Mercurial > libavcodec.hg
annotate snow.h @ 12197:fbf4d5b1b664 libavcodec
Remove FF_MM_SSE2/3 flags for CPUs where this is generally not faster than
regular MMX code. Examples of this are the Core1 CPU. Instead, set a new flag,
FF_MM_SSE2/3SLOW, which can be checked for particular SSE2/3 functions that
have been checked specifically on such CPUs and are actually faster than
their MMX counterparts.
In addition, use this flag to enable particular VP8 and LPC SSE2 functions
that are faster than their MMX counterparts.
Based on a patch by Loren Merritt <lorenm AT u washington edu>.
author | rbultje |
---|---|
date | Mon, 19 Jul 2010 22:38:23 +0000 |
parents | 0f0cd6b5791f |
children |
rev | line source |
---|---|
3198 | 1 /* |
2 * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at> | |
3 * Copyright (C) 2006 Robert Edele <yartrebo@earthlink.net> | |
4 * | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
5 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
6 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
7 * FFmpeg is free software; you can redistribute it and/or |
3198 | 8 * modify it under the terms of the GNU Lesser General Public |
9 * 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:
3582
diff
changeset
|
10 * version 2.1 of the License, or (at your option) any later version. |
3198 | 11 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3582
diff
changeset
|
12 * FFmpeg is distributed in the hope that it will be useful, |
3198 | 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. | |
16 * | |
17 * 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:
3582
diff
changeset
|
18 * License along with FFmpeg; if not, write to the Free Software |
3198 | 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 */ | |
21 | |
7760 | 22 #ifndef AVCODEC_SNOW_H |
23 #define AVCODEC_SNOW_H | |
3198 | 24 |
25 #include "dsputil.h" | |
11485 | 26 #include "dwt.h" |
3198 | 27 |
28 #define MID_STATE 128 | |
29 | |
30 #define MAX_PLANES 4 | |
31 #define QSHIFT 5 | |
32 #define QROOT (1<<QSHIFT) | |
33 #define LOSSLESS_QLOG -128 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
34 #define FRAC_BITS 4 |
3314 | 35 #define MAX_REF_FRAMES 8 |
3198 | 36 |
3206 | 37 #define LOG2_OBMC_MAX 8 |
3198 | 38 #define OBMC_MAX (1<<(LOG2_OBMC_MAX)) |
39 | |
3223 | 40 /* C bits used by mmx/sse2/altivec */ |
41 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
42 static av_always_inline void snow_interleave_line_header(int * i, int width, IDWTELEM * low, IDWTELEM * high){ |
3223 | 43 (*i) = (width) - 2; |
44 | |
45 if (width & 1){ | |
46 low[(*i)+1] = low[((*i)+1)>>1]; | |
47 (*i)--; | |
48 } | |
49 } | |
50 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
51 static av_always_inline void snow_interleave_line_footer(int * i, IDWTELEM * low, IDWTELEM * high){ |
3223 | 52 for (; (*i)>=0; (*i)-=2){ |
53 low[(*i)+1] = high[(*i)>>1]; | |
54 low[*i] = low[(*i)>>1]; | |
55 } | |
56 } | |
57 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
58 static av_always_inline void snow_horizontal_compose_lift_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w, int lift_high, int mul, int add, int shift){ |
3223 | 59 for(; i<w; i++){ |
60 dst[i] = src[i] - ((mul * (ref[i] + ref[i + 1]) + add) >> shift); | |
61 } | |
62 | |
63 if((width^lift_high)&1){ | |
64 dst[w] = src[w] - ((mul * 2 * ref[w] + add) >> shift); | |
65 } | |
66 } | |
67 | |
5587
3ae03eacbe9f
use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
michael
parents:
5565
diff
changeset
|
68 static av_always_inline void snow_horizontal_compose_liftS_lead_out(int i, IDWTELEM * dst, IDWTELEM * src, IDWTELEM * ref, int width, int w){ |
3223 | 69 for(; i<w; i++){ |
5565
93082c591c8b
Change rounding of the horizontal DWT to match the vertical one.
michael
parents:
5552
diff
changeset
|
70 dst[i] = src[i] + ((ref[i] + ref[(i+1)]+W_BO + 4 * src[i]) >> W_BS); |
3223 | 71 } |
72 | |
73 if(width&1){ | |
5565
93082c591c8b
Change rounding of the horizontal DWT to match the vertical one.
michael
parents:
5552
diff
changeset
|
74 dst[w] = src[w] + ((2 * ref[w] + W_BO + 4 * src[w]) >> W_BS); |
3223 | 75 } |
76 } | |
77 | |
7760 | 78 #endif /* AVCODEC_SNOW_H */ |