3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * This file is part of Libav.
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 * note, many functions in here may use MMX which trashes the FPU state, it is
27 * absolutely necessary to call emms_c() between dsp & float/double code
30 #ifndef AVCODEC_DSPUTIL_H
31 #define AVCODEC_DSPUTIL_H
33 #include "libavutil/intreadwrite.h"
40 void ff_fdct_ifast(int16_t *data);
41 void ff_fdct_ifast248(int16_t *data);
42 void ff_jpeg_fdct_islow_8(int16_t *data);
43 void ff_jpeg_fdct_islow_10(int16_t *data);
44 void ff_fdct248_islow_8(int16_t *data);
45 void ff_fdct248_islow_10(int16_t *data);
47 void ff_j_rev_dct(int16_t *data);
49 void ff_fdct_mmx(int16_t *block);
50 void ff_fdct_mmxext(int16_t *block);
51 void ff_fdct_sse2(int16_t *block);
53 #define H264_IDCT(depth) \
54 void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, int16_t *block, int stride);\
55 void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, int16_t *block, int stride);\
56 void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, int16_t *block, int stride);\
57 void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, int16_t *block, int stride);\
58 void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[6*8]);\
59 void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[6*8]);\
60 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[6*8]);\
61 void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[6*8]);\
62 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[6*8]);\
63 void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(int16_t *output, int16_t *input, int qmul);\
64 void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(int16_t *block, int qmul);\
65 void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(int16_t *block, int qmul);
71 void ff_svq3_luma_dc_dequant_idct_c(int16_t *output, int16_t *input, int qp);
72 void ff_svq3_add_idct_c(uint8_t *dst, int16_t *block, int stride, int qp, int dc);
75 extern const uint8_t ff_alternate_horizontal_scan[64];
76 extern const uint8_t ff_alternate_vertical_scan[64];
77 extern const uint8_t ff_zigzag_direct[64];
78 extern const uint8_t ff_zigzag248_direct[64];
80 /* pixel operations */
81 #define MAX_NEG_CROP 1024
84 extern uint32_t ff_squareTbl[512];
85 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
87 #define PUTAVG_PIXELS(depth)\
88 void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
89 void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
90 void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
91 void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);
97 #define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
98 #define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
99 #define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
100 #define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c
103 void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
104 void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
105 void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
106 void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
108 /* 1/2^n downscaling functions from imgconvert.c */
109 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
110 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
111 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
113 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
114 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
116 /* minimum alignment rules ;)
117 If you notice errors in the align stuff, need more alignment for some ASM code
118 for some CPU or need to use a function with less aligned data then send a mail
119 to the libav-devel mailing list, ...
121 !warning These alignments might not match reality, (missing attribute((align))
122 stuff somewhere possible).
123 I (Michael) did not check them, these are just the alignments which I think
124 could be reached easily ...
126 !future video codecs might need functions with less strict alignment
130 void get_pixels_c(int16_t *block, const uint8_t *pixels, int line_size);
131 void diff_pixels_c(int16_t *block, const uint8_t *s1, const uint8_t *s2, int stride);
132 void put_pixels_clamped_c(const int16_t *block, uint8_t *pixels, int line_size);
133 void add_pixels_clamped_c(const int16_t *block, uint8_t *pixels, int line_size);
134 void clear_blocks_c(int16_t *blocks);
137 /* add and put pixel (decoding) */
138 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
139 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
140 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
141 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);
142 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
143 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
145 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
147 #define DEF_OLD_QPEL(name)\
148 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
149 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
150 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
152 DEF_OLD_QPEL(qpel16_mc11_old_c)
153 DEF_OLD_QPEL(qpel16_mc31_old_c)
154 DEF_OLD_QPEL(qpel16_mc12_old_c)
155 DEF_OLD_QPEL(qpel16_mc32_old_c)
156 DEF_OLD_QPEL(qpel16_mc13_old_c)
157 DEF_OLD_QPEL(qpel16_mc33_old_c)
158 DEF_OLD_QPEL(qpel8_mc11_old_c)
159 DEF_OLD_QPEL(qpel8_mc31_old_c)
160 DEF_OLD_QPEL(qpel8_mc12_old_c)
161 DEF_OLD_QPEL(qpel8_mc32_old_c)
162 DEF_OLD_QPEL(qpel8_mc13_old_c)
163 DEF_OLD_QPEL(qpel8_mc33_old_c)
165 #define CALL_2X_PIXELS(a, b, n)\
166 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
167 b(block , pixels , line_size, h);\
168 b(block+n, pixels+n, line_size, h);\
171 /* motion estimation */
172 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
173 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
174 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))*/;
179 typedef struct ScanTable{
180 const uint8_t *scantable;
181 uint8_t permutated[64];
182 uint8_t raster_end[64];
185 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
186 void ff_init_scantable_permutation(uint8_t *idct_permutation,
187 int idct_permutation_type);
192 typedef struct DSPContext {
194 * Size of DCT coefficients.
198 /* pixel ops : interface with DCT */
199 void (*get_pixels)(int16_t *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
200 void (*diff_pixels)(int16_t *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
201 void (*put_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
202 void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
203 void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
204 void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
205 void (*add_pixels4)(uint8_t *pixels, int16_t *block, int line_size);
206 int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
208 * translational global motion compensation.
210 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
212 * global motion compensation.
214 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
215 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
216 void (*clear_block)(int16_t *block/*align 16*/);
217 void (*clear_blocks)(int16_t *blocks/*align 16*/);
218 int (*pix_sum)(uint8_t * pix, int line_size);
219 int (*pix_norm1)(uint8_t * pix, int line_size);
220 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
222 me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
224 me_cmp_func hadamard8_diff[6];
225 me_cmp_func dct_sad[6];
226 me_cmp_func quant_psnr[6];
232 me_cmp_func dct_max[6];
233 me_cmp_func dct264_sad[6];
235 me_cmp_func me_pre_cmp[6];
236 me_cmp_func me_cmp[6];
237 me_cmp_func me_sub_cmp[6];
238 me_cmp_func mb_cmp[6];
239 me_cmp_func ildct_cmp[6]; //only width 16 used
240 me_cmp_func frame_skip_cmp[6]; //only width 8 used
242 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
246 * Halfpel motion compensation with rounding (a+b+1)>>1.
247 * this is an array[4][4] of motion compensation functions for 4
248 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
249 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
250 * @param block destination where the result is stored
251 * @param pixels source
252 * @param line_size number of bytes in a horizontal line of block
255 op_pixels_func put_pixels_tab[4][4];
258 * Halfpel motion compensation with rounding (a+b+1)>>1.
259 * This is an array[4][4] of motion compensation functions for 4
260 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
261 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
262 * @param block destination into which the result is averaged (a+b+1)>>1
263 * @param pixels source
264 * @param line_size number of bytes in a horizontal line of block
267 op_pixels_func avg_pixels_tab[4][4];
270 * Halfpel motion compensation with no rounding (a+b)>>1.
271 * this is an array[2][4] of motion compensation functions for 2
272 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
273 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
274 * @param block destination where the result is stored
275 * @param pixels source
276 * @param line_size number of bytes in a horizontal line of block
279 op_pixels_func put_no_rnd_pixels_tab[4][4];
282 * Halfpel motion compensation with no rounding (a+b)>>1.
283 * this is an array[2][4] of motion compensation functions for 2
284 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
285 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
286 * @param block destination into which the result is averaged (a+b)>>1
287 * @param pixels source
288 * @param line_size number of bytes in a horizontal line of block
291 op_pixels_func avg_no_rnd_pixels_tab[4][4];
294 * Thirdpel motion compensation with rounding (a+b+1)>>1.
295 * this is an array[12] of motion compensation functions for the 9 thirdpe
297 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
298 * @param block destination where the result is stored
299 * @param pixels source
300 * @param line_size number of bytes in a horizontal line of block
303 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
304 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
306 qpel_mc_func put_qpel_pixels_tab[2][16];
307 qpel_mc_func avg_qpel_pixels_tab[2][16];
308 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
309 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
310 qpel_mc_func put_mspel_pixels_tab[8];
315 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
316 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
318 qpel_mc_func put_h264_qpel_pixels_tab[4][16];
319 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
321 me_cmp_func pix_abs[2][4];
323 /* huffyuv specific */
324 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
325 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
327 * subtract huffyuv's variant of median prediction
328 * note, this might read from src1[-1], src2[-1]
330 void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
331 void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
332 int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
333 void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
334 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
335 void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
337 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
338 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
340 void (*h261_loop_filter)(uint8_t *src, int stride);
342 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
343 void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
346 void (*fdct)(int16_t *block/* align 16*/);
347 void (*fdct248)(int16_t *block/* align 16*/);
350 void (*idct)(int16_t *block/* align 16*/);
353 * block -> idct -> clip to unsigned 8 bit -> dest.
354 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
355 * @param line_size size in bytes of a horizontal line of dest
357 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
360 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
361 * @param line_size size in bytes of a horizontal line of dest
363 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
366 * idct input permutation.
367 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
369 * this permutation must be performed before the idct_put/add, note, normally this can be merged
370 * with the zigzag/alternate scan<br>
371 * an example to avoid confusion:
372 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
373 * - (x -> reference dct -> reference idct -> x)
374 * - (x -> reference dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
375 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
377 uint8_t idct_permutation[64];
378 int idct_permutation_type;
379 #define FF_NO_IDCT_PERM 1
380 #define FF_LIBMPEG2_IDCT_PERM 2
381 #define FF_SIMPLE_IDCT_PERM 3
382 #define FF_TRANSPOSE_IDCT_PERM 4
383 #define FF_PARTTRANS_IDCT_PERM 5
384 #define FF_SSE2_IDCT_PERM 6
386 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
387 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
388 #define BASIS_SHIFT 16
389 #define RECON_SHIFT 6
391 void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
392 #define EDGE_WIDTH 16
394 #define EDGE_BOTTOM 2
396 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
399 * Calculate scalar product of two vectors.
400 * @param len length of vectors, should be multiple of 16
402 int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
405 * Calculate scalar product of v1 and v2,
406 * and v1[i] += v3[i] * mul
407 * @param len length of vectors, should be multiple of 16
409 int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
412 * Apply symmetric window in 16-bit fixed-point.
413 * @param output destination array
414 * constraints: 16-byte aligned
415 * @param input source array
416 * constraints: 16-byte aligned
417 * @param window window array
418 * constraints: 16-byte aligned, at least len/2 elements
419 * @param len full window length
420 * constraints: multiple of ? greater than zero
422 void (*apply_window_int16)(int16_t *output, const int16_t *input,
423 const int16_t *window, unsigned int len);
426 * Clip each element in an array of int32_t to a given minimum and maximum value.
427 * @param dst destination array
428 * constraints: 16-byte aligned
429 * @param src source array
430 * constraints: 16-byte aligned
431 * @param min minimum value
432 * constraints: must be in the range [-(1 << 24), 1 << 24]
433 * @param max maximum value
434 * constraints: must be in the range [-(1 << 24), 1 << 24]
435 * @param len number of elements in the array
436 * constraints: multiple of 32 greater than zero
438 void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
439 int32_t max, unsigned int len);
441 op_fill_func fill_block_tab[2];
444 void ff_dsputil_static_init(void);
445 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
447 int ff_check_alignment(void);
450 * permute block according to permuatation.
451 * @param last last non zero element in scantable order
453 void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
455 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
457 #define BYTE_VEC32(c) ((c)*0x01010101UL)
458 #define BYTE_VEC64(c) ((c)*0x0001000100010001UL)
460 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
462 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
465 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
467 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
470 static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
472 return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
475 static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
477 return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
480 static inline int get_penalty_factor(int lambda, int lambda2, int type){
484 return lambda>>FF_LAMBDA_SHIFT;
486 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
489 return (2*lambda)>>FF_LAMBDA_SHIFT;
494 return lambda2>>FF_LAMBDA_SHIFT;
500 void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
501 void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
502 void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
503 void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
504 void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
505 void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
506 void ff_dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
508 #if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMX
509 # define STRIDE_ALIGN 16
511 # define STRIDE_ALIGN 8
514 // Some broken preprocessors need a second expansion
515 // to be forced to tokenize __VA_ARGS__
518 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...) \
519 uint8_t la_##v[sizeof(t s o) + (a)]; \
520 t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
522 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) \
523 DECLARE_ALIGNED(a, t, la_##v) s o; \
526 #define LOCAL_ALIGNED(a, t, v, ...) E(LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,))
528 #if HAVE_LOCAL_ALIGNED_8
529 # define LOCAL_ALIGNED_8(t, v, ...) E(LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,))
531 # define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
534 #if HAVE_LOCAL_ALIGNED_16
535 # define LOCAL_ALIGNED_16(t, v, ...) E(LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,))
537 # define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
540 #define WRAPPER8_16_SQ(name8, name16)\
541 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
543 score +=name8(s, dst , src , stride, 8);\
544 score +=name8(s, dst+8 , src+8 , stride, 8);\
548 score +=name8(s, dst , src , stride, 8);\
549 score +=name8(s, dst+8 , src+8 , stride, 8);\
555 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
560 AV_COPY16U(dst, src);
566 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
571 AV_COPY32U(dst, src);
577 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
582 AV_COPY64U(dst, src);
588 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
593 AV_COPY64U(dst, src);
600 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
605 AV_COPY128U(dst, src);
611 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
616 AV_COPY128U(dst, src);
623 #endif /* AVCODEC_DSPUTIL_H */