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
|
;
; Copyright (c) 2016, Alliance for Open Media. All rights reserved
;
; This source code is subject to the terms of the BSD 2 Clause License and
; the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
; was not distributed with this source code in the LICENSE file, you can
; obtain it at www.aomedia.org/license/software. If the Alliance for Open
; Media Patent License 1.0 was not distributed with this source code in the
; PATENTS file, you can obtain it at www.aomedia.org/license/patent.
;
;
%include "third_party/x86inc/x86inc.asm"
SECTION .text
; Macro Arguments
; Arg 1: Width
; Arg 2: Height
; Arg 3: Number of general purpose registers: 5 for 32-bit build, 6 for 64-bit
; Arg 4: Type of function: if 0, normal sad; if 1, avg; if 2, skip rows
%macro SAD_FN 4
%if %4 == 0 ; normal sad
%if %3 == 5
cglobal sad%1x%2, 4, %3, 5, src, src_stride, ref, ref_stride, n_rows
%else ; %3 == 7
cglobal sad%1x%2, 4, %3, 6, src, src_stride, ref, ref_stride, \
src_stride3, ref_stride3, n_rows
%endif ; %3 == 5/7
%elif %4 == 2 ; skip
%if %3 == 5
cglobal sad_skip_%1x%2, 4, %3, 5, src, src_stride, ref, ref_stride, n_rows
%else ; %3 == 7
cglobal sad_skip_%1x%2, 4, %3, 6, src, src_stride, ref, ref_stride, \
src_stride3, ref_stride3, n_rows
%endif ; %3 == 5/7
%else
%if %3 == 5
cglobal sad%1x%2_avg, 5, 1 + %3, 5, src, src_stride, ref, ref_stride, \
second_pred, n_rows
%else ; %3 == 7
cglobal sad%1x%2_avg, 5, AOM_ARCH_X86_64 + %3, 6, src, src_stride, \
ref, ref_stride, \
second_pred, \
src_stride3, ref_stride3
%if AOM_ARCH_X86_64
%define n_rowsd r7d
%else ; x86-32
%define n_rowsd dword r0m
%endif ; x86-32/64
%endif ; %3 == 5/7
%endif ; sad/avg/skip
%if %4 == 2; skip rows so double the stride
lea src_strided, [src_strided*2]
lea ref_strided, [ref_strided*2]
%endif ; %4 skip
movsxdifnidn src_strideq, src_strided
movsxdifnidn ref_strideq, ref_strided
%if %3 == 7
lea src_stride3q, [src_strideq*3]
lea ref_stride3q, [ref_strideq*3]
%endif ; %3 == 7
%endmacro
; unsigned int aom_sad128x128_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD128XN 1-2 0
SAD_FN 128, %1, 5, %2
%if %2 == 2
mov n_rowsd, %1/2
%else
mov n_rowsd, %1
%endif
pxor m0, m0
.loop:
movu m1, [refq]
movu m2, [refq+16]
movu m3, [refq+32]
movu m4, [refq+48]
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
pavgb m2, [second_predq+mmsize*1]
pavgb m3, [second_predq+mmsize*2]
pavgb m4, [second_predq+mmsize*3]
%endif
psadbw m1, [srcq]
psadbw m2, [srcq+16]
psadbw m3, [srcq+32]
psadbw m4, [srcq+48]
paddd m1, m2
paddd m3, m4
paddd m0, m1
paddd m0, m3
movu m1, [refq+64]
movu m2, [refq+80]
movu m3, [refq+96]
movu m4, [refq+112]
%if %2 == 1
pavgb m1, [second_predq+mmsize*4]
pavgb m2, [second_predq+mmsize*5]
pavgb m3, [second_predq+mmsize*6]
pavgb m4, [second_predq+mmsize*7]
lea second_predq, [second_predq+mmsize*8]
%endif
psadbw m1, [srcq+64]
psadbw m2, [srcq+80]
psadbw m3, [srcq+96]
psadbw m4, [srcq+112]
add refq, ref_strideq
add srcq, src_strideq
paddd m1, m2
paddd m3, m4
paddd m0, m1
paddd m0, m3
sub n_rowsd, 1
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD128XN 128 ; sad128x128_sse2
SAD128XN 128, 1 ; sad128x128_avg_sse2
SAD128XN 128, 2 ; sad128x128_skip_sse2
SAD128XN 64 ; sad128x64_sse2
SAD128XN 64, 1 ; sad128x64_avg_sse2
SAD128XN 64, 2 ; sad128x64_skip_sse2
; unsigned int aom_sad64x64_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD64XN 1-2 0
SAD_FN 64, %1, 5, %2
%if %2 == 2
mov n_rowsd, %1/2
%else
mov n_rowsd, %1
%endif
pxor m0, m0
.loop:
movu m1, [refq]
movu m2, [refq+16]
movu m3, [refq+32]
movu m4, [refq+48]
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
pavgb m2, [second_predq+mmsize*1]
pavgb m3, [second_predq+mmsize*2]
pavgb m4, [second_predq+mmsize*3]
lea second_predq, [second_predq+mmsize*4]
%endif
psadbw m1, [srcq]
psadbw m2, [srcq+16]
psadbw m3, [srcq+32]
psadbw m4, [srcq+48]
paddd m1, m2
paddd m3, m4
add refq, ref_strideq
paddd m0, m1
add srcq, src_strideq
paddd m0, m3
dec n_rowsd
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD64XN 128 ; sad64x128_sse2
SAD64XN 64 ; sad64x64_sse2
SAD64XN 32 ; sad64x32_sse2
SAD64XN 16 ; sad64x16_sse2
SAD64XN 128, 1 ; sad64x128_avg_sse2
SAD64XN 64, 1 ; sad64x64_avg_sse2
SAD64XN 32, 1 ; sad64x32_avg_sse2
SAD64XN 16, 1 ; sad64x16_avg_sse2
SAD64XN 128, 2 ; sad64x128_skip_sse2
SAD64XN 64, 2 ; sad64x64_skip_sse2
SAD64XN 32, 2 ; sad64x32_skip_sse2
SAD64XN 16, 2 ; sad64x16_skip_sse2
; unsigned int aom_sad32x32_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD32XN 1-2 0
SAD_FN 32, %1, 5, %2
%if %2 == 2
mov n_rowsd, %1/4
%else
mov n_rowsd, %1/2
%endif
pxor m0, m0
.loop:
movu m1, [refq]
movu m2, [refq+16]
movu m3, [refq+ref_strideq]
movu m4, [refq+ref_strideq+16]
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
pavgb m2, [second_predq+mmsize*1]
pavgb m3, [second_predq+mmsize*2]
pavgb m4, [second_predq+mmsize*3]
lea second_predq, [second_predq+mmsize*4]
%endif
psadbw m1, [srcq]
psadbw m2, [srcq+16]
psadbw m3, [srcq+src_strideq]
psadbw m4, [srcq+src_strideq+16]
paddd m1, m2
paddd m3, m4
lea refq, [refq+ref_strideq*2]
paddd m0, m1
lea srcq, [srcq+src_strideq*2]
paddd m0, m3
dec n_rowsd
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD32XN 64 ; sad32x64_sse2
SAD32XN 32 ; sad32x32_sse2
SAD32XN 16 ; sad32x16_sse2
SAD32XN 8 ; sad_32x8_sse2
SAD32XN 64, 1 ; sad32x64_avg_sse2
SAD32XN 32, 1 ; sad32x32_avg_sse2
SAD32XN 16, 1 ; sad32x16_avg_sse2
SAD32XN 8, 1 ; sad_32x8_avg_sse2
SAD32XN 64, 2 ; sad32x64_skip_sse2
SAD32XN 32, 2 ; sad32x32_skip_sse2
SAD32XN 16, 2 ; sad32x16_skip_sse2
SAD32XN 8, 2 ; sad_32x8_skip_sse2
; unsigned int aom_sad16x{8,16}_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD16XN 1-2 0
SAD_FN 16, %1, 7, %2
%if %2 == 2
mov n_rowsd, %1/8
%else
mov n_rowsd, %1/4
%endif
pxor m0, m0
.loop:
movu m1, [refq]
movu m2, [refq+ref_strideq]
movu m3, [refq+ref_strideq*2]
movu m4, [refq+ref_stride3q]
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
pavgb m2, [second_predq+mmsize*1]
pavgb m3, [second_predq+mmsize*2]
pavgb m4, [second_predq+mmsize*3]
lea second_predq, [second_predq+mmsize*4]
%endif
psadbw m1, [srcq]
psadbw m2, [srcq+src_strideq]
psadbw m3, [srcq+src_strideq*2]
psadbw m4, [srcq+src_stride3q]
paddd m1, m2
paddd m3, m4
lea refq, [refq+ref_strideq*4]
paddd m0, m1
lea srcq, [srcq+src_strideq*4]
paddd m0, m3
dec n_rowsd
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD16XN 64 ; sad_16x64_sse2
SAD16XN 32 ; sad16x32_sse2
SAD16XN 16 ; sad16x16_sse2
SAD16XN 8 ; sad16x8_sse2
SAD16XN 4 ; sad_16x4_sse2
SAD16XN 64, 1 ; sad_16x64_avg_sse2
SAD16XN 32, 1 ; sad16x32_avg_sse2
SAD16XN 16, 1 ; sad16x16_avg_sse2
SAD16XN 8, 1 ; sad16x8_avg_sse2
SAD16XN 4, 1 ; sad_16x4_avg_sse2
SAD16XN 64, 2 ; sad_16x64_skip_sse2
SAD16XN 32, 2 ; sad16x32_skip_sse2
SAD16XN 16, 2 ; sad16x16_skip_sse2
SAD16XN 8, 2 ; sad16x8_skip_sse2
; unsigned int aom_sad8x{8,16}_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD8XN 1-2 0
SAD_FN 8, %1, 7, %2
%if %2 == 2
mov n_rowsd, %1/8
%else
mov n_rowsd, %1/4
%endif
pxor m0, m0
.loop:
movh m1, [refq]
movhps m1, [refq+ref_strideq]
movh m2, [refq+ref_strideq*2]
movhps m2, [refq+ref_stride3q]
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
pavgb m2, [second_predq+mmsize*1]
lea second_predq, [second_predq+mmsize*2]
%endif
movh m3, [srcq]
movhps m3, [srcq+src_strideq]
movh m4, [srcq+src_strideq*2]
movhps m4, [srcq+src_stride3q]
psadbw m1, m3
psadbw m2, m4
lea refq, [refq+ref_strideq*4]
paddd m0, m1
lea srcq, [srcq+src_strideq*4]
paddd m0, m2
dec n_rowsd
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD8XN 32 ; sad_8x32_sse2
SAD8XN 16 ; sad8x16_sse2
SAD8XN 8 ; sad8x8_sse2
SAD8XN 4 ; sad8x4_sse2
SAD8XN 32, 1 ; sad_8x32_avg_sse2
SAD8XN 16, 1 ; sad8x16_avg_sse2
SAD8XN 8, 1 ; sad8x8_avg_sse2
SAD8XN 4, 1 ; sad8x4_avg_sse2
SAD8XN 32, 2 ; sad_8x32_skip_sse2
SAD8XN 16, 2 ; sad8x16_skip_sse2
SAD8XN 8, 2 ; sad8x8_skip_sse2
; unsigned int aom_sad4x{4, 8}_sse2(uint8_t *src, int src_stride,
; uint8_t *ref, int ref_stride);
%macro SAD4XN 1-2 0
SAD_FN 4, %1, 7, %2
%if %2 == 2
mov n_rowsd, %1/8
%else
mov n_rowsd, %1/4
%endif
pxor m0, m0
.loop:
movd m1, [refq]
movd m2, [refq+ref_strideq]
movd m3, [refq+ref_strideq*2]
movd m4, [refq+ref_stride3q]
punpckldq m1, m2
punpckldq m3, m4
movlhps m1, m3
%if %2 == 1
pavgb m1, [second_predq+mmsize*0]
lea second_predq, [second_predq+mmsize*1]
%endif
movd m2, [srcq]
movd m5, [srcq+src_strideq]
movd m4, [srcq+src_strideq*2]
movd m3, [srcq+src_stride3q]
punpckldq m2, m5
punpckldq m4, m3
movlhps m2, m4
psadbw m1, m2
lea refq, [refq+ref_strideq*4]
paddd m0, m1
lea srcq, [srcq+src_strideq*4]
dec n_rowsd
jg .loop
movhlps m1, m0
paddd m0, m1
%if %2 == 2 ; we skipped rows, so now we need to double the sad
pslld m0, 1
%endif
movd eax, m0
RET
%endmacro
INIT_XMM sse2
SAD4XN 16 ; sad_4x16_sse2
SAD4XN 8 ; sad4x8_sse
SAD4XN 4 ; sad4x4_sse
SAD4XN 16, 1 ; sad_4x16_avg_sse2
SAD4XN 8, 1 ; sad4x8_avg_sse
SAD4XN 4, 1 ; sad4x4_avg_sse
SAD4XN 16, 2 ; sad_4x16_skip_sse2
SAD4XN 8, 2 ; sad4x8_skip_sse
|
