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
|
;
; 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.
;
EXPORT |vp8_loop_filter_simple_horizontal_edge_armv6|
EXPORT |vp8_loop_filter_simple_vertical_edge_armv6|
AREA |.text|, CODE, READONLY ; name this block of code
MACRO
TRANSPOSE_MATRIX $a0, $a1, $a2, $a3, $b0, $b1, $b2, $b3
; input: $a0, $a1, $a2, $a3; output: $b0, $b1, $b2, $b3
; a0: 03 02 01 00
; a1: 13 12 11 10
; a2: 23 22 21 20
; a3: 33 32 31 30
; b3 b2 b1 b0
uxtb16 $b1, $a1 ; xx 12 xx 10
uxtb16 $b0, $a0 ; xx 02 xx 00
uxtb16 $b3, $a3 ; xx 32 xx 30
uxtb16 $b2, $a2 ; xx 22 xx 20
orr $b1, $b0, $b1, lsl #8 ; 12 02 10 00
orr $b3, $b2, $b3, lsl #8 ; 32 22 30 20
uxtb16 $a1, $a1, ror #8 ; xx 13 xx 11
uxtb16 $a3, $a3, ror #8 ; xx 33 xx 31
uxtb16 $a0, $a0, ror #8 ; xx 03 xx 01
uxtb16 $a2, $a2, ror #8 ; xx 23 xx 21
orr $a0, $a0, $a1, lsl #8 ; 13 03 11 01
orr $a2, $a2, $a3, lsl #8 ; 33 23 31 21
pkhtb $b2, $b3, $b1, asr #16 ; 32 22 12 02 -- p1
pkhbt $b0, $b1, $b3, lsl #16 ; 30 20 10 00 -- p3
pkhtb $b3, $a2, $a0, asr #16 ; 33 23 13 03 -- p0
pkhbt $b1, $a0, $a2, lsl #16 ; 31 21 11 01 -- p2
MEND
src RN r0
pstep RN r1
;r0 unsigned char *src_ptr,
;r1 int src_pixel_step,
;r2 const char *blimit
;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
|vp8_loop_filter_simple_horizontal_edge_armv6| PROC
;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
stmdb sp!, {r4 - r11, lr}
ldrb r12, [r2] ; blimit
ldr r3, [src, -pstep, lsl #1] ; p1
ldr r4, [src, -pstep] ; p0
ldr r5, [src] ; q0
ldr r6, [src, pstep] ; q1
orr r12, r12, r12, lsl #8 ; blimit
ldr r2, c0x80808080
orr r12, r12, r12, lsl #16 ; blimit
mov r9, #4 ; double the count. we're doing 4 at a time
mov lr, #0 ; need 0 in a couple places
|simple_hnext8|
; vp8_simple_filter_mask()
uqsub8 r7, r3, r6 ; p1 - q1
uqsub8 r8, r6, r3 ; q1 - p1
uqsub8 r10, r4, r5 ; p0 - q0
uqsub8 r11, r5, r4 ; q0 - p0
orr r8, r8, r7 ; abs(p1 - q1)
orr r10, r10, r11 ; abs(p0 - q0)
uqadd8 r10, r10, r10 ; abs(p0 - q0) * 2
uhadd8 r8, r8, lr ; abs(p1 - q2) >> 1
uqadd8 r10, r10, r8 ; abs(p0 - q0)*2 + abs(p1 - q1)/2
mvn r8, #0
usub8 r10, r12, r10 ; compare to flimit. usub8 sets GE flags
sel r10, r8, lr ; filter mask: F or 0
cmp r10, #0
beq simple_hskip_filter ; skip filtering if all masks are 0x00
;vp8_simple_filter()
eor r3, r3, r2 ; p1 offset to convert to a signed value
eor r6, r6, r2 ; q1 offset to convert to a signed value
eor r4, r4, r2 ; p0 offset to convert to a signed value
eor r5, r5, r2 ; q0 offset to convert to a signed value
qsub8 r3, r3, r6 ; vp8_filter = p1 - q1
qsub8 r6, r5, r4 ; q0 - p0
qadd8 r3, r3, r6 ; += q0 - p0
ldr r7, c0x04040404
qadd8 r3, r3, r6 ; += q0 - p0
ldr r8, c0x03030303
qadd8 r3, r3, r6 ; vp8_filter = p1-q1 + 3*(q0-p0))
;STALL
and r3, r3, r10 ; vp8_filter &= mask
qadd8 r7 , r3 , r7 ; Filter1 = vp8_filter + 4
qadd8 r8 , r3 , r8 ; Filter2 = vp8_filter + 3
shadd8 r7 , r7 , lr
shadd8 r8 , r8 , lr
shadd8 r7 , r7 , lr
shadd8 r8 , r8 , lr
shadd8 r7 , r7 , lr ; Filter1 >>= 3
shadd8 r8 , r8 , lr ; Filter2 >>= 3
qsub8 r5 ,r5, r7 ; u = q0 - Filter1
qadd8 r4, r4, r8 ; u = p0 + Filter2
eor r5, r5, r2 ; *oq0 = u^0x80
str r5, [src] ; store oq0 result
eor r4, r4, r2 ; *op0 = u^0x80
str r4, [src, -pstep] ; store op0 result
|simple_hskip_filter|
subs r9, r9, #1
addne src, src, #4 ; next row
ldrne r3, [src, -pstep, lsl #1] ; p1
ldrne r4, [src, -pstep] ; p0
ldrne r5, [src] ; q0
ldrne r6, [src, pstep] ; q1
bne simple_hnext8
ldmia sp!, {r4 - r11, pc}
ENDP ; |vp8_loop_filter_simple_horizontal_edge_armv6|
;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
|vp8_loop_filter_simple_vertical_edge_armv6| PROC
;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
stmdb sp!, {r4 - r11, lr}
ldrb r12, [r2] ; r12: blimit
ldr r2, c0x80808080
orr r12, r12, r12, lsl #8
; load soure data to r7, r8, r9, r10
ldrh r3, [src, #-2]
pld [src, #23] ; preload for next block
ldrh r4, [src], pstep
orr r12, r12, r12, lsl #16
ldrh r5, [src, #-2]
pld [src, #23]
ldrh r6, [src], pstep
pkhbt r7, r3, r4, lsl #16
ldrh r3, [src, #-2]
pld [src, #23]
ldrh r4, [src], pstep
pkhbt r8, r5, r6, lsl #16
ldrh r5, [src, #-2]
pld [src, #23]
ldrh r6, [src], pstep
mov r11, #4 ; double the count. we're doing 4 at a time
|simple_vnext8|
; vp8_simple_filter_mask() function
pkhbt r9, r3, r4, lsl #16
pkhbt r10, r5, r6, lsl #16
;transpose r7, r8, r9, r10 to r3, r4, r5, r6
TRANSPOSE_MATRIX r7, r8, r9, r10, r3, r4, r5, r6
uqsub8 r7, r3, r6 ; p1 - q1
uqsub8 r8, r6, r3 ; q1 - p1
uqsub8 r9, r4, r5 ; p0 - q0
uqsub8 r10, r5, r4 ; q0 - p0
orr r7, r7, r8 ; abs(p1 - q1)
orr r9, r9, r10 ; abs(p0 - q0)
mov r8, #0
uqadd8 r9, r9, r9 ; abs(p0 - q0) * 2
uhadd8 r7, r7, r8 ; abs(p1 - q1) / 2
uqadd8 r7, r7, r9 ; abs(p0 - q0)*2 + abs(p1 - q1)/2
mvn r10, #0 ; r10 == -1
usub8 r7, r12, r7 ; compare to flimit
sel lr, r10, r8 ; filter mask
cmp lr, #0
beq simple_vskip_filter ; skip filtering
;vp8_simple_filter() function
eor r3, r3, r2 ; p1 offset to convert to a signed value
eor r6, r6, r2 ; q1 offset to convert to a signed value
eor r4, r4, r2 ; p0 offset to convert to a signed value
eor r5, r5, r2 ; q0 offset to convert to a signed value
qsub8 r3, r3, r6 ; vp8_filter = p1 - q1
qsub8 r6, r5, r4 ; q0 - p0
qadd8 r3, r3, r6 ; vp8_filter += q0 - p0
ldr r9, c0x03030303 ; r9 = 3
qadd8 r3, r3, r6 ; vp8_filter += q0 - p0
ldr r7, c0x04040404
qadd8 r3, r3, r6 ; vp8_filter = p1-q1 + 3*(q0-p0))
;STALL
and r3, r3, lr ; vp8_filter &= mask
qadd8 r9 , r3 , r9 ; Filter2 = vp8_filter + 3
qadd8 r3 , r3 , r7 ; Filter1 = vp8_filter + 4
shadd8 r9 , r9 , r8
shadd8 r3 , r3 , r8
shadd8 r9 , r9 , r8
shadd8 r3 , r3 , r8
shadd8 r9 , r9 , r8 ; Filter2 >>= 3
shadd8 r3 , r3 , r8 ; Filter1 >>= 3
;calculate output
sub src, src, pstep, lsl #2
qadd8 r4, r4, r9 ; u = p0 + Filter2
qsub8 r5, r5, r3 ; u = q0 - Filter1
eor r4, r4, r2 ; *op0 = u^0x80
eor r5, r5, r2 ; *oq0 = u^0x80
strb r4, [src, #-1] ; store the result
mov r4, r4, lsr #8
strb r5, [src], pstep
mov r5, r5, lsr #8
strb r4, [src, #-1]
mov r4, r4, lsr #8
strb r5, [src], pstep
mov r5, r5, lsr #8
strb r4, [src, #-1]
mov r4, r4, lsr #8
strb r5, [src], pstep
mov r5, r5, lsr #8
strb r4, [src, #-1]
strb r5, [src], pstep
|simple_vskip_filter|
subs r11, r11, #1
; load soure data to r7, r8, r9, r10
ldrneh r3, [src, #-2]
pld [src, #23] ; preload for next block
ldrneh r4, [src], pstep
ldrneh r5, [src, #-2]
pld [src, #23]
ldrneh r6, [src], pstep
pkhbt r7, r3, r4, lsl #16
ldrneh r3, [src, #-2]
pld [src, #23]
ldrneh r4, [src], pstep
pkhbt r8, r5, r6, lsl #16
ldrneh r5, [src, #-2]
pld [src, #23]
ldrneh r6, [src], pstep
bne simple_vnext8
ldmia sp!, {r4 - r11, pc}
ENDP ; |vp8_loop_filter_simple_vertical_edge_armv6|
; Constant Pool
c0x80808080 DCD 0x80808080
c0x03030303 DCD 0x03030303
c0x04040404 DCD 0x04040404
END
|
