX */
404 #define MM_3DNOW 0x0004 /* AMD 3DNOW */
405 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
406 #define MM_SSE 0x0008 /* SSE functions */
407 #define MM_SSE2 0x0010 /* PIV SSE2 functions */
408 #define MM_3DNOWEXT 0x0020 /* AMD 3DNowExt */
412 void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
413 void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
414 void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
416 static inline void emms(void)
418 __asm __volatile ("emms;":::"memory");
424 if (mm_flags & MM_MMX)\
429 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
431 #define DECLARE_ALIGNED_8(t,v) __declspec(align(8)) t v
434 #define STRIDE_ALIGN 8
436 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
437 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
439 #elif defined(ARCH_ARMV4L)
441 /* This is to use 4 bytes read to the IDCT pointers for some 'zero'
442 line optimizations */
443 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (4)))
444 #define STRIDE_ALIGN 4
446 #define MM_IWMMXT 0x0100 /* XScale IWMMXT */
450 void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx);
452 #elif defined(HAVE_MLIB)
454 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
455 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
456 #define STRIDE_ALIGN 8
458 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
460 #elif defined(ARCH_SPARC)
462 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
463 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
464 #define STRIDE_ALIGN 8
465 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
467 #elif defined(ARCH_ALPHA)
469 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
470 #define STRIDE_ALIGN 8
472 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
474 #elif defined(ARCH_POWERPC)
476 #define MM_ALTIVEC 0x0001 /* standard AltiVec */
480 #if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
481 #define pixel altivec_pixel
486 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (16)))
487 #define STRIDE_ALIGN 16
489 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
491 #elif defined(HAVE_MMI)
493 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (16)))
494 #define STRIDE_ALIGN 16
496 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
498 #elif defined(ARCH_SH4)
500 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
501 #define STRIDE_ALIGN 8
503 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
507 #define DECLARE_ALIGNED_8(t,v) t v __attribute__ ((aligned (8)))
508 #define STRIDE_ALIGN 8
514 struct unaligned_64 { uint64_t l; } __attribute__((packed));
515 struct unaligned_32 { uint32_t l; } __attribute__((packed));
516 struct unaligned_16 { uint16_t l; } __attribute__((packed));
518 #define LD16(a) (((const struct unaligned_16 *) (a))->l)
519 #define LD32(a) (((const struct unaligned_32 *) (a))->l)
520 #define LD64(a) (((const struct unaligned_64 *) (a))->l)
522 #define ST16(a, b) (((struct unaligned_16 *) (a))->l) = (b)
523 #define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
527 #define LD16(a) (*((uint16_t*)(a)))
528 #define LD32(a) (*((uint32_t*)(a)))
529 #define LD64(a) (*((uint64_t*)(a)))
531 #define ST32(a, b) *((uint32_t*)(a)) = (b)
533 #endif /* !__GNUC__ */
536 void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
537 int orig_linesize[3], int coded_linesize,
538 AVCodecContext *avctx);
540 /* FFT computation */
542 /* NOTE: soon integer code will be added, so you must use the
544 typedef float FFTSample;
546 typedef struct FFTComplex {
550 typedef struct FFTContext {
555 FFTComplex *exptab1; /* only used by SSE code */
556 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
559 int ff_fft_init(FFTContext *s, int nbits, int inverse);
560 void ff_fft_permute(FFTContext *s, FFTComplex *z);
561 void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
562 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
563 void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z);
565 static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)
569 void ff_fft_end(FFTContext *s);
571 /* MDCT computation */
573 typedef struct MDCTContext {
574 int n; /* size of MDCT (i.e. number of input data * 2) */
575 int nbits; /* n = 2^nbits */
576 /* pre/post rotation tables */
582 int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
583 void ff_imdct_calc(MDCTContext *s, FFTSample *output,
584 const FFTSample *input, FFTSample *tmp);
585 void ff_mdct_calc(MDCTContext *s, FFTSample *out,
586 const FFTSample *input, FFTSample *tmp);
587 void ff_mdct_end(MDCTContext *s);
589 #define WARPER8_16(name8, name16)\
590 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
591 return name8(s, dst , src , stride, h)\
592 +name8(s, dst+8 , src+8 , stride, h);\
595 #define WARPER8_16_SQ(name8, name16)\
596 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
598 score +=name8(s, dst , src , stride, 8);\
599 score +=name8(s, dst+8 , src+8 , stride, 8);\
603 score +=name8(s, dst , src , stride, 8);\
604 score +=name8(s, dst+8 , src+8 , stride, 8);\