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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
|
/*
* Alpha optimized DSP utils
* Copyright (c) 2002 Falk Hueffner <falk@debian.org>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include "asm.h"
#include "../dsputil.h"
void simple_idct_axp(DCTELEM *block);
void put_pixels_axp_asm(uint8_t *block, const uint8_t *pixels,
int line_size, int h);
void put_pixels_clamped_mvi_asm(const DCTELEM *block, uint8_t *pixels,
int line_size);
void add_pixels_clamped_mvi_asm(const DCTELEM *block, uint8_t *pixels,
int line_size);
void get_pixels_mvi(DCTELEM *restrict block,
const uint8_t *restrict pixels, int line_size);
void diff_pixels_mvi(DCTELEM *block, const uint8_t *s1, const uint8_t *s2,
int stride);
int pix_abs8x8_mvi(uint8_t *pix1, uint8_t *pix2, int line_size);
int pix_abs16x16_mvi_asm(uint8_t *pix1, uint8_t *pix2, int line_size);
int pix_abs16x16_x2_mvi(uint8_t *pix1, uint8_t *pix2, int line_size);
int pix_abs16x16_y2_mvi(uint8_t *pix1, uint8_t *pix2, int line_size);
int pix_abs16x16_xy2_mvi(uint8_t *pix1, uint8_t *pix2, int line_size);
#if 0
/* These functions were the base for the optimized assembler routines,
and remain here for documentation purposes. */
static void put_pixels_clamped_mvi(const DCTELEM *block, uint8_t *pixels,
int line_size)
{
int i = 8;
uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
ASM_ACCEPT_MVI;
do {
uint64_t shorts0, shorts1;
shorts0 = ldq(block);
shorts0 = maxsw4(shorts0, 0);
shorts0 = minsw4(shorts0, clampmask);
stl(pkwb(shorts0), pixels);
shorts1 = ldq(block + 4);
shorts1 = maxsw4(shorts1, 0);
shorts1 = minsw4(shorts1, clampmask);
stl(pkwb(shorts1), pixels + 4);
pixels += line_size;
block += 8;
} while (--i);
}
void add_pixels_clamped_mvi(const DCTELEM *block, uint8_t *pixels,
int line_size)
{
int h = 8;
/* Keep this function a leaf function by generating the constants
manually (mainly for the hack value ;-). */
uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
uint64_t signmask = zap(-1, 0x33);
signmask ^= signmask >> 1; /* 0x8000800080008000 */
ASM_ACCEPT_MVI;
do {
uint64_t shorts0, pix0, signs0;
uint64_t shorts1, pix1, signs1;
shorts0 = ldq(block);
shorts1 = ldq(block + 4);
pix0 = unpkbw(ldl(pixels));
/* Signed subword add (MMX paddw). */
signs0 = shorts0 & signmask;
shorts0 &= ~signmask;
shorts0 += pix0;
shorts0 ^= signs0;
/* Clamp. */
shorts0 = maxsw4(shorts0, 0);
shorts0 = minsw4(shorts0, clampmask);
/* Next 4. */
pix1 = unpkbw(ldl(pixels + 4));
signs1 = shorts1 & signmask;
shorts1 &= ~signmask;
shorts1 += pix1;
shorts1 ^= signs1;
shorts1 = maxsw4(shorts1, 0);
shorts1 = minsw4(shorts1, clampmask);
stl(pkwb(shorts0), pixels);
stl(pkwb(shorts1), pixels + 4);
pixels += line_size;
block += 8;
} while (--h);
}
#endif
static void clear_blocks_axp(DCTELEM *blocks) {
uint64_t *p = (uint64_t *) blocks;
int n = sizeof(DCTELEM) * 6 * 64;
do {
p[0] = 0;
p[1] = 0;
p[2] = 0;
p[3] = 0;
p[4] = 0;
p[5] = 0;
p[6] = 0;
p[7] = 0;
p += 8;
n -= 8 * 8;
} while (n);
}
static inline uint64_t avg2_no_rnd(uint64_t a, uint64_t b)
{
return (a & b) + (((a ^ b) & BYTE_VEC(0xfe)) >> 1);
}
static inline uint64_t avg2(uint64_t a, uint64_t b)
{
return (a | b) - (((a ^ b) & BYTE_VEC(0xfe)) >> 1);
}
#if 0
/* The XY2 routines basically utilize this scheme, but reuse parts in
each iteration. */
static inline uint64_t avg4(uint64_t l1, uint64_t l2, uint64_t l3, uint64_t l4)
{
uint64_t r1 = ((l1 & ~BYTE_VEC(0x03)) >> 2)
+ ((l2 & ~BYTE_VEC(0x03)) >> 2)
+ ((l3 & ~BYTE_VEC(0x03)) >> 2)
+ ((l4 & ~BYTE_VEC(0x03)) >> 2);
uint64_t 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;
}
#endif
#define OP(LOAD, STORE) \
do { \
STORE(LOAD(pixels), block); \
pixels += line_size; \
block += line_size; \
} while (--h)
#define OP_X2(LOAD, STORE) \
do { \
uint64_t pix1, pix2; \
\
pix1 = LOAD(pixels); \
pix2 = pix1 >> 8 | ((uint64_t) pixels[8] << 56); \
STORE(AVG2(pix1, pix2), block); \
pixels += line_size; \
block += line_size; \
} while (--h)
#define OP_Y2(LOAD, STORE) \
do { \
uint64_t pix = LOAD(pixels); \
do { \
uint64_t next_pix; \
\
pixels += line_size; \
next_pix = LOAD(pixels); \
STORE(AVG2(pix, next_pix), block); \
block += line_size; \
pix = next_pix; \
} while (--h); \
} while (0)
#define OP_XY2(LOAD, STORE) \
do { \
uint64_t pix1 = LOAD(pixels); \
uint64_t pix2 = pix1 >> 8 | ((uint64_t) pixels[8] << 56); \
uint64_t pix_l = (pix1 & BYTE_VEC(0x03)) \
+ (pix2 & BYTE_VEC(0x03)); \
uint64_t pix_h = ((pix1 & ~BYTE_VEC(0x03)) >> 2) \
+ ((pix2 & ~BYTE_VEC(0x03)) >> 2); \
\
do { \
uint64_t npix1, npix2; \
uint64_t npix_l, npix_h; \
uint64_t avg; \
\
pixels += line_size; \
npix1 = LOAD(pixels); \
npix2 = npix1 >> 8 | ((uint64_t) pixels[8] << 56); \
npix_l = (npix1 & BYTE_VEC(0x03)) \
+ (npix2 & BYTE_VEC(0x03)); \
npix_h = ((npix1 & ~BYTE_VEC(0x03)) >> 2) \
+ ((npix2 & ~BYTE_VEC(0x03)) >> 2); \
avg = (((pix_l + npix_l + AVG4_ROUNDER) >> 2) & BYTE_VEC(0x03)) \
+ pix_h + npix_h; \
STORE(avg, block); \
\
block += line_size; \
pix_l = npix_l; \
pix_h = npix_h; \
} while (--h); \
} while (0)
#define MAKE_OP(OPNAME, SUFF, OPKIND, STORE) \
static void OPNAME ## _pixels ## SUFF ## _axp \
(uint8_t *restrict block, const uint8_t *restrict pixels, \
int line_size, int h) \
{ \
if ((size_t) pixels & 0x7) { \
OPKIND(uldq, STORE); \
} else { \
OPKIND(ldq, STORE); \
} \
} \
\
static void OPNAME ## _pixels16 ## SUFF ## _axp \
(uint8_t *restrict block, const uint8_t *restrict pixels, \
int line_size, int h) \
{ \
OPNAME ## _pixels ## SUFF ## _axp(block, pixels, line_size, h); \
OPNAME ## _pixels ## SUFF ## _axp(block + 8, pixels + 8, line_size, h); \
}
#define PIXOP(OPNAME, STORE) \
MAKE_OP(OPNAME, , OP, STORE) \
MAKE_OP(OPNAME, _x2, OP_X2, STORE) \
MAKE_OP(OPNAME, _y2, OP_Y2, STORE) \
MAKE_OP(OPNAME, _xy2, OP_XY2, STORE)
/* Rounding primitives. */
#define AVG2 avg2
#define AVG4 avg4
#define AVG4_ROUNDER BYTE_VEC(0x02)
#define STORE(l, b) stq(l, b)
PIXOP(put, STORE);
#undef STORE
#define STORE(l, b) stq(AVG2(l, ldq(b)), b);
PIXOP(avg, STORE);
/* Not rounding primitives. */
#undef AVG2
#undef AVG4
#undef AVG4_ROUNDER
#undef STORE
#define AVG2 avg2_no_rnd
#define AVG4 avg4_no_rnd
#define AVG4_ROUNDER BYTE_VEC(0x01)
#define STORE(l, b) stq(l, b)
PIXOP(put_no_rnd, STORE);
#undef STORE
#define STORE(l, b) stq(AVG2(l, ldq(b)), b);
PIXOP(avg_no_rnd, STORE);
void put_pixels16_axp_asm(uint8_t *block, const uint8_t *pixels,
int line_size, int h)
{
put_pixels_axp_asm(block, pixels, line_size, h);
put_pixels_axp_asm(block + 8, pixels + 8, line_size, h);
}
void dsputil_init_alpha(void)
{
put_pixels_tab[0][0] = put_pixels16_axp_asm;
put_pixels_tab[0][1] = put_pixels16_x2_axp;
put_pixels_tab[0][2] = put_pixels16_y2_axp;
put_pixels_tab[0][3] = put_pixels16_xy2_axp;
put_no_rnd_pixels_tab[0][0] = put_pixels16_axp_asm;
put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_axp;
put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_axp;
put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_axp;
avg_pixels_tab[0][0] = avg_pixels16_axp;
avg_pixels_tab[0][1] = avg_pixels16_x2_axp;
avg_pixels_tab[0][2] = avg_pixels16_y2_axp;
avg_pixels_tab[0][3] = avg_pixels16_xy2_axp;
avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_axp;
avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_axp;
avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_axp;
avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_axp;
put_pixels_tab[1][0] = put_pixels_axp_asm;
put_pixels_tab[1][1] = put_pixels_x2_axp;
put_pixels_tab[1][2] = put_pixels_y2_axp;
put_pixels_tab[1][3] = put_pixels_xy2_axp;
put_no_rnd_pixels_tab[1][0] = put_pixels_axp_asm;
put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels_x2_axp;
put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels_y2_axp;
put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels_xy2_axp;
avg_pixels_tab[1][0] = avg_pixels_axp;
avg_pixels_tab[1][1] = avg_pixels_x2_axp;
avg_pixels_tab[1][2] = avg_pixels_y2_axp;
avg_pixels_tab[1][3] = avg_pixels_xy2_axp;
avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels_axp;
avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels_x2_axp;
avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels_y2_axp;
avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels_xy2_axp;
clear_blocks = clear_blocks_axp;
/* amask clears all bits that correspond to present features. */
if (amask(AMASK_MVI) == 0) {
put_pixels_clamped = put_pixels_clamped_mvi_asm;
add_pixels_clamped = add_pixels_clamped_mvi_asm;
get_pixels = get_pixels_mvi;
diff_pixels = diff_pixels_mvi;
pix_abs8x8 = pix_abs8x8_mvi;
pix_abs16x16 = pix_abs16x16_mvi_asm;
pix_abs16x16_x2 = pix_abs16x16_x2_mvi;
pix_abs16x16_y2 = pix_abs16x16_y2_mvi;
pix_abs16x16_xy2 = pix_abs16x16_xy2_mvi;
}
}
|
