media/libvpx/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm@b8a032363ba2 (annotated)
media/libvpx/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm
Thu, 22 Jan 2015 13:21:57 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Thu, 22 Jan 2015 13:21:57 +0100
- branch
- TOR_BUG_9701
- changeset 15
- b8a032363ba2
- permissions
- -rw-r--r--
Incorporate requested changes from Mozilla in review:
https://bugzilla.mozilla.org/show_bug.cgi?id=1123480#c6
| 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_mse16x16_neon| |
| michael@0 | 13 | EXPORT |vp8_get4x4sse_cs_neon| |
| michael@0 | 14 | |
| michael@0 | 15 | ARM |
| michael@0 | 16 | REQUIRE8 |
| michael@0 | 17 | PRESERVE8 |
| michael@0 | 18 | |
| michael@0 | 19 | AREA ||.text||, CODE, READONLY, ALIGN=2 |
| michael@0 | 20 | ;============================ |
| michael@0 | 21 | ; r0 unsigned char *src_ptr |
| michael@0 | 22 | ; r1 int source_stride |
| michael@0 | 23 | ; r2 unsigned char *ref_ptr |
| michael@0 | 24 | ; r3 int recon_stride |
| michael@0 | 25 | ; stack unsigned int *sse |
| michael@0 | 26 | ;note: in this function, sum is never used. So, we can remove this part of calculation |
| michael@0 | 27 | ;from vp8_variance(). |
| michael@0 | 28 | |
| michael@0 | 29 | |vp8_mse16x16_neon| PROC |
| michael@0 | 30 | vmov.i8 q7, #0 ;q7, q8, q9, q10 - sse |
| michael@0 | 31 | vmov.i8 q8, #0 |
| michael@0 | 32 | vmov.i8 q9, #0 |
| michael@0 | 33 | vmov.i8 q10, #0 |
| michael@0 | 34 | |
| michael@0 | 35 | mov r12, #8 |
| michael@0 | 36 | |
| michael@0 | 37 | mse16x16_neon_loop |
| michael@0 | 38 | vld1.8 {q0}, [r0], r1 ;Load up source and reference |
| michael@0 | 39 | vld1.8 {q2}, [r2], r3 |
| michael@0 | 40 | vld1.8 {q1}, [r0], r1 |
| michael@0 | 41 | vld1.8 {q3}, [r2], r3 |
| michael@0 | 42 | |
| michael@0 | 43 | vsubl.u8 q11, d0, d4 |
| michael@0 | 44 | vsubl.u8 q12, d1, d5 |
| michael@0 | 45 | vsubl.u8 q13, d2, d6 |
| michael@0 | 46 | vsubl.u8 q14, d3, d7 |
| michael@0 | 47 | |
| michael@0 | 48 | vmlal.s16 q7, d22, d22 |
| michael@0 | 49 | vmlal.s16 q8, d23, d23 |
| michael@0 | 50 | |
| michael@0 | 51 | subs r12, r12, #1 |
| michael@0 | 52 | |
| michael@0 | 53 | vmlal.s16 q9, d24, d24 |
| michael@0 | 54 | vmlal.s16 q10, d25, d25 |
| michael@0 | 55 | vmlal.s16 q7, d26, d26 |
| michael@0 | 56 | vmlal.s16 q8, d27, d27 |
| michael@0 | 57 | vmlal.s16 q9, d28, d28 |
| michael@0 | 58 | vmlal.s16 q10, d29, d29 |
| michael@0 | 59 | |
| michael@0 | 60 | bne mse16x16_neon_loop |
| michael@0 | 61 | |
| michael@0 | 62 | vadd.u32 q7, q7, q8 |
| michael@0 | 63 | vadd.u32 q9, q9, q10 |
| michael@0 | 64 | |
| michael@0 | 65 | ldr r12, [sp] ;load *sse from stack |
| michael@0 | 66 | |
| michael@0 | 67 | vadd.u32 q10, q7, q9 |
| michael@0 | 68 | vpaddl.u32 q1, q10 |
| michael@0 | 69 | vadd.u64 d0, d2, d3 |
| michael@0 | 70 | |
| michael@0 | 71 | vst1.32 {d0[0]}, [r12] |
| michael@0 | 72 | vmov.32 r0, d0[0] |
| michael@0 | 73 | |
| michael@0 | 74 | bx lr |
| michael@0 | 75 | |
| michael@0 | 76 | ENDP |
| michael@0 | 77 | |
| michael@0 | 78 | |
| michael@0 | 79 | ;============================= |
| michael@0 | 80 | ; r0 unsigned char *src_ptr, |
| michael@0 | 81 | ; r1 int source_stride, |
| michael@0 | 82 | ; r2 unsigned char *ref_ptr, |
| michael@0 | 83 | ; r3 int recon_stride |
| michael@0 | 84 | |vp8_get4x4sse_cs_neon| PROC |
| michael@0 | 85 | vld1.8 {d0}, [r0], r1 ;Load up source and reference |
| michael@0 | 86 | vld1.8 {d4}, [r2], r3 |
| michael@0 | 87 | vld1.8 {d1}, [r0], r1 |
| michael@0 | 88 | vld1.8 {d5}, [r2], r3 |
| michael@0 | 89 | vld1.8 {d2}, [r0], r1 |
| michael@0 | 90 | vld1.8 {d6}, [r2], r3 |
| michael@0 | 91 | vld1.8 {d3}, [r0], r1 |
| michael@0 | 92 | vld1.8 {d7}, [r2], r3 |
| michael@0 | 93 | |
| michael@0 | 94 | vsubl.u8 q11, d0, d4 |
| michael@0 | 95 | vsubl.u8 q12, d1, d5 |
| michael@0 | 96 | vsubl.u8 q13, d2, d6 |
| michael@0 | 97 | vsubl.u8 q14, d3, d7 |
| michael@0 | 98 | |
| michael@0 | 99 | vmull.s16 q7, d22, d22 |
| michael@0 | 100 | vmull.s16 q8, d24, d24 |
| michael@0 | 101 | vmull.s16 q9, d26, d26 |
| michael@0 | 102 | vmull.s16 q10, d28, d28 |
| michael@0 | 103 | |
| michael@0 | 104 | vadd.u32 q7, q7, q8 |
| michael@0 | 105 | vadd.u32 q9, q9, q10 |
| michael@0 | 106 | vadd.u32 q9, q7, q9 |
| michael@0 | 107 | |
| michael@0 | 108 | vpaddl.u32 q1, q9 |
| michael@0 | 109 | vadd.u64 d0, d2, d3 |
| michael@0 | 110 | |
| michael@0 | 111 | vmov.32 r0, d0[0] |
| michael@0 | 112 | bx lr |
| michael@0 | 113 | |
| michael@0 | 114 | ENDP |
| michael@0 | 115 | |
| michael@0 | 116 | END |
