1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
|
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
%define VP9_FILTER_WEIGHT 128
%define VP9_FILTER_SHIFT 7
;void vp9_post_proc_down_and_across_mmx
;(
; unsigned char *src_ptr,
; unsigned char *dst_ptr,
; int src_pixels_per_line,
; int dst_pixels_per_line,
; int rows,
; int cols,
; int flimit
;)
global sym(vp9_post_proc_down_and_across_mmx) PRIVATE
sym(vp9_post_proc_down_and_across_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 7
GET_GOT rbx
push rsi
push rdi
; end prolog
%if ABI_IS_32BIT=1 && CONFIG_PIC=1
; move the global rd onto the stack, since we don't have enough registers
; to do PIC addressing
movq mm0, [GLOBAL(rd)]
sub rsp, 8
movq [rsp], mm0
%define RD [rsp]
%else
%define RD [GLOBAL(rd)]
%endif
push rbx
lea rbx, [GLOBAL(Blur)]
movd mm2, dword ptr arg(6) ;flimit
punpcklwd mm2, mm2
punpckldq mm2, mm2
mov rsi, arg(0) ;src_ptr
mov rdi, arg(1) ;dst_ptr
movsxd rcx, DWORD PTR arg(4) ;rows
movsxd rax, DWORD PTR arg(2) ;src_pixels_per_line ; destination pitch?
pxor mm0, mm0 ; mm0 = 00000000
.nextrow:
xor rdx, rdx ; clear out rdx for use as loop counter
.nextcol:
pxor mm7, mm7 ; mm7 = 00000000
movq mm6, [rbx + 32 ] ; mm6 = kernel 2 taps
movq mm3, [rsi] ; mm4 = r0 p0..p7
punpcklbw mm3, mm0 ; mm3 = p0..p3
movq mm1, mm3 ; mm1 = p0..p3
pmullw mm3, mm6 ; mm3 *= kernel 2 modifiers
movq mm6, [rbx + 48] ; mm6 = kernel 3 taps
movq mm5, [rsi + rax] ; mm4 = r1 p0..p7
punpcklbw mm5, mm0 ; mm5 = r1 p0..p3
pmullw mm6, mm5 ; mm6 *= p0..p3 * kernel 3 modifiers
paddusw mm3, mm6 ; mm3 += mm6
; thresholding
movq mm7, mm1 ; mm7 = r0 p0..p3
psubusw mm7, mm5 ; mm7 = r0 p0..p3 - r1 p0..p3
psubusw mm5, mm1 ; mm5 = r1 p0..p3 - r0 p0..p3
paddusw mm7, mm5 ; mm7 = abs(r0 p0..p3 - r1 p0..p3)
pcmpgtw mm7, mm2
movq mm6, [rbx + 64 ] ; mm6 = kernel 4 modifiers
movq mm5, [rsi + 2*rax] ; mm4 = r2 p0..p7
punpcklbw mm5, mm0 ; mm5 = r2 p0..p3
pmullw mm6, mm5 ; mm5 *= kernel 4 modifiers
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = r0 p0..p3
psubusw mm6, mm5 ; mm6 = r0 p0..p3 - r2 p0..p3
psubusw mm5, mm1 ; mm5 = r2 p0..p3 - r2 p0..p3
paddusw mm6, mm5 ; mm6 = abs(r0 p0..p3 - r2 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
neg rax
movq mm6, [rbx ] ; kernel 0 taps
movq mm5, [rsi+2*rax] ; mm4 = r-2 p0..p7
punpcklbw mm5, mm0 ; mm5 = r-2 p0..p3
pmullw mm6, mm5 ; mm5 *= kernel 0 modifiers
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = r0 p0..p3
psubusw mm6, mm5 ; mm6 = p0..p3 - r-2 p0..p3
psubusw mm5, mm1 ; mm5 = r-2 p0..p3 - p0..p3
paddusw mm6, mm5 ; mm6 = abs(r0 p0..p3 - r-2 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
movq mm6, [rbx + 16] ; kernel 1 taps
movq mm4, [rsi+rax] ; mm4 = r-1 p0..p7
punpcklbw mm4, mm0 ; mm4 = r-1 p0..p3
pmullw mm6, mm4 ; mm4 *= kernel 1 modifiers.
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = r0 p0..p3
psubusw mm6, mm4 ; mm6 = p0..p3 - r-2 p0..p3
psubusw mm4, mm1 ; mm5 = r-1 p0..p3 - p0..p3
paddusw mm6, mm4 ; mm6 = abs(r0 p0..p3 - r-1 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
paddusw mm3, RD ; mm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; mm3 /= 128
pand mm1, mm7 ; mm1 select vals > thresh from source
pandn mm7, mm3 ; mm7 select vals < thresh from blurred result
paddusw mm1, mm7 ; combination
packuswb mm1, mm0 ; pack to bytes
movd [rdi], mm1 ;
neg rax ; pitch is positive
add rsi, 4
add rdi, 4
add rdx, 4
cmp edx, dword ptr arg(5) ;cols
jl .nextcol
; done with the all cols, start the across filtering in place
sub rsi, rdx
sub rdi, rdx
push rax
xor rdx, rdx
mov rax, [rdi-4];
.acrossnextcol:
pxor mm7, mm7 ; mm7 = 00000000
movq mm6, [rbx + 32 ] ;
movq mm4, [rdi+rdx] ; mm4 = p0..p7
movq mm3, mm4 ; mm3 = p0..p7
punpcklbw mm3, mm0 ; mm3 = p0..p3
movq mm1, mm3 ; mm1 = p0..p3
pmullw mm3, mm6 ; mm3 *= kernel 2 modifiers
movq mm6, [rbx + 48]
psrlq mm4, 8 ; mm4 = p1..p7
movq mm5, mm4 ; mm5 = p1..p7
punpcklbw mm5, mm0 ; mm5 = p1..p4
pmullw mm6, mm5 ; mm6 *= p1..p4 * kernel 3 modifiers
paddusw mm3, mm6 ; mm3 += mm6
; thresholding
movq mm7, mm1 ; mm7 = p0..p3
psubusw mm7, mm5 ; mm7 = p0..p3 - p1..p4
psubusw mm5, mm1 ; mm5 = p1..p4 - p0..p3
paddusw mm7, mm5 ; mm7 = abs(p0..p3 - p1..p4)
pcmpgtw mm7, mm2
movq mm6, [rbx + 64 ]
psrlq mm4, 8 ; mm4 = p2..p7
movq mm5, mm4 ; mm5 = p2..p7
punpcklbw mm5, mm0 ; mm5 = p2..p5
pmullw mm6, mm5 ; mm5 *= kernel 4 modifiers
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = p0..p3
psubusw mm6, mm5 ; mm6 = p0..p3 - p1..p4
psubusw mm5, mm1 ; mm5 = p1..p4 - p0..p3
paddusw mm6, mm5 ; mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
movq mm6, [rbx ]
movq mm4, [rdi+rdx-2] ; mm4 = p-2..p5
movq mm5, mm4 ; mm5 = p-2..p5
punpcklbw mm5, mm0 ; mm5 = p-2..p1
pmullw mm6, mm5 ; mm5 *= kernel 0 modifiers
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = p0..p3
psubusw mm6, mm5 ; mm6 = p0..p3 - p1..p4
psubusw mm5, mm1 ; mm5 = p1..p4 - p0..p3
paddusw mm6, mm5 ; mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
movq mm6, [rbx + 16]
psrlq mm4, 8 ; mm4 = p-1..p5
punpcklbw mm4, mm0 ; mm4 = p-1..p2
pmullw mm6, mm4 ; mm4 *= kernel 1 modifiers.
paddusw mm3, mm6 ; mm3 += mm5
; thresholding
movq mm6, mm1 ; mm6 = p0..p3
psubusw mm6, mm4 ; mm6 = p0..p3 - p1..p4
psubusw mm4, mm1 ; mm5 = p1..p4 - p0..p3
paddusw mm6, mm4 ; mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ; accumulate thresholds
paddusw mm3, RD ; mm3 += round value
psraw mm3, VP9_FILTER_SHIFT ; mm3 /= 128
pand mm1, mm7 ; mm1 select vals > thresh from source
pandn mm7, mm3 ; mm7 select vals < thresh from blurred result
paddusw mm1, mm7 ; combination
packuswb mm1, mm0 ; pack to bytes
mov DWORD PTR [rdi+rdx-4], eax ; store previous four bytes
movd eax, mm1
add rdx, 4
cmp edx, dword ptr arg(5) ;cols
jl .acrossnextcol;
mov DWORD PTR [rdi+rdx-4], eax
pop rax
; done with this rwo
add rsi,rax ; next line
movsxd rax, dword ptr arg(3) ;dst_pixels_per_line ; destination pitch?
add rdi,rax ; next destination
movsxd rax, dword ptr arg(2) ;src_pixels_per_line ; destination pitch?
dec rcx ; decrement count
jnz .nextrow ; next row
pop rbx
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
%undef RD
;void vp9_mbpost_proc_down_mmx(unsigned char *dst,
; int pitch, int rows, int cols,int flimit)
extern sym(vp9_rv)
global sym(vp9_mbpost_proc_down_mmx) PRIVATE
sym(vp9_mbpost_proc_down_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 5
GET_GOT rbx
push rsi
push rdi
; end prolog
ALIGN_STACK 16, rax
sub rsp, 136
; unsigned char d[16][8] at [rsp]
; create flimit2 at [rsp+128]
mov eax, dword ptr arg(4) ;flimit
mov [rsp+128], eax
mov [rsp+128+4], eax
%define flimit2 [rsp+128]
%if ABI_IS_32BIT=0
lea r8, [GLOBAL(sym(vp9_rv))]
%endif
;rows +=8;
add dword ptr arg(2), 8
;for(c=0; c<cols; c+=4)
.loop_col:
mov rsi, arg(0) ;s
pxor mm0, mm0 ;
movsxd rax, dword ptr arg(1) ;pitch ;
neg rax ; rax = -pitch
lea rsi, [rsi + rax*8]; ; rdi = s[-pitch*8]
neg rax
pxor mm5, mm5
pxor mm6, mm6 ;
pxor mm7, mm7 ;
mov rdi, rsi
mov rcx, 15 ;
.loop_initvar:
movd mm1, DWORD PTR [rdi];
punpcklbw mm1, mm0 ;
paddw mm5, mm1 ;
pmullw mm1, mm1 ;
movq mm2, mm1 ;
punpcklwd mm1, mm0 ;
punpckhwd mm2, mm0 ;
paddd mm6, mm1 ;
paddd mm7, mm2 ;
lea rdi, [rdi+rax] ;
dec rcx
jne .loop_initvar
;save the var and sum
xor rdx, rdx
.loop_row:
movd mm1, DWORD PTR [rsi] ; [s-pitch*8]
movd mm2, DWORD PTR [rdi] ; [s+pitch*7]
punpcklbw mm1, mm0
punpcklbw mm2, mm0
paddw mm5, mm2
psubw mm5, mm1
pmullw mm2, mm2
movq mm4, mm2
punpcklwd mm2, mm0
punpckhwd mm4, mm0
paddd mm6, mm2
paddd mm7, mm4
pmullw mm1, mm1
movq mm2, mm1
punpcklwd mm1, mm0
psubd mm6, mm1
punpckhwd mm2, mm0
psubd mm7, mm2
movq mm3, mm6
pslld mm3, 4
psubd mm3, mm6
movq mm1, mm5
movq mm4, mm5
pmullw mm1, mm1
pmulhw mm4, mm4
movq mm2, mm1
punpcklwd mm1, mm4
punpckhwd mm2, mm4
movq mm4, mm7
pslld mm4, 4
psubd mm4, mm7
psubd mm3, mm1
psubd mm4, mm2
psubd mm3, flimit2
psubd mm4, flimit2
psrad mm3, 31
psrad mm4, 31
packssdw mm3, mm4
packsswb mm3, mm0
movd mm1, DWORD PTR [rsi+rax*8]
movq mm2, mm1
punpcklbw mm1, mm0
paddw mm1, mm5
mov rcx, rdx
and rcx, 127
%if ABI_IS_32BIT=1 && CONFIG_PIC=1
push rax
lea rax, [GLOBAL(sym(vp9_rv))]
movq mm4, [rax + rcx*2] ;vp9_rv[rcx*2]
pop rax
%elif ABI_IS_32BIT=0
movq mm4, [r8 + rcx*2] ;vp9_rv[rcx*2]
%else
movq mm4, [sym(vp9_rv) + rcx*2]
%endif
paddw mm1, mm4
;paddw xmm1, eight8s
psraw mm1, 4
packuswb mm1, mm0
pand mm1, mm3
pandn mm3, mm2
por mm1, mm3
and rcx, 15
movd DWORD PTR [rsp+rcx*4], mm1 ;d[rcx*4]
mov rcx, rdx
sub rcx, 8
and rcx, 15
movd mm1, DWORD PTR [rsp+rcx*4] ;d[rcx*4]
movd [rsi], mm1
lea rsi, [rsi+rax]
lea rdi, [rdi+rax]
add rdx, 1
cmp edx, dword arg(2) ;rows
jl .loop_row
add dword arg(0), 4 ; s += 4
sub dword arg(3), 4 ; cols -= 4
cmp dword arg(3), 0
jg .loop_col
add rsp, 136
pop rsp
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
%undef flimit2
;void vp9_plane_add_noise_mmx (unsigned char *start, unsigned char *noise,
; unsigned char blackclamp[16],
; unsigned char whiteclamp[16],
; unsigned char bothclamp[16],
; unsigned int width, unsigned int height, int pitch)
extern sym(rand)
global sym(vp9_plane_add_noise_mmx) PRIVATE
sym(vp9_plane_add_noise_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 8
GET_GOT rbx
push rsi
push rdi
; end prolog
.addnoise_loop:
call sym(rand) WRT_PLT
mov rcx, arg(1) ;noise
and rax, 0xff
add rcx, rax
; we rely on the fact that the clamping vectors are stored contiguously
; in black/white/both order. Note that we have to reload this here because
; rdx could be trashed by rand()
mov rdx, arg(2) ; blackclamp
mov rdi, rcx
movsxd rcx, dword arg(5) ;[Width]
mov rsi, arg(0) ;Pos
xor rax,rax
.addnoise_nextset:
movq mm1,[rsi+rax] ; get the source
psubusb mm1, [rdx] ;blackclamp ; clamp both sides so we don't outrange adding noise
paddusb mm1, [rdx+32] ;bothclamp
psubusb mm1, [rdx+16] ;whiteclamp
movq mm2,[rdi+rax] ; get the noise for this line
paddb mm1,mm2 ; add it in
movq [rsi+rax],mm1 ; store the result
add rax,8 ; move to the next line
cmp rax, rcx
jl .addnoise_nextset
movsxd rax, dword arg(7) ; Pitch
add arg(0), rax ; Start += Pitch
sub dword arg(6), 1 ; Height -= 1
jg .addnoise_loop
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
align 16
Blur:
times 16 dw 16
times 8 dw 64
times 16 dw 16
times 8 dw 0
rd:
times 4 dw 0x40
|
