Mercurial > libavcodec.hg
comparison arm/aac.h @ 10881:798c62217427 libavcodec
AAC: ARM/NEON asm for VMUL2/4 functions
author | mru |
---|---|
date | Fri, 15 Jan 2010 02:58:24 +0000 |
parents | |
children | a69d6f292f95 |
comparison
equal
deleted
inserted
replaced
10880:1184d065c26a | 10881:798c62217427 |
---|---|
1 /* | |
2 * Copyright (c) 2010 Mans Rullgard <mans@mansr.com> | |
3 * | |
4 * This file is part of FFmpeg. | |
5 * | |
6 * FFmpeg is free software; you can redistribute it and/or | |
7 * modify it under the terms of the GNU Lesser General Public | |
8 * License as published by the Free Software Foundation; either | |
9 * version 2.1 of the License, or (at your option) any later version. | |
10 * | |
11 * FFmpeg is distributed in the hope that it will be useful, | |
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 * Lesser General Public License for more details. | |
15 * | |
16 * You should have received a copy of the GNU Lesser General Public | |
17 * License along with FFmpeg; if not, write to the Free Software | |
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
19 */ | |
20 | |
21 #if HAVE_NEON && HAVE_INLINE_ASM | |
22 | |
23 #define VMUL2 VMUL2 | |
24 static inline float *VMUL2(float *dst, const float *v, unsigned idx, | |
25 const float *scale) | |
26 { | |
27 unsigned v0, v1; | |
28 __asm__ volatile ("ubfx %0, %4, #0, #4 \n\t" | |
29 "ubfx %1, %4, #4, #4 \n\t" | |
30 "ldr %0, [%3, %0, lsl #2] \n\t" | |
31 "ldr %1, [%3, %1, lsl #2] \n\t" | |
32 "vld1.32 {d1[]}, [%5,:32] \n\t" | |
33 "vmov d0, %0, %1 \n\t" | |
34 "vmul.f32 d0, d0, d1 \n\t" | |
35 "vst1.32 {d0}, [%2,:64]! \n\t" | |
36 : "=&r"(v0), "=&r"(v1), "+r"(dst) | |
37 : "r"(v), "r"(idx), "r"(scale) | |
38 : "d0", "d1"); | |
39 return dst; | |
40 } | |
41 | |
42 #define VMUL4 VMUL4 | |
43 static inline float *VMUL4(float *dst, const float *v, unsigned idx, | |
44 const float *scale) | |
45 { | |
46 unsigned v0, v1, v2, v3; | |
47 __asm__ volatile ("ubfx %0, %6, #0, #2 \n\t" | |
48 "ubfx %1, %6, #2, #2 \n\t" | |
49 "ldr %0, [%5, %0, lsl #2] \n\t" | |
50 "ubfx %2, %6, #4, #2 \n\t" | |
51 "ldr %1, [%5, %1, lsl #2] \n\t" | |
52 "ubfx %3, %6, #6, #2 \n\t" | |
53 "ldr %2, [%5, %2, lsl #2] \n\t" | |
54 "vmov d0, %0, %1 \n\t" | |
55 "ldr %3, [%5, %3, lsl #2] \n\t" | |
56 "vld1.32 {d2[],d3[]},[%7,:32] \n\t" | |
57 "vmov d1, %2, %3 \n\t" | |
58 "vmul.f32 q0, q0, q1 \n\t" | |
59 "vst1.32 {q0}, [%4,:128]! \n\t" | |
60 : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst) | |
61 : "r"(v), "r"(idx), "r"(scale) | |
62 : "d0", "d1", "d2", "d3"); | |
63 return dst; | |
64 } | |
65 | |
66 #define VMUL2S VMUL2S | |
67 static inline float *VMUL2S(float *dst, const float *v, unsigned idx, | |
68 unsigned sign, const float *scale) | |
69 { | |
70 unsigned v0, v1, v2, v3; | |
71 __asm__ volatile ("ubfx %0, %6, #0, #4 \n\t" | |
72 "ubfx %1, %6, #4, #4 \n\t" | |
73 "ldr %0, [%5, %0, lsl #2] \n\t" | |
74 "lsl %2, %8, #30 \n\t" | |
75 "ldr %1, [%5, %1, lsl #2] \n\t" | |
76 "lsl %3, %8, #31 \n\t" | |
77 "vmov d0, %0, %1 \n\t" | |
78 "bic %2, %2, #1<<30 \n\t" | |
79 "vld1.32 {d1[]}, [%7,:32] \n\t" | |
80 "vmov d2, %2, %3 \n\t" | |
81 "veor d0, d0, d2 \n\t" | |
82 "vmul.f32 d0, d0, d1 \n\t" | |
83 "vst1.32 {d0}, [%4,:64]! \n\t" | |
84 : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst) | |
85 : "r"(v), "r"(idx), "r"(scale), "r"(sign) | |
86 : "d0", "d1", "d2"); | |
87 return dst; | |
88 } | |
89 | |
90 #define VMUL4S VMUL4S | |
91 static inline float *VMUL4S(float *dst, const float *v, unsigned idx, | |
92 unsigned sign, const float *scale) | |
93 { | |
94 unsigned v0, v1, v2, v3, nz; | |
95 __asm__ volatile ("vld1.32 {d2[],d3[]},[%9,:32] \n\t" | |
96 "ubfx %0, %8, #0, #2 \n\t" | |
97 "ubfx %1, %8, #2, #2 \n\t" | |
98 "ldr %0, [%7, %0, lsl #2] \n\t" | |
99 "ubfx %2, %8, #4, #2 \n\t" | |
100 "ldr %1, [%7, %1, lsl #2] \n\t" | |
101 "ubfx %3, %8, #6, #2 \n\t" | |
102 "ldr %2, [%7, %2, lsl #2] \n\t" | |
103 "vmov d0, %0, %1 \n\t" | |
104 "ldr %3, [%7, %3, lsl #2] \n\t" | |
105 "lsr %6, %8, #12 \n\t" | |
106 "rbit %6, %6 \n\t" | |
107 "vmov d1, %2, %3 \n\t" | |
108 "lsls %6, %6, #1 \n\t" | |
109 "and %0, %5, #1<<31 \n\t" | |
110 "lslcs %5, %5, #1 \n\t" | |
111 "lsls %6, %6, #1 \n\t" | |
112 "and %1, %5, #1<<31 \n\t" | |
113 "lslcs %5, %5, #1 \n\t" | |
114 "lsls %6, %6, #1 \n\t" | |
115 "and %2, %5, #1<<31 \n\t" | |
116 "lslcs %5, %5, #1 \n\t" | |
117 "vmov d4, %0, %1 \n\t" | |
118 "and %3, %5, #1<<31 \n\t" | |
119 "vmov d5, %2, %3 \n\t" | |
120 "veor q0, q0, q2 \n\t" | |
121 "vmul.f32 q0, q0, q1 \n\t" | |
122 "vst1.32 {q0}, [%4,:128]! \n\t" | |
123 : "=&r"(v0), "=&r"(v1), "=&r"(v2), "=&r"(v3), "+r"(dst), | |
124 "+r"(sign), "=r"(nz) | |
125 : "r"(v), "r"(idx), "r"(scale) | |
126 : "d0", "d1", "d2", "d3", "d4", "d5"); | |
127 return dst; | |
128 } | |
129 | |
130 #endif /* HAVE_NEON && HAVE_INLINE_ASM */ |