annotate avr32/mathops.h @ 11541:a3790cd1c9ca libavcodec

h264: Simplify decode_cabac_residual() specialization Gives more consistent inlining with some compilers (such as llvm).
author astrange
date Fri, 26 Mar 2010 03:29:31 +0000
parents 25136467a218
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
10065
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
1 /*
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
2 * Simple math operations
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
3 * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
4 *
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
5 * This file is part of FFmpeg.
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
6 *
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
7 * FFmpeg is free software; you can redistribute it and/or
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
8 * modify it under the terms of the GNU Lesser General Public
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
9 * License as published by the Free Software Foundation; either
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
10 * version 2.1 of the License, or (at your option) any later version.
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
11 *
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
12 * FFmpeg is distributed in the hope that it will be useful,
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
15 * Lesser General Public License for more details.
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
16 *
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
17 * You should have received a copy of the GNU Lesser General Public
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
18 * License along with FFmpeg; if not, write to the Free Software
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
20 */
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
21
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
22 #ifndef AVCODEC_AVR32_MATHOPS_H
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
23 #define AVCODEC_AVR32_MATHOPS_H
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
24
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
25 #include <stdint.h>
10080
25136467a218 Add necessary #include for config.h.
diego
parents: 10065
diff changeset
26 #include "config.h"
10065
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
27 #include "libavutil/common.h"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
28
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
29 #if HAVE_INLINE_ASM
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
30
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
31 #define MULL MULL
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
32 static inline av_const int MULL(int a, int b, unsigned shift)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
33 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
34 union { int64_t x; int hl[2]; } x;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
35 __asm__ ("muls.d %0, %1, %2 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
36 "lsr %0, %3 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
37 "or %0, %0, %m0<<%4 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
38 : "=r"(x) : "r"(b), "r"(a), "i"(shift), "i"(32-shift));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
39 return x.hl[1];
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
40 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
41
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
42 #define MULH MULH
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
43 static inline av_const int MULH(int a, int b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
44 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
45 union { int64_t x; int hl[2]; } x;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
46 __asm__ ("muls.d %0, %1, %2" : "=r"(x.x) : "r"(a), "r"(b));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
47 return x.hl[0];
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
48 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
49
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
50 #define MUL64 MUL64
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
51 static inline av_const int64_t MUL64(int a, int b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
52 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
53 int64_t x;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
54 __asm__ ("muls.d %0, %1, %2" : "=r"(x) : "r"(a), "r"(b));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
55 return x;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
56 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
57
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
58 static inline av_const int64_t MAC64(int64_t d, int a, int b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
59 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
60 __asm__ ("macs.d %0, %1, %2" : "+r"(d) : "r"(a), "r"(b));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
61 return d;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
62 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
63 #define MAC64(d, a, b) ((d) = MAC64(d, a, b))
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
64 #define MLS64(d, a, b) MAC64(d, -(a), b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
65
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
66 static inline av_const int MAC16(int d, int a, int b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
67 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
68 __asm__ ("machh.w %0, %1:b, %2:b" : "+r"(d) : "r"(a), "r"(b));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
69 return d;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
70 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
71 #define MAC16(d, a, b) ((d) = MAC16(d, a, b))
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
72 #define MLS16(d, a, b) MAC16(d, -(a), b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
73
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
74 #define MUL16 MUL16
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
75 static inline av_const int MUL16(int a, int b)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
76 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
77 int d;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
78 __asm__ ("mulhh.w %0, %1:b, %2:b" : "=r"(d) : "r"(a), "r"(b));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
79 return d;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
80 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
81
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
82 #define mid_pred mid_pred
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
83 static inline av_const int mid_pred(int a, int b, int c)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
84 {
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
85 int m;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
86 __asm__ ("mov %0, %2 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
87 "cp.w %1, %2 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
88 "movgt %0, %1 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
89 "movgt %1, %2 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
90 "cp.w %1, %3 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
91 "movle %1, %3 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
92 "cp.w %0, %1 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
93 "movgt %0, %1 \n\t"
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
94 : "=&r"(m), "+r"(a)
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
95 : "r"(b), "r"(c));
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
96 return m;
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
97 }
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
98
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
99 #endif /* HAVE_INLINE_ASM */
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
100
42a126c78744 AVR32: optimised mathops.h
mru
parents:
diff changeset
101 #endif /* AVCODEC_AVR32_MATHOPS_H */