Mercurial > libavcodec.hg
annotate mathops.h @ 11352:6e0af2cfdcfe libavcodec
Do MC and IDCT in coding (hilbert) order
This increases the slice size to 64 pixels, due to having to decode an
entire chroma superblock row per slice.
This can be up to 6% slower depending on clip and CPU, but is necessary
for future optimizations that gain significantly more than was lost.
author | conrad |
---|---|
date | Wed, 03 Mar 2010 23:27:40 +0000 |
parents | c975efe95584 |
children | 969a7d792b79 |
rev | line source |
---|---|
3733 | 1 /* |
2 * simple math operations | |
8629
04423b2f6e0b
cosmetics: Remove pointless period after copyright statement non-sentences.
diego
parents:
8627
diff
changeset
|
3 * Copyright (c) 2001, 2002 Fabrice Bellard |
3733 | 4 * Copyright (c) 2006 Michael Niedermayer <michaelni@gmx.at> et al |
5 * | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
6 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
7 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
8 * FFmpeg is free software; you can redistribute it and/or |
3733 | 9 * modify it under the terms of the GNU Lesser General Public |
10 * License as published by the Free Software Foundation; either | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
11 * version 2.1 of the License, or (at your option) any later version. |
3733 | 12 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
13 * FFmpeg is distributed in the hope that it will be useful, |
3733 | 14 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 * Lesser General Public License for more details. | |
17 * | |
18 * You should have received a copy of the GNU Lesser General Public | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3733
diff
changeset
|
19 * License along with FFmpeg; if not, write to the Free Software |
3733 | 20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
21 */ | |
7760 | 22 #ifndef AVCODEC_MATHOPS_H |
23 #define AVCODEC_MATHOPS_H | |
3733 | 24 |
6763 | 25 #include "libavutil/common.h" |
5162 | 26 |
9679 | 27 #if ARCH_ARM |
9678 | 28 # include "arm/mathops.h" |
10065 | 29 #elif ARCH_AVR32 |
30 # include "avr32/mathops.h" | |
9679 | 31 #elif ARCH_BFIN |
32 # include "bfin/mathops.h" | |
9729 | 33 #elif ARCH_MIPS |
34 # include "mips/mathops.h" | |
8590 | 35 #elif ARCH_PPC |
9678 | 36 # include "ppc/mathops.h" |
9679 | 37 #elif ARCH_X86 |
38 # include "x86/mathops.h" | |
3733 | 39 #endif |
40 | |
41 /* generic implementation */ | |
42 | |
43 #ifndef MULL | |
8201 | 44 # define MULL(a,b,s) (((int64_t)(a) * (int64_t)(b)) >> (s)) |
3733 | 45 #endif |
46 | |
47 #ifndef MULH | |
48 //gcc 3.4 creates an incredibly bloated mess out of this | |
49 //# define MULH(a,b) (((int64_t)(a) * (int64_t)(b))>>32) | |
50 | |
4283
d6f83e2f8804
rename always_inline to av_always_inline and move to common.h
mru
parents:
4112
diff
changeset
|
51 static av_always_inline int MULH(int a, int b){ |
3733 | 52 return ((int64_t)(a) * (int64_t)(b))>>32; |
53 } | |
54 #endif | |
55 | |
11018 | 56 #ifndef UMULH |
57 static av_always_inline unsigned UMULH(unsigned a, unsigned b){ | |
58 return ((uint64_t)(a) * (uint64_t)(b))>>32; | |
59 } | |
60 #endif | |
61 | |
3733 | 62 #ifndef MUL64 |
63 # define MUL64(a,b) ((int64_t)(a) * (int64_t)(b)) | |
64 #endif | |
65 | |
7255 | 66 #ifndef MAC64 |
67 # define MAC64(d, a, b) ((d) += MUL64(a, b)) | |
68 #endif | |
69 | |
70 #ifndef MLS64 | |
71 # define MLS64(d, a, b) ((d) -= MUL64(a, b)) | |
72 #endif | |
73 | |
3733 | 74 /* signed 16x16 -> 32 multiply add accumulate */ |
75 #ifndef MAC16 | |
76 # define MAC16(rt, ra, rb) rt += (ra) * (rb) | |
77 #endif | |
78 | |
79 /* signed 16x16 -> 32 multiply */ | |
80 #ifndef MUL16 | |
81 # define MUL16(ra, rb) ((ra) * (rb)) | |
82 #endif | |
83 | |
7255 | 84 #ifndef MLS16 |
85 # define MLS16(rt, ra, rb) ((rt) -= (ra) * (rb)) | |
86 #endif | |
87 | |
8627
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
88 /* median of 3 */ |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
89 #ifndef mid_pred |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
90 #define mid_pred mid_pred |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
91 static inline av_const int mid_pred(int a, int b, int c) |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
92 { |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
93 #if 0 |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
94 int t= (a-b)&((a-b)>>31); |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
95 a-=t; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
96 b+=t; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
97 b-= (b-c)&((b-c)>>31); |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
98 b+= (a-b)&((a-b)>>31); |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
99 |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
100 return b; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
101 #else |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
102 if(a>b){ |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
103 if(c>b){ |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
104 if(c>a) b=a; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
105 else b=c; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
106 } |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
107 }else{ |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
108 if(b>c){ |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
109 if(c>a) b=c; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
110 else b=a; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
111 } |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
112 } |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
113 return b; |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
114 #endif |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
115 } |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
116 #endif |
d6bab465b82c
moves mid_pred() into mathops.h (with arch specific code split by directory)
aurel
parents:
8590
diff
changeset
|
117 |
9096 | 118 #ifndef sign_extend |
119 static inline av_const int sign_extend(int val, unsigned bits) | |
120 { | |
121 return (val << (INT_BIT - bits)) >> (INT_BIT - bits); | |
122 } | |
123 #endif | |
124 | |
11237 | 125 #ifndef zero_extend |
126 static inline av_const unsigned zero_extend(unsigned val, unsigned bits) | |
127 { | |
128 return (val << (INT_BIT - bits)) >> (INT_BIT - bits); | |
129 } | |
130 #endif | |
131 | |
10951 | 132 #ifndef COPY3_IF_LT |
133 #define COPY3_IF_LT(x, y, a, b, c, d)\ | |
134 if ((y) < (x)) {\ | |
135 (x) = (y);\ | |
136 (a) = (b);\ | |
137 (c) = (d);\ | |
138 } | |
139 #endif | |
140 | |
11200 | 141 #ifndef NEG_SSR32 |
142 # define NEG_SSR32(a,s) ((( int32_t)(a))>>(32-(s))) | |
143 #endif | |
144 | |
145 #ifndef NEG_USR32 | |
146 # define NEG_USR32(a,s) (((uint32_t)(a))>>(32-(s))) | |
147 #endif | |
148 | |
7760 | 149 #endif /* AVCODEC_MATHOPS_H */ |
3733 | 150 |