| blob | 617c4bd3185bab85f6a041f74ca601448be8d1f5 |
1 --- include/attributes.h 2006-06-16 20:12:26.000000000 +0200
2 +++ libmpeg2/attributes.h 2006-06-16 20:12:50.000000000 +0200
3 @@ -25,7 +29,7 @@
4 #ifdef ATTRIBUTE_ALIGNED_MAX
5 #define ATTR_ALIGN(align) __attribute__ ((__aligned__ ((ATTRIBUTE_ALIGNED_MAX < align) ? ATTRIBUTE_ALIGNED_MAX : align)))
6 #else
7 -#define ATTR_ALIGN(align)
8 +#define ATTR_ALIGN(align) __attribute__ ((__aligned__ ((16 < align) ? 16 : align)))
9 #endif
11 #ifdef HAVE_BUILTIN_EXPECT
12 --- libmpeg2/cpu_accel.c 2006-06-16 20:12:26.000000000 +0200
13 +++ libmpeg2/cpu_accel.c 2006-06-16 20:12:50.000000000 +0200
14 @@ -22,6 +26,7 @@
15 */
17 #include "config.h"
18 +#include "cpudetect.h"
20 #include <inttypes.h>
22 @@ -30,9 +35,17 @@
23 #include "mpeg2_internal.h"
25 #ifdef ACCEL_DETECT
26 -#ifdef ARCH_X86
27 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
28 +
29 +/* MPlayer imports libmpeg2 as decoder, which detects MMX / 3DNow!
30 + * instructions via assembly. However, it is regarded as duplicaed work
31 + * in MPlayer, so that we enforce to use MPlayer's implementation.
32 + */
33 +#define USE_MPLAYER_CPUDETECT
34 +
35 static inline uint32_t arch_accel (void)
36 {
37 +#if !defined(USE_MPLAYER_CPUDETECT)
38 uint32_t eax, ebx, ecx, edx;
39 int AMD;
40 uint32_t caps;
41 @@ -105,10 +118,24 @@
42 caps |= MPEG2_ACCEL_X86_MMXEXT;
44 return caps;
45 +#else /* USE_MPLAYER_CPUDETECT: Use MPlayer's cpu capability property */
46 + caps = 0;
47 + if (gCpuCaps.hasMMX)
48 + caps |= MPEG2_ACCEL_X86_MMX;
49 + if (gCpuCaps.hasSSE2)
50 + caps |= MPEG2_ACCEL_X86_SSE2;
51 + if (gCpuCaps.hasMMX2)
52 + caps |= MPEG2_ACCEL_X86_MMXEXT;
53 + if (gCpuCaps.has3DNow)
54 + caps |= MPEG2_ACCEL_X86_3DNOW;
55 +
56 + return caps;
57 +
58 +#endif /* USE_MPLAYER_CPUDETECT */
59 }
60 -#endif /* ARCH_X86 */
61 +#endif /* ARCH_X86 || ARCH_X86_64 */
63 -#if defined(ARCH_PPC) || defined(ARCH_SPARC)
64 +#if defined(ARCH_PPC) || (defined(ARCH_SPARC) && defined(HAVE_VIS))
65 #include <signal.h>
66 #include <setjmp.h>
68 @@ -166,10 +166,10 @@
70 canjump = 1;
72 -#ifdef HAVE_ALTIVEC_H /* gnu */
73 -#define VAND(a,b,c) "vand " #a "," #b "," #c "\n\t"
74 -#else /* apple */
75 +#if defined( __APPLE_CC__ ) && defined( __APPLE_ALTIVEC__ ) /* apple */
76 #define VAND(a,b,c) "vand v" #a ",v" #b ",v" #c "\n\t"
77 +#else /* gnu */
78 +#define VAND(a,b,c) "vand " #a "," #b "," #c "\n\t"
79 #endif
80 asm volatile ("mtspr 256, %0\n\t"
81 VAND (0, 0, 0)
82 @@ -195,6 +222,7 @@
83 #ifdef ARCH_ALPHA
84 static inline uint32_t arch_accel (void)
85 {
86 +#ifdef CAN_COMPILE_ALPHA_MVI
87 uint64_t no_mvi;
89 asm volatile ("amask %1, %0"
90 @@ -202,6 +230,9 @@
91 : "rI" (256)); /* AMASK_MVI */
92 return no_mvi ? MPEG2_ACCEL_ALPHA : (MPEG2_ACCEL_ALPHA |
93 MPEG2_ACCEL_ALPHA_MVI);
94 +#else
95 + return MPEG2_ACCEL_ALPHA;
96 +#endif
97 }
98 #endif /* ARCH_ALPHA */
99 #endif /* ACCEL_DETECT */
100 @@ -212,7 +243,7 @@
102 accel = 0;
103 #ifdef ACCEL_DETECT
104 -#if defined (ARCH_X86) || defined (ARCH_PPC) || defined (ARCH_ALPHA) || defined (ARCH_SPARC)
105 +#if defined (ARCH_X86) || defined (ARCH_X86_64) || defined (ARCH_PPC) || defined (ARCH_ALPHA) || defined (ARCH_SPARC)
106 accel = arch_accel ();
107 #endif
108 #endif
109 --- libmpeg2/cpu_state.c 2006-06-16 20:12:26.000000000 +0200
110 +++ libmpeg2/cpu_state.c 2006-06-16 20:12:50.000000000 +0200
111 @@ -29,14 +33,14 @@
112 #include "mpeg2.h"
113 #include "attributes.h"
114 #include "mpeg2_internal.h"
115 -#ifdef ARCH_X86
116 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
117 #include "mmx.h"
118 #endif
120 void (* mpeg2_cpu_state_save) (cpu_state_t * state) = NULL;
121 void (* mpeg2_cpu_state_restore) (cpu_state_t * state) = NULL;
123 -#ifdef ARCH_X86
124 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
125 static void state_restore_mmx (cpu_state_t * state)
126 {
127 emms ();
128 @@ -48,18 +48,18 @@
129 #endif
131 -#ifdef ARCH_PPC
132 +#if defined(ARCH_PPC) && defined(HAVE_ALTIVEC)
133 -#ifdef HAVE_ALTIVEC_H /* gnu */
134 -#define LI(a,b) "li " #a "," #b "\n\t"
135 -#define STVX0(a,b,c) "stvx " #a ",0," #c "\n\t"
136 -#define STVX(a,b,c) "stvx " #a "," #b "," #c "\n\t"
137 -#define LVX0(a,b,c) "lvx " #a ",0," #c "\n\t"
138 -#define LVX(a,b,c) "lvx " #a "," #b "," #c "\n\t"
139 -#else /* apple */
140 +#if defined( __APPLE_CC__ ) && defined( __APPLE_ALTIVEC__ ) /* apple */
141 #define LI(a,b) "li r" #a "," #b "\n\t"
142 #define STVX0(a,b,c) "stvx v" #a ",0,r" #c "\n\t"
143 #define STVX(a,b,c) "stvx v" #a ",r" #b ",r" #c "\n\t"
144 #define LVX0(a,b,c) "lvx v" #a ",0,r" #c "\n\t"
145 #define LVX(a,b,c) "lvx v" #a ",r" #b ",r" #c "\n\t"
146 +#else /* gnu */
147 +#define LI(a,b) "li " #a "," #b "\n\t"
148 +#define STVX0(a,b,c) "stvx " #a ",0," #c "\n\t"
149 +#define STVX(a,b,c) "stvx " #a "," #b "," #c "\n\t"
150 +#define LVX0(a,b,c) "lvx " #a ",0," #c "\n\t"
151 +#define LVX(a,b,c) "lvx " #a "," #b "," #c "\n\t"
152 #endif
154 static void state_save_altivec (cpu_state_t * state)
155 @@ -115,12 +119,12 @@
157 void mpeg2_cpu_state_init (uint32_t accel)
158 {
159 -#ifdef ARCH_X86
160 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
161 if (accel & MPEG2_ACCEL_X86_MMX) {
162 mpeg2_cpu_state_restore = state_restore_mmx;
163 }
164 #endif
165 -#ifdef ARCH_PPC
166 +#if defined(ARCH_PPC) && defined(HAVE_ALTIVEC)
167 if (accel & MPEG2_ACCEL_PPC_ALTIVEC) {
168 mpeg2_cpu_state_save = state_save_altivec;
169 mpeg2_cpu_state_restore = state_restore_altivec;
170 --- libmpeg2/decode.c 2006-06-16 20:12:26.000000000 +0200
171 +++ libmpeg2/decode.c 2006-06-16 20:12:50.000000000 +0200
172 @@ -351,6 +355,15 @@
173 fbuf->buf[1] = buf[1];
174 fbuf->buf[2] = buf[2];
175 fbuf->id = id;
176 + // HACK! FIXME! At first I frame, copy pointers to prediction frame too!
177 + if (mpeg2dec->custom_fbuf && !mpeg2dec->fbuf[1]->buf[0]){
178 + mpeg2dec->fbuf[1]->buf[0]=buf[0];
179 + mpeg2dec->fbuf[1]->buf[1]=buf[1];
180 + mpeg2dec->fbuf[1]->buf[2]=buf[2];
181 + mpeg2dec->fbuf[1]->id=NULL;
182 + }
183 +// printf("libmpeg2: FBUF 0:%p 1:%p 2:%p\n",
184 +// mpeg2dec->fbuf[0]->buf[0],mpeg2dec->fbuf[1]->buf[0],mpeg2dec->fbuf[2]->buf[0]);
185 }
187 void mpeg2_custom_fbuf (mpeg2dec_t * mpeg2dec, int custom_fbuf)
188 --- libmpeg2/header.c 2006-06-16 20:12:26.000000000 +0200
189 +++ libmpeg2/header.c 2006-06-16 20:12:50.000000000 +0200
190 @@ -100,6 +104,9 @@
191 mpeg2dec->decoder.convert = NULL;
192 mpeg2dec->decoder.convert_id = NULL;
193 mpeg2dec->picture = mpeg2dec->pictures;
194 + memset(&mpeg2dec->fbuf_alloc[0].fbuf, 0, sizeof(mpeg2_fbuf_t));
195 + memset(&mpeg2dec->fbuf_alloc[1].fbuf, 0, sizeof(mpeg2_fbuf_t));
196 + memset(&mpeg2dec->fbuf_alloc[2].fbuf, 0, sizeof(mpeg2_fbuf_t));
197 mpeg2dec->fbuf[0] = &mpeg2dec->fbuf_alloc[0].fbuf;
198 mpeg2dec->fbuf[1] = &mpeg2dec->fbuf_alloc[1].fbuf;
199 mpeg2dec->fbuf[2] = &mpeg2dec->fbuf_alloc[2].fbuf;
200 @@ -551,6 +558,7 @@
201 if (!(mpeg2dec->sequence.flags & SEQ_FLAG_PROGRESSIVE_SEQUENCE)) {
202 picture->nb_fields = (buffer[3] & 2) ? 3 : 2;
203 flags |= (buffer[3] & 128) ? PIC_FLAG_TOP_FIELD_FIRST : 0;
204 + flags |= (buffer[3] & 2) ? PIC_FLAG_REPEAT_FIRST_FIELD : 0;
205 } else
206 picture->nb_fields = (buffer[3]&2) ? ((buffer[3]&128) ? 6 : 4) : 2;
207 break;
208 @@ -799,6 +807,7 @@
209 mpeg2dec->scaled[index] = mpeg2dec->q_scale_type;
210 for (i = 0; i < 32; i++) {
211 k = mpeg2dec->q_scale_type ? non_linear_scale[i] : (i << 1);
212 + decoder->quantizer_scales[i] = k;
213 for (j = 0; j < 64; j++)
214 decoder->quantizer_prescale[index][i][j] =
215 k * mpeg2dec->quantizer_matrix[index][j];
216 --- libmpeg2/idct.c 2006-06-16 20:12:26.000000000 +0200
217 +++ libmpeg2/idct.c 2006-06-16 20:12:50.000000000 +0200
218 @@ -239,12 +239,15 @@
220 void mpeg2_idct_init (uint32_t accel)
221 {
222 -#ifdef ARCH_X86
223 +#ifdef HAVE_MMX2
224 if (accel & MPEG2_ACCEL_X86_MMXEXT) {
225 mpeg2_idct_copy = mpeg2_idct_copy_mmxext;
226 mpeg2_idct_add = mpeg2_idct_add_mmxext;
227 mpeg2_idct_mmx_init ();
228 - } else if (accel & MPEG2_ACCEL_X86_MMX) {
229 + } else
230 +#endif
231 +#ifdef HAVE_MMX
232 + if (accel & MPEG2_ACCEL_X86_MMX) {
233 mpeg2_idct_copy = mpeg2_idct_copy_mmx;
234 mpeg2_idct_add = mpeg2_idct_add_mmx;
235 mpeg2_idct_mmx_init ();
236 @@ -254,11 +261,14 @@
237 } else
238 #endif
239 #ifdef ARCH_ALPHA
240 +#ifdef CAN_COMPILE_ALPHA_MVI
241 if (accel & MPEG2_ACCEL_ALPHA_MVI) {
242 mpeg2_idct_copy = mpeg2_idct_copy_mvi;
243 mpeg2_idct_add = mpeg2_idct_add_mvi;
244 mpeg2_idct_alpha_init ();
245 - } else if (accel & MPEG2_ACCEL_ALPHA) {
246 + } else
247 +#endif
248 + if (accel & MPEG2_ACCEL_ALPHA) {
249 int i;
251 mpeg2_idct_copy = mpeg2_idct_copy_alpha;
252 --- libmpeg2/idct_alpha.c 2006-06-16 20:12:26.000000000 +0200
253 +++ libmpeg2/idct_alpha.c 2006-06-16 20:12:50.000000000 +0200
254 @@ -157,6 +161,7 @@
255 block[8*7] = (a0 - b0) >> 17;
256 }
258 +#ifdef CAN_COMPILE_ALPHA_MVI
259 void mpeg2_idct_copy_mvi (int16_t * block, uint8_t * dest, const int stride)
260 {
261 uint64_t clampmask;
262 @@ -289,6 +294,7 @@
263 stq (p7, dest + 7 * stride);
264 }
265 }
266 +#endif
268 void mpeg2_idct_copy_alpha (int16_t * block, uint8_t * dest, const int stride)
269 {
270 --- libmpeg2/idct_mmx.c 2006-06-16 20:12:26.000000000 +0200
271 +++ libmpeg2/idct_mmx.c 2006-06-16 20:12:50.000000000 +0200
272 @@ -23,7 +27,7 @@
274 #include "config.h"
276 -#ifdef ARCH_X86
277 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
279 #include <inttypes.h>
281 --- libmpeg2/motion_comp.c 2006-06-16 20:12:26.000000000 +0200
282 +++ libmpeg2/motion_comp.c 2006-06-16 20:12:50.000000000 +0200
283 @@ -37,16 +37,22 @@
285 void mpeg2_mc_init (uint32_t accel)
286 {
287 -#ifdef ARCH_X86
288 +#ifdef HAVE_MMX2
289 if (accel & MPEG2_ACCEL_X86_MMXEXT)
290 mpeg2_mc = mpeg2_mc_mmxext;
291 - else if (accel & MPEG2_ACCEL_X86_3DNOW)
292 + else
293 +#endif
294 +#ifdef HAVE_3DNOW
295 + if (accel & MPEG2_ACCEL_X86_3DNOW)
296 mpeg2_mc = mpeg2_mc_3dnow;
297 - else if (accel & MPEG2_ACCEL_X86_MMX)
298 + else
299 +#endif
300 +#ifdef HAVE_MMX
301 + if (accel & MPEG2_ACCEL_X86_MMX)
302 mpeg2_mc = mpeg2_mc_mmx;
303 else
304 #endif
305 -#ifdef ARCH_PPC
306 +#if defined(ARCH_PPC) && defined(HAVE_ALTIVEC)
307 if (accel & MPEG2_ACCEL_PPC_ALTIVEC)
308 mpeg2_mc = mpeg2_mc_altivec;
309 else
310 @@ -52,7 +62,7 @@
311 mpeg2_mc = mpeg2_mc_alpha;
312 else
313 #endif
314 -#ifdef ARCH_SPARC
315 +#if defined(ARCH_SPARC) && defined(HAVE_VIS)
316 if (accel & MPEG2_ACCEL_SPARC_VIS)
317 mpeg2_mc = mpeg2_mc_vis;
318 else
319 @@ -67,6 +67,16 @@
320 mpeg2_mc = mpeg2_mc_vis;
321 else
322 #endif
323 +#ifdef ARCH_ARM
324 + if (1 /*accel & MPEG2_ACCEL_ARM*/) {
325 +#ifdef HAVE_IWMMXT
326 + if (1 /*accel & MPEG2_ACCEL_ARM_IWMMXT*/)
327 + mpeg2_mc = mpeg2_mc_iwmmxt;
328 + else
329 +#endif
330 + mpeg2_mc = mpeg2_mc_arm;
331 + } else
332 +#endif
333 mpeg2_mc = mpeg2_mc_c;
334 }
336 --- libmpeg2/motion_comp_mmx.c 2006-06-16 20:12:26.000000000 +0200
337 +++ libmpeg2/motion_comp_mmx.c 2006-06-16 20:12:50.000000000 +0200
338 @@ -23,7 +27,7 @@
340 #include "config.h"
342 -#ifdef ARCH_X86
343 +#if defined(ARCH_X86) || defined(ARCH_X86_64)
345 #include <inttypes.h>
347 --- include/mpeg2.h 2006-06-16 20:12:26.000000000 +0200
348 +++ libmpeg2/mpeg2.h 2006-06-16 20:12:50.000000000 +0200
349 @@ -82,6 +86,7 @@
350 #define PIC_FLAG_COMPOSITE_DISPLAY 32
351 #define PIC_FLAG_SKIP 64
352 #define PIC_FLAG_TAGS 128
353 +#define PIC_FLAG_REPEAT_FIRST_FIELD 256
354 #define PIC_MASK_COMPOSITE_DISPLAY 0xfffff000
356 typedef struct mpeg2_picture_s {
357 @@ -154,6 +159,7 @@
358 #define MPEG2_ACCEL_X86_MMX 1
359 #define MPEG2_ACCEL_X86_3DNOW 2
360 #define MPEG2_ACCEL_X86_MMXEXT 4
361 +#define MPEG2_ACCEL_X86_SSE2 8
362 #define MPEG2_ACCEL_PPC_ALTIVEC 1
363 #define MPEG2_ACCEL_ALPHA 1
364 #define MPEG2_ACCEL_ALPHA_MVI 2
365 --- libmpeg2/mpeg2_internal.h 2006-06-16 20:12:26.000000000 +0200
366 +++ libmpeg2/mpeg2_internal.h 2006-06-16 20:12:50.000000000 +0200
367 @@ -144,6 +148,12 @@
368 int second_field;
370 int mpeg1;
371 +
372 + /* for MPlayer: */
373 + int quantizer_scales[32];
374 + int quantizer_scale;
375 + char* quant_store;
376 + int quant_stride;
377 };
379 typedef struct {
380 @@ -214,6 +224,10 @@
381 int8_t q_scale_type, scaled[4];
382 uint8_t quantizer_matrix[4][64];
383 uint8_t new_quantizer_matrix[4][64];
384 +
385 + /* for MPlayer: */
386 + unsigned char *pending_buffer;
387 + int pending_length;
388 };
390 typedef struct {
391 @@ -312,3 +312,5 @@
392 extern mpeg2_mc_t mpeg2_mc_altivec;
393 extern mpeg2_mc_t mpeg2_mc_alpha;
394 extern mpeg2_mc_t mpeg2_mc_vis;
395 +extern mpeg2_mc_t mpeg2_mc_arm;
396 +extern mpeg2_mc_t mpeg2_mc_iwmmxt;
397 --- libmpeg2/slice.c 2006-06-16 20:12:26.000000000 +0200
398 +++ libmpeg2/slice.c 2006-06-16 20:12:50.000000000 +0200
399 @@ -142,6 +146,7 @@
401 quantizer_scale_code = UBITS (bit_buf, 5);
402 DUMPBITS (bit_buf, bits, 5);
403 + decoder->quantizer_scale = decoder->quantizer_scales[quantizer_scale_code];
405 decoder->quantizer_matrix[0] =
406 decoder->quantizer_prescale[0][quantizer_scale_code];
407 @@ -1568,6 +1569,18 @@
409 #define NEXT_MACROBLOCK \
410 do { \
411 + if(decoder->quant_store) { \
412 + if (decoder->picture_structure == TOP_FIELD) \
413 + decoder->quant_store[2*decoder->quant_stride*(decoder->v_offset>>4) \
414 + +(decoder->offset>>4)] = decoder->quantizer_scale; \
415 + else if (decoder->picture_structure == BOTTOM_FIELD) \
416 + decoder->quant_store[2*decoder->quant_stride*(decoder->v_offset>>4) \
417 + + decoder->quant_stride \
418 + +(decoder->offset>>4)] = decoder->quantizer_scale; \
419 + else \
420 + decoder->quant_store[decoder->quant_stride*(decoder->v_offset>>4) \
421 + +(decoder->offset>>4)] = decoder->quantizer_scale; \
422 + } \
423 decoder->offset += 16; \
424 if (decoder->offset == decoder->width) { \
425 do { /* just so we can use the break statement */ \
426 @@ -1604,6 +1604,12 @@
427 } \
428 } while (0)
430 +static void motion_dummy (mpeg2_decoder_t * const decoder,
431 + motion_t * const motion,
432 + mpeg2_mc_fct * const * const table)
433 +{
434 +}
435 +
436 void mpeg2_init_fbuf (mpeg2_decoder_t * decoder, uint8_t * current_fbuf[3],
437 uint8_t * forward_fbuf[3], uint8_t * backward_fbuf[3])
438 {
439 @@ -1661,7 +1667,9 @@
441 if (decoder->mpeg1) {
442 decoder->motion_parser[0] = motion_zero_420;
443 + decoder->motion_parser[MC_FIELD] = motion_dummy;
444 decoder->motion_parser[MC_FRAME] = motion_mp1;
445 + decoder->motion_parser[MC_DMV] = motion_dummy;
446 decoder->motion_parser[4] = motion_reuse_420;
447 } else if (decoder->picture_structure == FRAME_PICTURE) {
448 if (decoder->chroma_format == 0) {
449 --- libmpeg2/idct.c 2006-06-16 20:12:26.000000000 +0200
450 +++ libmpeg2/idct.c 2006-06-16 20:12:50.000000000 +0200
451 @@ -253,7 +253,7 @@
452 mpeg2_idct_mmx_init ();
453 } else
454 #endif
455 -#ifdef ARCH_PPC
456 +#if defined(ARCH_PPC) && defined(HAVE_ALTIVEC)
457 if (accel & MPEG2_ACCEL_PPC_ALTIVEC) {
458 mpeg2_idct_copy = mpeg2_idct_copy_altivec;
459 mpeg2_idct_add = mpeg2_idct_add_altivec;
460 --- libmpeg2/idct_altivec.c 2004/08/02 11:26:43 12933
461 +++ libmpeg2/idct_altivec.c 2005/05/15 20:11:34 15484
462 @@ -41,7 +41,7 @@
463 typedef vector signed int vector_s32_t;
464 typedef vector unsigned int vector_u32_t;
466 -#if defined(HAVE_ALTIVEC_H) && (__GNUC__ * 100 + __GNUC_MINOR__ < 303)
467 +#if defined( HAVE_ALTIVEC_H ) && !defined( __APPLE_ALTIVEC__ ) && (__GNUC__ * 100 + __GNUC_MINOR__ < 303)
468 /* work around gcc <3.3 vec_mergel bug */
469 static inline vector_s16_t my_vec_mergel (vector_s16_t const A,
470 vector_s16_t const B)
471 @@ -56,10 +56,10 @@
472 #define vec_mergel my_vec_mergel
473 #endif
475 -#ifdef HAVE_ALTIVEC_H /* gnu */
476 -#define VEC_S16(a,b,c,d,e,f,g,h) {a, b, c, d, e, f, g, h}
477 -#else /* apple */
478 +#if defined( __APPLE_CC__ ) && defined( __APPLE_ALTIVEC__ ) /* apple */
479 #define VEC_S16(a,b,c,d,e,f,g,h) (vector_s16_t) (a, b, c, d, e, f, g, h)
480 +#else /* gnu */
481 +#define VEC_S16(a,b,c,d,e,f,g,h) {a, b, c, d, e, f, g, h}
482 #endif
484 static const vector_s16_t constants ATTR_ALIGN(16) =
485 Index: libmpeg2/motion_comp_arm.c
486 ===================================================================
487 --- libmpeg2/motion_comp_arm.c (revision 0)
488 +++ libmpeg2/motion_comp_arm.c (revision 0)
489 @@ -0,0 +1,187 @@
490 +/*
491 + * motion_comp_arm.c
492 + * Copyright (C) 2004 AGAWA Koji <i (AT) atty (DOT) jp>
493 + *
494 + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
495 + * See http://libmpeg2.sourceforge.net/ for updates.
496 + *
497 + * mpeg2dec is free software; you can redistribute it and/or modify
498 + * it under the terms of the GNU General Public License as published by
499 + * the Free Software Foundation; either version 2 of the License, or
500 + * (at your option) any later version.
501 + *
502 + * mpeg2dec is distributed in the hope that it will be useful,
503 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
504 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
505 + * GNU General Public License for more details.
506 + *
507 + * You should have received a copy of the GNU General Public License
508 + * along with this program; if not, write to the Free Software
509 + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
510 + */
511 +
512 +#include "config.h"
513 +
514 +#ifdef ARCH_ARM
515 +
516 +#include <inttypes.h>
517 +
518 +#include "mpeg2.h"
519 +#include "attributes.h"
520 +#include "mpeg2_internal.h"
521 +
522 +#define avg2(a,b) ((a+b+1)>>1)
523 +#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
524 +
525 +#define predict_o(i) (ref[i])
526 +#define predict_x(i) (avg2 (ref[i], ref[i+1]))
527 +#define predict_y(i) (avg2 (ref[i], (ref+stride)[i]))
528 +#define predict_xy(i) (avg4 (ref[i], ref[i+1], \
529 + (ref+stride)[i], (ref+stride)[i+1]))
530 +
531 +#define put(predictor,i) dest[i] = predictor (i)
532 +#define avg(predictor,i) dest[i] = avg2 (predictor (i), dest[i])
533 +
534 +/* mc function template */
535 +
536 +#define MC_FUNC(op,xy) \
537 +static void inline MC_##op##_##xy##_16_c (uint8_t * dest, const uint8_t * ref, \
538 + const int stride, int height) \
539 +{ \
540 + do { \
541 + op (predict_##xy, 0); \
542 + op (predict_##xy, 1); \
543 + op (predict_##xy, 2); \
544 + op (predict_##xy, 3); \
545 + op (predict_##xy, 4); \
546 + op (predict_##xy, 5); \
547 + op (predict_##xy, 6); \
548 + op (predict_##xy, 7); \
549 + op (predict_##xy, 8); \
550 + op (predict_##xy, 9); \
551 + op (predict_##xy, 10); \
552 + op (predict_##xy, 11); \
553 + op (predict_##xy, 12); \
554 + op (predict_##xy, 13); \
555 + op (predict_##xy, 14); \
556 + op (predict_##xy, 15); \
557 + ref += stride; \
558 + dest += stride; \
559 + } while (--height); \
560 +} \
561 +static void MC_##op##_##xy##_8_c (uint8_t * dest, const uint8_t * ref, \
562 + const int stride, int height) \
563 +{ \
564 + do { \
565 + op (predict_##xy, 0); \
566 + op (predict_##xy, 1); \
567 + op (predict_##xy, 2); \
568 + op (predict_##xy, 3); \
569 + op (predict_##xy, 4); \
570 + op (predict_##xy, 5); \
571 + op (predict_##xy, 6); \
572 + op (predict_##xy, 7); \
573 + ref += stride; \
574 + dest += stride; \
575 + } while (--height); \
576 +} \
577 +/* definitions of the actual mc functions */
578 +
579 +MC_FUNC (put,o)
580 +MC_FUNC (avg,o)
581 +MC_FUNC (put,x)
582 +MC_FUNC (avg,x)
583 +MC_FUNC (put,y)
584 +MC_FUNC (avg,y)
585 +MC_FUNC (put,xy)
586 +MC_FUNC (avg,xy)
587 +
588 +
589 +extern void MC_put_o_16_arm (uint8_t * dest, const uint8_t * ref,
590 + int stride, int height);
591 +
592 +extern void MC_put_x_16_arm (uint8_t * dest, const uint8_t * ref,
593 + int stride, int height);
594 +
595 +
596 +static void MC_put_y_16_arm (uint8_t * dest, const uint8_t * ref,
597 + int stride, int height)
598 +{
599 + MC_put_y_16_c(dest, ref, stride, height);
600 +}
601 +
602 +static void MC_put_xy_16_arm (uint8_t * dest, const uint8_t * ref,
603 + int stride, int height)
604 +{
605 + MC_put_xy_16_c(dest, ref, stride, height);
606 +}
607 +
608 +extern void MC_put_o_8_arm (uint8_t * dest, const uint8_t * ref,
609 + int stride, int height);
610 +
611 +extern void MC_put_x_8_arm (uint8_t * dest, const uint8_t * ref,
612 + int stride, int height);
613 +
614 +static void MC_put_y_8_arm (uint8_t * dest, const uint8_t * ref,
615 + int stride, int height)
616 +{
617 + MC_put_y_8_c(dest, ref, stride, height);
618 +}
619 +
620 +static void MC_put_xy_8_arm (uint8_t * dest, const uint8_t * ref,
621 + int stride, int height)
622 +{
623 + MC_put_xy_8_c(dest, ref, stride, height);
624 +}
625 +
626 +static void MC_avg_o_16_arm (uint8_t * dest, const uint8_t * ref,
627 + int stride, int height)
628 +{
629 + MC_avg_o_16_c(dest, ref, stride, height);
630 +}
631 +
632 +static void MC_avg_x_16_arm (uint8_t * dest, const uint8_t * ref,
633 + int stride, int height)
634 +{
635 + MC_avg_x_16_c(dest, ref, stride, height);
636 +}
637 +
638 +static void MC_avg_y_16_arm (uint8_t * dest, const uint8_t * ref,
639 + int stride, int height)
640 +{
641 + MC_avg_y_16_c(dest, ref, stride, height);
642 +}
643 +
644 +static void MC_avg_xy_16_arm (uint8_t * dest, const uint8_t * ref,
645 + int stride, int height)
646 +{
647 + MC_avg_xy_16_c(dest, ref, stride, height);
648 +}
649 +
650 +static void MC_avg_o_8_arm (uint8_t * dest, const uint8_t * ref,
651 + int stride, int height)
652 +{
653 + MC_avg_o_8_c(dest, ref, stride, height);
654 +}
655 +
656 +static void MC_avg_x_8_arm (uint8_t * dest, const uint8_t * ref,
657 + int stride, int height)
658 +{
659 + MC_avg_x_8_c(dest, ref, stride, height);
660 +}
661 +
662 +static void MC_avg_y_8_arm (uint8_t * dest, const uint8_t * ref,
663 + int stride, int height)
664 +{
665 + MC_avg_y_8_c(dest, ref, stride, height);
666 +}
667 +
668 +static void MC_avg_xy_8_arm (uint8_t * dest, const uint8_t * ref,
669 + int stride, int height)
670 +{
671 + MC_avg_xy_8_c(dest, ref, stride, height);
672 +}
673 +
674 +MPEG2_MC_EXTERN (arm)
675 +
676 +#endif
677 Index: libmpeg2/motion_comp_arm_s.S
678 ===================================================================
679 --- libmpeg2/motion_comp_arm_s.S (revision 0)
680 +++ libmpeg2/motion_comp_arm_s.S (revision 0)
681 @@ -0,0 +1,322 @@
682 +@ motion_comp_arm_s.S
683 +@ Copyright (C) 2004 AGAWA Koji <i (AT) atty (DOT) jp>
684 +@
685 +@ This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
686 +@ See http://libmpeg2.sourceforge.net/ for updates.
687 +@
688 +@ mpeg2dec is free software; you can redistribute it and/or modify
689 +@ it under the terms of the GNU General Public License as published by
690 +@ the Free Software Foundation; either version 2 of the License, or
691 +@ (at your option) any later version.
692 +@
693 +@ mpeg2dec is distributed in the hope that it will be useful,
694 +@ but WITHOUT ANY WARRANTY; without even the implied warranty of
695 +@ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
696 +@ GNU General Public License for more details.
697 +@
698 +@ You should have received a copy of the GNU General Public License
699 +@ along with this program; if not, write to the Free Software
700 +@ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
701 +
702 + .text
703 +
704 +@ ----------------------------------------------------------------
705 + .align
706 + .global MC_put_o_16_arm
707 +MC_put_o_16_arm:
708 + @@ void func(uint8_t * dest, const uint8_t * ref, int stride, int height)
709 + pld [r1]
710 + stmfd sp!, {r4-r11, lr} @ R14 is also called LR
711 + and r4, r1, #3
712 + adr r5, MC_put_o_16_arm_align_jt
713 + add r5, r5, r4, lsl #2
714 + ldr pc, [r5]
715 +
716 +MC_put_o_16_arm_align0:
717 + ldmia r1, {r4-r7}
718 + add r1, r1, r2
719 + pld [r1]
720 + stmia r0, {r4-r7}
721 + subs r3, r3, #1
722 + add r0, r0, r2
723 + bne MC_put_o_16_arm_align0
724 + ldmfd sp!, {r4-r11, pc} @@ update PC with LR content.
725 +
726 +.macro PROC shift
727 + ldmia r1, {r4-r8}
728 + add r1, r1, r2
729 + mov r9, r4, lsr #(\shift)
730 + pld [r1]
731 + mov r10, r5, lsr #(\shift)
732 + orr r9, r9, r5, lsl #(32-\shift)
733 + mov r11, r6, lsr #(\shift)
734 + orr r10, r10, r6, lsl #(32-\shift)
735 + mov r12, r7, lsr #(\shift)
736 + orr r11, r11, r7, lsl #(32-\shift)
737 + orr r12, r12, r8, lsl #(32-\shift)
738 + stmia r0, {r9-r12}
739 + subs r3, r3, #1
740 + add r0, r0, r2
741 +.endm
742 +
743 +MC_put_o_16_arm_align1:
744 + and r1, r1, #0xFFFFFFFC
745 +1: PROC(8)
746 + bne 1b
747 + ldmfd sp!, {r4-r11, pc} @@ update PC with LR content.
748 +MC_put_o_16_arm_align2:
749 + and r1, r1, #0xFFFFFFFC
750 +1: PROC(16)
751 + bne 1b
752 + ldmfd sp!, {r4-r11, pc} @@ update PC with LR content.
753 +MC_put_o_16_arm_align3:
754 + and r1, r1, #0xFFFFFFFC
755 +1: PROC(24)
756 + bne 1b
757 + ldmfd sp!, {r4-r11, pc} @@ update PC with LR content.
758 +MC_put_o_16_arm_align_jt:
759 + .word MC_put_o_16_arm_align0
760 + .word MC_put_o_16_arm_align1
761 + .word MC_put_o_16_arm_align2
762 + .word MC_put_o_16_arm_align3
763 +
764 +@ ----------------------------------------------------------------
765 + .align
766 + .global MC_put_o_8_arm
767 +MC_put_o_8_arm:
768 + @@ void func(uint8_t * dest, const uint8_t * ref, int stride, int height)
769 + pld [r1]
770 + stmfd sp!, {r4-r10, lr} @ R14 is also called LR
771 + and r4, r1, #3
772 + adr r5, MC_put_o_8_arm_align_jt
773 + add r5, r5, r4, lsl #2
774 + ldr pc, [r5]
775 +MC_put_o_8_arm_align0:
776 + ldmia r1, {r4-r5}
777 + add r1, r1, r2
778 + pld [r1]
779 + stmia r0, {r4-r5}
780 + add r0, r0, r2
781 + subs r3, r3, #1
782 + bne MC_put_o_8_arm_align0
783 + ldmfd sp!, {r4-r10, pc} @@ update PC with LR content.
784 +
785 +.macro PROC8 shift
786 + ldmia r1, {r4-r6}
787 + add r1, r1, r2
788 + mov r9, r4, lsr #(\shift)
789 + pld [r1]
790 + mov r10, r5, lsr #(\shift)
791 + orr r9, r9, r5, lsl #(32-\shift)
792 + orr r10, r10, r6, lsl #(32-\shift)
793 + stmia r0, {r9-r10}
794 + subs r3, r3, #1
795 + add r0, r0, r2
796 +.endm
797 +
798 +MC_put_o_8_arm_align1:
799 + and r1, r1, #0xFFFFFFFC
800 +1: PROC8(8)
801 + bne 1b
802 + ldmfd sp!, {r4-r10, pc} @@ update PC with LR content.
803 +
804 +MC_put_o_8_arm_align2:
805 + and r1, r1, #0xFFFFFFFC
806 +1: PROC8(16)
807 + bne 1b
808 + ldmfd sp!, {r4-r10, pc} @@ update PC with LR content.
809 +
810 +MC_put_o_8_arm_align3:
811 + and r1, r1, #0xFFFFFFFC
812 +1: PROC8(24)
813 + bne 1b
814 + ldmfd sp!, {r4-r10, pc} @@ update PC with LR content.
815 +
816 +MC_put_o_8_arm_align_jt:
817 + .word MC_put_o_8_arm_align0
818 + .word MC_put_o_8_arm_align1
819 + .word MC_put_o_8_arm_align2
820 + .word MC_put_o_8_arm_align3
821 +
822 +@ ----------------------------------------------------------------
823 +.macro AVG_PW rW1, rW2
824 + mov \rW2, \rW2, lsl #24
825 + orr \rW2, \rW2, \rW1, lsr #8
826 + eor r9, \rW1, \rW2
827 + and \rW2, \rW1, \rW2
828 + and r10, r9, r12
829 + add \rW2, \rW2, r10, lsr #1
830 + and r10, r9, r11
831 + add \rW2, \rW2, r10
832 +.endm
833 +
834 + .align
835 + .global MC_put_x_16_arm
836 +MC_put_x_16_arm:
837 + @@ void func(uint8_t * dest, const uint8_t * ref, int stride, int height)
838 + pld [r1]
839 + stmfd sp!, {r4-r11,lr} @ R14 is also called LR
840 + and r4, r1, #3
841 + adr r5, MC_put_x_16_arm_align_jt
842 + ldr r11, [r5]
843 + mvn r12, r11
844 + add r5, r5, r4, lsl #2
845 + ldr pc, [r5, #4]
846 +
847 +.macro ADJ_ALIGN_QW shift, R0, R1, R2, R3, R4
848 + mov \R0, \R0, lsr #(\shift)
849 + orr \R0, \R0, \R1, lsl #(32 - \shift)
850 + mov \R1, \R1, lsr #(\shift)
851 + orr \R1, \R1, \R2, lsl #(32 - \shift)
852 + mov \R2, \R2, lsr #(\shift)
853 + orr \R2, \R2, \R3, lsl #(32 - \shift)
854 + mov \R3, \R3, lsr #(\shift)
855 + orr \R3, \R3, \R4, lsl #(32 - \shift)
856 + mov \R4, \R4, lsr #(\shift)
857 +@ and \R4, \R4, #0xFF
858 +.endm
859 +
860 +MC_put_x_16_arm_align0:
861 + ldmia r1, {r4-r8}
862 + add r1, r1, r2
863 + pld [r1]
864 + AVG_PW r7, r8
865 + AVG_PW r6, r7
866 + AVG_PW r5, r6
867 + AVG_PW r4, r5
868 + stmia r0, {r5-r8}
869 + subs r3, r3, #1
870 + add r0, r0, r2
871 + bne MC_put_x_16_arm_align0
872 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
873 +MC_put_x_16_arm_align1:
874 + and r1, r1, #0xFFFFFFFC
875 +1: ldmia r1, {r4-r8}
876 + add r1, r1, r2
877 + pld [r1]
878 + ADJ_ALIGN_QW 8, r4, r5, r6, r7, r8
879 + AVG_PW r7, r8
880 + AVG_PW r6, r7
881 + AVG_PW r5, r6
882 + AVG_PW r4, r5
883 + stmia r0, {r5-r8}
884 + subs r3, r3, #1
885 + add r0, r0, r2
886 + bne 1b
887 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
888 +MC_put_x_16_arm_align2:
889 + and r1, r1, #0xFFFFFFFC
890 +1: ldmia r1, {r4-r8}
891 + add r1, r1, r2
892 + pld [r1]
893 + ADJ_ALIGN_QW 16, r4, r5, r6, r7, r8
894 + AVG_PW r7, r8
895 + AVG_PW r6, r7
896 + AVG_PW r5, r6
897 + AVG_PW r4, r5
898 + stmia r0, {r5-r8}
899 + subs r3, r3, #1
900 + add r0, r0, r2
901 + bne 1b
902 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
903 +MC_put_x_16_arm_align3:
904 + and r1, r1, #0xFFFFFFFC
905 +1: ldmia r1, {r4-r8}
906 + add r1, r1, r2
907 + pld [r1]
908 + ADJ_ALIGN_QW 24, r4, r5, r6, r7, r8
909 + AVG_PW r7, r8
910 + AVG_PW r6, r7
911 + AVG_PW r5, r6
912 + AVG_PW r4, r5
913 + stmia r0, {r5-r8}
914 + subs r3, r3, #1
915 + add r0, r0, r2
916 + bne 1b
917 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
918 +MC_put_x_16_arm_align_jt:
919 + .word 0x01010101
920 + .word MC_put_x_16_arm_align0
921 + .word MC_put_x_16_arm_align1
922 + .word MC_put_x_16_arm_align2
923 + .word MC_put_x_16_arm_align3
924 +
925 +@ ----------------------------------------------------------------
926 + .align
927 + .global MC_put_x_8_arm
928 +MC_put_x_8_arm:
929 + @@ void func(uint8_t * dest, const uint8_t * ref, int stride, int height)
930 + pld [r1]
931 + stmfd sp!, {r4-r11,lr} @ R14 is also called LR
932 + and r4, r1, #3
933 + adr r5, MC_put_x_8_arm_align_jt
934 + ldr r11, [r5]
935 + mvn r12, r11
936 + add r5, r5, r4, lsl #2
937 + ldr pc, [r5, #4]
938 +
939 +.macro ADJ_ALIGN_DW shift, R0, R1, R2
940 + mov \R0, \R0, lsr #(\shift)
941 + orr \R0, \R0, \R1, lsl #(32 - \shift)
942 + mov \R1, \R1, lsr #(\shift)
943 + orr \R1, \R1, \R2, lsl #(32 - \shift)
944 + mov \R2, \R2, lsr #(\shift)
945 +@ and \R4, \R4, #0xFF
946 +.endm
947 +
948 +MC_put_x_8_arm_align0:
949 + ldmia r1, {r4-r6}
950 + add r1, r1, r2
951 + pld [r1]
952 + AVG_PW r5, r6
953 + AVG_PW r4, r5
954 + stmia r0, {r5-r6}
955 + subs r3, r3, #1
956 + add r0, r0, r2
957 + bne MC_put_x_8_arm_align0
958 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
959 +MC_put_x_8_arm_align1:
960 + and r1, r1, #0xFFFFFFFC
961 +1: ldmia r1, {r4-r6}
962 + add r1, r1, r2
963 + pld [r1]
964 + ADJ_ALIGN_DW 8, r4, r5, r6
965 + AVG_PW r5, r6
966 + AVG_PW r4, r5
967 + stmia r0, {r5-r6}
968 + subs r3, r3, #1
969 + add r0, r0, r2
970 + bne 1b
971 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
972 +MC_put_x_8_arm_align2:
973 + and r1, r1, #0xFFFFFFFC
974 +1: ldmia r1, {r4-r6}
975 + add r1, r1, r2
976 + pld [r1]
977 + ADJ_ALIGN_DW 16, r4, r5, r6
978 + AVG_PW r5, r6
979 + AVG_PW r4, r5
980 + stmia r0, {r5-r6}
981 + subs r3, r3, #1
982 + add r0, r0, r2
983 + bne 1b
984 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
985 +MC_put_x_8_arm_align3:
986 + and r1, r1, #0xFFFFFFFC
987 +1: ldmia r1, {r4-r6}
988 + add r1, r1, r2
989 + pld [r1]
990 + ADJ_ALIGN_DW 24, r4, r5, r6
991 + AVG_PW r5, r6
992 + AVG_PW r4, r5
993 + stmia r0, {r5-r6}
994 + subs r3, r3, #1
995 + add r0, r0, r2
996 + bne 1b
997 + ldmfd sp!, {r4-r11,pc} @@ update PC with LR content.
998 +MC_put_x_8_arm_align_jt:
999 + .word 0x01010101
1000 + .word MC_put_x_8_arm_align0
1001 + .word MC_put_x_8_arm_align1
1002 + .word MC_put_x_8_arm_align2
1003 + .word MC_put_x_8_arm_align3
1004 Index: libmpeg2/motion_comp_iwmmxt.c
1005 ===================================================================
1006 --- libmpeg2/motion_comp_iwmmxt.c (revision 0)
1007 +++ libmpeg2/motion_comp_iwmmxt.c (revision 0)
1008 @@ -0,0 +1,61 @@
1009 +/*
1010 + * motion_comp_iwmmxt.c
1011 + * Copyright (C) 2004 AGAWA Koji <i (AT) atty (DOT) jp>
1012 + *
1013 + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
1014 + * See http://libmpeg2.sourceforge.net/ for updates.
1015 + *
1016 + * mpeg2dec is free software; you can redistribute it and/or modify
1017 + * it under the terms of the GNU General Public License as published by
1018 + * the Free Software Foundation; either version 2 of the License, or
1019 + * (at your option) any later version.
1020 + *
1021 + * mpeg2dec is distributed in the hope that it will be useful,
1022 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
1023 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1024 + * GNU General Public License for more details.
1025 + *
1026 + * You should have received a copy of the GNU General Public License
1027 + * along with this program; if not, write to the Free Software
1028 + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
1029 + */
1030 +
1031 +#include "config.h"
1032 +
1033 +#ifdef ARCH_ARM
1034 +#ifdef HAVE_IWMMXT
1035 +
1036 +#include <inttypes.h>
1037 +
1038 +#include "mpeg2.h"
1039 +#include "attributes.h"
1040 +#include "mpeg2_internal.h"
1041 +
1042 +/* defined in libavcodec */
1043 +
1044 +extern void put_pixels16_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1045 +extern void put_pixels16_x2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1046 +extern void put_pixels16_y2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1047 +extern void put_pixels16_xy2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1048 +extern void put_pixels8_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1049 +extern void put_pixels8_x2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1050 +extern void put_pixels8_y2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1051 +extern void put_pixels8_xy2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1052 +extern void avg_pixels16_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1053 +extern void avg_pixels16_x2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1054 +extern void avg_pixels16_y2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1055 +extern void avg_pixels16_xy2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1056 +extern void avg_pixels8_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1057 +extern void avg_pixels8_x2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1058 +extern void avg_pixels8_y2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1059 +extern void avg_pixels8_xy2_iwmmxt(uint8_t * dest, const uint8_t * ref, const int stride, int height);
1060 +
1061 +mpeg2_mc_t mpeg2_mc_iwmmxt = {
1062 + {put_pixels16_iwmmxt, put_pixels16_x2_iwmmxt, put_pixels16_y2_iwmmxt, put_pixels16_xy2_iwmmxt,
1063 + put_pixels8_iwmmxt, put_pixels8_x2_iwmmxt, put_pixels8_y2_iwmmxt, put_pixels8_xy2_iwmmxt}, \
1064 + {avg_pixels16_iwmmxt, avg_pixels16_x2_iwmmxt, avg_pixels16_y2_iwmmxt, avg_pixels16_xy2_iwmmxt,
1065 + avg_pixels8_iwmmxt, avg_pixels8_x2_iwmmxt, avg_pixels8_y2_iwmmxt, avg_pixels8_xy2_iwmmxt}, \
1066 +};
1067 +
1068 +#endif
1069 +#endif