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
|
;
; Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license and patent
; grant that can be found in the LICENSE file in the root of the source
; tree. All contributing project authors may be found in the AUTHORS
; file in the root of the source tree.
;
EXPORT |vp8_bilinear_predict8x4_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
; r0 unsigned char *src_ptr,
; r1 int src_pixels_per_line,
; r2 int xoffset,
; r3 int yoffset,
; r4 unsigned char *dst_ptr,
; stack(lr) int dst_pitch
|vp8_bilinear_predict8x4_neon| PROC
push {r4, lr}
ldr r12, _bifilter8x4_coeff_
ldr r4, [sp, #8] ;load parameters from stack
ldr lr, [sp, #12] ;load parameters from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq skip_firstpass_filter
;First pass: output_height lines x output_width columns (5x8)
add r2, r12, r2, lsl #3 ;calculate filter location
vld1.u8 {q1}, [r0], r1 ;load src data
vld1.u32 {d31}, [r2] ;load first_pass filter
vld1.u8 {q2}, [r0], r1
vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
vld1.u8 {q3}, [r0], r1
vdup.8 d1, d31[4]
vld1.u8 {q4}, [r0], r1
vmull.u8 q6, d2, d0 ;(src_ptr[0] * vp8_filter[0])
vld1.u8 {q5}, [r0], r1
vmull.u8 q7, d4, d0
vmull.u8 q8, d6, d0
vmull.u8 q9, d8, d0
vmull.u8 q10, d10, d0
vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
vext.8 d5, d4, d5, #1
vext.8 d7, d6, d7, #1
vext.8 d9, d8, d9, #1
vext.8 d11, d10, d11, #1
vmlal.u8 q6, d3, d1 ;(src_ptr[1] * vp8_filter[1])
vmlal.u8 q7, d5, d1
vmlal.u8 q8, d7, d1
vmlal.u8 q9, d9, d1
vmlal.u8 q10, d11, d1
vqrshrn.u16 d22, q6, #7 ;shift/round/saturate to u8
vqrshrn.u16 d23, q7, #7
vqrshrn.u16 d24, q8, #7
vqrshrn.u16 d25, q9, #7
vqrshrn.u16 d26, q10, #7
;Second pass: 4x8
secondpass_filter
cmp r3, #0 ;skip second_pass filter if yoffset=0
beq skip_secondpass_filter
add r3, r12, r3, lsl #3
add r0, r4, lr
vld1.u32 {d31}, [r3] ;load second_pass filter
add r1, r0, lr
vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
vdup.8 d1, d31[4]
vmull.u8 q1, d22, d0 ;(src_ptr[0] * vp8_filter[0])
vmull.u8 q2, d23, d0
vmull.u8 q3, d24, d0
vmull.u8 q4, d25, d0
vmlal.u8 q1, d23, d1 ;(src_ptr[pixel_step] * vp8_filter[1])
vmlal.u8 q2, d24, d1
vmlal.u8 q3, d25, d1
vmlal.u8 q4, d26, d1
add r2, r1, lr
vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
vqrshrn.u16 d3, q2, #7
vqrshrn.u16 d4, q3, #7
vqrshrn.u16 d5, q4, #7
vst1.u8 {d2}, [r4] ;store result
vst1.u8 {d3}, [r0]
vst1.u8 {d4}, [r1]
vst1.u8 {d5}, [r2]
pop {r4, pc}
;--------------------
skip_firstpass_filter
vld1.u8 {d22}, [r0], r1 ;load src data
vld1.u8 {d23}, [r0], r1
vld1.u8 {d24}, [r0], r1
vld1.u8 {d25}, [r0], r1
vld1.u8 {d26}, [r0], r1
b secondpass_filter
;---------------------
skip_secondpass_filter
vst1.u8 {d22}, [r4], lr ;store result
vst1.u8 {d23}, [r4], lr
vst1.u8 {d24}, [r4], lr
vst1.u8 {d25}, [r4], lr
pop {r4, pc}
ENDP
;-----------------
AREA bifilters8x4_dat, DATA, READWRITE ;read/write by default
;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
;One word each is reserved. Label filter_coeff can be used to access the data.
;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
_bifilter8x4_coeff_
DCD bifilter8x4_coeff
bifilter8x4_coeff
DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
END
|
