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*/, ptrdiff_t 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);
144 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
146 #define DEF_OLD_QPEL(name)\
147 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
148 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
149 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
151 DEF_OLD_QPEL(qpel16_mc11_old_c)
152 DEF_OLD_QPEL(qpel16_mc31_old_c)
153 DEF_OLD_QPEL(qpel16_mc12_old_c)
154 DEF_OLD_QPEL(qpel16_mc32_old_c)
155 DEF_OLD_QPEL(qpel16_mc13_old_c)
156 DEF_OLD_QPEL(qpel16_mc33_old_c)
157 DEF_OLD_QPEL(qpel8_mc11_old_c)
158 DEF_OLD_QPEL(qpel8_mc31_old_c)
159 DEF_OLD_QPEL(qpel8_mc12_old_c)
160 DEF_OLD_QPEL(qpel8_mc32_old_c)
161 DEF_OLD_QPEL(qpel8_mc13_old_c)
162 DEF_OLD_QPEL(qpel8_mc33_old_c)
164 #define CALL_2X_PIXELS(a, b, n)\
165 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
166 b(block , pixels , line_size, h);\
167 b(block+n, pixels+n, line_size, h);\
170 /* motion estimation */
171 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
172 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
173 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))*/;
178 typedef struct ScanTable{
179 const uint8_t *scantable;
180 uint8_t permutated[64];
181 uint8_t raster_end[64];
184 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
185 void ff_init_scantable_permutation(uint8_t *idct_permutation,
186 int idct_permutation_type);
191 typedef struct DSPContext {
193 * Size of DCT coefficients.
197 /* pixel ops : interface with DCT */
198 void (*get_pixels)(int16_t *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
199 void (*diff_pixels)(int16_t *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
200 void (*put_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
201 void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
202 void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
203 void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
204 void (*add_pixels4)(uint8_t *pixels, int16_t *block, int line_size);
205 int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
207 * translational global motion compensation.
209 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
211 * global motion compensation.
213 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
214 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
215 void (*clear_block)(int16_t *block/*align 16*/);
216 void (*clear_blocks)(int16_t *blocks/*align 16*/);
217 int (*pix_sum)(uint8_t * pix, int line_size);
218 int (*pix_norm1)(uint8_t * pix, int line_size);
219 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
221 me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
223 me_cmp_func hadamard8_diff[6];
224 me_cmp_func dct_sad[6];
225 me_cmp_func quant_psnr[6];
231 me_cmp_func dct_max[6];
232 me_cmp_func dct264_sad[6];
234 me_cmp_func me_pre_cmp[6];
235 me_cmp_func me_cmp[6];
236 me_cmp_func me_sub_cmp[6];
237 me_cmp_func mb_cmp[6];
238 me_cmp_func ildct_cmp[6]; //only width 16 used
239 me_cmp_func frame_skip_cmp[6]; //only width 8 used
241 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
245 * Halfpel motion compensation with rounding (a+b+1)>>1.
246 * this is an array[4][4] of motion compensation functions for 4
247 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
248 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
249 * @param block destination where the result is stored
250 * @param pixels source
251 * @param line_size number of bytes in a horizontal line of block
254 op_pixels_func put_pixels_tab[4][4];
257 * Halfpel motion compensation with rounding (a+b+1)>>1.
258 * This is an array[4][4] of motion compensation functions for 4
259 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
260 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
261 * @param block destination into which the result is averaged (a+b+1)>>1
262 * @param pixels source
263 * @param line_size number of bytes in a horizontal line of block
266 op_pixels_func avg_pixels_tab[4][4];
269 * Halfpel motion compensation with no rounding (a+b)>>1.
270 * this is an array[2][4] of motion compensation functions for 2
271 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
272 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
273 * @param block destination where the result is stored
274 * @param pixels source
275 * @param line_size number of bytes in a horizontal line of block
278 op_pixels_func put_no_rnd_pixels_tab[2][4];
281 * Halfpel motion compensation with no rounding (a+b)>>1.
282 * this is an array[4] of motion compensation functions for 1
283 * horizontal blocksize (16) and the 4 halfpel positions<br>
284 * *pixels_tab[0][ xhalfpel + 2*yhalfpel ]
285 * @param block destination into which the result is averaged (a+b)>>1
286 * @param pixels source
287 * @param line_size number of bytes in a horizontal line of block
290 op_pixels_func avg_no_rnd_pixels_tab[4];
293 * Thirdpel motion compensation with rounding (a+b+1)>>1.
294 * this is an array[12] of motion compensation functions for the 9 thirdpe
296 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
297 * @param block destination where the result is stored
298 * @param pixels source
299 * @param line_size number of bytes in a horizontal line of block
302 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
303 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
305 qpel_mc_func put_qpel_pixels_tab[2][16];
306 qpel_mc_func avg_qpel_pixels_tab[2][16];
307 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
308 qpel_mc_func put_mspel_pixels_tab[8];
310 me_cmp_func pix_abs[2][4];
312 /* huffyuv specific */
313 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
314 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
316 * subtract huffyuv's variant of median prediction
317 * note, this might read from src1[-1], src2[-1]
319 void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
320 void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
321 int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
322 void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
323 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
324 void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
326 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
327 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
329 void (*h261_loop_filter)(uint8_t *src, int stride);
331 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
332 void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
335 void (*fdct)(int16_t *block/* align 16*/);
336 void (*fdct248)(int16_t *block/* align 16*/);
339 void (*idct)(int16_t *block/* align 16*/);
342 * block -> idct -> clip to unsigned 8 bit -> dest.
343 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
344 * @param line_size size in bytes of a horizontal line of dest
346 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
349 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
350 * @param line_size size in bytes of a horizontal line of dest
352 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
355 * idct input permutation.
356 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
358 * this permutation must be performed before the idct_put/add, note, normally this can be merged
359 * with the zigzag/alternate scan<br>
360 * an example to avoid confusion:
361 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
362 * - (x -> reference dct -> reference idct -> x)
363 * - (x -> reference dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
364 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
366 uint8_t idct_permutation[64];
367 int idct_permutation_type;
368 #define FF_NO_IDCT_PERM 1
369 #define FF_LIBMPEG2_IDCT_PERM 2
370 #define FF_SIMPLE_IDCT_PERM 3
371 #define FF_TRANSPOSE_IDCT_PERM 4
372 #define FF_PARTTRANS_IDCT_PERM 5
373 #define FF_SSE2_IDCT_PERM 6
375 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
376 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
377 #define BASIS_SHIFT 16
378 #define RECON_SHIFT 6
380 void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
381 #define EDGE_WIDTH 16
383 #define EDGE_BOTTOM 2
385 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
388 * Calculate scalar product of two vectors.
389 * @param len length of vectors, should be multiple of 16
391 int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
394 * Calculate scalar product of v1 and v2,
395 * and v1[i] += v3[i] * mul
396 * @param len length of vectors, should be multiple of 16
398 int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
401 * Apply symmetric window in 16-bit fixed-point.
402 * @param output destination array
403 * constraints: 16-byte aligned
404 * @param input source array
405 * constraints: 16-byte aligned
406 * @param window window array
407 * constraints: 16-byte aligned, at least len/2 elements
408 * @param len full window length
409 * constraints: multiple of ? greater than zero
411 void (*apply_window_int16)(int16_t *output, const int16_t *input,
412 const int16_t *window, unsigned int len);
415 * Clip each element in an array of int32_t to a given minimum and maximum value.
416 * @param dst destination array
417 * constraints: 16-byte aligned
418 * @param src source array
419 * constraints: 16-byte aligned
420 * @param min minimum value
421 * constraints: must be in the range [-(1 << 24), 1 << 24]
422 * @param max maximum value
423 * constraints: must be in the range [-(1 << 24), 1 << 24]
424 * @param len number of elements in the array
425 * constraints: multiple of 32 greater than zero
427 void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
428 int32_t max, unsigned int len);
430 op_fill_func fill_block_tab[2];
433 void ff_dsputil_static_init(void);
434 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
436 int ff_check_alignment(void);
439 * permute block according to permuatation.
440 * @param last last non zero element in scantable order
442 void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
444 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
446 #define BYTE_VEC32(c) ((c)*0x01010101UL)
447 #define BYTE_VEC64(c) ((c)*0x0001000100010001UL)
449 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
451 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
454 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
456 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
459 static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
461 return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
464 static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
466 return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
469 static inline int get_penalty_factor(int lambda, int lambda2, int type){
473 return lambda>>FF_LAMBDA_SHIFT;
475 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
478 return (2*lambda)>>FF_LAMBDA_SHIFT;
483 return lambda2>>FF_LAMBDA_SHIFT;
489 void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
490 void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
491 void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
492 void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
493 void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
494 void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
495 void ff_dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
497 #if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMX
498 # define STRIDE_ALIGN 16
500 # define STRIDE_ALIGN 8
503 // Some broken preprocessors need a second expansion
504 // to be forced to tokenize __VA_ARGS__
507 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...) \
508 uint8_t la_##v[sizeof(t s o) + (a)]; \
509 t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
511 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) \
512 DECLARE_ALIGNED(a, t, la_##v) s o; \
515 #define LOCAL_ALIGNED(a, t, v, ...) E(LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,))
517 #if HAVE_LOCAL_ALIGNED_8
518 # define LOCAL_ALIGNED_8(t, v, ...) E(LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,))
520 # define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
523 #if HAVE_LOCAL_ALIGNED_16
524 # define LOCAL_ALIGNED_16(t, v, ...) E(LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,))
526 # define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
529 #define WRAPPER8_16_SQ(name8, name16)\
530 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
532 score +=name8(s, dst , src , stride, 8);\
533 score +=name8(s, dst+8 , src+8 , stride, 8);\
537 score +=name8(s, dst , src , stride, 8);\
538 score +=name8(s, dst+8 , src+8 , stride, 8);\
544 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
549 AV_COPY16U(dst, src);
555 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
560 AV_COPY32U(dst, src);
566 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
571 AV_COPY64U(dst, src);
577 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
582 AV_COPY64U(dst, src);
589 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
594 AV_COPY128U(dst, src);
600 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
605 AV_COPY128U(dst, src);
612 #endif /* AVCODEC_DSPUTIL_H */