1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
|
/*
* Alpha optimized DSP utils
* Copyright (c) 2002 Falk Hueffner <falk@debian.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include "asm.h"
#include "../dsputil.h"
void simple_idct_axp(DCTELEM *block);
static void put_pixels_clamped_axp(const DCTELEM *block, UINT8 *pixels,
int line_size)
{
int i = 8;
do {
UINT64 shorts;
shorts = ldq(block);
shorts = maxsw4(shorts, 0);
shorts = minsw4(shorts, WORD_VEC(0x00ff));
stl(pkwb(shorts), pixels);
shorts = ldq(block + 4);
shorts = maxsw4(shorts, 0);
shorts = minsw4(shorts, WORD_VEC(0x00ff));
stl(pkwb(shorts), pixels + 4);
pixels += line_size;
block += 8;
} while (--i);
}
static void add_pixels_clamped_axp(const DCTELEM *block, UINT8 *pixels,
int line_size)
{
int i = 8;
do {
UINT64 shorts;
shorts = ldq(block);
shorts &= ~WORD_VEC(0x8000); /* clear highest bit to avoid overflow */
shorts += unpkbw(ldl(pixels));
shorts &= ~WORD_VEC(0x8000); /* hibit would be set for e. g. -2 + 3 */
shorts = minuw4(shorts, WORD_VEC(0x4000)); /* set neg. to 0x4000 */
shorts &= ~WORD_VEC(0x4000); /* ...and zap them */
shorts = minsw4(shorts, WORD_VEC(0x00ff)); /* clamp to 255 */
stl(pkwb(shorts), pixels);
/* next 4 */
shorts = ldq(block + 4);
shorts &= ~WORD_VEC(0x8000);
shorts += unpkbw(ldl(pixels + 4));
shorts &= ~WORD_VEC(0x8000);
shorts = minuw4(shorts, WORD_VEC(0x4000));
shorts &= ~WORD_VEC(0x4000);
shorts = minsw4(shorts, WORD_VEC(0x00ff));
stl(pkwb(shorts), pixels + 4);
pixels += line_size;
block += 8;
} while (--i);
}
/* Average 8 unsigned bytes in parallel: (b1 + b2) >> 1
Since the immediate result could be greater than 255, we do the
shift first. The result is too low by one if the bytes were both
odd, so we need to add (l1 & l2) & BYTE_VEC(0x01). */
static inline UINT64 avg2_no_rnd(UINT64 l1, UINT64 l2)
{
UINT64 correction = (l1 & l2) & BYTE_VEC(0x01);
l1 = (l1 & ~BYTE_VEC(0x01)) >> 1;
l2 = (l2 & ~BYTE_VEC(0x01)) >> 1;
return l1 + l2 + correction;
}
/* Average 8 bytes with rounding: (b1 + b2 + 1) >> 1
The '1' only has an effect when one byte is even and the other odd,
i. e. we also need to add (l1 ^ l2) & BYTE_VEC(0x01).
Incidentally, that is equivalent to (l1 | l2) & BYTE_VEC(0x01). */
static inline UINT64 avg2(UINT64 l1, UINT64 l2)
{
UINT64 correction = (l1 | l2) & BYTE_VEC(0x01);
l1 = (l1 & ~BYTE_VEC(0x01)) >> 1;
l2 = (l2 & ~BYTE_VEC(0x01)) >> 1;
return l1 + l2 + correction;
}
static inline UINT64 avg4(UINT64 l1, UINT64 l2, UINT64 l3, UINT64 l4)
{
UINT64 r1 = ((l1 & ~BYTE_VEC(0x03)) >> 2)
+ ((l2 & ~BYTE_VEC(0x03)) >> 2)
+ ((l3 & ~BYTE_VEC(0x03)) >> 2)
+ ((l4 & ~BYTE_VEC(0x03)) >> 2);
UINT64 r2 = (( (l1 & BYTE_VEC(0x03))
+ (l2 & BYTE_VEC(0x03))
+ (l3 & BYTE_VEC(0x03))
+ (l4 & BYTE_VEC(0x03))
+ BYTE_VEC(0x02)) >> 2) & BYTE_VEC(0x03);
return r1 + r2;
}
static inline UINT64 avg4_no_rnd(UINT64 l1, UINT64 l2, UINT64 l3, UINT64 l4)
{
UINT64 r1 = ((l1 & ~BYTE_VEC(0x03)) >> 2)
+ ((l2 & ~BYTE_VEC(0x03)) >> 2)
+ ((l3 & ~BYTE_VEC(0x03)) >> 2)
+ ((l4 & ~BYTE_VEC(0x03)) >> 2);
UINT64 r2 = (( (l1 & BYTE_VEC(0x03))
+ (l2 & BYTE_VEC(0x03))
+ (l3 & BYTE_VEC(0x03))
+ (l4 & BYTE_VEC(0x03))
+ BYTE_VEC(0x01)) >> 2) & BYTE_VEC(0x03);
return r1 + r2;
}
#define PIXOPNAME(suffix) put ## suffix
#define BTYPE UINT8
#define AVG2 avg2
#define AVG4 avg4
#define STORE(l, b) stq(l, b)
#include "pixops.h"
#undef PIXOPNAME
#undef BTYPE
#undef AVG2
#undef AVG4
#undef STORE
#define PIXOPNAME(suffix) put_no_rnd ## suffix
#define BTYPE UINT8
#define AVG2 avg2_no_rnd
#define AVG4 avg4_no_rnd
#define STORE(l, b) stq(l, b)
#include "pixops.h"
#undef PIXOPNAME
#undef BTYPE
#undef AVG2
#undef AVG4
#undef STORE
/* The following functions are untested. */
#if 0
#define PIXOPNAME(suffix) avg ## suffix
#define BTYPE UINT8
#define AVG2 avg2
#define AVG4 avg4
#define STORE(l, b) stq(AVG2(l, ldq(b)), b);
#include "pixops.h"
#undef PIXOPNAME
#undef BTYPE
#undef AVG2
#undef AVG4
#undef STORE
#define PIXOPNAME(suffix) avg_no_rnd ## suffix
#define BTYPE UINT8
#define AVG2 avg2_no_rnd
#define AVG4 avg4_no_rnd
#define STORE(l, b) stq(AVG2(l, ldq(b)), b);
#include "pixops.h"
#undef PIXOPNAME
#undef BTYPE
#undef AVG2
#undef AVG4
#undef STORE
#define PIXOPNAME(suffix) sub ## suffix
#define BTYPE DCTELEM
#define AVG2 avg2
#define AVG4 avg4
#define STORE(l, block) do { \
UINT64 xxx = l; \
(block)[0] -= (xxx >> 0) & 0xff; \
(block)[1] -= (xxx >> 8) & 0xff; \
(block)[2] -= (xxx >> 16) & 0xff; \
(block)[3] -= (xxx >> 24) & 0xff; \
(block)[4] -= (xxx >> 32) & 0xff; \
(block)[5] -= (xxx >> 40) & 0xff; \
(block)[6] -= (xxx >> 48) & 0xff; \
(block)[7] -= (xxx >> 56) & 0xff; \
} while (0)
#include "pixops.h"
#undef PIXOPNAME
#undef BTYPE
#undef AVG2
#undef AVG4
#undef STORE
#endif
void dsputil_init_alpha(void)
{
put_pixels_tab[0] = put_pixels_axp;
put_pixels_tab[1] = put_pixels_x2_axp;
put_pixels_tab[2] = put_pixels_y2_axp;
put_pixels_tab[3] = put_pixels_xy2_axp;
put_no_rnd_pixels_tab[0] = put_pixels_axp;
put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_axp;
put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_axp;
put_no_rnd_pixels_tab[3] = put_no_rnd_pixels_xy2_axp;
/* amask clears all bits that correspond to present features. */
if (amask(AMASK_MVI) == 0) {
fprintf(stderr, "MVI extension detected\n");
put_pixels_clamped = put_pixels_clamped_axp;
add_pixels_clamped = add_pixels_clamped_axp;
}
}
|
