gstvideoscaleorc-dist.c 54.1 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11

/* autogenerated from gstvideoscaleorc.orc */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#ifndef DISABLE_ORC
#include <orc/orc.h>
#endif
#include <glib.h>

12
13
14
15
16
17
18
19
20
21
22
23
#ifndef _ORC_INTEGER_TYPEDEFS_
#define _ORC_INTEGER_TYPEDEFS_
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
#include <stdint.h>
typedef int8_t orc_int8;
typedef int16_t orc_int16;
typedef int32_t orc_int32;
typedef int64_t orc_int64;
typedef uint8_t orc_uint8;
typedef uint16_t orc_uint16;
typedef uint32_t orc_uint32;
typedef uint64_t orc_uint64;
24
#define ORC_UINT64_C(x) UINT64_C(x)
25
26
27
28
29
30
31
32
33
#elif defined(_MSC_VER)
typedef signed __int8 orc_int8;
typedef signed __int16 orc_int16;
typedef signed __int32 orc_int32;
typedef signed __int64 orc_int64;
typedef unsigned __int8 orc_uint8;
typedef unsigned __int16 orc_uint16;
typedef unsigned __int32 orc_uint32;
typedef unsigned __int64 orc_uint64;
34
#define ORC_UINT64_C(x) (x##Ui64)
35
36
37
38
39
40
41
42
43
44
45
#else
#include <limits.h>
typedef signed char orc_int8;
typedef short orc_int16;
typedef int orc_int32;
typedef unsigned char orc_uint8;
typedef unsigned short orc_uint16;
typedef unsigned int orc_uint32;
#if INT_MAX == LONG_MAX
typedef long long orc_int64;
typedef unsigned long long orc_uint64;
46
#define ORC_UINT64_C(x) (x##ULL)
47
48
49
#else
typedef long orc_int64;
typedef unsigned long orc_uint64;
50
#define ORC_UINT64_C(x) (x##UL)
51
52
53
#endif
#endif
typedef union
54
55
56
57
58
{
  orc_int16 i;
  orc_int8 x2[2];
} orc_union16;
typedef union
59
60
61
{
  orc_int32 i;
  float f;
62
63
  orc_int16 x2[2];
  orc_int8 x4[4];
64
65
66
67
68
} orc_union32;
typedef union
{
  orc_int64 i;
  double f;
69
  orc_int32 x2[2];
70
  float x2f[2];
71
  orc_int16 x4[4];
72
73
74
75
} orc_union64;
#endif

void orc_merge_linear_u8 (orc_uint8 * d1, const orc_uint8 * s1,
76
    const orc_uint8 * s2, int p1, int n);
77
78
void orc_merge_linear_u16 (orc_uint16 * d1, const orc_uint16 * s1,
    const orc_uint16 * s2, int p1, int p2, int n);
79
80
void orc_splat_u16 (orc_uint16 * d1, int p1, int n);
void orc_splat_u32 (orc_uint32 * d1, int p1, int n);
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
void orc_downsample_u8 (guint8 * d1, const guint8 * s1, int n);
void orc_downsample_u16 (guint16 * d1, const guint16 * s1, int n);
void gst_videoscale_orc_downsample_u32 (guint8 * d1, const guint8 * s1, int n);
void gst_videoscale_orc_downsample_yuyv (guint8 * d1, const guint8 * s1, int n);
void gst_videoscale_orc_resample_nearest_u8 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n);
void gst_videoscale_orc_resample_bilinear_u8 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n);
void gst_videoscale_orc_resample_nearest_u32 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n);
void gst_videoscale_orc_resample_bilinear_u32 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n);
void gst_videoscale_orc_resample_merge_bilinear_u32 (guint8 * d1, guint8 * d2,
    const guint8 * s1, const guint8 * s2, int p1, int p2, int p3, int n);
void gst_videoscale_orc_merge_bicubic_u8 (guint8 * d1, const guint8 * s1,
    const guint8 * s2, const guint8 * s3, const guint8 * s4, int p1, int p2,
    int p3, int p4, int n);
98

99
100
void gst_videoscale_orc_init (void);

101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126

/* begin Orc C target preamble */
#define ORC_CLAMP(x,a,b) ((x)<(a) ? (a) : ((x)>(b) ? (b) : (x)))
#define ORC_ABS(a) ((a)<0 ? -(a) : (a))
#define ORC_MIN(a,b) ((a)<(b) ? (a) : (b))
#define ORC_MAX(a,b) ((a)>(b) ? (a) : (b))
#define ORC_SB_MAX 127
#define ORC_SB_MIN (-1-ORC_SB_MAX)
#define ORC_UB_MAX 255
#define ORC_UB_MIN 0
#define ORC_SW_MAX 32767
#define ORC_SW_MIN (-1-ORC_SW_MAX)
#define ORC_UW_MAX 65535
#define ORC_UW_MIN 0
#define ORC_SL_MAX 2147483647
#define ORC_SL_MIN (-1-ORC_SL_MAX)
#define ORC_UL_MAX 4294967295U
#define ORC_UL_MIN 0
#define ORC_CLAMP_SB(x) ORC_CLAMP(x,ORC_SB_MIN,ORC_SB_MAX)
#define ORC_CLAMP_UB(x) ORC_CLAMP(x,ORC_UB_MIN,ORC_UB_MAX)
#define ORC_CLAMP_SW(x) ORC_CLAMP(x,ORC_SW_MIN,ORC_SW_MAX)
#define ORC_CLAMP_UW(x) ORC_CLAMP(x,ORC_UW_MIN,ORC_UW_MAX)
#define ORC_CLAMP_SL(x) ORC_CLAMP(x,ORC_SL_MIN,ORC_SL_MAX)
#define ORC_CLAMP_UL(x) ORC_CLAMP(x,ORC_UL_MIN,ORC_UL_MAX)
#define ORC_SWAP_W(x) ((((x)&0xff)<<8) | (((x)&0xff00)>>8))
#define ORC_SWAP_L(x) ((((x)&0xff)<<24) | (((x)&0xff00)<<8) | (((x)&0xff0000)>>8) | (((x)&0xff000000)>>24))
127
#define ORC_SWAP_Q(x) ((((x)&ORC_UINT64_C(0xff))<<56) | (((x)&ORC_UINT64_C(0xff00))<<40) | (((x)&ORC_UINT64_C(0xff0000))<<24) | (((x)&ORC_UINT64_C(0xff000000))<<8) | (((x)&ORC_UINT64_C(0xff00000000))>>8) | (((x)&ORC_UINT64_C(0xff0000000000))>>24) | (((x)&ORC_UINT64_C(0xff000000000000))>>40) | (((x)&ORC_UINT64_C(0xff00000000000000))>>56))
128
#define ORC_PTR_OFFSET(ptr,offset) ((void *)(((unsigned char *)(ptr)) + (offset)))
129
130
#define ORC_DENORMAL(x) ((x) & ((((x)&0x7f800000) == 0) ? 0xff800000 : 0xffffffff))
#define ORC_ISNAN(x) ((((x)&0x7f800000) == 0x7f800000) && (((x)&0x007fffff) != 0))
131
132
#define ORC_DENORMAL_DOUBLE(x) ((x) & ((((x)&ORC_UINT64_C(0x7ff0000000000000)) == 0) ? ORC_UINT64_C(0xfff0000000000000) : ORC_UINT64_C(0xffffffffffffffff)))
#define ORC_ISNAN_DOUBLE(x) ((((x)&ORC_UINT64_C(0x7ff0000000000000)) == ORC_UINT64_C(0x7ff0000000000000)) && (((x)&ORC_UINT64_C(0x000fffffffffffff)) != 0))
133
134
135
136
137
138
139
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
#define ORC_RESTRICT restrict
#elif defined(__GNUC__) && __GNUC__ >= 4
#define ORC_RESTRICT __restrict__
#else
#define ORC_RESTRICT
#endif
140
141
142
143
144
145
146
/* end Orc C target preamble */



/* orc_merge_linear_u8 */
#ifdef DISABLE_ORC
void
147
orc_merge_linear_u8 (orc_uint8 * d1, const orc_uint8 * s1, const orc_uint8 * s2,
148
    int p1, int n)
149
150
{
  int i;
151
152
153
154
155
156
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  const orc_int8 *ORC_RESTRICT ptr5;
  orc_int8 var36;
  orc_int8 var37;
  orc_union16 var38;
157
158
159
  orc_union16 var39;
  orc_int8 var40;
  orc_int8 var41;
160
161
162
  orc_union16 var42;
  orc_union16 var43;
  orc_union16 var44;
163
164
165
  orc_union16 var45;
  orc_union16 var46;
  orc_int8 var47;
166
167
168
169

  ptr0 = (orc_int8 *) d1;
  ptr4 = (orc_int8 *) s1;
  ptr5 = (orc_int8 *) s2;
170

171
172
173
174
  /* 6: loadpw */
  var38.i = p1;
  /* 8: loadpw */
  var39.i = 0x00000080;         /* 128 or 6.32404e-322f */
175

176
  for (i = 0; i < n; i++) {
177
    /* 0: loadb */
178
179
180
181
182
    var41 = ptr4[i];
    /* 1: loadb */
    var36 = ptr4[i];
    /* 2: convubw */
    var42.i = (orc_uint8) var36;
183
    /* 3: loadb */
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
    var37 = ptr5[i];
    /* 4: convubw */
    var43.i = (orc_uint8) var37;
    /* 5: subw */
    var44.i = var43.i - var42.i;
    /* 7: mullw */
    var45.i = (var44.i * var38.i) & 0xffff;
    /* 9: addw */
    var46.i = var45.i + var39.i;
    /* 10: convhwb */
    var47 = ((orc_uint16) var46.i) >> 8;
    /* 11: addb */
    var40 = var47 + var41;
    /* 12: storeb */
    ptr0[i] = var40;
199
200
201
202
203
204
  }

}

#else
static void
205
_backup_orc_merge_linear_u8 (OrcExecutor * ORC_RESTRICT ex)
206
207
208
{
  int i;
  int n = ex->n;
209
210
211
212
213
214
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  const orc_int8 *ORC_RESTRICT ptr5;
  orc_int8 var36;
  orc_int8 var37;
  orc_union16 var38;
215
216
217
  orc_union16 var39;
  orc_int8 var40;
  orc_int8 var41;
218
219
220
  orc_union16 var42;
  orc_union16 var43;
  orc_union16 var44;
221
222
223
  orc_union16 var45;
  orc_union16 var46;
  orc_int8 var47;
224
225
226
227

  ptr0 = (orc_int8 *) ex->arrays[0];
  ptr4 = (orc_int8 *) ex->arrays[4];
  ptr5 = (orc_int8 *) ex->arrays[5];
228

229
230
231
232
  /* 6: loadpw */
  var38.i = ex->params[24];
  /* 8: loadpw */
  var39.i = 0x00000080;         /* 128 or 6.32404e-322f */
233

234
  for (i = 0; i < n; i++) {
235
    /* 0: loadb */
236
237
238
239
240
    var41 = ptr4[i];
    /* 1: loadb */
    var36 = ptr4[i];
    /* 2: convubw */
    var42.i = (orc_uint8) var36;
241
    /* 3: loadb */
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
    var37 = ptr5[i];
    /* 4: convubw */
    var43.i = (orc_uint8) var37;
    /* 5: subw */
    var44.i = var43.i - var42.i;
    /* 7: mullw */
    var45.i = (var44.i * var38.i) & 0xffff;
    /* 9: addw */
    var46.i = var45.i + var39.i;
    /* 10: convhwb */
    var47 = ((orc_uint16) var46.i) >> 8;
    /* 11: addb */
    var40 = var47 + var41;
    /* 12: storeb */
    ptr0[i] = var40;
257
258
259
260
  }

}

261
static OrcProgram *_orc_program_orc_merge_linear_u8;
262
void
263
orc_merge_linear_u8 (orc_uint8 * d1, const orc_uint8 * s1, const orc_uint8 * s2,
264
    int p1, int n)
265
266
{
  OrcExecutor _ex, *ex = &_ex;
267
  OrcProgram *p = _orc_program_orc_merge_linear_u8;
268
269
270
271
272
273
274
275
276
277
278
279
280
281
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->arrays[ORC_VAR_S2] = (void *) s2;
  ex->params[ORC_VAR_P1] = p1;

  func = p->code_exec;
  func (ex);
}
#endif
282
283
284
285
286
287
288
289
290


/* orc_merge_linear_u16 */
#ifdef DISABLE_ORC
void
orc_merge_linear_u16 (orc_uint16 * d1, const orc_uint16 * s1,
    const orc_uint16 * s2, int p1, int p2, int n)
{
  int i;
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
  orc_union16 *ORC_RESTRICT ptr0;
  const orc_union16 *ORC_RESTRICT ptr4;
  const orc_union16 *ORC_RESTRICT ptr5;
  orc_union16 var34;
  orc_union16 var35;
  orc_union16 var36;
  orc_union16 var37;
  orc_union16 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union32 var41;
  orc_union32 var42;

  ptr0 = (orc_union16 *) d1;
  ptr4 = (orc_union16 *) s1;
  ptr5 = (orc_union16 *) s2;

  /* 1: loadpw */
  var35.i = p1;
  /* 4: loadpw */
  var37.i = p2;
312
313

  for (i = 0; i < n; i++) {
314
315
316
    /* 0: loadw */
    var34 = ptr4[i];
    /* 2: muluwl */
David Schleef's avatar
David Schleef committed
317
    var39.i = ((orc_uint16) var34.i) * ((orc_uint16) var35.i);
318
319
320
    /* 3: loadw */
    var36 = ptr5[i];
    /* 5: muluwl */
David Schleef's avatar
David Schleef committed
321
    var40.i = ((orc_uint16) var36.i) * ((orc_uint16) var37.i);
322
323
324
325
326
327
328
329
    /* 6: addl */
    var41.i = var39.i + var40.i;
    /* 7: shrul */
    var42.i = ((orc_uint32) var41.i) >> 16;
    /* 8: convlw */
    var38.i = var42.i;
    /* 9: storew */
    ptr0[i] = var38;
330
331
332
333
334
335
  }

}

#else
static void
336
_backup_orc_merge_linear_u16 (OrcExecutor * ORC_RESTRICT ex)
337
338
339
{
  int i;
  int n = ex->n;
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
  orc_union16 *ORC_RESTRICT ptr0;
  const orc_union16 *ORC_RESTRICT ptr4;
  const orc_union16 *ORC_RESTRICT ptr5;
  orc_union16 var34;
  orc_union16 var35;
  orc_union16 var36;
  orc_union16 var37;
  orc_union16 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union32 var41;
  orc_union32 var42;

  ptr0 = (orc_union16 *) ex->arrays[0];
  ptr4 = (orc_union16 *) ex->arrays[4];
  ptr5 = (orc_union16 *) ex->arrays[5];

  /* 1: loadpw */
  var35.i = ex->params[24];
  /* 4: loadpw */
  var37.i = ex->params[25];
361
362

  for (i = 0; i < n; i++) {
363
364
365
    /* 0: loadw */
    var34 = ptr4[i];
    /* 2: muluwl */
David Schleef's avatar
David Schleef committed
366
    var39.i = ((orc_uint16) var34.i) * ((orc_uint16) var35.i);
367
368
369
    /* 3: loadw */
    var36 = ptr5[i];
    /* 5: muluwl */
David Schleef's avatar
David Schleef committed
370
    var40.i = ((orc_uint16) var36.i) * ((orc_uint16) var37.i);
371
372
373
374
375
376
377
378
    /* 6: addl */
    var41.i = var39.i + var40.i;
    /* 7: shrul */
    var42.i = ((orc_uint32) var41.i) >> 16;
    /* 8: convlw */
    var38.i = var42.i;
    /* 9: storew */
    ptr0[i] = var38;
379
380
381
382
  }

}

383
static OrcProgram *_orc_program_orc_merge_linear_u16;
384
385
386
387
388
void
orc_merge_linear_u16 (orc_uint16 * d1, const orc_uint16 * s1,
    const orc_uint16 * s2, int p1, int p2, int n)
{
  OrcExecutor _ex, *ex = &_ex;
389
  OrcProgram *p = _orc_program_orc_merge_linear_u16;
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->arrays[ORC_VAR_S2] = (void *) s2;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;

  func = p->code_exec;
  func (ex);
}
#endif
405
406
407
408
409
410
411
412


/* orc_splat_u16 */
#ifdef DISABLE_ORC
void
orc_splat_u16 (orc_uint16 * d1, int p1, int n)
{
  int i;
413
414
415
416
417
  orc_union16 *ORC_RESTRICT ptr0;
  orc_union16 var32;
  orc_union16 var33;

  ptr0 = (orc_union16 *) d1;
418

419
420
  /* 0: loadpw */
  var32.i = p1;
421
422

  for (i = 0; i < n; i++) {
423
424
425
426
    /* 1: copyw */
    var33.i = var32.i;
    /* 2: storew */
    ptr0[i] = var33;
427
428
429
430
431
432
  }

}

#else
static void
433
_backup_orc_splat_u16 (OrcExecutor * ORC_RESTRICT ex)
434
435
436
{
  int i;
  int n = ex->n;
437
438
439
  orc_union16 *ORC_RESTRICT ptr0;
  orc_union16 var32;
  orc_union16 var33;
440

441
442
443
444
  ptr0 = (orc_union16 *) ex->arrays[0];

  /* 0: loadpw */
  var32.i = ex->params[24];
445
446

  for (i = 0; i < n; i++) {
447
448
449
450
    /* 1: copyw */
    var33.i = var32.i;
    /* 2: storew */
    ptr0[i] = var33;
451
452
453
454
  }

}

455
static OrcProgram *_orc_program_orc_splat_u16;
456
457
458
459
void
orc_splat_u16 (orc_uint16 * d1, int p1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
460
  OrcProgram *p = _orc_program_orc_splat_u16;
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->params[ORC_VAR_P1] = p1;

  func = p->code_exec;
  func (ex);
}
#endif


/* orc_splat_u32 */
#ifdef DISABLE_ORC
void
orc_splat_u32 (orc_uint32 * d1, int p1, int n)
{
  int i;
481
482
483
  orc_union32 *ORC_RESTRICT ptr0;
  orc_union32 var32;
  orc_union32 var33;
484
485
486

  ptr0 = (orc_union32 *) d1;

487
488
489
  /* 0: loadpl */
  var32.i = p1;

490
  for (i = 0; i < n; i++) {
491
492
493
494
    /* 1: copyl */
    var33.i = var32.i;
    /* 2: storel */
    ptr0[i] = var33;
495
496
497
498
499
500
  }

}

#else
static void
501
_backup_orc_splat_u32 (OrcExecutor * ORC_RESTRICT ex)
502
503
504
{
  int i;
  int n = ex->n;
505
506
507
  orc_union32 *ORC_RESTRICT ptr0;
  orc_union32 var32;
  orc_union32 var33;
508
509
510

  ptr0 = (orc_union32 *) ex->arrays[0];

511
512
513
  /* 0: loadpl */
  var32.i = ex->params[24];

514
  for (i = 0; i < n; i++) {
515
516
517
518
    /* 1: copyl */
    var33.i = var32.i;
    /* 2: storel */
    ptr0[i] = var33;
519
520
521
522
  }

}

523
static OrcProgram *_orc_program_orc_splat_u32;
524
525
526
527
void
orc_splat_u32 (orc_uint32 * d1, int p1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
528
  OrcProgram *p = _orc_program_orc_splat_u32;
529
530
531
532
533
534
535
536
537
538
539
540
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->params[ORC_VAR_P1] = p1;

  func = p->code_exec;
  func (ex);
}
#endif
541
542


543
544
/* orc_downsample_u8 */
#ifdef DISABLE_ORC
545
void
546
orc_downsample_u8 (guint8 * d1, const guint8 * s1, int n)
547
{
548
549
550
551
552
553
554
  int i;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_union16 *ORC_RESTRICT ptr4;
  orc_union16 var34;
  orc_int8 var35;
  orc_int8 var36;
  orc_int8 var37;
555

556
557
  ptr0 = (orc_int8 *) d1;
  ptr4 = (orc_union16 *) s1;
558
559


560
561
562
563
564
565
566
567
568
569
570
  for (i = 0; i < n; i++) {
    /* 0: loadw */
    var34 = ptr4[i];
    /* 1: splitwb */
    var36 = (var34.i >> 8) & 0xff;
    var37 = var34.i & 0xff;
    /* 2: avgub */
    var35 = ((orc_uint8) var36 + (orc_uint8) var37 + 1) >> 1;
    /* 3: storeb */
    ptr0[i] = var35;
  }
571

572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
}

#else
static void
_backup_orc_downsample_u8 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_union16 *ORC_RESTRICT ptr4;
  orc_union16 var34;
  orc_int8 var35;
  orc_int8 var36;
  orc_int8 var37;

  ptr0 = (orc_int8 *) ex->arrays[0];
  ptr4 = (orc_union16 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: loadw */
    var34 = ptr4[i];
    /* 1: splitwb */
    var36 = (var34.i >> 8) & 0xff;
    var37 = var34.i & 0xff;
    /* 2: avgub */
    var35 = ((orc_uint8) var36 + (orc_uint8) var37 + 1) >> 1;
    /* 3: storeb */
    ptr0[i] = var35;
601
602
  }

603
}
604

605
606
607
608
609
610
611
static OrcProgram *_orc_program_orc_downsample_u8;
void
orc_downsample_u8 (guint8 * d1, const guint8 * s1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_orc_downsample_u8;
  void (*func) (OrcExecutor *);
612

613
  ex->program = p;
614

615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;

  func = p->code_exec;
  func (ex);
}
#endif


/* orc_downsample_u16 */
#ifdef DISABLE_ORC
void
orc_downsample_u16 (guint16 * d1, const guint16 * s1, int n)
{
  int i;
  orc_union16 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var34;
  orc_union16 var35;
  orc_union16 var36;
  orc_union16 var37;

  ptr0 = (orc_union16 *) d1;
  ptr4 = (orc_union32 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: loadl */
    var34 = ptr4[i];
    /* 1: splitlw */
    var36.i = (var34.i >> 16) & 0xffff;
    var37.i = var34.i & 0xffff;
    /* 2: avguw */
    var35.i = ((orc_uint16) var36.i + (orc_uint16) var37.i + 1) >> 1;
    /* 3: storew */
    ptr0[i] = var35;
652
653
  }

654
}
655

656
657
658
659
660
661
662
663
664
665
666
667
#else
static void
_backup_orc_downsample_u16 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union16 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var34;
  orc_union16 var35;
  orc_union16 var36;
  orc_union16 var37;
668

669
670
  ptr0 = (orc_union16 *) ex->arrays[0];
  ptr4 = (orc_union32 *) ex->arrays[4];
671

672
673
674
675
676
677
678
679
680
681
682

  for (i = 0; i < n; i++) {
    /* 0: loadl */
    var34 = ptr4[i];
    /* 1: splitlw */
    var36.i = (var34.i >> 16) & 0xffff;
    var37.i = var34.i & 0xffff;
    /* 2: avguw */
    var35.i = ((orc_uint16) var36.i + (orc_uint16) var37.i + 1) >> 1;
    /* 3: storew */
    ptr0[i] = var35;
683
684
  }

685
}
686

687
688
689
690
691
692
693
static OrcProgram *_orc_program_orc_downsample_u16;
void
orc_downsample_u16 (guint16 * d1, const guint16 * s1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_orc_downsample_u16;
  void (*func) (OrcExecutor *);
694

695
  ex->program = p;
696

697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_downsample_u32 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_downsample_u32 (guint8 * d1, const guint8 * s1, int n)
{
  int i;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union64 *ORC_RESTRICT ptr4;
  orc_union64 var34;
  orc_union32 var35;
  orc_union32 var36;
  orc_union32 var37;

  ptr0 = (orc_union32 *) d1;
  ptr4 = (orc_union64 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: loadq */
    var34 = ptr4[i];
    /* 1: splitql */
    var36.i = (var34.i >> 32) & 0xffffffff;
    var37.i = var34.i & 0xffffffff;
    /* 2: avgub */
    var35.x4[0] = ((orc_uint8) var36.x4[0] + (orc_uint8) var37.x4[0] + 1) >> 1;
    var35.x4[1] = ((orc_uint8) var36.x4[1] + (orc_uint8) var37.x4[1] + 1) >> 1;
    var35.x4[2] = ((orc_uint8) var36.x4[2] + (orc_uint8) var37.x4[2] + 1) >> 1;
    var35.x4[3] = ((orc_uint8) var36.x4[3] + (orc_uint8) var37.x4[3] + 1) >> 1;
    /* 3: storel */
    ptr0[i] = var35;
  }

}

#else
static void
_backup_gst_videoscale_orc_downsample_u32 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union64 *ORC_RESTRICT ptr4;
  orc_union64 var34;
  orc_union32 var35;
  orc_union32 var36;
  orc_union32 var37;

  ptr0 = (orc_union32 *) ex->arrays[0];
  ptr4 = (orc_union64 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: loadq */
    var34 = ptr4[i];
    /* 1: splitql */
    var36.i = (var34.i >> 32) & 0xffffffff;
    var37.i = var34.i & 0xffffffff;
    /* 2: avgub */
    var35.x4[0] = ((orc_uint8) var36.x4[0] + (orc_uint8) var37.x4[0] + 1) >> 1;
    var35.x4[1] = ((orc_uint8) var36.x4[1] + (orc_uint8) var37.x4[1] + 1) >> 1;
    var35.x4[2] = ((orc_uint8) var36.x4[2] + (orc_uint8) var37.x4[2] + 1) >> 1;
    var35.x4[3] = ((orc_uint8) var36.x4[3] + (orc_uint8) var37.x4[3] + 1) >> 1;
    /* 3: storel */
    ptr0[i] = var35;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_downsample_u32;
void
gst_videoscale_orc_downsample_u32 (guint8 * d1, const guint8 * s1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_downsample_u32;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_downsample_yuyv */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_downsample_yuyv (guint8 * d1, const guint8 * s1, int n)
{
  int i;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union64 *ORC_RESTRICT ptr4;
  orc_union64 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union32 var41;
  orc_union16 var42;
  orc_union16 var43;
  orc_union16 var44;
  orc_union16 var45;
  orc_union16 var46;
  orc_union16 var47;

  ptr0 = (orc_union32 *) d1;
  ptr4 = (orc_union64 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: loadq */
    var38 = ptr4[i];
    /* 1: splitwb */
    var40.x4[0] = (var38.x4[0] >> 8) & 0xff;
    var41.x4[0] = var38.x4[0] & 0xff;
    var40.x4[1] = (var38.x4[1] >> 8) & 0xff;
    var41.x4[1] = var38.x4[1] & 0xff;
    var40.x4[2] = (var38.x4[2] >> 8) & 0xff;
    var41.x4[2] = var38.x4[2] & 0xff;
    var40.x4[3] = (var38.x4[3] >> 8) & 0xff;
    var41.x4[3] = var38.x4[3] & 0xff;
    /* 2: splitwb */
    var42.x2[0] = (var40.x2[0] >> 8) & 0xff;
    var43.x2[0] = var40.x2[0] & 0xff;
    var42.x2[1] = (var40.x2[1] >> 8) & 0xff;
    var43.x2[1] = var40.x2[1] & 0xff;
    /* 3: avgub */
    var44.x2[0] = ((orc_uint8) var42.x2[0] + (orc_uint8) var43.x2[0] + 1) >> 1;
    var44.x2[1] = ((orc_uint8) var42.x2[1] + (orc_uint8) var43.x2[1] + 1) >> 1;
    /* 4: splitlw */
    var45.i = (var41.i >> 16) & 0xffff;
    var46.i = var41.i & 0xffff;
    /* 5: avgub */
    var47.x2[0] = ((orc_uint8) var45.x2[0] + (orc_uint8) var46.x2[0] + 1) >> 1;
    var47.x2[1] = ((orc_uint8) var45.x2[1] + (orc_uint8) var46.x2[1] + 1) >> 1;
    /* 6: mergebw */
David Schleef's avatar
David Schleef committed
845
846
847
848
    var39.x2[0] =
        ((orc_uint8) var44.x2[0] & 0x00ff) | ((orc_uint8) var47.x2[0] << 8);
    var39.x2[1] =
        ((orc_uint8) var44.x2[1] & 0x00ff) | ((orc_uint8) var47.x2[1] << 8);
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
    /* 7: storel */
    ptr0[i] = var39;
  }

}

#else
static void
_backup_gst_videoscale_orc_downsample_yuyv (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union64 *ORC_RESTRICT ptr4;
  orc_union64 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union32 var41;
  orc_union16 var42;
  orc_union16 var43;
  orc_union16 var44;
  orc_union16 var45;
  orc_union16 var46;
  orc_union16 var47;

  ptr0 = (orc_union32 *) ex->arrays[0];
  ptr4 = (orc_union64 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: loadq */
    var38 = ptr4[i];
    /* 1: splitwb */
    var40.x4[0] = (var38.x4[0] >> 8) & 0xff;
    var41.x4[0] = var38.x4[0] & 0xff;
    var40.x4[1] = (var38.x4[1] >> 8) & 0xff;
    var41.x4[1] = var38.x4[1] & 0xff;
    var40.x4[2] = (var38.x4[2] >> 8) & 0xff;
    var41.x4[2] = var38.x4[2] & 0xff;
    var40.x4[3] = (var38.x4[3] >> 8) & 0xff;
    var41.x4[3] = var38.x4[3] & 0xff;
    /* 2: splitwb */
    var42.x2[0] = (var40.x2[0] >> 8) & 0xff;
    var43.x2[0] = var40.x2[0] & 0xff;
    var42.x2[1] = (var40.x2[1] >> 8) & 0xff;
    var43.x2[1] = var40.x2[1] & 0xff;
    /* 3: avgub */
    var44.x2[0] = ((orc_uint8) var42.x2[0] + (orc_uint8) var43.x2[0] + 1) >> 1;
    var44.x2[1] = ((orc_uint8) var42.x2[1] + (orc_uint8) var43.x2[1] + 1) >> 1;
    /* 4: splitlw */
    var45.i = (var41.i >> 16) & 0xffff;
    var46.i = var41.i & 0xffff;
    /* 5: avgub */
    var47.x2[0] = ((orc_uint8) var45.x2[0] + (orc_uint8) var46.x2[0] + 1) >> 1;
    var47.x2[1] = ((orc_uint8) var45.x2[1] + (orc_uint8) var46.x2[1] + 1) >> 1;
    /* 6: mergebw */
David Schleef's avatar
David Schleef committed
905
906
907
908
    var39.x2[0] =
        ((orc_uint8) var44.x2[0] & 0x00ff) | ((orc_uint8) var47.x2[0] << 8);
    var39.x2[1] =
        ((orc_uint8) var44.x2[1] & 0x00ff) | ((orc_uint8) var47.x2[1] << 8);
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
    /* 7: storel */
    ptr0[i] = var39;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_downsample_yuyv;
void
gst_videoscale_orc_downsample_yuyv (guint8 * d1, const guint8 * s1, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_downsample_yuyv;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_resample_nearest_u8 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_resample_nearest_u8 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  int i;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  orc_int8 var32;

  ptr0 = (orc_int8 *) d1;
  ptr4 = (orc_int8 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: ldresnearb */
    var32 = ptr4[(p1 + i * p2) >> 16];
    /* 1: storeb */
    ptr0[i] = var32;
  }

}

#else
static void
_backup_gst_videoscale_orc_resample_nearest_u8 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  orc_int8 var32;

  ptr0 = (orc_int8 *) ex->arrays[0];
  ptr4 = (orc_int8 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: ldresnearb */
    var32 = ptr4[(ex->params[24] + i * ex->params[25]) >> 16];
    /* 1: storeb */
    ptr0[i] = var32;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_resample_nearest_u8;
void
gst_videoscale_orc_resample_nearest_u8 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_resample_nearest_u8;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_resample_bilinear_u8 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_resample_bilinear_u8 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  int i;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  orc_int8 var32;

  ptr0 = (orc_int8 *) d1;
  ptr4 = (orc_int8 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: ldreslinb */
    {
      int tmp = p1 + i * p2;
      var32 =
          ((orc_uint8) ptr4[tmp >> 16] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) ptr4[(tmp >> 16) + 1] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storeb */
    ptr0[i] = var32;
  }

}

#else
static void
_backup_gst_videoscale_orc_resample_bilinear_u8 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  orc_int8 var32;

  ptr0 = (orc_int8 *) ex->arrays[0];
  ptr4 = (orc_int8 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: ldreslinb */
    {
      int tmp = ex->params[24] + i * ex->params[25];
      var32 =
          ((orc_uint8) ptr4[tmp >> 16] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) ptr4[(tmp >> 16) + 1] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storeb */
    ptr0[i] = var32;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_resample_bilinear_u8;
void
gst_videoscale_orc_resample_bilinear_u8 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_resample_bilinear_u8;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_resample_nearest_u32 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_resample_nearest_u32 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  int i;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var32;

  ptr0 = (orc_union32 *) d1;
  ptr4 = (orc_union32 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: ldresnearl */
    var32 = ptr4[(p1 + i * p2) >> 16];
    /* 1: storel */
    ptr0[i] = var32;
  }

}

#else
static void
_backup_gst_videoscale_orc_resample_nearest_u32 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var32;

  ptr0 = (orc_union32 *) ex->arrays[0];
  ptr4 = (orc_union32 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: ldresnearl */
    var32 = ptr4[(ex->params[24] + i * ex->params[25]) >> 16];
    /* 1: storel */
    ptr0[i] = var32;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_resample_nearest_u32;
void
gst_videoscale_orc_resample_nearest_u32 (guint8 * d1, const guint8 * s1, int p1,
    int p2, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_resample_nearest_u32;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_resample_bilinear_u32 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_resample_bilinear_u32 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n)
{
  int i;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var32;

  ptr0 = (orc_union32 *) d1;
  ptr4 = (orc_union32 *) s1;


  for (i = 0; i < n; i++) {
    /* 0: ldreslinl */
    {
      int tmp = p1 + i * p2;
      orc_union32 a = ptr4[tmp >> 16];
      orc_union32 b = ptr4[(tmp >> 16) + 1];
      var32.x4[0] =
          ((orc_uint8) a.x4[0] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[0] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[1] =
          ((orc_uint8) a.x4[1] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[1] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[2] =
          ((orc_uint8) a.x4[2] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[2] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[3] =
          ((orc_uint8) a.x4[3] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[3] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storel */
    ptr0[i] = var32;
  }

}

#else
static void
_backup_gst_videoscale_orc_resample_bilinear_u32 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union32 *ORC_RESTRICT ptr0;
  const orc_union32 *ORC_RESTRICT ptr4;
  orc_union32 var32;

  ptr0 = (orc_union32 *) ex->arrays[0];
  ptr4 = (orc_union32 *) ex->arrays[4];


  for (i = 0; i < n; i++) {
    /* 0: ldreslinl */
    {
      int tmp = ex->params[24] + i * ex->params[25];
      orc_union32 a = ptr4[tmp >> 16];
      orc_union32 b = ptr4[(tmp >> 16) + 1];
      var32.x4[0] =
          ((orc_uint8) a.x4[0] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[0] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[1] =
          ((orc_uint8) a.x4[1] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[1] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[2] =
          ((orc_uint8) a.x4[2] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[2] * ((tmp >> 8) & 0xff)) >> 8;
      var32.x4[3] =
          ((orc_uint8) a.x4[3] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[3] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storel */
    ptr0[i] = var32;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_resample_bilinear_u32;
void
gst_videoscale_orc_resample_bilinear_u32 (guint8 * d1, const guint8 * s1,
    int p1, int p2, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_resample_bilinear_u32;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_resample_merge_bilinear_u32 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_resample_merge_bilinear_u32 (guint8 * d1, guint8 * d2,
    const guint8 * s1, const guint8 * s2, int p1, int p2, int p3, int n)
{
  int i;
  orc_union32 *ORC_RESTRICT ptr0;
  orc_union32 *ORC_RESTRICT ptr1;
  const orc_union32 *ORC_RESTRICT ptr4;
  const orc_union32 *ORC_RESTRICT ptr5;
  orc_union64 var37;
  orc_union32 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union64 var41;
  orc_union64 var42;
  orc_union64 var43;
  orc_union64 var44;
  orc_union32 var45;

  ptr0 = (orc_union32 *) d1;
  ptr1 = (orc_union32 *) d2;
  ptr4 = (orc_union32 *) s1;
  ptr5 = (orc_union32 *) s2;

  /* 6: loadpw */
  var37.x4[0] = p1;
  var37.x4[1] = p1;
  var37.x4[2] = p1;
  var37.x4[3] = p1;

  for (i = 0; i < n; i++) {
    /* 0: ldreslinl */
    {
      int tmp = p2 + i * p3;
      orc_union32 a = ptr5[tmp >> 16];
      orc_union32 b = ptr5[(tmp >> 16) + 1];
      var39.x4[0] =
          ((orc_uint8) a.x4[0] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[0] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[1] =
          ((orc_uint8) a.x4[1] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[1] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[2] =
          ((orc_uint8) a.x4[2] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[2] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[3] =
          ((orc_uint8) a.x4[3] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[3] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storel */
    ptr1[i] = var39;
1310
    /* 2: loadl */
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
    var40 = ptr4[i];
    /* 3: convubw */
    var41.x4[0] = (orc_uint8) var40.x4[0];
    var41.x4[1] = (orc_uint8) var40.x4[1];
    var41.x4[2] = (orc_uint8) var40.x4[2];
    var41.x4[3] = (orc_uint8) var40.x4[3];
    /* 4: convubw */
    var42.x4[0] = (orc_uint8) var39.x4[0];
    var42.x4[1] = (orc_uint8) var39.x4[1];
    var42.x4[2] = (orc_uint8) var39.x4[2];
    var42.x4[3] = (orc_uint8) var39.x4[3];
    /* 5: subw */
    var43.x4[0] = var42.x4[0] - var41.x4[0];
    var43.x4[1] = var42.x4[1] - var41.x4[1];
    var43.x4[2] = var42.x4[2] - var41.x4[2];
    var43.x4[3] = var42.x4[3] - var41.x4[3];
    /* 7: mullw */
    var44.x4[0] = (var43.x4[0] * var37.x4[0]) & 0xffff;
    var44.x4[1] = (var43.x4[1] * var37.x4[1]) & 0xffff;
    var44.x4[2] = (var43.x4[2] * var37.x4[2]) & 0xffff;
    var44.x4[3] = (var43.x4[3] * var37.x4[3]) & 0xffff;
    /* 8: convhwb */
    var45.x4[0] = ((orc_uint16) var44.x4[0]) >> 8;
    var45.x4[1] = ((orc_uint16) var44.x4[1]) >> 8;
    var45.x4[2] = ((orc_uint16) var44.x4[2]) >> 8;
    var45.x4[3] = ((orc_uint16) var44.x4[3]) >> 8;
    /* 9: addb */
    var38.x4[0] = var45.x4[0] + var40.x4[0];
    var38.x4[1] = var45.x4[1] + var40.x4[1];
    var38.x4[2] = var45.x4[2] + var40.x4[2];
    var38.x4[3] = var45.x4[3] + var40.x4[3];
    /* 10: storel */
    ptr0[i] = var38;
  }

}

#else
static void
_backup_gst_videoscale_orc_resample_merge_bilinear_u32 (OrcExecutor *
    ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_union32 *ORC_RESTRICT ptr0;
  orc_union32 *ORC_RESTRICT ptr1;
  const orc_union32 *ORC_RESTRICT ptr4;
  const orc_union32 *ORC_RESTRICT ptr5;
  orc_union64 var37;
  orc_union32 var38;
  orc_union32 var39;
  orc_union32 var40;
  orc_union64 var41;
  orc_union64 var42;
  orc_union64 var43;
  orc_union64 var44;
  orc_union32 var45;

  ptr0 = (orc_union32 *) ex->arrays[0];
  ptr1 = (orc_union32 *) ex->arrays[1];
  ptr4 = (orc_union32 *) ex->arrays[4];
  ptr5 = (orc_union32 *) ex->arrays[5];

  /* 6: loadpw */
  var37.x4[0] = ex->params[24];
  var37.x4[1] = ex->params[24];
  var37.x4[2] = ex->params[24];
  var37.x4[3] = ex->params[24];

  for (i = 0; i < n; i++) {
    /* 0: ldreslinl */
    {
      int tmp = ex->params[25] + i * ex->params[26];
      orc_union32 a = ptr5[tmp >> 16];
      orc_union32 b = ptr5[(tmp >> 16) + 1];
      var39.x4[0] =
          ((orc_uint8) a.x4[0] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[0] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[1] =
          ((orc_uint8) a.x4[1] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[1] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[2] =
          ((orc_uint8) a.x4[2] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[2] * ((tmp >> 8) & 0xff)) >> 8;
      var39.x4[3] =
          ((orc_uint8) a.x4[3] * (256 - ((tmp >> 8) & 0xff)) +
          (orc_uint8) b.x4[3] * ((tmp >> 8) & 0xff)) >> 8;
    }
    /* 1: storel */
    ptr1[i] = var39;
1401
    /* 2: loadl */
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
    var40 = ptr4[i];
    /* 3: convubw */
    var41.x4[0] = (orc_uint8) var40.x4[0];
    var41.x4[1] = (orc_uint8) var40.x4[1];
    var41.x4[2] = (orc_uint8) var40.x4[2];
    var41.x4[3] = (orc_uint8) var40.x4[3];
    /* 4: convubw */
    var42.x4[0] = (orc_uint8) var39.x4[0];
    var42.x4[1] = (orc_uint8) var39.x4[1];
    var42.x4[2] = (orc_uint8) var39.x4[2];
    var42.x4[3] = (orc_uint8) var39.x4[3];
    /* 5: subw */
    var43.x4[0] = var42.x4[0] - var41.x4[0];
    var43.x4[1] = var42.x4[1] - var41.x4[1];
    var43.x4[2] = var42.x4[2] - var41.x4[2];
    var43.x4[3] = var42.x4[3] - var41.x4[3];
    /* 7: mullw */
    var44.x4[0] = (var43.x4[0] * var37.x4[0]) & 0xffff;
    var44.x4[1] = (var43.x4[1] * var37.x4[1]) & 0xffff;
    var44.x4[2] = (var43.x4[2] * var37.x4[2]) & 0xffff;
    var44.x4[3] = (var43.x4[3] * var37.x4[3]) & 0xffff;
    /* 8: convhwb */
    var45.x4[0] = ((orc_uint16) var44.x4[0]) >> 8;
    var45.x4[1] = ((orc_uint16) var44.x4[1]) >> 8;
    var45.x4[2] = ((orc_uint16) var44.x4[2]) >> 8;
    var45.x4[3] = ((orc_uint16) var44.x4[3]) >> 8;
    /* 9: addb */
    var38.x4[0] = var45.x4[0] + var40.x4[0];
    var38.x4[1] = var45.x4[1] + var40.x4[1];
    var38.x4[2] = var45.x4[2] + var40.x4[2];
    var38.x4[3] = var45.x4[3] + var40.x4[3];
    /* 10: storel */
    ptr0[i] = var38;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_resample_merge_bilinear_u32;
void
gst_videoscale_orc_resample_merge_bilinear_u32 (guint8 * d1, guint8 * d2,
    const guint8 * s1, const guint8 * s2, int p1, int p2, int p3, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_resample_merge_bilinear_u32;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_D2] = d2;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->arrays[ORC_VAR_S2] = (void *) s2;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;
  ex->params[ORC_VAR_P3] = p3;

  func = p->code_exec;
  func (ex);
}
#endif


/* gst_videoscale_orc_merge_bicubic_u8 */
#ifdef DISABLE_ORC
void
gst_videoscale_orc_merge_bicubic_u8 (guint8 * d1, const guint8 * s1,
    const guint8 * s2, const guint8 * s3, const guint8 * s4, int p1, int p2,
    int p3, int p4, int n)
{
  int i;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  const orc_int8 *ORC_RESTRICT ptr5;
  const orc_int8 *ORC_RESTRICT ptr6;
  const orc_int8 *ORC_RESTRICT ptr7;
  orc_int8 var34;
  orc_int8 var35;
  orc_int8 var36;
  orc_int8 var37;
  orc_int8 var38;
  orc_int8 var39;
  orc_int8 var40;
  orc_int8 var41;
  orc_union16 var42;
  orc_int8 var43;
  orc_union16 var44;
  orc_union16 var45;
  orc_union16 var46;
  orc_union16 var47;
  orc_union16 var48;
  orc_union16 var49;
  orc_union16 var50;
  orc_union16 var51;
  orc_union16 var52;

  ptr0 = (orc_int8 *) d1;
  ptr4 = (orc_int8 *) s1;
  ptr5 = (orc_int8 *) s2;
  ptr6 = (orc_int8 *) s3;
  ptr7 = (orc_int8 *) s4;

  /* 1: loadpb */
  var35 = p2;
  /* 4: loadpb */
  var37 = p3;
  /* 8: loadpb */
  var39 = p1;
  /* 12: loadpb */
  var41 = p4;
  /* 15: loadpw */
  var42.i = 0x00000020;         /* 32 or 1.58101e-322f */

  for (i = 0; i < n; i++) {
    /* 0: loadb */
    var34 = ptr5[i];
    /* 2: mulubw */
David Schleef's avatar
David Schleef committed
1519
    var44.i = ((orc_uint8) var34) * ((orc_uint8) var35);
1520
1521
1522
    /* 3: loadb */
    var36 = ptr6[i];
    /* 5: mulubw */
David Schleef's avatar
David Schleef committed
1523
    var45.i = ((orc_uint8) var36) * ((orc_uint8) var37);
1524
1525
1526
1527
1528
    /* 6: addw */
    var46.i = var44.i + var45.i;
    /* 7: loadb */
    var38 = ptr4[i];
    /* 9: mulubw */
David Schleef's avatar
David Schleef committed
1529
    var47.i = ((orc_uint8) var38) * ((orc_uint8) var39);
1530
1531
1532
1533
1534
    /* 10: subw */
    var48.i = var46.i - var47.i;
    /* 11: loadb */
    var40 = ptr7[i];
    /* 13: mulubw */
David Schleef's avatar
David Schleef committed
1535
    var49.i = ((orc_uint8) var40) * ((orc_uint8) var41);
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
    /* 14: subw */
    var50.i = var48.i - var49.i;
    /* 16: addw */
    var51.i = var50.i + var42.i;
    /* 17: shrsw */
    var52.i = var51.i >> 6;
    /* 18: convsuswb */
    var43 = ORC_CLAMP_UB (var52.i);
    /* 19: storeb */
    ptr0[i] = var43;
  }

}

#else
static void
_backup_gst_videoscale_orc_merge_bicubic_u8 (OrcExecutor * ORC_RESTRICT ex)
{
  int i;
  int n = ex->n;
  orc_int8 *ORC_RESTRICT ptr0;
  const orc_int8 *ORC_RESTRICT ptr4;
  const orc_int8 *ORC_RESTRICT ptr5;
  const orc_int8 *ORC_RESTRICT ptr6;
  const orc_int8 *ORC_RESTRICT ptr7;
  orc_int8 var34;
  orc_int8 var35;
  orc_int8 var36;
  orc_int8 var37;
  orc_int8 var38;
  orc_int8 var39;
  orc_int8 var40;
  orc_int8 var41;
  orc_union16 var42;
  orc_int8 var43;
  orc_union16 var44;
  orc_union16 var45;
  orc_union16 var46;
  orc_union16 var47;
  orc_union16 var48;
  orc_union16 var49;
  orc_union16 var50;
  orc_union16 var51;
  orc_union16 var52;

  ptr0 = (orc_int8 *) ex->arrays[0];
  ptr4 = (orc_int8 *) ex->arrays[4];
  ptr5 = (orc_int8 *) ex->arrays[5];
  ptr6 = (orc_int8 *) ex->arrays[6];
  ptr7 = (orc_int8 *) ex->arrays[7];

  /* 1: loadpb */
  var35 = ex->params[25];
  /* 4: loadpb */
  var37 = ex->params[26];
  /* 8: loadpb */
  var39 = ex->params[24];
  /* 12: loadpb */
  var41 = ex->params[27];
  /* 15: loadpw */
  var42.i = 0x00000020;         /* 32 or 1.58101e-322f */

  for (i = 0; i < n; i++) {
    /* 0: loadb */
    var34 = ptr5[i];
    /* 2: mulubw */
David Schleef's avatar
David Schleef committed
1602
    var44.i = ((orc_uint8) var34) * ((orc_uint8) var35);
1603
1604
1605
    /* 3: loadb */
    var36 = ptr6[i];
    /* 5: mulubw */
David Schleef's avatar
David Schleef committed
1606
    var45.i = ((orc_uint8) var36) * ((orc_uint8) var37);
1607
1608
1609
1610
1611
    /* 6: addw */
    var46.i = var44.i + var45.i;
    /* 7: loadb */
    var38 = ptr4[i];
    /* 9: mulubw */
David Schleef's avatar
David Schleef committed
1612
    var47.i = ((orc_uint8) var38) * ((orc_uint8) var39);
1613
1614
1615
1616
1617
    /* 10: subw */
    var48.i = var46.i - var47.i;
    /* 11: loadb */
    var40 = ptr7[i];
    /* 13: mulubw */
David Schleef's avatar
David Schleef committed
1618
    var49.i = ((orc_uint8) var40) * ((orc_uint8) var41);
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
    /* 14: subw */
    var50.i = var48.i - var49.i;
    /* 16: addw */
    var51.i = var50.i + var42.i;
    /* 17: shrsw */
    var52.i = var51.i >> 6;
    /* 18: convsuswb */
    var43 = ORC_CLAMP_UB (var52.i);
    /* 19: storeb */
    ptr0[i] = var43;
  }

}

static OrcProgram *_orc_program_gst_videoscale_orc_merge_bicubic_u8;
void
gst_videoscale_orc_merge_bicubic_u8 (guint8 * d1, const guint8 * s1,
    const guint8 * s2, const guint8 * s3, const guint8 * s4, int p1, int p2,
    int p3, int p4, int n)
{
  OrcExecutor _ex, *ex = &_ex;
  OrcProgram *p = _orc_program_gst_videoscale_orc_merge_bicubic_u8;
  void (*func) (OrcExecutor *);

  ex->program = p;

  ex->n = n;
  ex->arrays[ORC_VAR_D1] = d1;
  ex->arrays[ORC_VAR_S1] = (void *) s1;
  ex->arrays[ORC_VAR_S2] = (void *) s2;
  ex->arrays[ORC_VAR_S3] = (void *) s3;
  ex->arrays[ORC_VAR_S4] = (void *) s4;
  ex->params[ORC_VAR_P1] = p1;
  ex->params[ORC_VAR_P2] = p2;
  ex->params[ORC_VAR_P3] = p3;
  ex->params[ORC_VAR_P4] = p4;

  func = p->code_exec;
  func (ex);
}
#endif


void
gst_videoscale_orc_init (void)
{
#ifndef DISABLE_ORC
  {
    /* orc_merge_linear_u8 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_merge_linear_u8");
    orc_program_set_backup_function (p, _backup_orc_merge_linear_u8);
    orc_program_add_destination (p, 1, "d1");
    orc_program_add_source (p, 1, "s1");
    orc_program_add_source (p, 1, "s2");
    orc_program_add_constant (p, 4, 0x00000080, "c1");
    orc_program_add_parameter (p, 1, "p1");
    orc_program_add_temporary (p, 2, "t1");
    orc_program_add_temporary (p, 2, "t2");
    orc_program_add_temporary (p, 1, "t3");
    orc_program_add_temporary (p, 1, "t4");

    orc_program_append_2 (p, "loadb", 0, ORC_VAR_T3, ORC_VAR_S1, ORC_VAR_D1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "convubw", 0, ORC_VAR_T1, ORC_VAR_S1, ORC_VAR_D1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "convubw", 0, ORC_VAR_T2, ORC_VAR_S2, ORC_VAR_D1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "subw", 0, ORC_VAR_T2, ORC_VAR_T2, ORC_VAR_T1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "mullw", 0, ORC_VAR_T2, ORC_VAR_T2, ORC_VAR_P1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "addw", 0, ORC_VAR_T2, ORC_VAR_T2, ORC_VAR_C1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "convhwb", 0, ORC_VAR_T4, ORC_VAR_T2, ORC_VAR_D1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "addb", 0, ORC_VAR_D1, ORC_VAR_T4, ORC_VAR_T3,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_merge_linear_u8 = p;
  }
  {
    /* orc_merge_linear_u16 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_merge_linear_u16");
    orc_program_set_backup_function (p, _backup_orc_merge_linear_u16);
    orc_program_add_destination (p, 2, "d1");
    orc_program_add_source (p, 2, "s1");
    orc_program_add_source (p, 2, "s2");
    orc_program_add_constant (p, 4, 0x00000010, "c1");
    orc_program_add_parameter (p, 2, "p1");
    orc_program_add_parameter (p, 2, "p2");
    orc_program_add_temporary (p, 4, "t1");
    orc_program_add_temporary (p, 4, "t2");

    orc_program_append_2 (p, "muluwl", 0, ORC_VAR_T1, ORC_VAR_S1, ORC_VAR_P1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "muluwl", 0, ORC_VAR_T2, ORC_VAR_S2, ORC_VAR_P2,
        ORC_VAR_D1);
    orc_program_append_2 (p, "addl", 0, ORC_VAR_T1, ORC_VAR_T1, ORC_VAR_T2,
        ORC_VAR_D1);
    orc_program_append_2 (p, "shrul", 0, ORC_VAR_T1, ORC_VAR_T1, ORC_VAR_C1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "convlw", 0, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_D1,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_merge_linear_u16 = p;
  }
  {
    /* orc_splat_u16 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_splat_u16");
    orc_program_set_backup_function (p, _backup_orc_splat_u16);
    orc_program_add_destination (p, 2, "d1");
    orc_program_add_parameter (p, 2, "p1");

    orc_program_append_2 (p, "copyw", 0, ORC_VAR_D1, ORC_VAR_P1, ORC_VAR_D1,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_splat_u16 = p;
  }
  {
    /* orc_splat_u32 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_splat_u32");
    orc_program_set_backup_function (p, _backup_orc_splat_u32);
    orc_program_add_destination (p, 4, "d1");
    orc_program_add_parameter (p, 4, "p1");

    orc_program_append_2 (p, "copyl", 0, ORC_VAR_D1, ORC_VAR_P1, ORC_VAR_D1,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_splat_u32 = p;
  }
  {
    /* orc_downsample_u8 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_downsample_u8");
    orc_program_set_backup_function (p, _backup_orc_downsample_u8);
    orc_program_add_destination (p, 1, "d1");
    orc_program_add_source (p, 2, "s1");
    orc_program_add_temporary (p, 1, "t1");
    orc_program_add_temporary (p, 1, "t2");

    orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T1, ORC_VAR_T2, ORC_VAR_S1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "avgub", 0, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_downsample_u8 = p;
  }
  {
    /* orc_downsample_u16 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "orc_downsample_u16");
    orc_program_set_backup_function (p, _backup_orc_downsample_u16);
    orc_program_add_destination (p, 2, "d1");
    orc_program_add_source (p, 4, "s1");
    orc_program_add_temporary (p, 2, "t1");
    orc_program_add_temporary (p, 2, "t2");

    orc_program_append_2 (p, "splitlw", 0, ORC_VAR_T1, ORC_VAR_T2, ORC_VAR_S1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "avguw", 0, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_orc_downsample_u16 = p;
  }
  {
    /* gst_videoscale_orc_downsample_u32 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "gst_videoscale_orc_downsample_u32");
    orc_program_set_backup_function (p,
        _backup_gst_videoscale_orc_downsample_u32);
    orc_program_add_destination (p, 4, "d1");
    orc_program_add_source (p, 8, "s1");
    orc_program_add_temporary (p, 4, "t1");
    orc_program_add_temporary (p, 4, "t2");

    orc_program_append_2 (p, "splitql", 0, ORC_VAR_T1, ORC_VAR_T2, ORC_VAR_S1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "avgub", 2, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_gst_videoscale_orc_downsample_u32 = p;
  }
  {
    /* gst_videoscale_orc_downsample_yuyv */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "gst_videoscale_orc_downsample_yuyv");
    orc_program_set_backup_function (p,
        _backup_gst_videoscale_orc_downsample_yuyv);
    orc_program_add_destination (p, 4, "d1");
    orc_program_add_source (p, 8, "s1");
    orc_program_add_temporary (p, 4, "t1");
    orc_program_add_temporary (p, 4, "t2");
    orc_program_add_temporary (p, 2, "t3");
    orc_program_add_temporary (p, 2, "t4");
    orc_program_add_temporary (p, 2, "t5");
    orc_program_add_temporary (p, 2, "t6");

    orc_program_append_2 (p, "splitwb", 2, ORC_VAR_T1, ORC_VAR_T2, ORC_VAR_S1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "splitwb", 1, ORC_VAR_T3, ORC_VAR_T4, ORC_VAR_T1,
        ORC_VAR_D1);
    orc_program_append_2 (p, "avgub", 1, ORC_VAR_T5, ORC_VAR_T3, ORC_VAR_T4,
        ORC_VAR_D1);
    orc_program_append_2 (p, "splitlw", 0, ORC_VAR_T3, ORC_VAR_T4, ORC_VAR_T2,
        ORC_VAR_D1);
    orc_program_append_2 (p, "avgub", 1, ORC_VAR_T6, ORC_VAR_T3, ORC_VAR_T4,
        ORC_VAR_D1);
    orc_program_append_2 (p, "mergebw", 1, ORC_VAR_D1, ORC_VAR_T5, ORC_VAR_T6,
        ORC_VAR_D1);

    result = orc_program_compile (p);

    _orc_program_gst_videoscale_orc_downsample_yuyv = p;
  }
  {
    /* gst_videoscale_orc_resample_nearest_u8 */
    OrcProgram *p;
    OrcCompileResult result;

    p = orc_program_new ();
    orc_program_set_name (p, "gst_videoscale_orc_resample_nearest_u8");
    orc_program_set_backup_function (p,
        _backup_gst_videoscale_orc_resample_nearest_u8);
    orc_program_add_destination (p, 1<