media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm@ac0c01689b40 (annotated)
media/libvpx/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
Thu, 15 Jan 2015 15:59:08 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Thu, 15 Jan 2015 15:59:08 +0100
- branch
- TOR_BUG_9701
- changeset 10
- ac0c01689b40
- permissions
- -rw-r--r--
Implement a real Private Browsing Mode condition by changing the API/ABI;
This solves Tor bug #9701, complying with disk avoidance documented in
https://www.torproject.org/projects/torbrowser/design/#disk-avoidance.
| michael@0 | 1 | ; |
| michael@0 | 2 | ; Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
| michael@0 | 3 | ; |
| michael@0 | 4 | ; Use of this source code is governed by a BSD-style license |
| michael@0 | 5 | ; that can be found in the LICENSE file in the root of the source |
| michael@0 | 6 | ; tree. An additional intellectual property rights grant can be found |
| michael@0 | 7 | ; in the file PATENTS. All contributing project authors may |
| michael@0 | 8 | ; be found in the AUTHORS file in the root of the source tree. |
| michael@0 | 9 | ; |
| michael@0 | 10 | |
| michael@0 | 11 | |
| michael@0 | 12 | EXPORT |vp8_bilinear_predict4x4_neon| |
| michael@0 | 13 | ARM |
| michael@0 | 14 | REQUIRE8 |
| michael@0 | 15 | PRESERVE8 |
| michael@0 | 16 | |
| michael@0 | 17 | AREA ||.text||, CODE, READONLY, ALIGN=2 |
| michael@0 | 18 | ; r0 unsigned char *src_ptr, |
| michael@0 | 19 | ; r1 int src_pixels_per_line, |
| michael@0 | 20 | ; r2 int xoffset, |
| michael@0 | 21 | ; r3 int yoffset, |
| michael@0 | 22 | ; r4 unsigned char *dst_ptr, |
| michael@0 | 23 | ; stack(lr) int dst_pitch |
| michael@0 | 24 | |
| michael@0 | 25 | |vp8_bilinear_predict4x4_neon| PROC |
| michael@0 | 26 | push {r4, lr} |
| michael@0 | 27 | |
| michael@0 | 28 | adr r12, bifilter4_coeff |
| michael@0 | 29 | ldr r4, [sp, #8] ;load parameters from stack |
| michael@0 | 30 | ldr lr, [sp, #12] ;load parameters from stack |
| michael@0 | 31 | |
| michael@0 | 32 | cmp r2, #0 ;skip first_pass filter if xoffset=0 |
| michael@0 | 33 | beq skip_firstpass_filter |
| michael@0 | 34 | |
| michael@0 | 35 | ;First pass: output_height lines x output_width columns (5x4) |
| michael@0 | 36 | vld1.u8 {d2}, [r0], r1 ;load src data |
| michael@0 | 37 | add r2, r12, r2, lsl #3 ;calculate Hfilter location (2coeffsx4bytes=8bytes) |
| michael@0 | 38 | |
| michael@0 | 39 | vld1.u8 {d3}, [r0], r1 |
| michael@0 | 40 | vld1.u32 {d31}, [r2] ;first_pass filter |
| michael@0 | 41 | |
| michael@0 | 42 | vld1.u8 {d4}, [r0], r1 |
| michael@0 | 43 | vdup.8 d0, d31[0] ;first_pass filter (d0-d1) |
| michael@0 | 44 | vld1.u8 {d5}, [r0], r1 |
| michael@0 | 45 | vdup.8 d1, d31[4] |
| michael@0 | 46 | vld1.u8 {d6}, [r0], r1 |
| michael@0 | 47 | |
| michael@0 | 48 | vshr.u64 q4, q1, #8 ;construct src_ptr[1] |
| michael@0 | 49 | vshr.u64 q5, q2, #8 |
| michael@0 | 50 | vshr.u64 d12, d6, #8 |
| michael@0 | 51 | |
| michael@0 | 52 | vzip.32 d2, d3 ;put 2-line data in 1 register (src_ptr[0]) |
| michael@0 | 53 | vzip.32 d4, d5 |
| michael@0 | 54 | vzip.32 d8, d9 ;put 2-line data in 1 register (src_ptr[1]) |
| michael@0 | 55 | vzip.32 d10, d11 |
| michael@0 | 56 | |
| michael@0 | 57 | vmull.u8 q7, d2, d0 ;(src_ptr[0] * vp8_filter[0]) |
| michael@0 | 58 | vmull.u8 q8, d4, d0 |
| michael@0 | 59 | vmull.u8 q9, d6, d0 |
| michael@0 | 60 | |
| michael@0 | 61 | vmlal.u8 q7, d8, d1 ;(src_ptr[1] * vp8_filter[1]) |
| michael@0 | 62 | vmlal.u8 q8, d10, d1 |
| michael@0 | 63 | vmlal.u8 q9, d12, d1 |
| michael@0 | 64 | |
| michael@0 | 65 | vqrshrn.u16 d28, q7, #7 ;shift/round/saturate to u8 |
| michael@0 | 66 | vqrshrn.u16 d29, q8, #7 |
| michael@0 | 67 | vqrshrn.u16 d30, q9, #7 |
| michael@0 | 68 | |
| michael@0 | 69 | ;Second pass: 4x4 |
| michael@0 | 70 | secondpass_filter |
| michael@0 | 71 | cmp r3, #0 ;skip second_pass filter if yoffset=0 |
| michael@0 | 72 | beq skip_secondpass_filter |
| michael@0 | 73 | |
| michael@0 | 74 | add r3, r12, r3, lsl #3 ;calculate Vfilter location |
| michael@0 | 75 | vld1.u32 {d31}, [r3] ;load second_pass filter |
| michael@0 | 76 | |
| michael@0 | 77 | vdup.8 d0, d31[0] ;second_pass filter parameters (d0-d5) |
| michael@0 | 78 | vdup.8 d1, d31[4] |
| michael@0 | 79 | |
| michael@0 | 80 | vmull.u8 q1, d28, d0 |
| michael@0 | 81 | vmull.u8 q2, d29, d0 |
| michael@0 | 82 | |
| michael@0 | 83 | vext.8 d26, d28, d29, #4 ;construct src_ptr[pixel_step] |
| michael@0 | 84 | vext.8 d27, d29, d30, #4 |
| michael@0 | 85 | |
| michael@0 | 86 | vmlal.u8 q1, d26, d1 |
| michael@0 | 87 | vmlal.u8 q2, d27, d1 |
| michael@0 | 88 | |
| michael@0 | 89 | add r0, r4, lr |
| michael@0 | 90 | add r1, r0, lr |
| michael@0 | 91 | add r2, r1, lr |
| michael@0 | 92 | |
| michael@0 | 93 | vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8 |
| michael@0 | 94 | vqrshrn.u16 d3, q2, #7 |
| michael@0 | 95 | |
| michael@0 | 96 | vst1.32 {d2[0]}, [r4] ;store result |
| michael@0 | 97 | vst1.32 {d2[1]}, [r0] |
| michael@0 | 98 | vst1.32 {d3[0]}, [r1] |
| michael@0 | 99 | vst1.32 {d3[1]}, [r2] |
| michael@0 | 100 | |
| michael@0 | 101 | pop {r4, pc} |
| michael@0 | 102 | |
| michael@0 | 103 | ;-------------------- |
| michael@0 | 104 | skip_firstpass_filter |
| michael@0 | 105 | |
| michael@0 | 106 | vld1.32 {d28[0]}, [r0], r1 ;load src data |
| michael@0 | 107 | vld1.32 {d28[1]}, [r0], r1 |
| michael@0 | 108 | vld1.32 {d29[0]}, [r0], r1 |
| michael@0 | 109 | vld1.32 {d29[1]}, [r0], r1 |
| michael@0 | 110 | vld1.32 {d30[0]}, [r0], r1 |
| michael@0 | 111 | |
| michael@0 | 112 | b secondpass_filter |
| michael@0 | 113 | |
| michael@0 | 114 | ;--------------------- |
| michael@0 | 115 | skip_secondpass_filter |
| michael@0 | 116 | vst1.32 {d28[0]}, [r4], lr ;store result |
| michael@0 | 117 | vst1.32 {d28[1]}, [r4], lr |
| michael@0 | 118 | vst1.32 {d29[0]}, [r4], lr |
| michael@0 | 119 | vst1.32 {d29[1]}, [r4], lr |
| michael@0 | 120 | |
| michael@0 | 121 | pop {r4, pc} |
| michael@0 | 122 | |
| michael@0 | 123 | ENDP |
| michael@0 | 124 | |
| michael@0 | 125 | ;----------------- |
| michael@0 | 126 | |
| michael@0 | 127 | bifilter4_coeff |
| michael@0 | 128 | DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112 |
| michael@0 | 129 | |
| michael@0 | 130 | END |
