2 ; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
4 ; Use of this source code is governed by a BSD-style license
5 ; that can be found in the LICENSE file in the root of the source
6 ; tree. An additional intellectual property rights grant can be found
7 ; in the file PATENTS. All contributing project authors may
8 ; be found in the AUTHORS file in the root of the source tree.
12 EXPORT |vp8_bilinear_predict8x8_neon|
17 AREA ||.text||, CODE, READONLY, ALIGN=2
18 ; r0 unsigned char *src_ptr,
19 ; r1 int src_pixels_per_line,
22 ; r4 unsigned char *dst_ptr,
23 ; stack(lr) int dst_pitch
25 |vp8_bilinear_predict8x8_neon| PROC
28 adr r12, bifilter8_coeff
29 ldr r4, [sp, #8] ;load parameters from stack
30 ldr lr, [sp, #12] ;load parameters from stack
32 cmp r2, #0 ;skip first_pass filter if xoffset=0
33 beq skip_firstpass_filter
35 ;First pass: output_height lines x output_width columns (9x8)
36 add r2, r12, r2, lsl #3 ;calculate filter location
38 vld1.u8 {q1}, [r0], r1 ;load src data
39 vld1.u32 {d31}, [r2] ;load first_pass filter
40 vld1.u8 {q2}, [r0], r1
41 vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
42 vld1.u8 {q3}, [r0], r1
44 vld1.u8 {q4}, [r0], r1
46 vmull.u8 q6, d2, d0 ;(src_ptr[0] * vp8_filter[0])
51 vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
56 vmlal.u8 q6, d3, d1 ;(src_ptr[1] * vp8_filter[1])
61 vld1.u8 {q1}, [r0], r1 ;load src data
62 vqrshrn.u16 d22, q6, #7 ;shift/round/saturate to u8
63 vld1.u8 {q2}, [r0], r1
64 vqrshrn.u16 d23, q7, #7
65 vld1.u8 {q3}, [r0], r1
66 vqrshrn.u16 d24, q8, #7
67 vld1.u8 {q4}, [r0], r1
68 vqrshrn.u16 d25, q9, #7
70 ;first_pass filtering on the rest 5-line data
71 vld1.u8 {q5}, [r0], r1
73 vmull.u8 q6, d2, d0 ;(src_ptr[0] * vp8_filter[0])
79 vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
83 vext.8 d11, d10, d11, #1
85 vmlal.u8 q6, d3, d1 ;(src_ptr[1] * vp8_filter[1])
91 vqrshrn.u16 d26, q6, #7 ;shift/round/saturate to u8
92 vqrshrn.u16 d27, q7, #7
93 vqrshrn.u16 d28, q8, #7
94 vqrshrn.u16 d29, q9, #7
95 vqrshrn.u16 d30, q10, #7
99 cmp r3, #0 ;skip second_pass filter if yoffset=0
100 beq skip_secondpass_filter
102 add r3, r12, r3, lsl #3
105 vld1.u32 {d31}, [r3] ;load second_pass filter
108 vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
111 vmull.u8 q1, d22, d0 ;(src_ptr[0] * vp8_filter[0])
120 vmlal.u8 q1, d23, d1 ;(src_ptr[pixel_step] * vp8_filter[1])
129 vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
130 vqrshrn.u16 d3, q2, #7
131 vqrshrn.u16 d4, q3, #7
132 vqrshrn.u16 d5, q4, #7
133 vqrshrn.u16 d6, q5, #7
134 vqrshrn.u16 d7, q6, #7
135 vqrshrn.u16 d8, q7, #7
136 vqrshrn.u16 d9, q8, #7
138 vst1.u8 {d2}, [r4] ;store result
140 vst1.u8 {d4}, [r1], lr
141 vst1.u8 {d5}, [r1], lr
142 vst1.u8 {d6}, [r1], lr
143 vst1.u8 {d7}, [r1], lr
144 vst1.u8 {d8}, [r1], lr
145 vst1.u8 {d9}, [r1], lr
149 ;--------------------
150 skip_firstpass_filter
151 vld1.u8 {d22}, [r0], r1 ;load src data
152 vld1.u8 {d23}, [r0], r1
153 vld1.u8 {d24}, [r0], r1
154 vld1.u8 {d25}, [r0], r1
155 vld1.u8 {d26}, [r0], r1
156 vld1.u8 {d27}, [r0], r1
157 vld1.u8 {d28}, [r0], r1
158 vld1.u8 {d29}, [r0], r1
159 vld1.u8 {d30}, [r0], r1
163 ;---------------------
164 skip_secondpass_filter
165 vst1.u8 {d22}, [r4], lr ;store result
166 vst1.u8 {d23}, [r4], lr
167 vst1.u8 {d24}, [r4], lr
168 vst1.u8 {d25}, [r4], lr
169 vst1.u8 {d26}, [r4], lr
170 vst1.u8 {d27}, [r4], lr
171 vst1.u8 {d28}, [r4], lr
172 vst1.u8 {d29}, [r4], lr
181 DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112