1 |
/* |
/***************************************************************************** |
2 |
|
* |
3 |
Copyright (C) 2002 Benjamin Herrenschmidt <benh@kernel.crashing.org> |
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - altivec sum of absolute difference (C version) |
5 |
This program is free software; you can redistribute it and/or modify |
* |
6 |
it under the terms of the GNU General Public License as published by |
* Copyright (C) 2002 Benjamin Herrenschmidt <benh@kernel.crashing.org> |
7 |
the Free Software Foundation; either version 2 of the License, or |
* |
8 |
(at your option) any later version. |
* This program is an implementation of a part of one or more MPEG-4 |
9 |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
10 |
This program is distributed in the hope that it will be useful, |
* to use this software module in hardware or software products are |
11 |
but WITHOUT ANY WARRANTY; without even the implied warranty of |
* advised that its use may infringe existing patents or copyrights, and |
12 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
* any such use would be at such party's own risk. The original |
13 |
GNU General Public License for more details. |
* developer of this software module and his/her company, and subsequent |
14 |
|
* editors and their companies, will have no liability for use of this |
15 |
You should have received a copy of the GNU General Public License |
* software or modifications or derivatives thereof. |
16 |
along with this program; if not, write to the Free Software |
* |
17 |
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* This program is free software; you can redistribute it and/or modify |
18 |
|
* it under the terms of the GNU General Public License as published by |
19 |
|
* the Free Software Foundation; either version 2 of the License, or |
20 |
$Id: sad_altivec.c,v 1.2 2002-04-11 10:18:40 canard Exp $ |
* (at your option) any later version. |
21 |
$Source: /home/xvid/cvs_copy/cvs-server-root/xvid/xvidcore/src/motion/ppc_asm/sad_altivec.c,v $ |
* |
22 |
$Date: 2002-04-11 10:18:40 $ |
* This program is distributed in the hope that it will be useful, |
23 |
$Author: canard $ |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
24 |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
25 |
*/ |
* GNU General Public License for more details. |
26 |
|
* |
27 |
|
* You should have received a copy of the GNU General Public License |
28 |
|
* along with this program; if not, write to the Free Software |
29 |
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
30 |
|
* |
31 |
|
* $Id: sad_altivec.c,v 1.4 2002-09-06 16:59:47 chl Exp $ |
32 |
|
* |
33 |
|
****************************************************************************/ |
34 |
|
|
35 |
#define G_REG |
#define G_REG |
36 |
|
|
47 |
static const vector unsigned char perms[2] = { |
static const vector unsigned char perms[2] = { |
48 |
(vector unsigned char)( /* Used when cur is aligned */ |
(vector unsigned char)( /* Used when cur is aligned */ |
49 |
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, |
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, |
50 |
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17 |
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17), |
|
), |
|
51 |
(vector unsigned char)( /* Used when cur is unaligned */ |
(vector unsigned char)( /* Used when cur is unaligned */ |
52 |
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, |
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, |
53 |
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f |
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f), |
|
), |
|
54 |
}; |
}; |
55 |
|
|
56 |
#ifdef G_REG |
#ifdef G_REG |
57 |
void sadInit_altivec(void) |
void |
58 |
|
sadInit_altivec(void) |
59 |
{ |
{ |
60 |
perm0 = perms[0]; |
perm0 = perms[0]; |
61 |
perm1 = perms[1]; |
perm1 = perms[1]; |
62 |
zerovec = (vector unsigned int)(0); |
zerovec = (vector unsigned int)(0); |
63 |
} |
} |
64 |
static inline const vector unsigned char get_perm(unsigned long i) |
static inline const vector unsigned char |
65 |
|
get_perm(unsigned long i) |
66 |
{ |
{ |
67 |
return i ? perm1 : perm0; |
return i ? perm1 : perm0; |
68 |
} |
} |
69 |
|
|
70 |
#define ZERODEF |
#define ZERODEF |
71 |
#define ZEROVEC zerovec |
#define ZEROVEC zerovec |
72 |
#else |
#else |
73 |
void sadInit_altivec(void) { } |
void |
74 |
static inline const vector unsigned char get_perm(unsigned long i) |
sadInit_altivec(void) |
75 |
|
{ |
76 |
|
} |
77 |
|
static inline const vector unsigned char |
78 |
|
get_perm(unsigned long i) |
79 |
{ |
{ |
80 |
return perms[i]; |
return perms[i]; |
81 |
} |
} |
82 |
|
|
83 |
#define ZERODEF vector unsigned int zerovec = (vector unsigned int)(0) |
#define ZERODEF vector unsigned int zerovec = (vector unsigned int)(0) |
84 |
#define ZEROVEC zerovec |
#define ZEROVEC zerovec |
85 |
#endif |
#endif |
107 |
vector unsigned int sad; |
vector unsigned int sad; |
108 |
vector signed int sumdiffs, best_vec; |
vector signed int sumdiffs, best_vec; |
109 |
unsigned long result; |
unsigned long result; |
110 |
|
|
111 |
ZERODEF; |
ZERODEF; |
112 |
|
|
113 |
#ifdef DEBUG |
#ifdef DEBUG |
185 |
vector unsigned char perm_cur; |
vector unsigned char perm_cur; |
186 |
vector unsigned char perm_ref1, perm_ref2; |
vector unsigned char perm_ref1, perm_ref2; |
187 |
unsigned long result; |
unsigned long result; |
188 |
|
|
189 |
ZERODEF; |
ZERODEF; |
190 |
|
|
191 |
#ifdef DEBUG |
#ifdef DEBUG |
238 |
vector unsigned char t2,t3,t4, mn; |
vector unsigned char t2,t3,t4, mn; |
239 |
vector unsigned int mean, dev; |
vector unsigned int mean, dev; |
240 |
vector signed int sumdiffs; |
vector signed int sumdiffs; |
241 |
vector unsigned char c0,c1,c2,c3,c4,c5,c6,c7,c8,c9,c10,c11,c12,c13,c14,c15; |
vector unsigned char c0, c1, c2, c3, c4, c5, c6, c7, c8, c9, c10, c11, c12, |
242 |
|
c13, c14, c15; |
243 |
unsigned long result; |
unsigned long result; |
244 |
|
|
245 |
ZERODEF; |
ZERODEF; |
246 |
|
|
247 |
mean = (vector unsigned int)(ZEROVEC); |
mean = (vector unsigned int)(ZEROVEC); |
266 |
MEAN16(15); |
MEAN16(15); |
267 |
|
|
268 |
sumdiffs = vec_sums((vector signed int) mean, (vector signed int) ZEROVEC); |
sumdiffs = vec_sums((vector signed int) mean, (vector signed int) ZEROVEC); |
269 |
mn = vec_perm((vector unsigned char)sumdiffs, (vector unsigned char)sumdiffs, |
mn = vec_perm((vector unsigned char) sumdiffs, |
270 |
(vector unsigned char)(14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14)); |
(vector unsigned char) sumdiffs, (vector unsigned char) (14, |
271 |
|
14, |
272 |
|
14, |
273 |
|
14, |
274 |
|
14, |
275 |
|
14, |
276 |
|
14, |
277 |
|
14, |
278 |
|
14, |
279 |
|
14, |
280 |
|
14, |
281 |
|
14, |
282 |
|
14, |
283 |
|
14, |
284 |
|
14, |
285 |
|
14)); |
286 |
DEV16(0); |
DEV16(0); |
287 |
DEV16(1); |
DEV16(1); |
288 |
DEV16(2); |
DEV16(2); |