|
yading@10
|
1 /*
|
|
yading@10
|
2 * ARM NEON optimised integer operations
|
|
yading@10
|
3 * Copyright (c) 2009 Kostya Shishkov
|
|
yading@10
|
4 *
|
|
yading@10
|
5 * This file is part of FFmpeg.
|
|
yading@10
|
6 *
|
|
yading@10
|
7 * FFmpeg is free software; you can redistribute it and/or
|
|
yading@10
|
8 * modify it under the terms of the GNU Lesser General Public
|
|
yading@10
|
9 * License as published by the Free Software Foundation; either
|
|
yading@10
|
10 * version 2.1 of the License, or (at your option) any later version.
|
|
yading@10
|
11 *
|
|
yading@10
|
12 * FFmpeg is distributed in the hope that it will be useful,
|
|
yading@10
|
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
yading@10
|
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
yading@10
|
15 * Lesser General Public License for more details.
|
|
yading@10
|
16 *
|
|
yading@10
|
17 * You should have received a copy of the GNU Lesser General Public
|
|
yading@10
|
18 * License along with FFmpeg; if not, write to the Free Software
|
|
yading@10
|
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
yading@10
|
20 */
|
|
yading@10
|
21
|
|
yading@10
|
22 #include "libavutil/arm/asm.S"
|
|
yading@10
|
23
|
|
yading@10
|
24 .fpu neon
|
|
yading@10
|
25
|
|
yading@10
|
26 function ff_scalarproduct_int16_neon, export=1
|
|
yading@10
|
27 vmov.i16 q0, #0
|
|
yading@10
|
28 vmov.i16 q1, #0
|
|
yading@10
|
29 vmov.i16 q2, #0
|
|
yading@10
|
30 vmov.i16 q3, #0
|
|
yading@10
|
31 1: vld1.16 {d16-d17}, [r0]!
|
|
yading@10
|
32 vld1.16 {d20-d21}, [r1,:128]!
|
|
yading@10
|
33 vmlal.s16 q0, d16, d20
|
|
yading@10
|
34 vld1.16 {d18-d19}, [r0]!
|
|
yading@10
|
35 vmlal.s16 q1, d17, d21
|
|
yading@10
|
36 vld1.16 {d22-d23}, [r1,:128]!
|
|
yading@10
|
37 vmlal.s16 q2, d18, d22
|
|
yading@10
|
38 vmlal.s16 q3, d19, d23
|
|
yading@10
|
39 subs r2, r2, #16
|
|
yading@10
|
40 bne 1b
|
|
yading@10
|
41
|
|
yading@10
|
42 vpadd.s32 d16, d0, d1
|
|
yading@10
|
43 vpadd.s32 d17, d2, d3
|
|
yading@10
|
44 vpadd.s32 d10, d4, d5
|
|
yading@10
|
45 vpadd.s32 d11, d6, d7
|
|
yading@10
|
46 vpadd.s32 d0, d16, d17
|
|
yading@10
|
47 vpadd.s32 d1, d10, d11
|
|
yading@10
|
48 vpadd.s32 d2, d0, d1
|
|
yading@10
|
49 vpaddl.s32 d3, d2
|
|
yading@10
|
50 vmov.32 r0, d3[0]
|
|
yading@10
|
51 bx lr
|
|
yading@10
|
52 endfunc
|
|
yading@10
|
53
|
|
yading@10
|
54 @ scalarproduct_and_madd_int16(/*aligned*/v0,v1,v2,order,mul)
|
|
yading@10
|
55 function ff_scalarproduct_and_madd_int16_neon, export=1
|
|
yading@10
|
56 vld1.16 {d28[],d29[]}, [sp]
|
|
yading@10
|
57 vmov.i16 q0, #0
|
|
yading@10
|
58 vmov.i16 q1, #0
|
|
yading@10
|
59 vmov.i16 q2, #0
|
|
yading@10
|
60 vmov.i16 q3, #0
|
|
yading@10
|
61 mov r12, r0
|
|
yading@10
|
62
|
|
yading@10
|
63 1: vld1.16 {d16-d17}, [r0,:128]!
|
|
yading@10
|
64 vld1.16 {d18-d19}, [r1]!
|
|
yading@10
|
65 vld1.16 {d20-d21}, [r2]!
|
|
yading@10
|
66 vld1.16 {d22-d23}, [r0,:128]!
|
|
yading@10
|
67 vld1.16 {d24-d25}, [r1]!
|
|
yading@10
|
68 vld1.16 {d26-d27}, [r2]!
|
|
yading@10
|
69 vmul.s16 q10, q10, q14
|
|
yading@10
|
70 vmul.s16 q13, q13, q14
|
|
yading@10
|
71 vmlal.s16 q0, d16, d18
|
|
yading@10
|
72 vmlal.s16 q1, d17, d19
|
|
yading@10
|
73 vadd.s16 q10, q8, q10
|
|
yading@10
|
74 vadd.s16 q13, q11, q13
|
|
yading@10
|
75 vmlal.s16 q2, d22, d24
|
|
yading@10
|
76 vmlal.s16 q3, d23, d25
|
|
yading@10
|
77 vst1.16 {q10}, [r12,:128]!
|
|
yading@10
|
78 subs r3, r3, #16
|
|
yading@10
|
79 vst1.16 {q13}, [r12,:128]!
|
|
yading@10
|
80 bne 1b
|
|
yading@10
|
81
|
|
yading@10
|
82 vpadd.s32 d16, d0, d1
|
|
yading@10
|
83 vpadd.s32 d17, d2, d3
|
|
yading@10
|
84 vpadd.s32 d10, d4, d5
|
|
yading@10
|
85 vpadd.s32 d11, d6, d7
|
|
yading@10
|
86 vpadd.s32 d0, d16, d17
|
|
yading@10
|
87 vpadd.s32 d1, d10, d11
|
|
yading@10
|
88 vpadd.s32 d2, d0, d1
|
|
yading@10
|
89 vpaddl.s32 d3, d2
|
|
yading@10
|
90 vmov.32 r0, d3[0]
|
|
yading@10
|
91 bx lr
|
|
yading@10
|
92 endfunc
|
