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
|
;
; 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
; HIGH_PROCESS_4x2x4 first, off_{first,second}_{src,ref}, advance_at_end
%macro HIGH_PROCESS_4x2x4 5-6 0
movh m0, [srcq +%2*2]
%if %1 == 1
movu m4, [ref1q+%3*2]
movu m5, [ref2q+%3*2]
movu m6, [ref3q+%3*2]
movu m7, [ref4q+%3*2]
movhps m0, [srcq +%4*2]
movhps m4, [ref1q+%5*2]
movhps m5, [ref2q+%5*2]
movhps m6, [ref3q+%5*2]
movhps m7, [ref4q+%5*2]
mova m3, m0
mova m2, m0
psubusw m3, m4
psubusw m2, m5
psubusw m4, m0
psubusw m5, m0
por m4, m3
por m5, m2
pmaddwd m4, m1
pmaddwd m5, m1
mova m3, m0
mova m2, m0
psubusw m3, m6
psubusw m2, m7
psubusw m6, m0
psubusw m7, m0
por m6, m3
por m7, m2
pmaddwd m6, m1
pmaddwd m7, m1
%else
movu m2, [ref1q+%3*2]
movhps m0, [srcq +%4*2]
movhps m2, [ref1q+%5*2]
mova m3, m0
psubusw m3, m2
psubusw m2, m0
por m2, m3
pmaddwd m2, m1
paddd m4, m2
movu m2, [ref2q+%3*2]
mova m3, m0
movhps m2, [ref2q+%5*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
pmaddwd m2, m1
paddd m5, m2
movu m2, [ref3q+%3*2]
mova m3, m0
movhps m2, [ref3q+%5*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
pmaddwd m2, m1
paddd m6, m2
movu m2, [ref4q+%3*2]
mova m3, m0
movhps m2, [ref4q+%5*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
pmaddwd m2, m1
paddd m7, m2
%endif
%if %6 == 1
lea srcq, [srcq +src_strideq*4]
lea ref1q, [ref1q+ref_strideq*4]
lea ref2q, [ref2q+ref_strideq*4]
lea ref3q, [ref3q+ref_strideq*4]
lea ref4q, [ref4q+ref_strideq*4]
%endif
%endmacro
; PROCESS_8x2x4 first, off_{first,second}_{src,ref}, advance_at_end
%macro HIGH_PROCESS_8x2x4 5-6 0
; 1st 8 px
mova m0, [srcq +%2*2]
%if %1 == 1
movu m4, [ref1q+%3*2]
movu m5, [ref2q+%3*2]
movu m6, [ref3q+%3*2]
movu m7, [ref4q+%3*2]
mova m3, m0
mova m2, m0
psubusw m3, m4
psubusw m2, m5
psubusw m4, m0
psubusw m5, m0
por m4, m3
por m5, m2
pmaddwd m4, m1
pmaddwd m5, m1
mova m3, m0
mova m2, m0
psubusw m3, m6
psubusw m2, m7
psubusw m6, m0
psubusw m7, m0
por m6, m3
por m7, m2
pmaddwd m6, m1
pmaddwd m7, m1
%else
mova m3, m0
movu m2, [ref1q+%3*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m4, m2
movu m2, [ref2q+%3*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m5, m2
movu m2, [ref3q+%3*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m6, m2
movu m2, [ref4q+%3*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
pmaddwd m2, m1
paddd m7, m2
%endif
; 2nd 8 px
mova m0, [srcq +(%4)*2]
mova m3, m0
movu m2, [ref1q+(%5)*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m4, m2
movu m2, [ref2q+(%5)*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m5, m2
movu m2, [ref3q+(%5)*2]
psubusw m3, m2
psubusw m2, m0
por m2, m3
mova m3, m0
pmaddwd m2, m1
paddd m6, m2
movu m2, [ref4q+(%5)*2]
psubusw m3, m2
psubusw m2, m0
%if %6 == 1
lea srcq, [srcq +src_strideq*4]
lea ref1q, [ref1q+ref_strideq*4]
lea ref2q, [ref2q+ref_strideq*4]
lea ref3q, [ref3q+ref_strideq*4]
lea ref4q, [ref4q+ref_strideq*4]
%endif
por m2, m3
pmaddwd m2, m1
paddd m7, m2
%endmacro
; HIGH_PROCESS_16x2x4 first, off_{first,second}_{src,ref}, advance_at_end
%macro HIGH_PROCESS_16x2x4 5-6 0
HIGH_PROCESS_8x2x4 %1, %2, %3, (%2 + 8), (%3 + 8)
HIGH_PROCESS_8x2x4 0, %4, %5, (%4 + 8), (%5 + 8), %6
%endmacro
; HIGH_PROCESS_32x2x4 first, off_{first,second}_{src,ref}, advance_at_end
%macro HIGH_PROCESS_32x2x4 5-6 0
HIGH_PROCESS_16x2x4 %1, %2, %3, (%2 + 16), (%3 + 16)
HIGH_PROCESS_16x2x4 0, %4, %5, (%4 + 16), (%5 + 16), %6
%endmacro
; HIGH_PROCESS_64x2x4 first, off_{first,second}_{src,ref}, advance_at_end
%macro HIGH_PROCESS_64x2x4 5-6 0
HIGH_PROCESS_32x2x4 %1, %2, %3, (%2 + 32), (%3 + 32)
HIGH_PROCESS_32x2x4 0, %4, %5, (%4 + 32), (%5 + 32), %6
%endmacro
; void aom_highbd_sadNxNx4d_sse2(uint8_t *src, int src_stride,
; uint8_t *ref[4], int ref_stride,
; uint32_t res[4]);
; where NxN = 64x64, 32x32, 16x16, 16x8, 8x16 or 8x8
%macro HIGH_SADNXN4D 2
%if UNIX64
cglobal highbd_sad%1x%2x4d, 5, 8, 8, src, src_stride, ref1, ref_stride, \
res, ref2, ref3, ref4
%else
cglobal highbd_sad%1x%2x4d, 4, 7, 8, src, src_stride, ref1, ref_stride, \
ref2, ref3, ref4
%endif
; set m1
push srcq
mov srcd, 0x00010001
movd m1, srcd
pshufd m1, m1, 0x0
pop srcq
movsxdifnidn src_strideq, src_strided
movsxdifnidn ref_strideq, ref_strided
mov ref2q, [ref1q+gprsize*1]
mov ref3q, [ref1q+gprsize*2]
mov ref4q, [ref1q+gprsize*3]
mov ref1q, [ref1q+gprsize*0]
; convert byte pointers to short pointers
shl srcq, 1
shl ref2q, 1
shl ref3q, 1
shl ref4q, 1
shl ref1q, 1
HIGH_PROCESS_%1x2x4 1, 0, 0, src_strideq, ref_strideq, 1
%rep (%2-4)/2
HIGH_PROCESS_%1x2x4 0, 0, 0, src_strideq, ref_strideq, 1
%endrep
HIGH_PROCESS_%1x2x4 0, 0, 0, src_strideq, ref_strideq, 0
; N.B. HIGH_PROCESS outputs dwords (32 bits)
; so in high bit depth even the smallest width (4) needs 128bits i.e. XMM
movhlps m0, m4
movhlps m1, m5
movhlps m2, m6
movhlps m3, m7
paddd m4, m0
paddd m5, m1
paddd m6, m2
paddd m7, m3
punpckldq m4, m5
punpckldq m6, m7
movhlps m0, m4
movhlps m1, m6
paddd m4, m0
paddd m6, m1
punpcklqdq m4, m6
movifnidn r4, r4mp
movu [r4], m4
RET
%endmacro
INIT_XMM sse2
HIGH_SADNXN4D 64, 64
HIGH_SADNXN4D 64, 32
HIGH_SADNXN4D 32, 64
HIGH_SADNXN4D 32, 32
HIGH_SADNXN4D 32, 16
HIGH_SADNXN4D 16, 32
HIGH_SADNXN4D 16, 16
HIGH_SADNXN4D 16, 8
HIGH_SADNXN4D 8, 16
HIGH_SADNXN4D 8, 8
HIGH_SADNXN4D 8, 4
HIGH_SADNXN4D 4, 8
HIGH_SADNXN4D 4, 4
HIGH_SADNXN4D 4, 16
HIGH_SADNXN4D 16, 4
HIGH_SADNXN4D 8, 32
HIGH_SADNXN4D 32, 8
HIGH_SADNXN4D 16, 64
HIGH_SADNXN4D 64, 16
|
