2 * ARM NEON optimised integer operations
3 * Copyright (c) 2009 Kostya Shishkov
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 function ff_scalarproduct_int16_neon, export=1
36 1: vld1.16 {d16-d17}, [r0]!
37 vld1.16 {d20-d21}, [r1,:128]!
38 vmull.s16 q12, d16, d20
39 vld1.16 {d18-d19}, [r0]!
40 vmull.s16 q13, d17, d21
41 vld1.16 {d22-d23}, [r1,:128]!
42 vmull.s16 q14, d18, d22
43 vmull.s16 q15, d19, d23
47 vshl.s32 q10, q14, q12
49 vshl.s32 q11, q15, q12
56 2: vld1.16 {d16-d17}, [r0]!
57 vld1.16 {d20-d21}, [r1,:128]!
58 vmlal.s16 q0, d16, d20
59 vld1.16 {d18-d19}, [r0]!
60 vmlal.s16 q1, d17, d21
61 vld1.16 {d22-d23}, [r1,:128]!
62 vmlal.s16 q2, d18, d22
63 vmlal.s16 q3, d19, d23
67 3: vpadd.s32 d16, d0, d1
71 vpadd.s32 d0, d16, d17
72 vpadd.s32 d1, d10, d11
79 @ scalarproduct_and_madd_int16(/*aligned*/v0,v1,v2,order,mul)
80 function ff_scalarproduct_and_madd_int16_neon, export=1
81 vld1.16 {d28[],d29[]}, [sp]
88 1: vld1.16 {d16-d17}, [r0,:128]!
89 vld1.16 {d18-d19}, [r1]!
90 vld1.16 {d20-d21}, [r2]!
91 vld1.16 {d22-d23}, [r0,:128]!
92 vld1.16 {d24-d25}, [r1]!
93 vld1.16 {d26-d27}, [r2]!
94 vmul.s16 q10, q10, q14
95 vmul.s16 q13, q13, q14
96 vmlal.s16 q0, d16, d18
97 vmlal.s16 q1, d17, d19
99 vadd.s16 q13, q11, q13
100 vmlal.s16 q2, d22, d24
101 vmlal.s16 q3, d23, d25
102 vst1.16 {q10}, [r12,:128]!
104 vst1.16 {q13}, [r12,:128]!
107 vpadd.s32 d16, d0, d1
108 vpadd.s32 d17, d2, d3
109 vpadd.s32 d10, d4, d5
110 vpadd.s32 d11, d6, d7
111 vpadd.s32 d0, d16, d17
112 vpadd.s32 d1, d10, d11