« back to all changes in this revision
Viewing changes to libdv/dct_block_mmx_x86_64.S
- browse files at revision 4
- compare with another revision
- download diff
- download tarball
- view history from revision 4
- Committer: Bazaar Package Importer
- Author(s): Daniel Kobras
- Date: 2006-09-26 14:22:15 UTC
- mfrom: (1.3.1 upstream) (3.1.4 edgy)
- Revision ID: james.westby@ubuntu.com-20060926142215-q2arp4stpw6lrb5p
Tags: 1.0.0-1
* New upstream version.
* Removed patches:
+ [01_changelog_update_CVS]
+ [10_amd64_linkage_fix_CVS]
+ [10_inline_fixes_CVS]
+ [20_no_exec_stack_CVS]
+ [30_unload_memleak_fix_CVS]
+ [40_playdv_exit_fix_CVS]
+ [50_gcc41_asm_constraint_fixes_CVS]
All of the above are part of the new upstream version.
* debian/control: In Build-Depends, remove alternative dependencies on
obsolete SDL and X packages.
* debian/control: Complies with version 3.7.2 of Debian policy.
* debian/libdv4.README.Debian: Document lack of position-independent code
in i386 version of libdv's shared library as mandated by recent versions
of Debian policy.
* Removed patches:
+ [01_changelog_update_CVS]
+ [10_amd64_linkage_fix_CVS]
+ [10_inline_fixes_CVS]
+ [20_no_exec_stack_CVS]
+ [30_unload_memleak_fix_CVS]
+ [40_playdv_exit_fix_CVS]
+ [50_gcc41_asm_constraint_fixes_CVS]
All of the above are part of the new upstream version.
* debian/control: In Build-Depends, remove alternative dependencies on
obsolete SDL and X packages.
* debian/control: Complies with version 3.7.2 of Debian policy.
* debian/libdv4.README.Debian: Document lack of position-independent code
in i386 version of libdv's shared library as mandated by recent versions
of Debian policy.
- files added:
- debian/libdv4.README.Debian
- files modified:
- AUTHORS
added
removed
libdv/dct_block_mmx_x86_64.S
1
/*
2
* dct_block_mmx_x86_64.S
3
*
4
* Copyright (C) Peter Schlaile - February 2001
5
*
6
* This file is part of libdv, a free DV (IEC 61834/SMPTE 314M)
7
* codec.
8
*
9
* libdv is free software; you can redistribute it and/or modify it
10
* under the terms of the GNU Lesser Public License as published by
11
* the Free Software Foundation; either version 2.1, or (at your
12
* option) any later version.
13
*
14
* libdv is distributed in the hope that it will be useful, but
15
* WITHOUT ANY WARRANTY; without even the implied warranty of
16
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17
* Lesser Public License for more details.
18
*
19
* You should have received a copy of the GNU Lesser Public License
20
* along with libdv; see the file COPYING. If not, write to
21
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
22
*
23
* The libdv homepage is http://libdv.sourceforge.net/.
24
*/
25
26
# rdi - input and output data pointer
27
# the input data each 16 bit element in the 8x8 matrix is left aligned:
28
# the output data is tranposed and each 16 bit element in the 8x8 matrix is left aligned:
29
# e.g. in 11...1110000 format
30
# israelh. 11/11/97 removed emms. moved to stub
31
# MMX implementation. Using MMX transpose
32
33
#define YUV_PRECISION 1
34
35
#define NSHIFT 15
36
#define PRESHIFT 1
37
#define WA4_SHIFT (NSHIFT-1)
38
#define WA5_SHIFT (NSHIFT+1)
39
40
41
.data
42
43
.align 8
44
45
WA1: .word 23171,23171,23171,23171 /* 0.70711 * 32768 */
46
WA2: .word 17734,17734,17734,17734 /* 0.54120 * 32768 */
47
WA3: .word 23171,23171,23171,23171 /* 0.70711 * 32768 */
48
WA4: .word 21407,21407,21407,21407 /* 1.30658 * 16384 */
49
WA5: .word 25079,25079,25079,25079 /* 0.38268 * 65536 */
50
51
scratch1: .quad 0
52
scratch2: .quad 0
53
scratch3: .quad 0
54
scratch4: .quad 0
55
56
.section .note.GNU-stack, "", @progbits
57
58
.text
59
60
.align 8
61
.global _dv_dct_88_block_mmx_x86_64
62
.hidden _dv_dct_88_block_mmx_x86_64
63
.type _dv_dct_88_block_mmx_x86_64,@function
64
_dv_dct_88_block_mmx_x86_64:
65
66
/* void _dv_dct_88_block_mmx_x86_64(int16_t* block); */
67
68
/* argument is block=rdi */
69
70
mov %rdi, %r11 # block
71
72
# Column 0
73
movq 16*0(%r11), %mm0 # v0
74
movq 16*7(%r11), %mm1 # v7
75
movq %mm0, %mm2 # duplicate v0
76
paddw %mm1, %mm0 # v00: v0+v7
77
psubw %mm1, %mm2 # v07: v0-v7
78
79
movq 16*1(%r11), %mm1 # v1
80
movq 16*6(%r11), %mm3 # v6
81
movq %mm1, %mm4 # duplicate v1
82
paddw %mm3, %mm1 # v01: v1+v6
83
psubw %mm3, %mm4 # v06: v1-v6
84
85
movq 16*2(%r11), %mm3 # v2
86
movq 16*5(%r11), %mm5 # v5
87
movq %mm3, %mm6 # duplicate v2
88
paddw %mm5, %mm3 # v02: v2+v5
89
psubw %mm5, %mm6 # v05: v2-v5
90
91
movq 16*3(%r11), %mm5 # v3
92
movq 16*4(%r11), %mm7 # v4
93
movq %mm7, scratch1(%rip) # scratch1: v4 ;
94
movq %mm5, %mm7 # duplicate v3
95
paddw scratch1(%rip), %mm5 # v03: v3+v4
96
psubw scratch1(%rip), %mm7 # v04: v3-v4
97
movq %mm5, scratch2(%rip) # scratch2: v03
98
movq %mm0, %mm5 # mm5: v00
99
100
paddw scratch2(%rip), %mm0 # v10: v00+v03
101
psubw scratch2(%rip), %mm5 # v13: v00-v03
102
movq %mm3, scratch3(%rip) # scratch3: v02
103
movq %mm1, %mm3 # duplicate v01
104
105
paddw scratch3(%rip), %mm1 # v11: v01+v02
106
psubw scratch3(%rip), %mm3 # v12: v01-v02
107
108
movq %mm6, scratch4(%rip) # scratch4: v05
109
movq %mm0, %mm6 # duplicate v10
110
111
paddw %mm1, %mm0 # v10+v11
112
psubw %mm1, %mm6 # v10-v11
113
114
movq %mm0, (%r11) # out0: v10+v11
115
movq %mm6, 16*4(%r11) # out4: v10-v11
116
117
movq %mm4, %mm0 # mm0: v06
118
paddw scratch4(%rip), %mm4 # v15: v05+v06
119
paddw %mm2, %mm0 # v16: v07+v06
120
121
pmulhw WA3(%rip), %mm4 # v35~: WA3*v15
122
psllw $1, %mm4 # v35: compensate the coeefient scale
123
124
movq %mm4, %mm6 # duplicate v35
125
paddw %mm2, %mm4 # v45: v07+v35
126
psubw %mm6, %mm2 # v47: v07-v35
127
128
paddw %mm5, %mm3 # v22: v12+v13
129
130
pmulhw WA1(%rip), %mm3 # v32~: WA1*v22
131
psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
132
movq %mm5, %mm6 # duplicate v13
133
134
paddw %mm3, %mm5 # v13+v32
135
psubw %mm3, %mm6 # v13-v32
136
137
movq %mm5, 16*2(%r11) # out2: v13+v32
138
movq %mm6, 16*6(%r11) # out6: v13-v32
139
140
141
paddw scratch4(%rip), %mm7 # v14n: v04+v05
142
movq %mm0, %mm5 # duplicate v16
143
144
psubw %mm7, %mm0 # va1: v16-v14n
145
pmulhw WA5(%rip), %mm0 # va0~: va1*WA5
146
pmulhw WA4(%rip), %mm5 # v36~~: v16*WA4
147
pmulhw WA2(%rip), %mm7 # v34~~: v14n*WA2
148
psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeefient scale
149
psllw $16-NSHIFT, %mm7 # v34: compensate the coeefient scale
150
151
psubw %mm0, %mm5 # v36~: v36~~-va0~
152
psubw %mm0, %mm7 # v34~: v34~~-va0~
153
154
movq %mm4, %mm0 # duplicate v45
155
paddw %mm5, %mm4 # v45+v36
156
psubw %mm5, %mm0 # v45-v36
157
158
movq %mm4, 16*1(%r11) # out1: v45+v36
159
movq %mm0, 16*7(%r11) # out7: v45-v36
160
161
movq %mm2, %mm5 # duplicate v47
162
paddw %mm7, %mm2 # v47+v34
163
psubw %mm7, %mm5 # v47-v34
164
165
movq %mm2, 16*5(%r11) # out5: v47+v34
166
movq %mm5, 16*3(%r11) # out3: v47-v34
167
168
169
# column 1
170
171
add $8, %r11 # point to the next 4 columns.
172
# it can be done by adding 8 to
173
# immediates but this is nicer
174
175
movq (%r11), %mm0 # v0
176
movq 16*7(%r11), %mm1 # v7
177
movq %mm0, %mm2 # duplicate v0
178
paddw %mm1, %mm0 # v00: v0+v7
179
psubw %mm1, %mm2 # v07: v0-v7
180
181
movq 16(%r11), %mm1 # v1
182
movq 16*6(%r11), %mm3 # v6
183
movq %mm1, %mm4 # duplicate v1
184
paddw %mm3, %mm1 # v01: v1+v6
185
psubw %mm3, %mm4 # v06: v1-v6
186
187
movq 16*2(%r11), %mm3 # v2
188
movq 16*5(%r11), %mm5 # v5
189
movq %mm3, %mm6 # duplicate v2
190
paddw %mm5, %mm3 # v02: v2+v5
191
psubw %mm5, %mm6 # v05: v2-v5
192
193
movq 16*3(%r11), %mm5 # v3
194
movq 16*4(%r11), %mm7 # v4
195
movq %mm7, scratch1(%rip) # scratch1: v4 ;
196
movq %mm5, %mm7 # duplicate v3
197
paddw scratch1(%rip), %mm5 # v03: v3+v4
198
psubw scratch1(%rip), %mm7 # v04: v3-v4
199
movq %mm5, scratch2(%rip) # scratch2: v03
200
movq %mm0, %mm5 # mm5: v00
201
202
paddw scratch2(%rip), %mm0 # v10: v00+v03
203
psubw scratch2(%rip), %mm5 # v13: v00-v03
204
movq %mm3, scratch3(%rip) # scratc3: v02
205
movq %mm1, %mm3 # duplicate v01
206
207
paddw scratch3(%rip), %mm1 # v11: v01+v02
208
psubw scratch3(%rip), %mm3 # v12: v01-v02
209
210
movq %mm6, scratch4(%rip) # scratc4: v05
211
movq %mm0, %mm6 # duplicate v10
212
213
paddw %mm1, %mm0 # v10+v11
214
psubw %mm1, %mm6 # v10-v11
215
216
movq %mm0, (%r11) # out0: v10+v11
217
movq %mm6, 16*4(%r11) # out4: v10-v11
218
219
movq %mm4, %mm0 # mm0: v06
220
paddw scratch4(%rip), %mm4 # v15: v05+v06
221
paddw %mm2, %mm0 # v16: v07+v06
222
223
pmulhw WA3(%rip), %mm4 # v35~: WA3*v15
224
psllw $16-NSHIFT, %mm4 # v35: compensate the coeefient scale
225
226
movq %mm4, %mm6 # duplicate v35
227
paddw %mm2, %mm4 # v45: v07+v35
228
psubw %mm6, %mm2 # v47: v07-v35
229
230
paddw %mm5, %mm3 # v22: v12+v13
231
232
pmulhw WA1(%rip), %mm3 # v32~: WA3*v15
233
psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
234
movq %mm5, %mm6 # duplicate v13
235
236
paddw %mm3, %mm5 # v13+v32
237
psubw %mm3, %mm6 # v13-v32
238
239
movq %mm5, 16*2(%r11) # out2: v13+v32
240
movq %mm6, 16*6(%r11) # out6: v13-v32
241
242
paddw scratch4(%rip), %mm7 # v14n: v04+v05
243
movq %mm0, %mm5 # duplicate v16
244
245
psubw %mm7, %mm0 # va1: v16-v14n
246
pmulhw WA2(%rip), %mm7 # v34~~: v14n*WA2
247
pmulhw WA5(%rip), %mm0 # va0~: va1*WA5
248
pmulhw WA4(%rip), %mm5 # v36~~: v16*WA4
249
psllw $16-NSHIFT, %mm7
250
psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeffient
251
# scale note that WA4 is shifted 1 bit less than the others
252
253
psubw %mm0, %mm5 # v36~: v36~~-va0~
254
psubw %mm0, %mm7 # v34~: v34~~-va0~
255
256
movq %mm4, %mm0 # duplicate v45
257
paddw %mm5, %mm4 # v45+v36
258
psubw %mm5, %mm0 # v45-v36
259
260
movq %mm4, 16*1(%r11) # out1: v45+v36
261
movq %mm0, 16*7(%r11) # out7: v45-v36
262
263
movq %mm2, %mm5 # duplicate v47
264
paddw %mm7, %mm2 # v47+v34
265
psubw %mm7, %mm5 # v47-v34
266
267
movq %mm2, 16*5(%r11) # out5: v47+v34
268
movq %mm5, 16*3(%r11) # out3: v47-v34
269
270
ret $0
271
272
/* do mmx postscaling and reordering... */
273
274
.align 8
275
.global _dv_dct_block_mmx_x86_64_postscale_88
276
.hidden _dv_dct_block_mmx_x86_64_postscale_88
277
.type _dv_dct_block_mmx_x86_64_postscale_88,@function
278
_dv_dct_block_mmx_x86_64_postscale_88:
279
280
/* void _dv_dct_block_mmx_x86_64_postscale_88(int16_t* block, int16_t* postscale_matrix); */
281
282
/* arguments are block=rdi, postscale=rsi */
283
284
push %rbx
285
push %r12
286
287
mov %rdi,%r11 # block matrix
288
mov %rsi,%r12 # postscale matrix
289
290
sub $128+2, %rsp # reserve some stack space in the redzone, plus 2 bytes
291
292
movq 0*8(%r11), %mm0
293
movq 1*8(%r11), %mm1
294
movq 2*8(%r11), %mm2
295
movq 3*8(%r11), %mm3
296
297
movq %mm0, %mm4
298
movq %mm1, %mm5
299
movq %mm2, %mm6
300
movq %mm3, %mm7
301
psraw $0xf, %mm4
302
psraw $0xf, %mm5
303
psraw $0xf, %mm6
304
psraw $0xf, %mm7
305
306
pxor %mm4, %mm0
307
pxor %mm5, %mm1
308
pxor %mm6, %mm2
309
pxor %mm7, %mm3
310
311
psubw %mm4, %mm0
312
psubw %mm5, %mm1
313
psubw %mm6, %mm2
314
psubw %mm7, %mm3
315
316
pmulhw 0*8(%r12), %mm0
317
pmulhw 1*8(%r12), %mm1
318
pmulhw 2*8(%r12), %mm2
319
pmulhw 3*8(%r12), %mm3
320
321
psraw $YUV_PRECISION, %mm0
322
psraw $YUV_PRECISION, %mm1
323
psraw $YUV_PRECISION, %mm2
324
psraw $YUV_PRECISION, %mm3
325
326
pxor %mm4, %mm0
327
pxor %mm5, %mm1
328
pxor %mm6, %mm2
329
pxor %mm7, %mm3
330
331
psubw %mm4, %mm0
332
psubw %mm5, %mm1
333
psubw %mm6, %mm2
334
psubw %mm7, %mm3
335
336
movq %mm0, 0*8(%r11)
337
movq %mm1, 1*8(%r11)
338
movq %mm2, 2*8(%r11)
339
movq %mm3, 3*8(%r11)
340
341
movq 4*8(%r11), %mm0
342
movq 5*8(%r11), %mm1
343
movq 6*8(%r11), %mm2
344
movq 7*8(%r11), %mm3
345
346
movq %mm0, %mm4
347
movq %mm1, %mm5
348
349
movl 0*2(%r11), %eax
350
movw 2*2(%r11), %bx
351
352
movq %mm2, %mm6
353
movq %mm3, %mm7
354
355
movl %eax, 1*2(%rsp)
356
movw %bx, 6*2(%rsp)
357
358
psraw $0xf, %mm4
359
psraw $0xf, %mm5
360
361
movw 3*2(%r11), %ax
362
movw 4*2(%r11), %bx
363
364
psraw $0xf, %mm6
365
psraw $0xf, %mm7
366
367
pxor %mm4, %mm0
368
pxor %mm5, %mm1
369
370
movw %ax, 7*2(%rsp)
371
movw %bx, 15*2(%rsp)
372
373
pxor %mm6, %mm2
374
pxor %mm7, %mm3
375
376
movw 5*2(%r11), %ax
377
movw 6*2(%r11), %bx
378
379
psubw %mm4, %mm0
380
psubw %mm5, %mm1
381
382
movw %ax, 16*2(%rsp)
383
movw %bx, 28*2(%rsp)
384
385
psubw %mm6, %mm2
386
psubw %mm7, %mm3
387
388
movw 7*2(%r11), %ax
389
movw 8*2(%r11), %bx
390
391
pmulhw 4*8(%r12), %mm0
392
pmulhw 5*8(%r12), %mm1
393
394
movw %ax, 29*2(%rsp)
395
movw %bx, 3*2(%rsp)
396
397
pmulhw 6*8(%r12), %mm2
398
pmulhw 7*8(%r12), %mm3
399
400
movw 9*2(%r11), %ax
401
movw 10*2(%r11), %bx
402
403
psraw $YUV_PRECISION, %mm0
404
psraw $YUV_PRECISION, %mm1
405
406
movw %ax, 5*2(%rsp)
407
movw %bx, 8*2(%rsp)
408
409
psraw $YUV_PRECISION, %mm2
410
psraw $YUV_PRECISION, %mm3
411
412
movw 11*2(%r11), %ax
413
movw 12*2(%r11), %bx
414
415
pxor %mm4, %mm0
416
pxor %mm5, %mm1
417
418
movw %ax, 14*2(%rsp)
419
movw %bx, 17*2(%rsp)
420
421
pxor %mm6, %mm2
422
pxor %mm7, %mm3
423
424
movw 13*2(%r11), %ax
425
movw 14*2(%r11), %bx
426
427
psubw %mm4, %mm0
428
psubw %mm5, %mm1
429
430
movw %ax, 27*2(%rsp)
431
movw %bx, 30*2(%rsp)
432
433
psubw %mm6, %mm2
434
psubw %mm7, %mm3
435
436
movq %mm0, 4*8(%r11)
437
movq %mm1, 5*8(%r11)
438
movq %mm2, 6*8(%r11)
439
movq %mm3, 7*8(%r11)
440
441
movq 8*8(%r11), %mm0
442
movq 9*8(%r11), %mm1
443
444
movw 15*2(%r11), %ax
445
movw 16*2(%r11), %bx
446
447
movq 10*8(%r11), %mm2
448
movq 11*8(%r11), %mm3
449
450
movw %ax, 43*2(%rsp)
451
movw %bx, 4*2(%rsp)
452
453
movq %mm0, %mm4
454
movq %mm1, %mm5
455
456
movw 17*2(%r11), %ax
457
movw 18*2(%r11), %bx
458
459
movq %mm2, %mm6
460
movq %mm3, %mm7
461
462
movw %ax, 9*2(%rsp)
463
movw %bx, 13*2(%rsp)
464
465
psraw $0xf, %mm4
466
psraw $0xf, %mm5
467
468
movw 19*2(%r11), %ax
469
movw 20*2(%r11), %bx
470
471
psraw $0xf, %mm6
472
psraw $0xf, %mm7
473
474
movw %ax, 18*2(%rsp)
475
movw %bx, 26*2(%rsp)
476
477
pxor %mm4, %mm0
478
pxor %mm5, %mm1
479
480
movw 21*2(%r11), %ax
481
movw 22*2(%r11), %bx
482
483
pxor %mm6, %mm2
484
pxor %mm7, %mm3
485
486
movw %ax, 31*2(%rsp)
487
movw %bx, 42*2(%rsp)
488
489
psubw %mm4, %mm0
490
psubw %mm5, %mm1
491
492
movw 23*2(%r11), %ax
493
movw 24*2(%r11), %bx
494
495
psubw %mm6, %mm2
496
psubw %mm7, %mm3
497
498
movw %ax, 44*2(%rsp)
499
movw %bx, 10*2(%rsp)
500
501
pmulhw 8*8(%r12), %mm0
502
pmulhw 9*8(%r12), %mm1
503
504
movw 25*2(%r11), %ax
505
movw 26*2(%r11), %bx
506
507
pmulhw 10*8(%r12), %mm2
508
pmulhw 11*8(%r12), %mm3
509
510
psraw $YUV_PRECISION, %mm0
511
psraw $YUV_PRECISION, %mm1
512
513
movw %ax, 12*2(%rsp)
514
movw %bx, 19*2(%rsp)
515
516
psraw $YUV_PRECISION, %mm2
517
psraw $YUV_PRECISION, %mm3
518
519
movw 27*2(%r11), %ax
520
movw 28*2(%r11), %bx
521
522
pxor %mm4, %mm0
523
pxor %mm5, %mm1
524
525
movw %ax, 25*2(%rsp)
526
movw %bx, 32*2(%rsp)
527
528
pxor %mm6, %mm2
529
pxor %mm7, %mm3
530
531
movw 29*2(%r11), %ax
532
movw 30*2(%r11), %bx
533
534
psubw %mm4, %mm0
535
psubw %mm5, %mm1
536
537
movw %ax, 41*2(%rsp)
538
movw %bx, 45*2(%rsp)
539
540
psubw %mm6, %mm2
541
psubw %mm7, %mm3
542
543
movq %mm0, 8*8(%r11)
544
movq %mm1, 9*8(%r11)
545
movq %mm2, 10*8(%r11)
546
movq %mm3, 11*8(%r11)
547
548
549
movq 12*8(%r11), %mm0
550
movq 13*8(%r11), %mm1
551
movq 14*8(%r11), %mm2
552
movq 15*8(%r11), %mm3
553
554
movq %mm0, %mm4
555
movq %mm1, %mm5
556
557
movw 31*2(%r11), %ax
558
movw 32*2(%r11), %bx
559
560
movq %mm2, %mm6
561
movq %mm3, %mm7
562
563
movw %ax, 54*2(%rsp)
564
movw %bx, 11*2(%rsp)
565
566
psraw $0xf, %mm4
567
psraw $0xf, %mm5
568
569
movw 33*2(%r11), %ax
570
movw 34*2(%r11), %bx
571
572
psraw $0xf, %mm6
573
psraw $0xf, %mm7
574
575
movw %ax, 20*2(%rsp)
576
movw %bx, 24*2(%rsp)
577
578
pxor %mm4, %mm0
579
pxor %mm5, %mm1
580
581
movw 35*2(%r11), %ax
582
movw 36*2(%r11), %bx
583
584
pxor %mm6, %mm2
585
pxor %mm7, %mm3
586
587
movw %ax, 33*2(%rsp)
588
movw %bx, 40*2(%rsp)
589
590
psubw %mm4, %mm0
591
psubw %mm5, %mm1
592
593
movw 37*2(%r11), %ax
594
movw 38*2(%r11), %bx
595
596
psubw %mm6, %mm2
597
psubw %mm7, %mm3
598
599
movw %ax, 46*2(%rsp)
600
movw %bx, 53*2(%rsp)
601
602
pmulhw 12*8(%r12), %mm0
603
pmulhw 13*8(%r12), %mm1
604
605
movw 39*2(%r11), %ax
606
movw 40*2(%r11), %bx
607
608
pmulhw 14*8(%r12), %mm2
609
pmulhw 15*8(%r12), %mm3
610
611
movw %ax, 55*2(%rsp)
612
movw %bx, 21*2(%rsp)
613
614
psraw $YUV_PRECISION, %mm0
615
psraw $YUV_PRECISION, %mm1
616
617
movw 41*2(%r11), %ax
618
movw 42*2(%r11), %bx
619
620
psraw $YUV_PRECISION, %mm2
621
psraw $YUV_PRECISION, %mm3
622
623
movw %ax, 23*2(%rsp)
624
movw %bx, 34*2(%rsp)
625
626
pxor %mm4, %mm0
627
pxor %mm5, %mm1
628
629
movw 43*2(%r11), %ax
630
movw 44*2(%r11), %bx
631
632
pxor %mm6, %mm2
633
pxor %mm7, %mm3
634
635
movw %ax, 39*2(%rsp)
636
movw %bx, 47*2(%rsp)
637
638
psubw %mm4, %mm0
639
psubw %mm5, %mm1
640
641
movw 45*2(%r11), %ax
642
movw 46*2(%r11), %bx
643
644
psubw %mm6, %mm2
645
psubw %mm7, %mm3
646
647
movw %ax, 52*2(%rsp)
648
movw %bx, 56*2(%rsp)
649
650
movq %mm0, 12*8(%r11)
651
movq %mm1, 13*8(%r11)
652
653
movw 47*2(%r11), %ax
654
movw 48*2(%r11), %bx
655
656
movq %mm2, 14*8(%r11)
657
movq %mm3, 15*8(%r11)
658
659
movw %ax, 61*2(%rsp)
660
movw %bx, 22*2(%rsp)
661
662
movw 49*2(%r11), %ax
663
movw 50*2(%r11), %bx
664
665
666
movw %ax, 35*2(%rsp)
667
movw %bx, 38*2(%rsp)
668
669
670
movw 51*2(%r11), %ax
671
movw 52*2(%r11), %bx
672
673
movw %ax, 48*2(%rsp)
674
movw %bx, 51*2(%rsp)
675
676
677
movw 53*2(%r11), %ax
678
movw 54*2(%r11), %bx
679
680
movw %ax, 57*2(%rsp)
681
movw %bx, 60*2(%rsp)
682
683
movw 55*2(%r11), %ax
684
movw 56*2(%r11), %bx
685
686
687
movw %ax, 62*2(%rsp)
688
movw %bx, 36*2(%rsp)
689
690
movw 57*2(%r11), %ax
691
movw 58*2(%r11), %bx
692
693
movw %ax, 37*2(%rsp)
694
movw %bx, 49*2(%rsp)
695
696
movw 59*2(%r11), %ax
697
movw 60*2(%r11), %bx
698
699
movw %ax, 50*2(%rsp)
700
movw %bx, 58*2(%rsp)
701
702
movw 61*2(%r11), %ax
703
movl 62*2(%r11), %ebx
704
705
movw %ax, 59*2(%rsp)
706
movl %ebx, 63*2(%rsp)
707
708
add $2, %rsp /* give back 2 bytes of stack space */
709
710
movq (%rsp) , %mm0
711
movq 8(%rsp) , %mm1
712
movq 16(%rsp), %mm2
713
movq 24(%rsp), %mm3
714
715
movq %mm0, (%r11)
716
movq %mm1, 8(%r11)
717
movq %mm2, 16(%r11)
718
movq %mm3, 24(%r11)
719
720
movq 32(%rsp) , %mm0
721
movq 32+8(%rsp) , %mm1
722
movq 32+16(%rsp), %mm2
723
movq 32+24(%rsp), %mm3
724
725
movq %mm0, 32(%r11)
726
movq %mm1, 32+8(%r11)
727
movq %mm2, 32+16(%r11)
728
movq %mm3, 32+24(%r11)
729
730
movq 64(%rsp) , %mm0
731
movq 64+8(%rsp) , %mm1
732
movq 64+16(%rsp), %mm2
733
movq 64+24(%rsp), %mm3
734
735
movq %mm0, 64(%r11)
736
movq %mm1, 64+8(%r11)
737
movq %mm2, 64+16(%r11)
738
movq %mm3, 64+24(%r11)
739
740
movq 96(%rsp) , %mm0
741
movq 96+8(%rsp) , %mm1
742
movq 96+16(%rsp), %mm2
743
movq 96+24(%rsp), %mm3
744
745
add $128, %rsp /* restore the stack pointer */
746
747
pop %r12
748
pop %rbx
749
ret
750
751
.align 8
752
.global _dv_dct_248_block_mmx_x86_64
753
.hidden _dv_dct_248_block_mmx_x86_64
754
.type _dv_dct_248_block_mmx_x86_64,@function
755
_dv_dct_248_block_mmx_x86_64:
756
757
/* void _dv_dct_248_block_mmx_x86_64(int16_t* block); */
758
759
/* argument is block=rdi */
760
761
mov %rdi, %r11 # source
762
763
# column 0
764
765
movq 16*0(%r11), %mm0 # v00
766
movq 16*2(%r11), %mm1 # v01
767
movq 16*4(%r11), %mm2 # v02
768
movq 16*6(%r11), %mm3 # v03
769
770
movq %mm0, %mm4 # v00
771
movq %mm1, %mm5 # v01
772
773
paddw %mm3, %mm0 # v10: v00+v03
774
psubw %mm3, %mm4 # v13: v00-v03
775
776
paddw %mm2, %mm1 # v11: v01+v02
777
psubw %mm2, %mm5 # v12: v01-v02
778
779
movq %mm0, %mm3 # v10
780
781
paddw %mm4, %mm5 # v22: v12+v13
782
paddw %mm1, %mm0 # v20: v10+v11
783
psubw %mm1, %mm3 # v21: v10-v11
784
785
pmulhw WA1(%rip), %mm5 # v32~: WA1*v22
786
movq %mm4, %mm2
787
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
788
789
movq %mm0, 16*0(%r11)
790
791
psubw %mm5, %mm2 # v43: v13-v32
792
paddw %mm4, %mm5 # v42: v32+v13
793
794
movq %mm2, 16*6(%r11)
795
movq %mm3, 16*4(%r11)
796
movq %mm5, 16*2(%r11)
797
798
# column 1
799
800
add $8, %r11 # point to the next 4 columns.
801
# it can be done by adding 8 to immediates
802
# but this is nicer
803
804
movq 16*0(%r11), %mm0 # v00
805
movq 16*2(%r11), %mm1 # v01
806
movq 16*4(%r11), %mm2 # v02
807
movq 16*6(%r11), %mm3 # v03
808
809
movq %mm0, %mm4
810
movq %mm1, %mm5
811
812
paddw %mm3, %mm0 # v10: v00+v03
813
psubw %mm3, %mm4 # v13: v00-v03
814
815
paddw %mm2, %mm1 # v11: v01+v02
816
psubw %mm2, %mm5 # v12: v01-v02
817
818
movq %mm0, %mm3
819
820
paddw %mm4, %mm5 # v22: v12+v13
821
paddw %mm1, %mm0 # v20: v10+v11
822
psubw %mm1, %mm3 # v21: v10-v11
823
824
pmulhw WA1(%rip), %mm5 # v32~: WA1*v22
825
movq %mm4, %mm2
826
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
827
828
movq %mm0, 16*0(%r11)
829
830
psubw %mm5, %mm2 # v43: v13-v32
831
paddw %mm4, %mm5 # v42: v32+v13
832
833
movq %mm2, 16*6(%r11)
834
movq %mm3, 16*4(%r11)
835
movq %mm5, 16*2(%r11)
836
837
# column 0 ... second line
838
839
add $8, %r11
840
841
movq 16*0(%r11), %mm0 # v00
842
movq 16*2(%r11), %mm1 # v01
843
movq 16*4(%r11), %mm2 # v02
844
movq 16*6(%r11), %mm3 # v03
845
846
movq %mm0, %mm4
847
movq %mm1, %mm5
848
849
paddw %mm3, %mm0 # v10: v00+v03
850
psubw %mm3, %mm4 # v13: v00-v03
851
852
paddw %mm2, %mm1 # v11: v01+v02
853
psubw %mm2, %mm5 # v12: v01-v02
854
855
movq %mm0, %mm3
856
857
paddw %mm4, %mm5 # v22: v12+v13
858
paddw %mm1, %mm0 # v20: v10+v11
859
psubw %mm1, %mm3 # v21: v10-v11
860
861
pmulhw WA1(%rip), %mm5 # v32~: WA1*v22
862
movq %mm4, %mm2
863
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
864
865
movq %mm0, 16*0(%r11)
866
867
psubw %mm5, %mm2 # v43: v13-v32
868
paddw %mm4, %mm5 # v42: v32+v13
869
870
movq %mm2, 16*6(%r11)
871
movq %mm3, 16*4(%r11)
872
movq %mm5, 16*2(%r11)
873
874
# column 1 ... second line
875
876
add $8, %r11
877
878
movq 16*0(%r11), %mm0 # v00
879
movq 16*2(%r11), %mm1 # v01
880
movq 16*4(%r11), %mm2 # v02
881
movq 16*6(%r11), %mm3 # v03
882
883
movq %mm0, %mm4
884
movq %mm1, %mm5
885
886
paddw %mm3, %mm0 # v10: v00+v03
887
psubw %mm3, %mm4 # v13: v00-v03
888
889
paddw %mm2, %mm1 # v11: v01+v02
890
psubw %mm2, %mm5 # v12: v01-v02
891
892
movq %mm0, %mm3
893
894
paddw %mm4, %mm5 # v22: v12+v13
895
paddw %mm1, %mm0 # v20: v10+v11
896
psubw %mm1, %mm3 # v21: v10-v11
897
898
pmulhw WA1(%rip), %mm5 # v32~: WA1*v22
899
movq %mm4, %mm2
900
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
901
902
movq %mm0, 16*0(%r11)
903
904
psubw %mm5, %mm2 # v43: v13-v32
905
paddw %mm4, %mm5 # v42: v32+v13
906
907
movq %mm2, 16*6(%r11)
908
movq %mm3, 16*4(%r11)
909
movq %mm5, 16*2(%r11)
910
911
ret
912
913
.align 8
914
.global _dv_dct_248_block_mmx_x86_64_post_sum
915
.hidden _dv_dct_248_block_mmx_x86_64_post_sum
916
.type _dv_dct_248_block_mmx_x86_64_post_sum,@function
917
_dv_dct_248_block_mmx_x86_64_post_sum:
918
919
/* void _dv_dct_248_block_mmx_x86_64_post_sum(int16_t* out_block); */
920
921
/* argument is block=rdi */
922
923
mov %rdi, %r11 # source
924
925
movq 16*0(%r11), %mm0
926
movq 16*1(%r11), %mm1
927
movq 16*2(%r11), %mm2
928
movq 16*3(%r11), %mm3
929
movq 16*4(%r11), %mm4
930
movq 16*5(%r11), %mm5
931
movq 16*6(%r11), %mm6
932
movq 16*7(%r11), %mm7
933
934
psubw %mm1, %mm0
935
paddw 16*0(%r11), %mm1
936
movq %mm0, 16*4(%r11)
937
movq %mm1, 16*0(%r11)
938
939
movq %mm4, %mm1
940
movq %mm2, %mm0
941
paddw %mm5, %mm1
942
psubw %mm5, %mm4
943
paddw %mm3, %mm0
944
psubw %mm3, %mm2
945
946
movq %mm0, 16*1(%r11)
947
movq %mm2, 16*5(%r11)
948
movq %mm1, 16*2(%r11)
949
movq %mm4, 16*6(%r11)
950
951
paddw %mm6, %mm7
952
psubw 16*7(%r11), %mm6
953
954
movq %mm7, 16*3(%r11)
955
movq %mm6, 16*7(%r11)
956
957
add $8, %r11
958
959
movq 16*0(%r11), %mm0
960
movq 16*1(%r11), %mm1
961
movq 16*2(%r11), %mm2
962
movq 16*3(%r11), %mm3
963
movq 16*4(%r11), %mm4
964
movq 16*5(%r11), %mm5
965
movq 16*6(%r11), %mm6
966
movq 16*7(%r11), %mm7
967
968
psubw %mm1, %mm0
969
paddw 16*0(%r11), %mm1
970
movq %mm0, 16*4(%r11)
971
movq %mm1, 16*0(%r11)
972
973
movq %mm4, %mm1
974
movq %mm2, %mm0
975
paddw %mm5, %mm1
976
psubw %mm5, %mm4
977
paddw %mm3, %mm0
978
psubw %mm3, %mm2
979
980
movq %mm0, 16*1(%r11)
981
movq %mm2, 16*5(%r11)
982
movq %mm1, 16*2(%r11)
983
movq %mm4, 16*6(%r11)
984
985
paddw %mm6, %mm7
986
psubw 16*7(%r11), %mm6
987
988
movq %mm7, 16*3(%r11)
989
movq %mm6, 16*7(%r11)
990
991
ret
992
993
.align 8
994
.global _dv_dct_block_mmx_x86_64_postscale_248
995
.hidden _dv_dct_block_mmx_x86_64_postscale_248
996
.type _dv_dct_block_mmx_x86_64_postscale_248,@function
997
_dv_dct_block_mmx_x86_64_postscale_248:
998
999
/* void _dv_dct_block_mmx_x86_64_postscale_248(int16_t* block, int16_t* postscale_matrix); */
1000
1001
/* arguments are block=rdi, postscale=rsi */
1002
1003
push %r12
1004
1005
mov %rdi,%r11 # block matrix
1006
mov %rsi, %r12 # postscale matrix
1007
1008
movq 0*8(%r11), %mm0
1009
movq 1*8(%r11), %mm1
1010
movq 2*8(%r11), %mm2
1011
movq 3*8(%r11), %mm3
1012
1013
movq %mm0, %mm4
1014
movq %mm1, %mm5
1015
movq %mm2, %mm6
1016
movq %mm3, %mm7
1017
psraw $0xf, %mm4
1018
psraw $0xf, %mm5
1019
psraw $0xf, %mm6
1020
psraw $0xf, %mm7
1021
1022
pxor %mm4, %mm0
1023
pxor %mm5, %mm1
1024
pxor %mm6, %mm2
1025
pxor %mm7, %mm3
1026
1027
psubw %mm4, %mm0
1028
psubw %mm5, %mm1
1029
psubw %mm6, %mm2
1030
psubw %mm7, %mm3
1031
1032
pmulhw 0*8(%r12), %mm0
1033
pmulhw 1*8(%r12), %mm1
1034
pmulhw 2*8(%r12), %mm2
1035
pmulhw 3*8(%r12), %mm3
1036
1037
psraw $YUV_PRECISION, %mm0
1038
psraw $YUV_PRECISION, %mm1
1039
psraw $YUV_PRECISION, %mm2
1040
psraw $YUV_PRECISION, %mm3
1041
1042
pxor %mm4, %mm0
1043
pxor %mm5, %mm1
1044
pxor %mm6, %mm2
1045
pxor %mm7, %mm3
1046
1047
psubw %mm4, %mm0
1048
psubw %mm5, %mm1
1049
psubw %mm6, %mm2
1050
psubw %mm7, %mm3
1051
1052
movq %mm0, 0*8(%r11)
1053
movq %mm1, 1*8(%r11)
1054
movq %mm2, 2*8(%r11)
1055
movq %mm3, 3*8(%r11)
1056
1057
1058
movq 4*8(%r11), %mm0
1059
movq 5*8(%r11), %mm1
1060
movq 6*8(%r11), %mm2
1061
movq 7*8(%r11), %mm3
1062
1063
movq %mm0, %mm4
1064
movq %mm1, %mm5
1065
movq %mm2, %mm6
1066
movq %mm3, %mm7
1067
psraw $0xf, %mm4
1068
psraw $0xf, %mm5
1069
psraw $0xf, %mm6
1070
psraw $0xf, %mm7
1071
1072
pxor %mm4, %mm0
1073
pxor %mm5, %mm1
1074
pxor %mm6, %mm2
1075
pxor %mm7, %mm3
1076
1077
psubw %mm4, %mm0
1078
psubw %mm5, %mm1
1079
psubw %mm6, %mm2
1080
psubw %mm7, %mm3
1081
1082
pmulhw 4*8(%r12), %mm0
1083
pmulhw 5*8(%r12), %mm1
1084
pmulhw 6*8(%r12), %mm2
1085
pmulhw 7*8(%r12), %mm3
1086
1087
psraw $YUV_PRECISION, %mm0
1088
psraw $YUV_PRECISION, %mm1
1089
psraw $YUV_PRECISION, %mm2
1090
psraw $YUV_PRECISION, %mm3
1091
1092
pxor %mm4, %mm0
1093
pxor %mm5, %mm1
1094
pxor %mm6, %mm2
1095
pxor %mm7, %mm3
1096
1097
psubw %mm4, %mm0
1098
psubw %mm5, %mm1
1099
psubw %mm6, %mm2
1100
psubw %mm7, %mm3
1101
1102
movq %mm0, 4*8(%r11)
1103
movq %mm1, 5*8(%r11)
1104
movq %mm2, 6*8(%r11)
1105
movq %mm3, 7*8(%r11)
1106
1107
1108
movq 8*8(%r11), %mm0
1109
movq 9*8(%r11), %mm1
1110
movq 10*8(%r11), %mm2
1111
movq 11*8(%r11), %mm3
1112
1113
movq %mm0, %mm4
1114
movq %mm1, %mm5
1115
movq %mm2, %mm6
1116
movq %mm3, %mm7
1117
psraw $0xf, %mm4
1118
psraw $0xf, %mm5
1119
psraw $0xf, %mm6
1120
psraw $0xf, %mm7
1121
1122
pxor %mm4, %mm0
1123
pxor %mm5, %mm1
1124
pxor %mm6, %mm2
1125
pxor %mm7, %mm3
1126
1127
psubw %mm4, %mm0
1128
psubw %mm5, %mm1
1129
psubw %mm6, %mm2
1130
psubw %mm7, %mm3
1131
1132
pmulhw 8*8(%r12), %mm0
1133
pmulhw 9*8(%r12), %mm1
1134
pmulhw 10*8(%r12), %mm2
1135
pmulhw 11*8(%r12), %mm3
1136
1137
psraw $YUV_PRECISION, %mm0
1138
psraw $YUV_PRECISION, %mm1
1139
psraw $YUV_PRECISION, %mm2
1140
psraw $YUV_PRECISION, %mm3
1141
1142
pxor %mm4, %mm0
1143
pxor %mm5, %mm1
1144
pxor %mm6, %mm2
1145
pxor %mm7, %mm3
1146
1147
psubw %mm4, %mm0
1148
psubw %mm5, %mm1
1149
psubw %mm6, %mm2
1150
psubw %mm7, %mm3
1151
1152
movq %mm0, 8*8(%r11)
1153
movq %mm1, 9*8(%r11)
1154
movq %mm2, 10*8(%r11)
1155
movq %mm3, 11*8(%r11)
1156
1157
1158
movq 12*8(%r11), %mm0
1159
movq 13*8(%r11), %mm1
1160
movq 14*8(%r11), %mm2
1161
movq 15*8(%r11), %mm3
1162
1163
movq %mm0, %mm4
1164
movq %mm1, %mm5
1165
movq %mm2, %mm6
1166
movq %mm3, %mm7
1167
psraw $0xf, %mm4
1168
psraw $0xf, %mm5
1169
psraw $0xf, %mm6
1170
psraw $0xf, %mm7
1171
1172
pxor %mm4, %mm0
1173
pxor %mm5, %mm1
1174
pxor %mm6, %mm2
1175
pxor %mm7, %mm3
1176
1177
psubw %mm4, %mm0
1178
psubw %mm5, %mm1
1179
psubw %mm6, %mm2
1180
psubw %mm7, %mm3
1181
1182
pmulhw 12*8(%r12), %mm0
1183
pmulhw 13*8(%r12), %mm1
1184
pmulhw 14*8(%r12), %mm2
1185
pmulhw 15*8(%r12), %mm3
1186
1187
psraw $YUV_PRECISION, %mm0
1188
psraw $YUV_PRECISION, %mm1
1189
psraw $YUV_PRECISION, %mm2
1190
psraw $YUV_PRECISION, %mm3
1191
1192
pxor %mm4, %mm0
1193
pxor %mm5, %mm1
1194
pxor %mm6, %mm2
1195
pxor %mm7, %mm3
1196
1197
psubw %mm4, %mm0
1198
psubw %mm5, %mm1
1199
psubw %mm6, %mm2
1200
psubw %mm7, %mm3
1201
1202
movq %mm0, 12*8(%r11)
1203
movq %mm1, 13*8(%r11)
1204
movq %mm2, 14*8(%r11)
1205
movq %mm3, 15*8(%r11)
1206
1207
pop %r12
1208
1209
ret
Loggerhead is a web-based interface for Breezy
Version: 2.0.1