2 * ARM-NEON-optimized IDCT functions
3 * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/arm/asm.S"
24 function ff_put_pixels_clamped_neon, export=1
25 vld1.16 {d16-d19}, [r0,:128]!
27 vld1.16 {d20-d23}, [r0,:128]!
29 vld1.16 {d24-d27}, [r0,:128]!
31 vld1.16 {d28-d31}, [r0,:128]!
33 vst1.8 {d0}, [r1,:64], r2
35 vst1.8 {d1}, [r1,:64], r2
37 vst1.8 {d2}, [r1,:64], r2
39 vst1.8 {d3}, [r1,:64], r2
41 vst1.8 {d4}, [r1,:64], r2
42 vst1.8 {d5}, [r1,:64], r2
43 vst1.8 {d6}, [r1,:64], r2
44 vst1.8 {d7}, [r1,:64], r2
48 function ff_put_signed_pixels_clamped_neon, export=1
50 vld1.16 {d16-d17}, [r0,:128]!
52 vld1.16 {d18-d19}, [r0,:128]!
54 vld1.16 {d16-d17}, [r0,:128]!
56 vld1.16 {d18-d19}, [r0,:128]!
58 vld1.16 {d20-d21}, [r0,:128]!
60 vld1.16 {d22-d23}, [r0,:128]!
62 vst1.8 {d0}, [r1,:64], r2
64 vst1.8 {d1}, [r1,:64], r2
66 vst1.8 {d2}, [r1,:64], r2
68 vld1.16 {d24-d25}, [r0,:128]!
70 vld1.16 {d26-d27}, [r0,:128]!
73 vst1.8 {d3}, [r1,:64], r2
75 vst1.8 {d4}, [r1,:64], r2
77 vst1.8 {d5}, [r1,:64], r2
80 vst1.8 {d6}, [r1,:64], r2
81 vst1.8 {d7}, [r1,:64], r2
85 function ff_add_pixels_clamped_neon, export=1
87 vld1.8 {d16}, [r1,:64], r2
88 vld1.16 {d0-d1}, [r0,:128]!
90 vld1.8 {d17}, [r1,:64], r2
91 vld1.16 {d2-d3}, [r0,:128]!
93 vld1.8 {d18}, [r1,:64], r2
95 vld1.16 {d4-d5}, [r0,:128]!
97 vst1.8 {d0}, [r3,:64], r2
99 vld1.8 {d19}, [r1,:64], r2
100 vld1.16 {d6-d7}, [r0,:128]!
103 vst1.8 {d2}, [r3,:64], r2
104 vld1.8 {d16}, [r1,:64], r2
106 vld1.16 {d0-d1}, [r0,:128]!
108 vst1.8 {d4}, [r3,:64], r2
109 vld1.8 {d17}, [r1,:64], r2
110 vld1.16 {d2-d3}, [r0,:128]!
112 vst1.8 {d6}, [r3,:64], r2
114 vld1.8 {d18}, [r1,:64], r2
115 vld1.16 {d4-d5}, [r0,:128]!
117 vst1.8 {d0}, [r3,:64], r2
119 vld1.8 {d19}, [r1,:64], r2
121 vld1.16 {d6-d7}, [r0,:128]!
123 vst1.8 {d2}, [r3,:64], r2
125 vst1.8 {d4}, [r3,:64], r2
126 vst1.8 {d6}, [r3,:64], r2