ffmpeg / libavcodec / sparc / dsputil_vis.c @ 8dbe5856
History | View | Annotate | Download (118 KB)
| 1 |
/*
|
|---|---|
| 2 |
* Copyright (C) 2003 David S. Miller <davem@redhat.com>
|
| 3 |
*
|
| 4 |
* This file is part of FFmpeg.
|
| 5 |
*
|
| 6 |
* FFmpeg is free software; you can redistribute it and/or
|
| 7 |
* modify it under the terms of the GNU Lesser General Public
|
| 8 |
* License as published by the Free Software Foundation; either
|
| 9 |
* version 2.1 of the License, or (at your option) any later version.
|
| 10 |
*
|
| 11 |
* FFmpeg is distributed in the hope that it will be useful,
|
| 12 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 13 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 14 |
* Lesser General Public License for more details.
|
| 15 |
*
|
| 16 |
* You should have received a copy of the GNU Lesser General Public
|
| 17 |
* License along with FFmpeg; if not, write to the Free Software
|
| 18 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
| 19 |
*/
|
| 20 |
|
| 21 |
/* The *no_round* functions have been added by James A. Morrison, 2003,2004.
|
| 22 |
The vis code from libmpeg2 was adapted for ffmpeg by James A. Morrison.
|
| 23 |
*/
|
| 24 |
|
| 25 |
#include "config.h" |
| 26 |
|
| 27 |
#include <inttypes.h> |
| 28 |
|
| 29 |
#include "libavcodec/dsputil.h" |
| 30 |
#include "dsputil_vis.h" |
| 31 |
|
| 32 |
#include "vis.h" |
| 33 |
|
| 34 |
/* The trick used in some of this file is the formula from the MMX
|
| 35 |
* motion comp code, which is:
|
| 36 |
*
|
| 37 |
* (x+y+1)>>1 == (x|y)-((x^y)>>1)
|
| 38 |
*
|
| 39 |
* This allows us to average 8 bytes at a time in a 64-bit FPU reg.
|
| 40 |
* We avoid overflows by masking before we do the shift, and we
|
| 41 |
* implement the shift by multiplying by 1/2 using mul8x16. So in
|
| 42 |
* VIS this is (assume 'x' is in f0, 'y' is in f2, a repeating mask
|
| 43 |
* of '0xfe' is in f4, a repeating mask of '0x7f' is in f6, and
|
| 44 |
* the value 0x80808080 is in f8):
|
| 45 |
*
|
| 46 |
* fxor f0, f2, f10
|
| 47 |
* fand f10, f4, f10
|
| 48 |
* fmul8x16 f8, f10, f10
|
| 49 |
* fand f10, f6, f10
|
| 50 |
* for f0, f2, f12
|
| 51 |
* fpsub16 f12, f10, f10
|
| 52 |
*/
|
| 53 |
|
| 54 |
#define DUP4(x) {x, x, x, x}
|
| 55 |
#define DUP8(x) {x, x, x, x, x, x, x, x}
|
| 56 |
DECLARE_ALIGNED(8, static const int16_t, constants1)[] = DUP4 (1); |
| 57 |
DECLARE_ALIGNED(8, static const int16_t, constants2)[] = DUP4 (2); |
| 58 |
DECLARE_ALIGNED(8, static const int16_t, constants3)[] = DUP4 (3); |
| 59 |
DECLARE_ALIGNED(8, static const int16_t, constants6)[] = DUP4 (6); |
| 60 |
DECLARE_ALIGNED(8, static const int8_t, constants_fe)[] = DUP8 (0xfe); |
| 61 |
DECLARE_ALIGNED(8, static const int8_t, constants_7f)[] = DUP8 (0x7f); |
| 62 |
DECLARE_ALIGNED(8, static const int8_t, constants128)[] = DUP8 (128); |
| 63 |
DECLARE_ALIGNED(8, static const int16_t, constants256_512)[] = |
| 64 |
{256, 512, 256, 512};
|
| 65 |
DECLARE_ALIGNED(8, static const int16_t, constants256_1024)[] = |
| 66 |
{256, 1024, 256, 1024};
|
| 67 |
|
| 68 |
#define REF_0 0 |
| 69 |
#define REF_0_1 1 |
| 70 |
#define REF_2 2 |
| 71 |
#define REF_2_1 3 |
| 72 |
#define REF_4 4 |
| 73 |
#define REF_4_1 5 |
| 74 |
#define REF_6 6 |
| 75 |
#define REF_6_1 7 |
| 76 |
#define REF_S0 8 |
| 77 |
#define REF_S0_1 9 |
| 78 |
#define REF_S2 10 |
| 79 |
#define REF_S2_1 11 |
| 80 |
#define REF_S4 12 |
| 81 |
#define REF_S4_1 13 |
| 82 |
#define REF_S6 14 |
| 83 |
#define REF_S6_1 15 |
| 84 |
#define DST_0 16 |
| 85 |
#define DST_1 17 |
| 86 |
#define DST_2 18 |
| 87 |
#define DST_3 19 |
| 88 |
#define CONST_1 20 |
| 89 |
#define CONST_2 20 |
| 90 |
#define CONST_3 20 |
| 91 |
#define CONST_6 20 |
| 92 |
#define MASK_fe 20 |
| 93 |
#define CONST_128 22 |
| 94 |
#define CONST_256 22 |
| 95 |
#define CONST_512 22 |
| 96 |
#define CONST_1024 22 |
| 97 |
#define TMP0 24 |
| 98 |
#define TMP1 25 |
| 99 |
#define TMP2 26 |
| 100 |
#define TMP3 27 |
| 101 |
#define TMP4 28 |
| 102 |
#define TMP5 29 |
| 103 |
#define ZERO 30 |
| 104 |
#define MASK_7f 30 |
| 105 |
|
| 106 |
#define TMP6 32 |
| 107 |
#define TMP8 34 |
| 108 |
#define TMP10 36 |
| 109 |
#define TMP12 38 |
| 110 |
#define TMP14 40 |
| 111 |
#define TMP16 42 |
| 112 |
#define TMP18 44 |
| 113 |
#define TMP20 46 |
| 114 |
#define TMP22 48 |
| 115 |
#define TMP24 50 |
| 116 |
#define TMP26 52 |
| 117 |
#define TMP28 54 |
| 118 |
#define TMP30 56 |
| 119 |
#define TMP32 58 |
| 120 |
|
| 121 |
static void MC_put_o_16_vis (uint8_t * dest, const uint8_t * ref, |
| 122 |
const int stride, int height) |
| 123 |
{
|
| 124 |
ref = vis_alignaddr(ref); |
| 125 |
do { /* 5 cycles */ |
| 126 |
vis_ld64(ref[0], TMP0);
|
| 127 |
|
| 128 |
vis_ld64_2(ref, 8, TMP2);
|
| 129 |
|
| 130 |
vis_ld64_2(ref, 16, TMP4);
|
| 131 |
ref += stride; |
| 132 |
|
| 133 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 134 |
vis_st64(REF_0, dest[0]);
|
| 135 |
|
| 136 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 137 |
vis_st64_2(REF_2, dest, 8);
|
| 138 |
dest += stride; |
| 139 |
} while (--height);
|
| 140 |
} |
| 141 |
|
| 142 |
static void MC_put_o_8_vis (uint8_t * dest, const uint8_t * ref, |
| 143 |
const int stride, int height) |
| 144 |
{
|
| 145 |
ref = vis_alignaddr(ref); |
| 146 |
do { /* 4 cycles */ |
| 147 |
vis_ld64(ref[0], TMP0);
|
| 148 |
|
| 149 |
vis_ld64(ref[8], TMP2);
|
| 150 |
ref += stride; |
| 151 |
|
| 152 |
/* stall */
|
| 153 |
|
| 154 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 155 |
vis_st64(REF_0, dest[0]);
|
| 156 |
dest += stride; |
| 157 |
} while (--height);
|
| 158 |
} |
| 159 |
|
| 160 |
|
| 161 |
static void MC_avg_o_16_vis (uint8_t * dest, const uint8_t * ref, |
| 162 |
const int stride, int height) |
| 163 |
{
|
| 164 |
int stride_8 = stride + 8; |
| 165 |
|
| 166 |
ref = vis_alignaddr(ref); |
| 167 |
|
| 168 |
vis_ld64(ref[0], TMP0);
|
| 169 |
|
| 170 |
vis_ld64(ref[8], TMP2);
|
| 171 |
|
| 172 |
vis_ld64(ref[16], TMP4);
|
| 173 |
|
| 174 |
vis_ld64(dest[0], DST_0);
|
| 175 |
|
| 176 |
vis_ld64(dest[8], DST_2);
|
| 177 |
|
| 178 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 179 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 180 |
|
| 181 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 182 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 183 |
|
| 184 |
vis_ld64(constants128[0], CONST_128);
|
| 185 |
|
| 186 |
ref += stride; |
| 187 |
height = (height >> 1) - 1; |
| 188 |
|
| 189 |
do { /* 24 cycles */ |
| 190 |
vis_ld64(ref[0], TMP0);
|
| 191 |
vis_xor(DST_0, REF_0, TMP6); |
| 192 |
|
| 193 |
vis_ld64_2(ref, 8, TMP2);
|
| 194 |
vis_and(TMP6, MASK_fe, TMP6); |
| 195 |
|
| 196 |
vis_ld64_2(ref, 16, TMP4);
|
| 197 |
ref += stride; |
| 198 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 199 |
vis_xor(DST_2, REF_2, TMP8); |
| 200 |
|
| 201 |
vis_and(TMP8, MASK_fe, TMP8); |
| 202 |
|
| 203 |
vis_or(DST_0, REF_0, TMP10); |
| 204 |
vis_ld64_2(dest, stride, DST_0); |
| 205 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 206 |
|
| 207 |
vis_or(DST_2, REF_2, TMP12); |
| 208 |
vis_ld64_2(dest, stride_8, DST_2); |
| 209 |
|
| 210 |
vis_ld64(ref[0], TMP14);
|
| 211 |
vis_and(TMP6, MASK_7f, TMP6); |
| 212 |
|
| 213 |
vis_and(TMP8, MASK_7f, TMP8); |
| 214 |
|
| 215 |
vis_psub16(TMP10, TMP6, TMP6); |
| 216 |
vis_st64(TMP6, dest[0]);
|
| 217 |
|
| 218 |
vis_psub16(TMP12, TMP8, TMP8); |
| 219 |
vis_st64_2(TMP8, dest, 8);
|
| 220 |
|
| 221 |
dest += stride; |
| 222 |
vis_ld64_2(ref, 8, TMP16);
|
| 223 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 224 |
|
| 225 |
vis_ld64_2(ref, 16, TMP18);
|
| 226 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 227 |
ref += stride; |
| 228 |
|
| 229 |
vis_xor(DST_0, REF_0, TMP20); |
| 230 |
|
| 231 |
vis_and(TMP20, MASK_fe, TMP20); |
| 232 |
|
| 233 |
vis_xor(DST_2, REF_2, TMP22); |
| 234 |
vis_mul8x16(CONST_128, TMP20, TMP20); |
| 235 |
|
| 236 |
vis_and(TMP22, MASK_fe, TMP22); |
| 237 |
|
| 238 |
vis_or(DST_0, REF_0, TMP24); |
| 239 |
vis_mul8x16(CONST_128, TMP22, TMP22); |
| 240 |
|
| 241 |
vis_or(DST_2, REF_2, TMP26); |
| 242 |
|
| 243 |
vis_ld64_2(dest, stride, DST_0); |
| 244 |
vis_faligndata(TMP14, TMP16, REF_0); |
| 245 |
|
| 246 |
vis_ld64_2(dest, stride_8, DST_2); |
| 247 |
vis_faligndata(TMP16, TMP18, REF_2); |
| 248 |
|
| 249 |
vis_and(TMP20, MASK_7f, TMP20); |
| 250 |
|
| 251 |
vis_and(TMP22, MASK_7f, TMP22); |
| 252 |
|
| 253 |
vis_psub16(TMP24, TMP20, TMP20); |
| 254 |
vis_st64(TMP20, dest[0]);
|
| 255 |
|
| 256 |
vis_psub16(TMP26, TMP22, TMP22); |
| 257 |
vis_st64_2(TMP22, dest, 8);
|
| 258 |
dest += stride; |
| 259 |
} while (--height);
|
| 260 |
|
| 261 |
vis_ld64(ref[0], TMP0);
|
| 262 |
vis_xor(DST_0, REF_0, TMP6); |
| 263 |
|
| 264 |
vis_ld64_2(ref, 8, TMP2);
|
| 265 |
vis_and(TMP6, MASK_fe, TMP6); |
| 266 |
|
| 267 |
vis_ld64_2(ref, 16, TMP4);
|
| 268 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 269 |
vis_xor(DST_2, REF_2, TMP8); |
| 270 |
|
| 271 |
vis_and(TMP8, MASK_fe, TMP8); |
| 272 |
|
| 273 |
vis_or(DST_0, REF_0, TMP10); |
| 274 |
vis_ld64_2(dest, stride, DST_0); |
| 275 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 276 |
|
| 277 |
vis_or(DST_2, REF_2, TMP12); |
| 278 |
vis_ld64_2(dest, stride_8, DST_2); |
| 279 |
|
| 280 |
vis_ld64(ref[0], TMP14);
|
| 281 |
vis_and(TMP6, MASK_7f, TMP6); |
| 282 |
|
| 283 |
vis_and(TMP8, MASK_7f, TMP8); |
| 284 |
|
| 285 |
vis_psub16(TMP10, TMP6, TMP6); |
| 286 |
vis_st64(TMP6, dest[0]);
|
| 287 |
|
| 288 |
vis_psub16(TMP12, TMP8, TMP8); |
| 289 |
vis_st64_2(TMP8, dest, 8);
|
| 290 |
|
| 291 |
dest += stride; |
| 292 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 293 |
|
| 294 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 295 |
|
| 296 |
vis_xor(DST_0, REF_0, TMP20); |
| 297 |
|
| 298 |
vis_and(TMP20, MASK_fe, TMP20); |
| 299 |
|
| 300 |
vis_xor(DST_2, REF_2, TMP22); |
| 301 |
vis_mul8x16(CONST_128, TMP20, TMP20); |
| 302 |
|
| 303 |
vis_and(TMP22, MASK_fe, TMP22); |
| 304 |
|
| 305 |
vis_or(DST_0, REF_0, TMP24); |
| 306 |
vis_mul8x16(CONST_128, TMP22, TMP22); |
| 307 |
|
| 308 |
vis_or(DST_2, REF_2, TMP26); |
| 309 |
|
| 310 |
vis_and(TMP20, MASK_7f, TMP20); |
| 311 |
|
| 312 |
vis_and(TMP22, MASK_7f, TMP22); |
| 313 |
|
| 314 |
vis_psub16(TMP24, TMP20, TMP20); |
| 315 |
vis_st64(TMP20, dest[0]);
|
| 316 |
|
| 317 |
vis_psub16(TMP26, TMP22, TMP22); |
| 318 |
vis_st64_2(TMP22, dest, 8);
|
| 319 |
} |
| 320 |
|
| 321 |
static void MC_avg_o_8_vis (uint8_t * dest, const uint8_t * ref, |
| 322 |
const int stride, int height) |
| 323 |
{
|
| 324 |
ref = vis_alignaddr(ref); |
| 325 |
|
| 326 |
vis_ld64(ref[0], TMP0);
|
| 327 |
|
| 328 |
vis_ld64(ref[8], TMP2);
|
| 329 |
|
| 330 |
vis_ld64(dest[0], DST_0);
|
| 331 |
|
| 332 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 333 |
|
| 334 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 335 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 336 |
|
| 337 |
vis_ld64(constants128[0], CONST_128);
|
| 338 |
|
| 339 |
ref += stride; |
| 340 |
height = (height >> 1) - 1; |
| 341 |
|
| 342 |
do { /* 12 cycles */ |
| 343 |
vis_ld64(ref[0], TMP0);
|
| 344 |
vis_xor(DST_0, REF_0, TMP4); |
| 345 |
|
| 346 |
vis_ld64(ref[8], TMP2);
|
| 347 |
vis_and(TMP4, MASK_fe, TMP4); |
| 348 |
|
| 349 |
vis_or(DST_0, REF_0, TMP6); |
| 350 |
vis_ld64_2(dest, stride, DST_0); |
| 351 |
ref += stride; |
| 352 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 353 |
|
| 354 |
vis_ld64(ref[0], TMP12);
|
| 355 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 356 |
|
| 357 |
vis_ld64(ref[8], TMP2);
|
| 358 |
vis_xor(DST_0, REF_0, TMP0); |
| 359 |
ref += stride; |
| 360 |
|
| 361 |
vis_and(TMP0, MASK_fe, TMP0); |
| 362 |
|
| 363 |
vis_and(TMP4, MASK_7f, TMP4); |
| 364 |
|
| 365 |
vis_psub16(TMP6, TMP4, TMP4); |
| 366 |
vis_st64(TMP4, dest[0]);
|
| 367 |
dest += stride; |
| 368 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 369 |
|
| 370 |
vis_or(DST_0, REF_0, TMP6); |
| 371 |
vis_ld64_2(dest, stride, DST_0); |
| 372 |
|
| 373 |
vis_faligndata(TMP12, TMP2, REF_0); |
| 374 |
|
| 375 |
vis_and(TMP0, MASK_7f, TMP0); |
| 376 |
|
| 377 |
vis_psub16(TMP6, TMP0, TMP4); |
| 378 |
vis_st64(TMP4, dest[0]);
|
| 379 |
dest += stride; |
| 380 |
} while (--height);
|
| 381 |
|
| 382 |
vis_ld64(ref[0], TMP0);
|
| 383 |
vis_xor(DST_0, REF_0, TMP4); |
| 384 |
|
| 385 |
vis_ld64(ref[8], TMP2);
|
| 386 |
vis_and(TMP4, MASK_fe, TMP4); |
| 387 |
|
| 388 |
vis_or(DST_0, REF_0, TMP6); |
| 389 |
vis_ld64_2(dest, stride, DST_0); |
| 390 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 391 |
|
| 392 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 393 |
|
| 394 |
vis_xor(DST_0, REF_0, TMP0); |
| 395 |
|
| 396 |
vis_and(TMP0, MASK_fe, TMP0); |
| 397 |
|
| 398 |
vis_and(TMP4, MASK_7f, TMP4); |
| 399 |
|
| 400 |
vis_psub16(TMP6, TMP4, TMP4); |
| 401 |
vis_st64(TMP4, dest[0]);
|
| 402 |
dest += stride; |
| 403 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 404 |
|
| 405 |
vis_or(DST_0, REF_0, TMP6); |
| 406 |
|
| 407 |
vis_and(TMP0, MASK_7f, TMP0); |
| 408 |
|
| 409 |
vis_psub16(TMP6, TMP0, TMP4); |
| 410 |
vis_st64(TMP4, dest[0]);
|
| 411 |
} |
| 412 |
|
| 413 |
static void MC_put_x_16_vis (uint8_t * dest, const uint8_t * ref, |
| 414 |
const int stride, int height) |
| 415 |
{
|
| 416 |
unsigned long off = (unsigned long) ref & 0x7; |
| 417 |
unsigned long off_plus_1 = off + 1; |
| 418 |
|
| 419 |
ref = vis_alignaddr(ref); |
| 420 |
|
| 421 |
vis_ld64(ref[0], TMP0);
|
| 422 |
|
| 423 |
vis_ld64_2(ref, 8, TMP2);
|
| 424 |
|
| 425 |
vis_ld64_2(ref, 16, TMP4);
|
| 426 |
|
| 427 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 428 |
|
| 429 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 430 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 431 |
|
| 432 |
vis_ld64(constants128[0], CONST_128);
|
| 433 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 434 |
|
| 435 |
if (off != 0x7) { |
| 436 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 437 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 438 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 439 |
} else {
|
| 440 |
vis_src1(TMP2, REF_2); |
| 441 |
vis_src1(TMP4, REF_6); |
| 442 |
} |
| 443 |
|
| 444 |
ref += stride; |
| 445 |
height = (height >> 1) - 1; |
| 446 |
|
| 447 |
do { /* 34 cycles */ |
| 448 |
vis_ld64(ref[0], TMP0);
|
| 449 |
vis_xor(REF_0, REF_2, TMP6); |
| 450 |
|
| 451 |
vis_ld64_2(ref, 8, TMP2);
|
| 452 |
vis_xor(REF_4, REF_6, TMP8); |
| 453 |
|
| 454 |
vis_ld64_2(ref, 16, TMP4);
|
| 455 |
vis_and(TMP6, MASK_fe, TMP6); |
| 456 |
ref += stride; |
| 457 |
|
| 458 |
vis_ld64(ref[0], TMP14);
|
| 459 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 460 |
vis_and(TMP8, MASK_fe, TMP8); |
| 461 |
|
| 462 |
vis_ld64_2(ref, 8, TMP16);
|
| 463 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 464 |
vis_or(REF_0, REF_2, TMP10); |
| 465 |
|
| 466 |
vis_ld64_2(ref, 16, TMP18);
|
| 467 |
ref += stride; |
| 468 |
vis_or(REF_4, REF_6, TMP12); |
| 469 |
|
| 470 |
vis_alignaddr_g0((void *)off);
|
| 471 |
|
| 472 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 473 |
|
| 474 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 475 |
|
| 476 |
if (off != 0x7) { |
| 477 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 478 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 479 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 480 |
} else {
|
| 481 |
vis_src1(TMP2, REF_2); |
| 482 |
vis_src1(TMP4, REF_6); |
| 483 |
} |
| 484 |
|
| 485 |
vis_and(TMP6, MASK_7f, TMP6); |
| 486 |
|
| 487 |
vis_and(TMP8, MASK_7f, TMP8); |
| 488 |
|
| 489 |
vis_psub16(TMP10, TMP6, TMP6); |
| 490 |
vis_st64(TMP6, dest[0]);
|
| 491 |
|
| 492 |
vis_psub16(TMP12, TMP8, TMP8); |
| 493 |
vis_st64_2(TMP8, dest, 8);
|
| 494 |
dest += stride; |
| 495 |
|
| 496 |
vis_xor(REF_0, REF_2, TMP6); |
| 497 |
|
| 498 |
vis_xor(REF_4, REF_6, TMP8); |
| 499 |
|
| 500 |
vis_and(TMP6, MASK_fe, TMP6); |
| 501 |
|
| 502 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 503 |
vis_and(TMP8, MASK_fe, TMP8); |
| 504 |
|
| 505 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 506 |
vis_or(REF_0, REF_2, TMP10); |
| 507 |
|
| 508 |
vis_or(REF_4, REF_6, TMP12); |
| 509 |
|
| 510 |
vis_alignaddr_g0((void *)off);
|
| 511 |
|
| 512 |
vis_faligndata(TMP14, TMP16, REF_0); |
| 513 |
|
| 514 |
vis_faligndata(TMP16, TMP18, REF_4); |
| 515 |
|
| 516 |
if (off != 0x7) { |
| 517 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 518 |
vis_faligndata(TMP14, TMP16, REF_2); |
| 519 |
vis_faligndata(TMP16, TMP18, REF_6); |
| 520 |
} else {
|
| 521 |
vis_src1(TMP16, REF_2); |
| 522 |
vis_src1(TMP18, REF_6); |
| 523 |
} |
| 524 |
|
| 525 |
vis_and(TMP6, MASK_7f, TMP6); |
| 526 |
|
| 527 |
vis_and(TMP8, MASK_7f, TMP8); |
| 528 |
|
| 529 |
vis_psub16(TMP10, TMP6, TMP6); |
| 530 |
vis_st64(TMP6, dest[0]);
|
| 531 |
|
| 532 |
vis_psub16(TMP12, TMP8, TMP8); |
| 533 |
vis_st64_2(TMP8, dest, 8);
|
| 534 |
dest += stride; |
| 535 |
} while (--height);
|
| 536 |
|
| 537 |
vis_ld64(ref[0], TMP0);
|
| 538 |
vis_xor(REF_0, REF_2, TMP6); |
| 539 |
|
| 540 |
vis_ld64_2(ref, 8, TMP2);
|
| 541 |
vis_xor(REF_4, REF_6, TMP8); |
| 542 |
|
| 543 |
vis_ld64_2(ref, 16, TMP4);
|
| 544 |
vis_and(TMP6, MASK_fe, TMP6); |
| 545 |
|
| 546 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 547 |
vis_and(TMP8, MASK_fe, TMP8); |
| 548 |
|
| 549 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 550 |
vis_or(REF_0, REF_2, TMP10); |
| 551 |
|
| 552 |
vis_or(REF_4, REF_6, TMP12); |
| 553 |
|
| 554 |
vis_alignaddr_g0((void *)off);
|
| 555 |
|
| 556 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 557 |
|
| 558 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 559 |
|
| 560 |
if (off != 0x7) { |
| 561 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 562 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 563 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 564 |
} else {
|
| 565 |
vis_src1(TMP2, REF_2); |
| 566 |
vis_src1(TMP4, REF_6); |
| 567 |
} |
| 568 |
|
| 569 |
vis_and(TMP6, MASK_7f, TMP6); |
| 570 |
|
| 571 |
vis_and(TMP8, MASK_7f, TMP8); |
| 572 |
|
| 573 |
vis_psub16(TMP10, TMP6, TMP6); |
| 574 |
vis_st64(TMP6, dest[0]);
|
| 575 |
|
| 576 |
vis_psub16(TMP12, TMP8, TMP8); |
| 577 |
vis_st64_2(TMP8, dest, 8);
|
| 578 |
dest += stride; |
| 579 |
|
| 580 |
vis_xor(REF_0, REF_2, TMP6); |
| 581 |
|
| 582 |
vis_xor(REF_4, REF_6, TMP8); |
| 583 |
|
| 584 |
vis_and(TMP6, MASK_fe, TMP6); |
| 585 |
|
| 586 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 587 |
vis_and(TMP8, MASK_fe, TMP8); |
| 588 |
|
| 589 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 590 |
vis_or(REF_0, REF_2, TMP10); |
| 591 |
|
| 592 |
vis_or(REF_4, REF_6, TMP12); |
| 593 |
|
| 594 |
vis_and(TMP6, MASK_7f, TMP6); |
| 595 |
|
| 596 |
vis_and(TMP8, MASK_7f, TMP8); |
| 597 |
|
| 598 |
vis_psub16(TMP10, TMP6, TMP6); |
| 599 |
vis_st64(TMP6, dest[0]);
|
| 600 |
|
| 601 |
vis_psub16(TMP12, TMP8, TMP8); |
| 602 |
vis_st64_2(TMP8, dest, 8);
|
| 603 |
} |
| 604 |
|
| 605 |
static void MC_put_x_8_vis (uint8_t * dest, const uint8_t * ref, |
| 606 |
const int stride, int height) |
| 607 |
{
|
| 608 |
unsigned long off = (unsigned long) ref & 0x7; |
| 609 |
unsigned long off_plus_1 = off + 1; |
| 610 |
|
| 611 |
ref = vis_alignaddr(ref); |
| 612 |
|
| 613 |
vis_ld64(ref[0], TMP0);
|
| 614 |
|
| 615 |
vis_ld64(ref[8], TMP2);
|
| 616 |
|
| 617 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 618 |
|
| 619 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 620 |
|
| 621 |
vis_ld64(constants128[0], CONST_128);
|
| 622 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 623 |
|
| 624 |
if (off != 0x7) { |
| 625 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 626 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 627 |
} else {
|
| 628 |
vis_src1(TMP2, REF_2); |
| 629 |
} |
| 630 |
|
| 631 |
ref += stride; |
| 632 |
height = (height >> 1) - 1; |
| 633 |
|
| 634 |
do { /* 20 cycles */ |
| 635 |
vis_ld64(ref[0], TMP0);
|
| 636 |
vis_xor(REF_0, REF_2, TMP4); |
| 637 |
|
| 638 |
vis_ld64_2(ref, 8, TMP2);
|
| 639 |
vis_and(TMP4, MASK_fe, TMP4); |
| 640 |
ref += stride; |
| 641 |
|
| 642 |
vis_ld64(ref[0], TMP8);
|
| 643 |
vis_or(REF_0, REF_2, TMP6); |
| 644 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 645 |
|
| 646 |
vis_alignaddr_g0((void *)off);
|
| 647 |
|
| 648 |
vis_ld64_2(ref, 8, TMP10);
|
| 649 |
ref += stride; |
| 650 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 651 |
|
| 652 |
if (off != 0x7) { |
| 653 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 654 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 655 |
} else {
|
| 656 |
vis_src1(TMP2, REF_2); |
| 657 |
} |
| 658 |
|
| 659 |
vis_and(TMP4, MASK_7f, TMP4); |
| 660 |
|
| 661 |
vis_psub16(TMP6, TMP4, DST_0); |
| 662 |
vis_st64(DST_0, dest[0]);
|
| 663 |
dest += stride; |
| 664 |
|
| 665 |
vis_xor(REF_0, REF_2, TMP12); |
| 666 |
|
| 667 |
vis_and(TMP12, MASK_fe, TMP12); |
| 668 |
|
| 669 |
vis_or(REF_0, REF_2, TMP14); |
| 670 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 671 |
|
| 672 |
vis_alignaddr_g0((void *)off);
|
| 673 |
vis_faligndata(TMP8, TMP10, REF_0); |
| 674 |
if (off != 0x7) { |
| 675 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 676 |
vis_faligndata(TMP8, TMP10, REF_2); |
| 677 |
} else {
|
| 678 |
vis_src1(TMP10, REF_2); |
| 679 |
} |
| 680 |
|
| 681 |
vis_and(TMP12, MASK_7f, TMP12); |
| 682 |
|
| 683 |
vis_psub16(TMP14, TMP12, DST_0); |
| 684 |
vis_st64(DST_0, dest[0]);
|
| 685 |
dest += stride; |
| 686 |
} while (--height);
|
| 687 |
|
| 688 |
vis_ld64(ref[0], TMP0);
|
| 689 |
vis_xor(REF_0, REF_2, TMP4); |
| 690 |
|
| 691 |
vis_ld64_2(ref, 8, TMP2);
|
| 692 |
vis_and(TMP4, MASK_fe, TMP4); |
| 693 |
|
| 694 |
vis_or(REF_0, REF_2, TMP6); |
| 695 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 696 |
|
| 697 |
vis_alignaddr_g0((void *)off);
|
| 698 |
|
| 699 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 700 |
|
| 701 |
if (off != 0x7) { |
| 702 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 703 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 704 |
} else {
|
| 705 |
vis_src1(TMP2, REF_2); |
| 706 |
} |
| 707 |
|
| 708 |
vis_and(TMP4, MASK_7f, TMP4); |
| 709 |
|
| 710 |
vis_psub16(TMP6, TMP4, DST_0); |
| 711 |
vis_st64(DST_0, dest[0]);
|
| 712 |
dest += stride; |
| 713 |
|
| 714 |
vis_xor(REF_0, REF_2, TMP12); |
| 715 |
|
| 716 |
vis_and(TMP12, MASK_fe, TMP12); |
| 717 |
|
| 718 |
vis_or(REF_0, REF_2, TMP14); |
| 719 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 720 |
|
| 721 |
vis_and(TMP12, MASK_7f, TMP12); |
| 722 |
|
| 723 |
vis_psub16(TMP14, TMP12, DST_0); |
| 724 |
vis_st64(DST_0, dest[0]);
|
| 725 |
dest += stride; |
| 726 |
} |
| 727 |
|
| 728 |
static void MC_avg_x_16_vis (uint8_t * dest, const uint8_t * ref, |
| 729 |
const int stride, int height) |
| 730 |
{
|
| 731 |
unsigned long off = (unsigned long) ref & 0x7; |
| 732 |
unsigned long off_plus_1 = off + 1; |
| 733 |
|
| 734 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 735 |
|
| 736 |
vis_ld64(constants3[0], CONST_3);
|
| 737 |
vis_fzero(ZERO); |
| 738 |
vis_ld64(constants256_512[0], CONST_256);
|
| 739 |
|
| 740 |
ref = vis_alignaddr(ref); |
| 741 |
do { /* 26 cycles */ |
| 742 |
vis_ld64(ref[0], TMP0);
|
| 743 |
|
| 744 |
vis_ld64(ref[8], TMP2);
|
| 745 |
|
| 746 |
vis_alignaddr_g0((void *)off);
|
| 747 |
|
| 748 |
vis_ld64(ref[16], TMP4);
|
| 749 |
|
| 750 |
vis_ld64(dest[0], DST_0);
|
| 751 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 752 |
|
| 753 |
vis_ld64(dest[8], DST_2);
|
| 754 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 755 |
|
| 756 |
if (off != 0x7) { |
| 757 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 758 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 759 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 760 |
} else {
|
| 761 |
vis_src1(TMP2, REF_2); |
| 762 |
vis_src1(TMP4, REF_6); |
| 763 |
} |
| 764 |
|
| 765 |
vis_mul8x16au(REF_0, CONST_256, TMP0); |
| 766 |
|
| 767 |
vis_pmerge(ZERO, REF_2, TMP4); |
| 768 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 769 |
|
| 770 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 771 |
|
| 772 |
vis_padd16(TMP0, TMP4, TMP0); |
| 773 |
|
| 774 |
vis_mul8x16al(DST_0, CONST_512, TMP4); |
| 775 |
vis_padd16(TMP2, TMP6, TMP2); |
| 776 |
|
| 777 |
vis_mul8x16al(DST_1, CONST_512, TMP6); |
| 778 |
|
| 779 |
vis_mul8x16au(REF_6, CONST_256, TMP12); |
| 780 |
|
| 781 |
vis_padd16(TMP0, TMP4, TMP0); |
| 782 |
vis_mul8x16au(REF_6_1, CONST_256, TMP14); |
| 783 |
|
| 784 |
vis_padd16(TMP2, TMP6, TMP2); |
| 785 |
vis_mul8x16au(REF_4, CONST_256, TMP16); |
| 786 |
|
| 787 |
vis_padd16(TMP0, CONST_3, TMP8); |
| 788 |
vis_mul8x16au(REF_4_1, CONST_256, TMP18); |
| 789 |
|
| 790 |
vis_padd16(TMP2, CONST_3, TMP10); |
| 791 |
vis_pack16(TMP8, DST_0); |
| 792 |
|
| 793 |
vis_pack16(TMP10, DST_1); |
| 794 |
vis_padd16(TMP16, TMP12, TMP0); |
| 795 |
|
| 796 |
vis_st64(DST_0, dest[0]);
|
| 797 |
vis_mul8x16al(DST_2, CONST_512, TMP4); |
| 798 |
vis_padd16(TMP18, TMP14, TMP2); |
| 799 |
|
| 800 |
vis_mul8x16al(DST_3, CONST_512, TMP6); |
| 801 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 802 |
|
| 803 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 804 |
|
| 805 |
vis_padd16(TMP0, TMP4, TMP0); |
| 806 |
|
| 807 |
vis_padd16(TMP2, TMP6, TMP2); |
| 808 |
vis_pack16(TMP0, DST_2); |
| 809 |
|
| 810 |
vis_pack16(TMP2, DST_3); |
| 811 |
vis_st64(DST_2, dest[8]);
|
| 812 |
|
| 813 |
ref += stride; |
| 814 |
dest += stride; |
| 815 |
} while (--height);
|
| 816 |
} |
| 817 |
|
| 818 |
static void MC_avg_x_8_vis (uint8_t * dest, const uint8_t * ref, |
| 819 |
const int stride, int height) |
| 820 |
{
|
| 821 |
unsigned long off = (unsigned long) ref & 0x7; |
| 822 |
unsigned long off_plus_1 = off + 1; |
| 823 |
int stride_times_2 = stride << 1; |
| 824 |
|
| 825 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 826 |
|
| 827 |
vis_ld64(constants3[0], CONST_3);
|
| 828 |
vis_fzero(ZERO); |
| 829 |
vis_ld64(constants256_512[0], CONST_256);
|
| 830 |
|
| 831 |
ref = vis_alignaddr(ref); |
| 832 |
height >>= 2;
|
| 833 |
do { /* 47 cycles */ |
| 834 |
vis_ld64(ref[0], TMP0);
|
| 835 |
|
| 836 |
vis_ld64_2(ref, 8, TMP2);
|
| 837 |
ref += stride; |
| 838 |
|
| 839 |
vis_alignaddr_g0((void *)off);
|
| 840 |
|
| 841 |
vis_ld64(ref[0], TMP4);
|
| 842 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 843 |
|
| 844 |
vis_ld64_2(ref, 8, TMP6);
|
| 845 |
ref += stride; |
| 846 |
|
| 847 |
vis_ld64(ref[0], TMP8);
|
| 848 |
|
| 849 |
vis_ld64_2(ref, 8, TMP10);
|
| 850 |
ref += stride; |
| 851 |
vis_faligndata(TMP4, TMP6, REF_4); |
| 852 |
|
| 853 |
vis_ld64(ref[0], TMP12);
|
| 854 |
|
| 855 |
vis_ld64_2(ref, 8, TMP14);
|
| 856 |
ref += stride; |
| 857 |
vis_faligndata(TMP8, TMP10, REF_S0); |
| 858 |
|
| 859 |
vis_faligndata(TMP12, TMP14, REF_S4); |
| 860 |
|
| 861 |
if (off != 0x7) { |
| 862 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 863 |
|
| 864 |
vis_ld64(dest[0], DST_0);
|
| 865 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 866 |
|
| 867 |
vis_ld64_2(dest, stride, DST_2); |
| 868 |
vis_faligndata(TMP4, TMP6, REF_6); |
| 869 |
|
| 870 |
vis_faligndata(TMP8, TMP10, REF_S2); |
| 871 |
|
| 872 |
vis_faligndata(TMP12, TMP14, REF_S6); |
| 873 |
} else {
|
| 874 |
vis_ld64(dest[0], DST_0);
|
| 875 |
vis_src1(TMP2, REF_2); |
| 876 |
|
| 877 |
vis_ld64_2(dest, stride, DST_2); |
| 878 |
vis_src1(TMP6, REF_6); |
| 879 |
|
| 880 |
vis_src1(TMP10, REF_S2); |
| 881 |
|
| 882 |
vis_src1(TMP14, REF_S6); |
| 883 |
} |
| 884 |
|
| 885 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 886 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 887 |
|
| 888 |
vis_pmerge(ZERO, REF_2, TMP4); |
| 889 |
vis_mul8x16au(REF_2_1, CONST_256, TMP6); |
| 890 |
|
| 891 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 892 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 893 |
|
| 894 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 895 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 896 |
|
| 897 |
vis_padd16(TMP0, TMP4, TMP0); |
| 898 |
vis_mul8x16au(REF_4, CONST_256, TMP8); |
| 899 |
|
| 900 |
vis_padd16(TMP2, TMP6, TMP2); |
| 901 |
vis_mul8x16au(REF_4_1, CONST_256, TMP10); |
| 902 |
|
| 903 |
vis_padd16(TMP0, TMP16, TMP0); |
| 904 |
vis_mul8x16au(REF_6, CONST_256, TMP12); |
| 905 |
|
| 906 |
vis_padd16(TMP2, TMP18, TMP2); |
| 907 |
vis_mul8x16au(REF_6_1, CONST_256, TMP14); |
| 908 |
|
| 909 |
vis_padd16(TMP8, CONST_3, TMP8); |
| 910 |
vis_mul8x16al(DST_2, CONST_512, TMP16); |
| 911 |
|
| 912 |
vis_padd16(TMP8, TMP12, TMP8); |
| 913 |
vis_mul8x16al(DST_3, CONST_512, TMP18); |
| 914 |
|
| 915 |
vis_padd16(TMP10, TMP14, TMP10); |
| 916 |
vis_pack16(TMP0, DST_0); |
| 917 |
|
| 918 |
vis_pack16(TMP2, DST_1); |
| 919 |
vis_st64(DST_0, dest[0]);
|
| 920 |
dest += stride; |
| 921 |
vis_padd16(TMP10, CONST_3, TMP10); |
| 922 |
|
| 923 |
vis_ld64_2(dest, stride, DST_0); |
| 924 |
vis_padd16(TMP8, TMP16, TMP8); |
| 925 |
|
| 926 |
vis_ld64_2(dest, stride_times_2, TMP4/*DST_2*/);
|
| 927 |
vis_padd16(TMP10, TMP18, TMP10); |
| 928 |
vis_pack16(TMP8, DST_2); |
| 929 |
|
| 930 |
vis_pack16(TMP10, DST_3); |
| 931 |
vis_st64(DST_2, dest[0]);
|
| 932 |
dest += stride; |
| 933 |
|
| 934 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP2); |
| 935 |
vis_pmerge(ZERO, REF_S0, TMP0); |
| 936 |
|
| 937 |
vis_pmerge(ZERO, REF_S2, TMP24); |
| 938 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP6); |
| 939 |
|
| 940 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 941 |
vis_mul8x16au(REF_S4, CONST_256, TMP8); |
| 942 |
|
| 943 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 944 |
vis_mul8x16au(REF_S4_1, CONST_256, TMP10); |
| 945 |
|
| 946 |
vis_padd16(TMP0, TMP24, TMP0); |
| 947 |
vis_mul8x16au(REF_S6, CONST_256, TMP12); |
| 948 |
|
| 949 |
vis_padd16(TMP2, TMP6, TMP2); |
| 950 |
vis_mul8x16au(REF_S6_1, CONST_256, TMP14); |
| 951 |
|
| 952 |
vis_padd16(TMP8, CONST_3, TMP8); |
| 953 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 954 |
|
| 955 |
vis_padd16(TMP10, CONST_3, TMP10); |
| 956 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 957 |
|
| 958 |
vis_padd16(TMP8, TMP12, TMP8); |
| 959 |
vis_mul8x16al(TMP4/*DST_2*/, CONST_512, TMP20);
|
| 960 |
|
| 961 |
vis_mul8x16al(TMP5/*DST_3*/, CONST_512, TMP22);
|
| 962 |
vis_padd16(TMP0, TMP16, TMP0); |
| 963 |
|
| 964 |
vis_padd16(TMP2, TMP18, TMP2); |
| 965 |
vis_pack16(TMP0, DST_0); |
| 966 |
|
| 967 |
vis_padd16(TMP10, TMP14, TMP10); |
| 968 |
vis_pack16(TMP2, DST_1); |
| 969 |
vis_st64(DST_0, dest[0]);
|
| 970 |
dest += stride; |
| 971 |
|
| 972 |
vis_padd16(TMP8, TMP20, TMP8); |
| 973 |
|
| 974 |
vis_padd16(TMP10, TMP22, TMP10); |
| 975 |
vis_pack16(TMP8, DST_2); |
| 976 |
|
| 977 |
vis_pack16(TMP10, DST_3); |
| 978 |
vis_st64(DST_2, dest[0]);
|
| 979 |
dest += stride; |
| 980 |
} while (--height);
|
| 981 |
} |
| 982 |
|
| 983 |
static void MC_put_y_16_vis (uint8_t * dest, const uint8_t * ref, |
| 984 |
const int stride, int height) |
| 985 |
{
|
| 986 |
ref = vis_alignaddr(ref); |
| 987 |
vis_ld64(ref[0], TMP0);
|
| 988 |
|
| 989 |
vis_ld64_2(ref, 8, TMP2);
|
| 990 |
|
| 991 |
vis_ld64_2(ref, 16, TMP4);
|
| 992 |
ref += stride; |
| 993 |
|
| 994 |
vis_ld64(ref[0], TMP6);
|
| 995 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 996 |
|
| 997 |
vis_ld64_2(ref, 8, TMP8);
|
| 998 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 999 |
|
| 1000 |
vis_ld64_2(ref, 16, TMP10);
|
| 1001 |
ref += stride; |
| 1002 |
|
| 1003 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 1004 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 1005 |
|
| 1006 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 1007 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 1008 |
|
| 1009 |
vis_ld64(constants128[0], CONST_128);
|
| 1010 |
height = (height >> 1) - 1; |
| 1011 |
do { /* 24 cycles */ |
| 1012 |
vis_ld64(ref[0], TMP0);
|
| 1013 |
vis_xor(REF_0, REF_2, TMP12); |
| 1014 |
|
| 1015 |
vis_ld64_2(ref, 8, TMP2);
|
| 1016 |
vis_xor(REF_4, REF_6, TMP16); |
| 1017 |
|
| 1018 |
vis_ld64_2(ref, 16, TMP4);
|
| 1019 |
ref += stride; |
| 1020 |
vis_or(REF_0, REF_2, TMP14); |
| 1021 |
|
| 1022 |
vis_ld64(ref[0], TMP6);
|
| 1023 |
vis_or(REF_4, REF_6, TMP18); |
| 1024 |
|
| 1025 |
vis_ld64_2(ref, 8, TMP8);
|
| 1026 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1027 |
|
| 1028 |
vis_ld64_2(ref, 16, TMP10);
|
| 1029 |
ref += stride; |
| 1030 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 1031 |
|
| 1032 |
vis_and(TMP12, MASK_fe, TMP12); |
| 1033 |
|
| 1034 |
vis_and(TMP16, MASK_fe, TMP16); |
| 1035 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 1036 |
|
| 1037 |
vis_mul8x16(CONST_128, TMP16, TMP16); |
| 1038 |
vis_xor(REF_0, REF_2, TMP0); |
| 1039 |
|
| 1040 |
vis_xor(REF_4, REF_6, TMP2); |
| 1041 |
|
| 1042 |
vis_or(REF_0, REF_2, TMP20); |
| 1043 |
|
| 1044 |
vis_and(TMP12, MASK_7f, TMP12); |
| 1045 |
|
| 1046 |
vis_and(TMP16, MASK_7f, TMP16); |
| 1047 |
|
| 1048 |
vis_psub16(TMP14, TMP12, TMP12); |
| 1049 |
vis_st64(TMP12, dest[0]);
|
| 1050 |
|
| 1051 |
vis_psub16(TMP18, TMP16, TMP16); |
| 1052 |
vis_st64_2(TMP16, dest, 8);
|
| 1053 |
dest += stride; |
| 1054 |
|
| 1055 |
vis_or(REF_4, REF_6, TMP18); |
| 1056 |
|
| 1057 |
vis_and(TMP0, MASK_fe, TMP0); |
| 1058 |
|
| 1059 |
vis_and(TMP2, MASK_fe, TMP2); |
| 1060 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 1061 |
|
| 1062 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 1063 |
vis_mul8x16(CONST_128, TMP2, TMP2); |
| 1064 |
|
| 1065 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 1066 |
|
| 1067 |
vis_and(TMP0, MASK_7f, TMP0); |
| 1068 |
|
| 1069 |
vis_and(TMP2, MASK_7f, TMP2); |
| 1070 |
|
| 1071 |
vis_psub16(TMP20, TMP0, TMP0); |
| 1072 |
vis_st64(TMP0, dest[0]);
|
| 1073 |
|
| 1074 |
vis_psub16(TMP18, TMP2, TMP2); |
| 1075 |
vis_st64_2(TMP2, dest, 8);
|
| 1076 |
dest += stride; |
| 1077 |
} while (--height);
|
| 1078 |
|
| 1079 |
vis_ld64(ref[0], TMP0);
|
| 1080 |
vis_xor(REF_0, REF_2, TMP12); |
| 1081 |
|
| 1082 |
vis_ld64_2(ref, 8, TMP2);
|
| 1083 |
vis_xor(REF_4, REF_6, TMP16); |
| 1084 |
|
| 1085 |
vis_ld64_2(ref, 16, TMP4);
|
| 1086 |
vis_or(REF_0, REF_2, TMP14); |
| 1087 |
|
| 1088 |
vis_or(REF_4, REF_6, TMP18); |
| 1089 |
|
| 1090 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1091 |
|
| 1092 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 1093 |
|
| 1094 |
vis_and(TMP12, MASK_fe, TMP12); |
| 1095 |
|
| 1096 |
vis_and(TMP16, MASK_fe, TMP16); |
| 1097 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 1098 |
|
| 1099 |
vis_mul8x16(CONST_128, TMP16, TMP16); |
| 1100 |
vis_xor(REF_0, REF_2, TMP0); |
| 1101 |
|
| 1102 |
vis_xor(REF_4, REF_6, TMP2); |
| 1103 |
|
| 1104 |
vis_or(REF_0, REF_2, TMP20); |
| 1105 |
|
| 1106 |
vis_and(TMP12, MASK_7f, TMP12); |
| 1107 |
|
| 1108 |
vis_and(TMP16, MASK_7f, TMP16); |
| 1109 |
|
| 1110 |
vis_psub16(TMP14, TMP12, TMP12); |
| 1111 |
vis_st64(TMP12, dest[0]);
|
| 1112 |
|
| 1113 |
vis_psub16(TMP18, TMP16, TMP16); |
| 1114 |
vis_st64_2(TMP16, dest, 8);
|
| 1115 |
dest += stride; |
| 1116 |
|
| 1117 |
vis_or(REF_4, REF_6, TMP18); |
| 1118 |
|
| 1119 |
vis_and(TMP0, MASK_fe, TMP0); |
| 1120 |
|
| 1121 |
vis_and(TMP2, MASK_fe, TMP2); |
| 1122 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 1123 |
|
| 1124 |
vis_mul8x16(CONST_128, TMP2, TMP2); |
| 1125 |
|
| 1126 |
vis_and(TMP0, MASK_7f, TMP0); |
| 1127 |
|
| 1128 |
vis_and(TMP2, MASK_7f, TMP2); |
| 1129 |
|
| 1130 |
vis_psub16(TMP20, TMP0, TMP0); |
| 1131 |
vis_st64(TMP0, dest[0]);
|
| 1132 |
|
| 1133 |
vis_psub16(TMP18, TMP2, TMP2); |
| 1134 |
vis_st64_2(TMP2, dest, 8);
|
| 1135 |
} |
| 1136 |
|
| 1137 |
static void MC_put_y_8_vis (uint8_t * dest, const uint8_t * ref, |
| 1138 |
const int stride, int height) |
| 1139 |
{
|
| 1140 |
ref = vis_alignaddr(ref); |
| 1141 |
vis_ld64(ref[0], TMP0);
|
| 1142 |
|
| 1143 |
vis_ld64_2(ref, 8, TMP2);
|
| 1144 |
ref += stride; |
| 1145 |
|
| 1146 |
vis_ld64(ref[0], TMP4);
|
| 1147 |
|
| 1148 |
vis_ld64_2(ref, 8, TMP6);
|
| 1149 |
ref += stride; |
| 1150 |
|
| 1151 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 1152 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1153 |
|
| 1154 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 1155 |
vis_faligndata(TMP4, TMP6, REF_2); |
| 1156 |
|
| 1157 |
vis_ld64(constants128[0], CONST_128);
|
| 1158 |
height = (height >> 1) - 1; |
| 1159 |
do { /* 12 cycles */ |
| 1160 |
vis_ld64(ref[0], TMP0);
|
| 1161 |
vis_xor(REF_0, REF_2, TMP4); |
| 1162 |
|
| 1163 |
vis_ld64_2(ref, 8, TMP2);
|
| 1164 |
ref += stride; |
| 1165 |
vis_and(TMP4, MASK_fe, TMP4); |
| 1166 |
|
| 1167 |
vis_or(REF_0, REF_2, TMP6); |
| 1168 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 1169 |
|
| 1170 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1171 |
vis_ld64(ref[0], TMP0);
|
| 1172 |
|
| 1173 |
vis_ld64_2(ref, 8, TMP2);
|
| 1174 |
ref += stride; |
| 1175 |
vis_xor(REF_0, REF_2, TMP12); |
| 1176 |
|
| 1177 |
vis_and(TMP4, MASK_7f, TMP4); |
| 1178 |
|
| 1179 |
vis_and(TMP12, MASK_fe, TMP12); |
| 1180 |
|
| 1181 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 1182 |
vis_or(REF_0, REF_2, TMP14); |
| 1183 |
|
| 1184 |
vis_psub16(TMP6, TMP4, DST_0); |
| 1185 |
vis_st64(DST_0, dest[0]);
|
| 1186 |
dest += stride; |
| 1187 |
|
| 1188 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 1189 |
|
| 1190 |
vis_and(TMP12, MASK_7f, TMP12); |
| 1191 |
|
| 1192 |
vis_psub16(TMP14, TMP12, DST_0); |
| 1193 |
vis_st64(DST_0, dest[0]);
|
| 1194 |
dest += stride; |
| 1195 |
} while (--height);
|
| 1196 |
|
| 1197 |
vis_ld64(ref[0], TMP0);
|
| 1198 |
vis_xor(REF_0, REF_2, TMP4); |
| 1199 |
|
| 1200 |
vis_ld64_2(ref, 8, TMP2);
|
| 1201 |
vis_and(TMP4, MASK_fe, TMP4); |
| 1202 |
|
| 1203 |
vis_or(REF_0, REF_2, TMP6); |
| 1204 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 1205 |
|
| 1206 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1207 |
|
| 1208 |
vis_xor(REF_0, REF_2, TMP12); |
| 1209 |
|
| 1210 |
vis_and(TMP4, MASK_7f, TMP4); |
| 1211 |
|
| 1212 |
vis_and(TMP12, MASK_fe, TMP12); |
| 1213 |
|
| 1214 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 1215 |
vis_or(REF_0, REF_2, TMP14); |
| 1216 |
|
| 1217 |
vis_psub16(TMP6, TMP4, DST_0); |
| 1218 |
vis_st64(DST_0, dest[0]);
|
| 1219 |
dest += stride; |
| 1220 |
|
| 1221 |
vis_and(TMP12, MASK_7f, TMP12); |
| 1222 |
|
| 1223 |
vis_psub16(TMP14, TMP12, DST_0); |
| 1224 |
vis_st64(DST_0, dest[0]);
|
| 1225 |
} |
| 1226 |
|
| 1227 |
static void MC_avg_y_16_vis (uint8_t * dest, const uint8_t * ref, |
| 1228 |
const int stride, int height) |
| 1229 |
{
|
| 1230 |
int stride_8 = stride + 8; |
| 1231 |
int stride_16 = stride + 16; |
| 1232 |
|
| 1233 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1234 |
|
| 1235 |
ref = vis_alignaddr(ref); |
| 1236 |
|
| 1237 |
vis_ld64(ref[ 0], TMP0);
|
| 1238 |
vis_fzero(ZERO); |
| 1239 |
|
| 1240 |
vis_ld64(ref[ 8], TMP2);
|
| 1241 |
|
| 1242 |
vis_ld64(ref[16], TMP4);
|
| 1243 |
|
| 1244 |
vis_ld64(constants3[0], CONST_3);
|
| 1245 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 1246 |
|
| 1247 |
vis_ld64(constants256_512[0], CONST_256);
|
| 1248 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 1249 |
height >>= 1;
|
| 1250 |
|
| 1251 |
do { /* 31 cycles */ |
| 1252 |
vis_ld64_2(ref, stride, TMP0); |
| 1253 |
vis_pmerge(ZERO, REF_2, TMP12); |
| 1254 |
vis_mul8x16au(REF_2_1, CONST_256, TMP14); |
| 1255 |
|
| 1256 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1257 |
vis_pmerge(ZERO, REF_6, TMP16); |
| 1258 |
vis_mul8x16au(REF_6_1, CONST_256, TMP18); |
| 1259 |
|
| 1260 |
vis_ld64_2(ref, stride_16, TMP4); |
| 1261 |
ref += stride; |
| 1262 |
|
| 1263 |
vis_ld64(dest[0], DST_0);
|
| 1264 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1265 |
|
| 1266 |
vis_ld64_2(dest, 8, DST_2);
|
| 1267 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 1268 |
|
| 1269 |
vis_ld64_2(ref, stride, TMP6); |
| 1270 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 1271 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 1272 |
|
| 1273 |
vis_ld64_2(ref, stride_8, TMP8); |
| 1274 |
vis_pmerge(ZERO, REF_4, TMP4); |
| 1275 |
|
| 1276 |
vis_ld64_2(ref, stride_16, TMP10); |
| 1277 |
ref += stride; |
| 1278 |
|
| 1279 |
vis_ld64_2(dest, stride, REF_S0/*DST_4*/);
|
| 1280 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 1281 |
vis_mul8x16au(REF_4_1, CONST_256, TMP6); |
| 1282 |
|
| 1283 |
vis_ld64_2(dest, stride_8, REF_S2/*DST_6*/);
|
| 1284 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 1285 |
vis_mul8x16al(DST_0, CONST_512, TMP20); |
| 1286 |
|
| 1287 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 1288 |
vis_mul8x16al(DST_1, CONST_512, TMP22); |
| 1289 |
|
| 1290 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 1291 |
vis_mul8x16al(DST_2, CONST_512, TMP24); |
| 1292 |
|
| 1293 |
vis_padd16(TMP4, CONST_3, TMP4); |
| 1294 |
vis_mul8x16al(DST_3, CONST_512, TMP26); |
| 1295 |
|
| 1296 |
vis_padd16(TMP6, CONST_3, TMP6); |
| 1297 |
|
| 1298 |
vis_padd16(TMP12, TMP20, TMP12); |
| 1299 |
vis_mul8x16al(REF_S0, CONST_512, TMP20); |
| 1300 |
|
| 1301 |
vis_padd16(TMP14, TMP22, TMP14); |
| 1302 |
vis_mul8x16al(REF_S0_1, CONST_512, TMP22); |
| 1303 |
|
| 1304 |
vis_padd16(TMP16, TMP24, TMP16); |
| 1305 |
vis_mul8x16al(REF_S2, CONST_512, TMP24); |
| 1306 |
|
| 1307 |
vis_padd16(TMP18, TMP26, TMP18); |
| 1308 |
vis_mul8x16al(REF_S2_1, CONST_512, TMP26); |
| 1309 |
|
| 1310 |
vis_padd16(TMP12, TMP0, TMP12); |
| 1311 |
vis_mul8x16au(REF_2, CONST_256, TMP28); |
| 1312 |
|
| 1313 |
vis_padd16(TMP14, TMP2, TMP14); |
| 1314 |
vis_mul8x16au(REF_2_1, CONST_256, TMP30); |
| 1315 |
|
| 1316 |
vis_padd16(TMP16, TMP4, TMP16); |
| 1317 |
vis_mul8x16au(REF_6, CONST_256, REF_S4); |
| 1318 |
|
| 1319 |
vis_padd16(TMP18, TMP6, TMP18); |
| 1320 |
vis_mul8x16au(REF_6_1, CONST_256, REF_S6); |
| 1321 |
|
| 1322 |
vis_pack16(TMP12, DST_0); |
| 1323 |
vis_padd16(TMP28, TMP0, TMP12); |
| 1324 |
|
| 1325 |
vis_pack16(TMP14, DST_1); |
| 1326 |
vis_st64(DST_0, dest[0]);
|
| 1327 |
vis_padd16(TMP30, TMP2, TMP14); |
| 1328 |
|
| 1329 |
vis_pack16(TMP16, DST_2); |
| 1330 |
vis_padd16(REF_S4, TMP4, TMP16); |
| 1331 |
|
| 1332 |
vis_pack16(TMP18, DST_3); |
| 1333 |
vis_st64_2(DST_2, dest, 8);
|
| 1334 |
dest += stride; |
| 1335 |
vis_padd16(REF_S6, TMP6, TMP18); |
| 1336 |
|
| 1337 |
vis_padd16(TMP12, TMP20, TMP12); |
| 1338 |
|
| 1339 |
vis_padd16(TMP14, TMP22, TMP14); |
| 1340 |
vis_pack16(TMP12, DST_0); |
| 1341 |
|
| 1342 |
vis_padd16(TMP16, TMP24, TMP16); |
| 1343 |
vis_pack16(TMP14, DST_1); |
| 1344 |
vis_st64(DST_0, dest[0]);
|
| 1345 |
|
| 1346 |
vis_padd16(TMP18, TMP26, TMP18); |
| 1347 |
vis_pack16(TMP16, DST_2); |
| 1348 |
|
| 1349 |
vis_pack16(TMP18, DST_3); |
| 1350 |
vis_st64_2(DST_2, dest, 8);
|
| 1351 |
dest += stride; |
| 1352 |
} while (--height);
|
| 1353 |
} |
| 1354 |
|
| 1355 |
static void MC_avg_y_8_vis (uint8_t * dest, const uint8_t * ref, |
| 1356 |
const int stride, int height) |
| 1357 |
{
|
| 1358 |
int stride_8 = stride + 8; |
| 1359 |
|
| 1360 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1361 |
|
| 1362 |
ref = vis_alignaddr(ref); |
| 1363 |
|
| 1364 |
vis_ld64(ref[ 0], TMP0);
|
| 1365 |
vis_fzero(ZERO); |
| 1366 |
|
| 1367 |
vis_ld64(ref[ 8], TMP2);
|
| 1368 |
|
| 1369 |
vis_ld64(constants3[0], CONST_3);
|
| 1370 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 1371 |
|
| 1372 |
vis_ld64(constants256_512[0], CONST_256);
|
| 1373 |
|
| 1374 |
height >>= 1;
|
| 1375 |
do { /* 20 cycles */ |
| 1376 |
vis_ld64_2(ref, stride, TMP0); |
| 1377 |
vis_pmerge(ZERO, REF_2, TMP8); |
| 1378 |
vis_mul8x16au(REF_2_1, CONST_256, TMP10); |
| 1379 |
|
| 1380 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1381 |
ref += stride; |
| 1382 |
|
| 1383 |
vis_ld64(dest[0], DST_0);
|
| 1384 |
|
| 1385 |
vis_ld64_2(dest, stride, DST_2); |
| 1386 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1387 |
|
| 1388 |
vis_ld64_2(ref, stride, TMP4); |
| 1389 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 1390 |
vis_pmerge(ZERO, REF_0, TMP12); |
| 1391 |
|
| 1392 |
vis_ld64_2(ref, stride_8, TMP6); |
| 1393 |
ref += stride; |
| 1394 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 1395 |
vis_pmerge(ZERO, REF_0_1, TMP14); |
| 1396 |
|
| 1397 |
vis_padd16(TMP12, CONST_3, TMP12); |
| 1398 |
vis_mul8x16al(DST_2, CONST_512, TMP24); |
| 1399 |
|
| 1400 |
vis_padd16(TMP14, CONST_3, TMP14); |
| 1401 |
vis_mul8x16al(DST_3, CONST_512, TMP26); |
| 1402 |
|
| 1403 |
vis_faligndata(TMP4, TMP6, REF_2); |
| 1404 |
|
| 1405 |
vis_padd16(TMP8, TMP12, TMP8); |
| 1406 |
|
| 1407 |
vis_padd16(TMP10, TMP14, TMP10); |
| 1408 |
vis_mul8x16au(REF_2, CONST_256, TMP20); |
| 1409 |
|
| 1410 |
vis_padd16(TMP8, TMP16, TMP0); |
| 1411 |
vis_mul8x16au(REF_2_1, CONST_256, TMP22); |
| 1412 |
|
| 1413 |
vis_padd16(TMP10, TMP18, TMP2); |
| 1414 |
vis_pack16(TMP0, DST_0); |
| 1415 |
|
| 1416 |
vis_pack16(TMP2, DST_1); |
| 1417 |
vis_st64(DST_0, dest[0]);
|
| 1418 |
dest += stride; |
| 1419 |
vis_padd16(TMP12, TMP20, TMP12); |
| 1420 |
|
| 1421 |
vis_padd16(TMP14, TMP22, TMP14); |
| 1422 |
|
| 1423 |
vis_padd16(TMP12, TMP24, TMP0); |
| 1424 |
|
| 1425 |
vis_padd16(TMP14, TMP26, TMP2); |
| 1426 |
vis_pack16(TMP0, DST_2); |
| 1427 |
|
| 1428 |
vis_pack16(TMP2, DST_3); |
| 1429 |
vis_st64(DST_2, dest[0]);
|
| 1430 |
dest += stride; |
| 1431 |
} while (--height);
|
| 1432 |
} |
| 1433 |
|
| 1434 |
static void MC_put_xy_16_vis (uint8_t * dest, const uint8_t * ref, |
| 1435 |
const int stride, int height) |
| 1436 |
{
|
| 1437 |
unsigned long off = (unsigned long) ref & 0x7; |
| 1438 |
unsigned long off_plus_1 = off + 1; |
| 1439 |
int stride_8 = stride + 8; |
| 1440 |
int stride_16 = stride + 16; |
| 1441 |
|
| 1442 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1443 |
|
| 1444 |
ref = vis_alignaddr(ref); |
| 1445 |
|
| 1446 |
vis_ld64(ref[ 0], TMP0);
|
| 1447 |
vis_fzero(ZERO); |
| 1448 |
|
| 1449 |
vis_ld64(ref[ 8], TMP2);
|
| 1450 |
|
| 1451 |
vis_ld64(ref[16], TMP4);
|
| 1452 |
|
| 1453 |
vis_ld64(constants2[0], CONST_2);
|
| 1454 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 1455 |
|
| 1456 |
vis_ld64(constants256_512[0], CONST_256);
|
| 1457 |
vis_faligndata(TMP2, TMP4, REF_S4); |
| 1458 |
|
| 1459 |
if (off != 0x7) { |
| 1460 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1461 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 1462 |
vis_faligndata(TMP2, TMP4, REF_S6); |
| 1463 |
} else {
|
| 1464 |
vis_src1(TMP2, REF_S2); |
| 1465 |
vis_src1(TMP4, REF_S6); |
| 1466 |
} |
| 1467 |
|
| 1468 |
height >>= 1;
|
| 1469 |
do {
|
| 1470 |
vis_ld64_2(ref, stride, TMP0); |
| 1471 |
vis_mul8x16au(REF_S0, CONST_256, TMP12); |
| 1472 |
vis_pmerge(ZERO, REF_S0_1, TMP14); |
| 1473 |
|
| 1474 |
vis_alignaddr_g0((void *)off);
|
| 1475 |
|
| 1476 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1477 |
vis_mul8x16au(REF_S2, CONST_256, TMP16); |
| 1478 |
vis_pmerge(ZERO, REF_S2_1, TMP18); |
| 1479 |
|
| 1480 |
vis_ld64_2(ref, stride_16, TMP4); |
| 1481 |
ref += stride; |
| 1482 |
vis_mul8x16au(REF_S4, CONST_256, TMP20); |
| 1483 |
vis_pmerge(ZERO, REF_S4_1, TMP22); |
| 1484 |
|
| 1485 |
vis_ld64_2(ref, stride, TMP6); |
| 1486 |
vis_mul8x16au(REF_S6, CONST_256, TMP24); |
| 1487 |
vis_pmerge(ZERO, REF_S6_1, TMP26); |
| 1488 |
|
| 1489 |
vis_ld64_2(ref, stride_8, TMP8); |
| 1490 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1491 |
|
| 1492 |
vis_ld64_2(ref, stride_16, TMP10); |
| 1493 |
ref += stride; |
| 1494 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 1495 |
|
| 1496 |
vis_faligndata(TMP6, TMP8, REF_S0); |
| 1497 |
|
| 1498 |
vis_faligndata(TMP8, TMP10, REF_S4); |
| 1499 |
|
| 1500 |
if (off != 0x7) { |
| 1501 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1502 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 1503 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 1504 |
vis_faligndata(TMP6, TMP8, REF_S2); |
| 1505 |
vis_faligndata(TMP8, TMP10, REF_S6); |
| 1506 |
} else {
|
| 1507 |
vis_src1(TMP2, REF_2); |
| 1508 |
vis_src1(TMP4, REF_6); |
| 1509 |
vis_src1(TMP8, REF_S2); |
| 1510 |
vis_src1(TMP10, REF_S6); |
| 1511 |
} |
| 1512 |
|
| 1513 |
vis_mul8x16au(REF_0, CONST_256, TMP0); |
| 1514 |
vis_pmerge(ZERO, REF_0_1, TMP2); |
| 1515 |
|
| 1516 |
vis_mul8x16au(REF_2, CONST_256, TMP4); |
| 1517 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 1518 |
|
| 1519 |
vis_padd16(TMP0, CONST_2, TMP8); |
| 1520 |
vis_mul8x16au(REF_4, CONST_256, TMP0); |
| 1521 |
|
| 1522 |
vis_padd16(TMP2, CONST_2, TMP10); |
| 1523 |
vis_mul8x16au(REF_4_1, CONST_256, TMP2); |
| 1524 |
|
| 1525 |
vis_padd16(TMP8, TMP4, TMP8); |
| 1526 |
vis_mul8x16au(REF_6, CONST_256, TMP4); |
| 1527 |
|
| 1528 |
vis_padd16(TMP10, TMP6, TMP10); |
| 1529 |
vis_mul8x16au(REF_6_1, CONST_256, TMP6); |
| 1530 |
|
| 1531 |
vis_padd16(TMP12, TMP8, TMP12); |
| 1532 |
|
| 1533 |
vis_padd16(TMP14, TMP10, TMP14); |
| 1534 |
|
| 1535 |
vis_padd16(TMP12, TMP16, TMP12); |
| 1536 |
|
| 1537 |
vis_padd16(TMP14, TMP18, TMP14); |
| 1538 |
vis_pack16(TMP12, DST_0); |
| 1539 |
|
| 1540 |
vis_pack16(TMP14, DST_1); |
| 1541 |
vis_st64(DST_0, dest[0]);
|
| 1542 |
vis_padd16(TMP0, CONST_2, TMP12); |
| 1543 |
|
| 1544 |
vis_mul8x16au(REF_S0, CONST_256, TMP0); |
| 1545 |
vis_padd16(TMP2, CONST_2, TMP14); |
| 1546 |
|
| 1547 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP2); |
| 1548 |
vis_padd16(TMP12, TMP4, TMP12); |
| 1549 |
|
| 1550 |
vis_mul8x16au(REF_S2, CONST_256, TMP4); |
| 1551 |
vis_padd16(TMP14, TMP6, TMP14); |
| 1552 |
|
| 1553 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP6); |
| 1554 |
vis_padd16(TMP20, TMP12, TMP20); |
| 1555 |
|
| 1556 |
vis_padd16(TMP22, TMP14, TMP22); |
| 1557 |
|
| 1558 |
vis_padd16(TMP20, TMP24, TMP20); |
| 1559 |
|
| 1560 |
vis_padd16(TMP22, TMP26, TMP22); |
| 1561 |
vis_pack16(TMP20, DST_2); |
| 1562 |
|
| 1563 |
vis_pack16(TMP22, DST_3); |
| 1564 |
vis_st64_2(DST_2, dest, 8);
|
| 1565 |
dest += stride; |
| 1566 |
vis_padd16(TMP0, TMP4, TMP24); |
| 1567 |
|
| 1568 |
vis_mul8x16au(REF_S4, CONST_256, TMP0); |
| 1569 |
vis_padd16(TMP2, TMP6, TMP26); |
| 1570 |
|
| 1571 |
vis_mul8x16au(REF_S4_1, CONST_256, TMP2); |
| 1572 |
vis_padd16(TMP24, TMP8, TMP24); |
| 1573 |
|
| 1574 |
vis_padd16(TMP26, TMP10, TMP26); |
| 1575 |
vis_pack16(TMP24, DST_0); |
| 1576 |
|
| 1577 |
vis_pack16(TMP26, DST_1); |
| 1578 |
vis_st64(DST_0, dest[0]);
|
| 1579 |
vis_pmerge(ZERO, REF_S6, TMP4); |
| 1580 |
|
| 1581 |
vis_pmerge(ZERO, REF_S6_1, TMP6); |
| 1582 |
|
| 1583 |
vis_padd16(TMP0, TMP4, TMP0); |
| 1584 |
|
| 1585 |
vis_padd16(TMP2, TMP6, TMP2); |
| 1586 |
|
| 1587 |
vis_padd16(TMP0, TMP12, TMP0); |
| 1588 |
|
| 1589 |
vis_padd16(TMP2, TMP14, TMP2); |
| 1590 |
vis_pack16(TMP0, DST_2); |
| 1591 |
|
| 1592 |
vis_pack16(TMP2, DST_3); |
| 1593 |
vis_st64_2(DST_2, dest, 8);
|
| 1594 |
dest += stride; |
| 1595 |
} while (--height);
|
| 1596 |
} |
| 1597 |
|
| 1598 |
static void MC_put_xy_8_vis (uint8_t * dest, const uint8_t * ref, |
| 1599 |
const int stride, int height) |
| 1600 |
{
|
| 1601 |
unsigned long off = (unsigned long) ref & 0x7; |
| 1602 |
unsigned long off_plus_1 = off + 1; |
| 1603 |
int stride_8 = stride + 8; |
| 1604 |
|
| 1605 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1606 |
|
| 1607 |
ref = vis_alignaddr(ref); |
| 1608 |
|
| 1609 |
vis_ld64(ref[ 0], TMP0);
|
| 1610 |
vis_fzero(ZERO); |
| 1611 |
|
| 1612 |
vis_ld64(ref[ 8], TMP2);
|
| 1613 |
|
| 1614 |
vis_ld64(constants2[0], CONST_2);
|
| 1615 |
|
| 1616 |
vis_ld64(constants256_512[0], CONST_256);
|
| 1617 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 1618 |
|
| 1619 |
if (off != 0x7) { |
| 1620 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1621 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 1622 |
} else {
|
| 1623 |
vis_src1(TMP2, REF_S2); |
| 1624 |
} |
| 1625 |
|
| 1626 |
height >>= 1;
|
| 1627 |
do { /* 26 cycles */ |
| 1628 |
vis_ld64_2(ref, stride, TMP0); |
| 1629 |
vis_mul8x16au(REF_S0, CONST_256, TMP8); |
| 1630 |
vis_pmerge(ZERO, REF_S2, TMP12); |
| 1631 |
|
| 1632 |
vis_alignaddr_g0((void *)off);
|
| 1633 |
|
| 1634 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1635 |
ref += stride; |
| 1636 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP10); |
| 1637 |
vis_pmerge(ZERO, REF_S2_1, TMP14); |
| 1638 |
|
| 1639 |
vis_ld64_2(ref, stride, TMP4); |
| 1640 |
|
| 1641 |
vis_ld64_2(ref, stride_8, TMP6); |
| 1642 |
ref += stride; |
| 1643 |
vis_faligndata(TMP0, TMP2, REF_S4); |
| 1644 |
|
| 1645 |
vis_pmerge(ZERO, REF_S4, TMP18); |
| 1646 |
|
| 1647 |
vis_pmerge(ZERO, REF_S4_1, TMP20); |
| 1648 |
|
| 1649 |
vis_faligndata(TMP4, TMP6, REF_S0); |
| 1650 |
|
| 1651 |
if (off != 0x7) { |
| 1652 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1653 |
vis_faligndata(TMP0, TMP2, REF_S6); |
| 1654 |
vis_faligndata(TMP4, TMP6, REF_S2); |
| 1655 |
} else {
|
| 1656 |
vis_src1(TMP2, REF_S6); |
| 1657 |
vis_src1(TMP6, REF_S2); |
| 1658 |
} |
| 1659 |
|
| 1660 |
vis_padd16(TMP18, CONST_2, TMP18); |
| 1661 |
vis_mul8x16au(REF_S6, CONST_256, TMP22); |
| 1662 |
|
| 1663 |
vis_padd16(TMP20, CONST_2, TMP20); |
| 1664 |
vis_mul8x16au(REF_S6_1, CONST_256, TMP24); |
| 1665 |
|
| 1666 |
vis_mul8x16au(REF_S0, CONST_256, TMP26); |
| 1667 |
vis_pmerge(ZERO, REF_S0_1, TMP28); |
| 1668 |
|
| 1669 |
vis_mul8x16au(REF_S2, CONST_256, TMP30); |
| 1670 |
vis_padd16(TMP18, TMP22, TMP18); |
| 1671 |
|
| 1672 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP32); |
| 1673 |
vis_padd16(TMP20, TMP24, TMP20); |
| 1674 |
|
| 1675 |
vis_padd16(TMP8, TMP18, TMP8); |
| 1676 |
|
| 1677 |
vis_padd16(TMP10, TMP20, TMP10); |
| 1678 |
|
| 1679 |
vis_padd16(TMP8, TMP12, TMP8); |
| 1680 |
|
| 1681 |
vis_padd16(TMP10, TMP14, TMP10); |
| 1682 |
vis_pack16(TMP8, DST_0); |
| 1683 |
|
| 1684 |
vis_pack16(TMP10, DST_1); |
| 1685 |
vis_st64(DST_0, dest[0]);
|
| 1686 |
dest += stride; |
| 1687 |
vis_padd16(TMP18, TMP26, TMP18); |
| 1688 |
|
| 1689 |
vis_padd16(TMP20, TMP28, TMP20); |
| 1690 |
|
| 1691 |
vis_padd16(TMP18, TMP30, TMP18); |
| 1692 |
|
| 1693 |
vis_padd16(TMP20, TMP32, TMP20); |
| 1694 |
vis_pack16(TMP18, DST_2); |
| 1695 |
|
| 1696 |
vis_pack16(TMP20, DST_3); |
| 1697 |
vis_st64(DST_2, dest[0]);
|
| 1698 |
dest += stride; |
| 1699 |
} while (--height);
|
| 1700 |
} |
| 1701 |
|
| 1702 |
static void MC_avg_xy_16_vis (uint8_t * dest, const uint8_t * ref, |
| 1703 |
const int stride, int height) |
| 1704 |
{
|
| 1705 |
unsigned long off = (unsigned long) ref & 0x7; |
| 1706 |
unsigned long off_plus_1 = off + 1; |
| 1707 |
int stride_8 = stride + 8; |
| 1708 |
int stride_16 = stride + 16; |
| 1709 |
|
| 1710 |
vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1711 |
|
| 1712 |
ref = vis_alignaddr(ref); |
| 1713 |
|
| 1714 |
vis_ld64(ref[ 0], TMP0);
|
| 1715 |
vis_fzero(ZERO); |
| 1716 |
|
| 1717 |
vis_ld64(ref[ 8], TMP2);
|
| 1718 |
|
| 1719 |
vis_ld64(ref[16], TMP4);
|
| 1720 |
|
| 1721 |
vis_ld64(constants6[0], CONST_6);
|
| 1722 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 1723 |
|
| 1724 |
vis_ld64(constants256_1024[0], CONST_256);
|
| 1725 |
vis_faligndata(TMP2, TMP4, REF_S4); |
| 1726 |
|
| 1727 |
if (off != 0x7) { |
| 1728 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1729 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 1730 |
vis_faligndata(TMP2, TMP4, REF_S6); |
| 1731 |
} else {
|
| 1732 |
vis_src1(TMP2, REF_S2); |
| 1733 |
vis_src1(TMP4, REF_S6); |
| 1734 |
} |
| 1735 |
|
| 1736 |
height >>= 1;
|
| 1737 |
do { /* 55 cycles */ |
| 1738 |
vis_ld64_2(ref, stride, TMP0); |
| 1739 |
vis_mul8x16au(REF_S0, CONST_256, TMP12); |
| 1740 |
vis_pmerge(ZERO, REF_S0_1, TMP14); |
| 1741 |
|
| 1742 |
vis_alignaddr_g0((void *)off);
|
| 1743 |
|
| 1744 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1745 |
vis_mul8x16au(REF_S2, CONST_256, TMP16); |
| 1746 |
vis_pmerge(ZERO, REF_S2_1, TMP18); |
| 1747 |
|
| 1748 |
vis_ld64_2(ref, stride_16, TMP4); |
| 1749 |
ref += stride; |
| 1750 |
vis_mul8x16au(REF_S4, CONST_256, TMP20); |
| 1751 |
vis_pmerge(ZERO, REF_S4_1, TMP22); |
| 1752 |
|
| 1753 |
vis_ld64_2(ref, stride, TMP6); |
| 1754 |
vis_mul8x16au(REF_S6, CONST_256, TMP24); |
| 1755 |
vis_pmerge(ZERO, REF_S6_1, TMP26); |
| 1756 |
|
| 1757 |
vis_ld64_2(ref, stride_8, TMP8); |
| 1758 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 1759 |
|
| 1760 |
vis_ld64_2(ref, stride_16, TMP10); |
| 1761 |
ref += stride; |
| 1762 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 1763 |
|
| 1764 |
vis_ld64(dest[0], DST_0);
|
| 1765 |
vis_faligndata(TMP6, TMP8, REF_S0); |
| 1766 |
|
| 1767 |
vis_ld64_2(dest, 8, DST_2);
|
| 1768 |
vis_faligndata(TMP8, TMP10, REF_S4); |
| 1769 |
|
| 1770 |
if (off != 0x7) { |
| 1771 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1772 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 1773 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 1774 |
vis_faligndata(TMP6, TMP8, REF_S2); |
| 1775 |
vis_faligndata(TMP8, TMP10, REF_S6); |
| 1776 |
} else {
|
| 1777 |
vis_src1(TMP2, REF_2); |
| 1778 |
vis_src1(TMP4, REF_6); |
| 1779 |
vis_src1(TMP8, REF_S2); |
| 1780 |
vis_src1(TMP10, REF_S6); |
| 1781 |
} |
| 1782 |
|
| 1783 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 1784 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 1785 |
|
| 1786 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 1787 |
vis_pmerge(ZERO, REF_0_1, TMP2); |
| 1788 |
|
| 1789 |
vis_mul8x16au(REF_2, CONST_256, TMP4); |
| 1790 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 1791 |
|
| 1792 |
vis_mul8x16al(DST_2, CONST_1024, REF_0); |
| 1793 |
vis_padd16(TMP0, CONST_6, TMP0); |
| 1794 |
|
| 1795 |
vis_mul8x16al(DST_3, CONST_1024, REF_2); |
| 1796 |
vis_padd16(TMP2, CONST_6, TMP2); |
| 1797 |
|
| 1798 |
vis_padd16(TMP0, TMP4, TMP0); |
| 1799 |
vis_mul8x16au(REF_4, CONST_256, TMP4); |
| 1800 |
|
| 1801 |
vis_padd16(TMP2, TMP6, TMP2); |
| 1802 |
vis_mul8x16au(REF_4_1, CONST_256, TMP6); |
| 1803 |
|
| 1804 |
vis_padd16(TMP12, TMP0, TMP12); |
| 1805 |
vis_mul8x16au(REF_6, CONST_256, TMP8); |
| 1806 |
|
| 1807 |
vis_padd16(TMP14, TMP2, TMP14); |
| 1808 |
vis_mul8x16au(REF_6_1, CONST_256, TMP10); |
| 1809 |
|
| 1810 |
vis_padd16(TMP12, TMP16, TMP12); |
| 1811 |
vis_mul8x16au(REF_S0, CONST_256, REF_4); |
| 1812 |
|
| 1813 |
vis_padd16(TMP14, TMP18, TMP14); |
| 1814 |
vis_mul8x16au(REF_S0_1, CONST_256, REF_6); |
| 1815 |
|
| 1816 |
vis_padd16(TMP12, TMP30, TMP12); |
| 1817 |
|
| 1818 |
vis_padd16(TMP14, TMP32, TMP14); |
| 1819 |
vis_pack16(TMP12, DST_0); |
| 1820 |
|
| 1821 |
vis_pack16(TMP14, DST_1); |
| 1822 |
vis_st64(DST_0, dest[0]);
|
| 1823 |
vis_padd16(TMP4, CONST_6, TMP4); |
| 1824 |
|
| 1825 |
vis_ld64_2(dest, stride, DST_0); |
| 1826 |
vis_padd16(TMP6, CONST_6, TMP6); |
| 1827 |
vis_mul8x16au(REF_S2, CONST_256, TMP12); |
| 1828 |
|
| 1829 |
vis_padd16(TMP4, TMP8, TMP4); |
| 1830 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP14); |
| 1831 |
|
| 1832 |
vis_padd16(TMP6, TMP10, TMP6); |
| 1833 |
|
| 1834 |
vis_padd16(TMP20, TMP4, TMP20); |
| 1835 |
|
| 1836 |
vis_padd16(TMP22, TMP6, TMP22); |
| 1837 |
|
| 1838 |
vis_padd16(TMP20, TMP24, TMP20); |
| 1839 |
|
| 1840 |
vis_padd16(TMP22, TMP26, TMP22); |
| 1841 |
|
| 1842 |
vis_padd16(TMP20, REF_0, TMP20); |
| 1843 |
vis_mul8x16au(REF_S4, CONST_256, REF_0); |
| 1844 |
|
| 1845 |
vis_padd16(TMP22, REF_2, TMP22); |
| 1846 |
vis_pack16(TMP20, DST_2); |
| 1847 |
|
| 1848 |
vis_pack16(TMP22, DST_3); |
| 1849 |
vis_st64_2(DST_2, dest, 8);
|
| 1850 |
dest += stride; |
| 1851 |
|
| 1852 |
vis_ld64_2(dest, 8, DST_2);
|
| 1853 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 1854 |
vis_pmerge(ZERO, REF_S4_1, REF_2); |
| 1855 |
|
| 1856 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 1857 |
vis_padd16(REF_4, TMP0, TMP8); |
| 1858 |
|
| 1859 |
vis_mul8x16au(REF_S6, CONST_256, REF_4); |
| 1860 |
vis_padd16(REF_6, TMP2, TMP10); |
| 1861 |
|
| 1862 |
vis_mul8x16au(REF_S6_1, CONST_256, REF_6); |
| 1863 |
vis_padd16(TMP8, TMP12, TMP8); |
| 1864 |
|
| 1865 |
vis_padd16(TMP10, TMP14, TMP10); |
| 1866 |
|
| 1867 |
vis_padd16(TMP8, TMP30, TMP8); |
| 1868 |
|
| 1869 |
vis_padd16(TMP10, TMP32, TMP10); |
| 1870 |
vis_pack16(TMP8, DST_0); |
| 1871 |
|
| 1872 |
vis_pack16(TMP10, DST_1); |
| 1873 |
vis_st64(DST_0, dest[0]);
|
| 1874 |
|
| 1875 |
vis_padd16(REF_0, TMP4, REF_0); |
| 1876 |
|
| 1877 |
vis_mul8x16al(DST_2, CONST_1024, TMP30); |
| 1878 |
vis_padd16(REF_2, TMP6, REF_2); |
| 1879 |
|
| 1880 |
vis_mul8x16al(DST_3, CONST_1024, TMP32); |
| 1881 |
vis_padd16(REF_0, REF_4, REF_0); |
| 1882 |
|
| 1883 |
vis_padd16(REF_2, REF_6, REF_2); |
| 1884 |
|
| 1885 |
vis_padd16(REF_0, TMP30, REF_0); |
| 1886 |
|
| 1887 |
/* stall */
|
| 1888 |
|
| 1889 |
vis_padd16(REF_2, TMP32, REF_2); |
| 1890 |
vis_pack16(REF_0, DST_2); |
| 1891 |
|
| 1892 |
vis_pack16(REF_2, DST_3); |
| 1893 |
vis_st64_2(DST_2, dest, 8);
|
| 1894 |
dest += stride; |
| 1895 |
} while (--height);
|
| 1896 |
} |
| 1897 |
|
| 1898 |
static void MC_avg_xy_8_vis (uint8_t * dest, const uint8_t * ref, |
| 1899 |
const int stride, int height) |
| 1900 |
{
|
| 1901 |
unsigned long off = (unsigned long) ref & 0x7; |
| 1902 |
unsigned long off_plus_1 = off + 1; |
| 1903 |
int stride_8 = stride + 8; |
| 1904 |
|
| 1905 |
vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT);
|
| 1906 |
|
| 1907 |
ref = vis_alignaddr(ref); |
| 1908 |
|
| 1909 |
vis_ld64(ref[0], TMP0);
|
| 1910 |
vis_fzero(ZERO); |
| 1911 |
|
| 1912 |
vis_ld64_2(ref, 8, TMP2);
|
| 1913 |
|
| 1914 |
vis_ld64(constants6[0], CONST_6);
|
| 1915 |
|
| 1916 |
vis_ld64(constants256_1024[0], CONST_256);
|
| 1917 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 1918 |
|
| 1919 |
if (off != 0x7) { |
| 1920 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1921 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 1922 |
} else {
|
| 1923 |
vis_src1(TMP2, REF_S2); |
| 1924 |
} |
| 1925 |
|
| 1926 |
height >>= 1;
|
| 1927 |
do { /* 31 cycles */ |
| 1928 |
vis_ld64_2(ref, stride, TMP0); |
| 1929 |
vis_mul8x16au(REF_S0, CONST_256, TMP8); |
| 1930 |
vis_pmerge(ZERO, REF_S0_1, TMP10); |
| 1931 |
|
| 1932 |
vis_ld64_2(ref, stride_8, TMP2); |
| 1933 |
ref += stride; |
| 1934 |
vis_mul8x16au(REF_S2, CONST_256, TMP12); |
| 1935 |
vis_pmerge(ZERO, REF_S2_1, TMP14); |
| 1936 |
|
| 1937 |
vis_alignaddr_g0((void *)off);
|
| 1938 |
|
| 1939 |
vis_ld64_2(ref, stride, TMP4); |
| 1940 |
vis_faligndata(TMP0, TMP2, REF_S4); |
| 1941 |
|
| 1942 |
vis_ld64_2(ref, stride_8, TMP6); |
| 1943 |
ref += stride; |
| 1944 |
|
| 1945 |
vis_ld64(dest[0], DST_0);
|
| 1946 |
vis_faligndata(TMP4, TMP6, REF_S0); |
| 1947 |
|
| 1948 |
vis_ld64_2(dest, stride, DST_2); |
| 1949 |
|
| 1950 |
if (off != 0x7) { |
| 1951 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 1952 |
vis_faligndata(TMP0, TMP2, REF_S6); |
| 1953 |
vis_faligndata(TMP4, TMP6, REF_S2); |
| 1954 |
} else {
|
| 1955 |
vis_src1(TMP2, REF_S6); |
| 1956 |
vis_src1(TMP6, REF_S2); |
| 1957 |
} |
| 1958 |
|
| 1959 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 1960 |
vis_pmerge(ZERO, REF_S4, TMP22); |
| 1961 |
|
| 1962 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 1963 |
vis_pmerge(ZERO, REF_S4_1, TMP24); |
| 1964 |
|
| 1965 |
vis_mul8x16au(REF_S6, CONST_256, TMP26); |
| 1966 |
vis_pmerge(ZERO, REF_S6_1, TMP28); |
| 1967 |
|
| 1968 |
vis_mul8x16au(REF_S0, CONST_256, REF_S4); |
| 1969 |
vis_padd16(TMP22, CONST_6, TMP22); |
| 1970 |
|
| 1971 |
vis_mul8x16au(REF_S0_1, CONST_256, REF_S6); |
| 1972 |
vis_padd16(TMP24, CONST_6, TMP24); |
| 1973 |
|
| 1974 |
vis_mul8x16al(DST_2, CONST_1024, REF_0); |
| 1975 |
vis_padd16(TMP22, TMP26, TMP22); |
| 1976 |
|
| 1977 |
vis_mul8x16al(DST_3, CONST_1024, REF_2); |
| 1978 |
vis_padd16(TMP24, TMP28, TMP24); |
| 1979 |
|
| 1980 |
vis_mul8x16au(REF_S2, CONST_256, TMP26); |
| 1981 |
vis_padd16(TMP8, TMP22, TMP8); |
| 1982 |
|
| 1983 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP28); |
| 1984 |
vis_padd16(TMP10, TMP24, TMP10); |
| 1985 |
|
| 1986 |
vis_padd16(TMP8, TMP12, TMP8); |
| 1987 |
|
| 1988 |
vis_padd16(TMP10, TMP14, TMP10); |
| 1989 |
|
| 1990 |
vis_padd16(TMP8, TMP30, TMP8); |
| 1991 |
|
| 1992 |
vis_padd16(TMP10, TMP32, TMP10); |
| 1993 |
vis_pack16(TMP8, DST_0); |
| 1994 |
|
| 1995 |
vis_pack16(TMP10, DST_1); |
| 1996 |
vis_st64(DST_0, dest[0]);
|
| 1997 |
dest += stride; |
| 1998 |
|
| 1999 |
vis_padd16(REF_S4, TMP22, TMP12); |
| 2000 |
|
| 2001 |
vis_padd16(REF_S6, TMP24, TMP14); |
| 2002 |
|
| 2003 |
vis_padd16(TMP12, TMP26, TMP12); |
| 2004 |
|
| 2005 |
vis_padd16(TMP14, TMP28, TMP14); |
| 2006 |
|
| 2007 |
vis_padd16(TMP12, REF_0, TMP12); |
| 2008 |
|
| 2009 |
vis_padd16(TMP14, REF_2, TMP14); |
| 2010 |
vis_pack16(TMP12, DST_2); |
| 2011 |
|
| 2012 |
vis_pack16(TMP14, DST_3); |
| 2013 |
vis_st64(DST_2, dest[0]);
|
| 2014 |
dest += stride; |
| 2015 |
} while (--height);
|
| 2016 |
} |
| 2017 |
|
| 2018 |
/* End of rounding code */
|
| 2019 |
|
| 2020 |
/* Start of no rounding code */
|
| 2021 |
/* The trick used in some of this file is the formula from the MMX
|
| 2022 |
* motion comp code, which is:
|
| 2023 |
*
|
| 2024 |
* (x+y)>>1 == (x&y)+((x^y)>>1)
|
| 2025 |
*
|
| 2026 |
* This allows us to average 8 bytes at a time in a 64-bit FPU reg.
|
| 2027 |
* We avoid overflows by masking before we do the shift, and we
|
| 2028 |
* implement the shift by multiplying by 1/2 using mul8x16. So in
|
| 2029 |
* VIS this is (assume 'x' is in f0, 'y' is in f2, a repeating mask
|
| 2030 |
* of '0xfe' is in f4, a repeating mask of '0x7f' is in f6, and
|
| 2031 |
* the value 0x80808080 is in f8):
|
| 2032 |
*
|
| 2033 |
* fxor f0, f2, f10
|
| 2034 |
* fand f10, f4, f10
|
| 2035 |
* fmul8x16 f8, f10, f10
|
| 2036 |
* fand f10, f6, f10
|
| 2037 |
* fand f0, f2, f12
|
| 2038 |
* fpadd16 f12, f10, f10
|
| 2039 |
*/
|
| 2040 |
|
| 2041 |
static void MC_put_no_round_o_16_vis (uint8_t * dest, const uint8_t * ref, |
| 2042 |
const int stride, int height) |
| 2043 |
{
|
| 2044 |
ref = vis_alignaddr(ref); |
| 2045 |
do { /* 5 cycles */ |
| 2046 |
vis_ld64(ref[0], TMP0);
|
| 2047 |
|
| 2048 |
vis_ld64_2(ref, 8, TMP2);
|
| 2049 |
|
| 2050 |
vis_ld64_2(ref, 16, TMP4);
|
| 2051 |
ref += stride; |
| 2052 |
|
| 2053 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2054 |
vis_st64(REF_0, dest[0]);
|
| 2055 |
|
| 2056 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 2057 |
vis_st64_2(REF_2, dest, 8);
|
| 2058 |
dest += stride; |
| 2059 |
} while (--height);
|
| 2060 |
} |
| 2061 |
|
| 2062 |
static void MC_put_no_round_o_8_vis (uint8_t * dest, const uint8_t * ref, |
| 2063 |
const int stride, int height) |
| 2064 |
{
|
| 2065 |
ref = vis_alignaddr(ref); |
| 2066 |
do { /* 4 cycles */ |
| 2067 |
vis_ld64(ref[0], TMP0);
|
| 2068 |
|
| 2069 |
vis_ld64(ref[8], TMP2);
|
| 2070 |
ref += stride; |
| 2071 |
|
| 2072 |
/* stall */
|
| 2073 |
|
| 2074 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2075 |
vis_st64(REF_0, dest[0]);
|
| 2076 |
dest += stride; |
| 2077 |
} while (--height);
|
| 2078 |
} |
| 2079 |
|
| 2080 |
|
| 2081 |
static void MC_avg_no_round_o_16_vis (uint8_t * dest, const uint8_t * ref, |
| 2082 |
const int stride, int height) |
| 2083 |
{
|
| 2084 |
int stride_8 = stride + 8; |
| 2085 |
|
| 2086 |
ref = vis_alignaddr(ref); |
| 2087 |
|
| 2088 |
vis_ld64(ref[0], TMP0);
|
| 2089 |
|
| 2090 |
vis_ld64(ref[8], TMP2);
|
| 2091 |
|
| 2092 |
vis_ld64(ref[16], TMP4);
|
| 2093 |
|
| 2094 |
vis_ld64(dest[0], DST_0);
|
| 2095 |
|
| 2096 |
vis_ld64(dest[8], DST_2);
|
| 2097 |
|
| 2098 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 2099 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2100 |
|
| 2101 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 2102 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 2103 |
|
| 2104 |
vis_ld64(constants128[0], CONST_128);
|
| 2105 |
|
| 2106 |
ref += stride; |
| 2107 |
height = (height >> 1) - 1; |
| 2108 |
|
| 2109 |
do { /* 24 cycles */ |
| 2110 |
vis_ld64(ref[0], TMP0);
|
| 2111 |
vis_xor(DST_0, REF_0, TMP6); |
| 2112 |
|
| 2113 |
vis_ld64_2(ref, 8, TMP2);
|
| 2114 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2115 |
|
| 2116 |
vis_ld64_2(ref, 16, TMP4);
|
| 2117 |
ref += stride; |
| 2118 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2119 |
vis_xor(DST_2, REF_2, TMP8); |
| 2120 |
|
| 2121 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2122 |
|
| 2123 |
vis_and(DST_0, REF_0, TMP10); |
| 2124 |
vis_ld64_2(dest, stride, DST_0); |
| 2125 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2126 |
|
| 2127 |
vis_and(DST_2, REF_2, TMP12); |
| 2128 |
vis_ld64_2(dest, stride_8, DST_2); |
| 2129 |
|
| 2130 |
vis_ld64(ref[0], TMP14);
|
| 2131 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2132 |
|
| 2133 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2134 |
|
| 2135 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2136 |
vis_st64(TMP6, dest[0]);
|
| 2137 |
|
| 2138 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2139 |
vis_st64_2(TMP8, dest, 8);
|
| 2140 |
|
| 2141 |
dest += stride; |
| 2142 |
vis_ld64_2(ref, 8, TMP16);
|
| 2143 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2144 |
|
| 2145 |
vis_ld64_2(ref, 16, TMP18);
|
| 2146 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 2147 |
ref += stride; |
| 2148 |
|
| 2149 |
vis_xor(DST_0, REF_0, TMP20); |
| 2150 |
|
| 2151 |
vis_and(TMP20, MASK_fe, TMP20); |
| 2152 |
|
| 2153 |
vis_xor(DST_2, REF_2, TMP22); |
| 2154 |
vis_mul8x16(CONST_128, TMP20, TMP20); |
| 2155 |
|
| 2156 |
vis_and(TMP22, MASK_fe, TMP22); |
| 2157 |
|
| 2158 |
vis_and(DST_0, REF_0, TMP24); |
| 2159 |
vis_mul8x16(CONST_128, TMP22, TMP22); |
| 2160 |
|
| 2161 |
vis_and(DST_2, REF_2, TMP26); |
| 2162 |
|
| 2163 |
vis_ld64_2(dest, stride, DST_0); |
| 2164 |
vis_faligndata(TMP14, TMP16, REF_0); |
| 2165 |
|
| 2166 |
vis_ld64_2(dest, stride_8, DST_2); |
| 2167 |
vis_faligndata(TMP16, TMP18, REF_2); |
| 2168 |
|
| 2169 |
vis_and(TMP20, MASK_7f, TMP20); |
| 2170 |
|
| 2171 |
vis_and(TMP22, MASK_7f, TMP22); |
| 2172 |
|
| 2173 |
vis_padd16(TMP24, TMP20, TMP20); |
| 2174 |
vis_st64(TMP20, dest[0]);
|
| 2175 |
|
| 2176 |
vis_padd16(TMP26, TMP22, TMP22); |
| 2177 |
vis_st64_2(TMP22, dest, 8);
|
| 2178 |
dest += stride; |
| 2179 |
} while (--height);
|
| 2180 |
|
| 2181 |
vis_ld64(ref[0], TMP0);
|
| 2182 |
vis_xor(DST_0, REF_0, TMP6); |
| 2183 |
|
| 2184 |
vis_ld64_2(ref, 8, TMP2);
|
| 2185 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2186 |
|
| 2187 |
vis_ld64_2(ref, 16, TMP4);
|
| 2188 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2189 |
vis_xor(DST_2, REF_2, TMP8); |
| 2190 |
|
| 2191 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2192 |
|
| 2193 |
vis_and(DST_0, REF_0, TMP10); |
| 2194 |
vis_ld64_2(dest, stride, DST_0); |
| 2195 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2196 |
|
| 2197 |
vis_and(DST_2, REF_2, TMP12); |
| 2198 |
vis_ld64_2(dest, stride_8, DST_2); |
| 2199 |
|
| 2200 |
vis_ld64(ref[0], TMP14);
|
| 2201 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2202 |
|
| 2203 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2204 |
|
| 2205 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2206 |
vis_st64(TMP6, dest[0]);
|
| 2207 |
|
| 2208 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2209 |
vis_st64_2(TMP8, dest, 8);
|
| 2210 |
|
| 2211 |
dest += stride; |
| 2212 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2213 |
|
| 2214 |
vis_faligndata(TMP2, TMP4, REF_2); |
| 2215 |
|
| 2216 |
vis_xor(DST_0, REF_0, TMP20); |
| 2217 |
|
| 2218 |
vis_and(TMP20, MASK_fe, TMP20); |
| 2219 |
|
| 2220 |
vis_xor(DST_2, REF_2, TMP22); |
| 2221 |
vis_mul8x16(CONST_128, TMP20, TMP20); |
| 2222 |
|
| 2223 |
vis_and(TMP22, MASK_fe, TMP22); |
| 2224 |
|
| 2225 |
vis_and(DST_0, REF_0, TMP24); |
| 2226 |
vis_mul8x16(CONST_128, TMP22, TMP22); |
| 2227 |
|
| 2228 |
vis_and(DST_2, REF_2, TMP26); |
| 2229 |
|
| 2230 |
vis_and(TMP20, MASK_7f, TMP20); |
| 2231 |
|
| 2232 |
vis_and(TMP22, MASK_7f, TMP22); |
| 2233 |
|
| 2234 |
vis_padd16(TMP24, TMP20, TMP20); |
| 2235 |
vis_st64(TMP20, dest[0]);
|
| 2236 |
|
| 2237 |
vis_padd16(TMP26, TMP22, TMP22); |
| 2238 |
vis_st64_2(TMP22, dest, 8);
|
| 2239 |
} |
| 2240 |
|
| 2241 |
static void MC_avg_no_round_o_8_vis (uint8_t * dest, const uint8_t * ref, |
| 2242 |
const int stride, int height) |
| 2243 |
{
|
| 2244 |
ref = vis_alignaddr(ref); |
| 2245 |
|
| 2246 |
vis_ld64(ref[0], TMP0);
|
| 2247 |
|
| 2248 |
vis_ld64(ref[8], TMP2);
|
| 2249 |
|
| 2250 |
vis_ld64(dest[0], DST_0);
|
| 2251 |
|
| 2252 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 2253 |
|
| 2254 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 2255 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2256 |
|
| 2257 |
vis_ld64(constants128[0], CONST_128);
|
| 2258 |
|
| 2259 |
ref += stride; |
| 2260 |
height = (height >> 1) - 1; |
| 2261 |
|
| 2262 |
do { /* 12 cycles */ |
| 2263 |
vis_ld64(ref[0], TMP0);
|
| 2264 |
vis_xor(DST_0, REF_0, TMP4); |
| 2265 |
|
| 2266 |
vis_ld64(ref[8], TMP2);
|
| 2267 |
vis_and(TMP4, MASK_fe, TMP4); |
| 2268 |
|
| 2269 |
vis_and(DST_0, REF_0, TMP6); |
| 2270 |
vis_ld64_2(dest, stride, DST_0); |
| 2271 |
ref += stride; |
| 2272 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 2273 |
|
| 2274 |
vis_ld64(ref[0], TMP12);
|
| 2275 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2276 |
|
| 2277 |
vis_ld64(ref[8], TMP2);
|
| 2278 |
vis_xor(DST_0, REF_0, TMP0); |
| 2279 |
ref += stride; |
| 2280 |
|
| 2281 |
vis_and(TMP0, MASK_fe, TMP0); |
| 2282 |
|
| 2283 |
vis_and(TMP4, MASK_7f, TMP4); |
| 2284 |
|
| 2285 |
vis_padd16(TMP6, TMP4, TMP4); |
| 2286 |
vis_st64(TMP4, dest[0]);
|
| 2287 |
dest += stride; |
| 2288 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 2289 |
|
| 2290 |
vis_and(DST_0, REF_0, TMP6); |
| 2291 |
vis_ld64_2(dest, stride, DST_0); |
| 2292 |
|
| 2293 |
vis_faligndata(TMP12, TMP2, REF_0); |
| 2294 |
|
| 2295 |
vis_and(TMP0, MASK_7f, TMP0); |
| 2296 |
|
| 2297 |
vis_padd16(TMP6, TMP0, TMP4); |
| 2298 |
vis_st64(TMP4, dest[0]);
|
| 2299 |
dest += stride; |
| 2300 |
} while (--height);
|
| 2301 |
|
| 2302 |
vis_ld64(ref[0], TMP0);
|
| 2303 |
vis_xor(DST_0, REF_0, TMP4); |
| 2304 |
|
| 2305 |
vis_ld64(ref[8], TMP2);
|
| 2306 |
vis_and(TMP4, MASK_fe, TMP4); |
| 2307 |
|
| 2308 |
vis_and(DST_0, REF_0, TMP6); |
| 2309 |
vis_ld64_2(dest, stride, DST_0); |
| 2310 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 2311 |
|
| 2312 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2313 |
|
| 2314 |
vis_xor(DST_0, REF_0, TMP0); |
| 2315 |
|
| 2316 |
vis_and(TMP0, MASK_fe, TMP0); |
| 2317 |
|
| 2318 |
vis_and(TMP4, MASK_7f, TMP4); |
| 2319 |
|
| 2320 |
vis_padd16(TMP6, TMP4, TMP4); |
| 2321 |
vis_st64(TMP4, dest[0]);
|
| 2322 |
dest += stride; |
| 2323 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 2324 |
|
| 2325 |
vis_and(DST_0, REF_0, TMP6); |
| 2326 |
|
| 2327 |
vis_and(TMP0, MASK_7f, TMP0); |
| 2328 |
|
| 2329 |
vis_padd16(TMP6, TMP0, TMP4); |
| 2330 |
vis_st64(TMP4, dest[0]);
|
| 2331 |
} |
| 2332 |
|
| 2333 |
static void MC_put_no_round_x_16_vis (uint8_t * dest, const uint8_t * ref, |
| 2334 |
const int stride, int height) |
| 2335 |
{
|
| 2336 |
unsigned long off = (unsigned long) ref & 0x7; |
| 2337 |
unsigned long off_plus_1 = off + 1; |
| 2338 |
|
| 2339 |
ref = vis_alignaddr(ref); |
| 2340 |
|
| 2341 |
vis_ld64(ref[0], TMP0);
|
| 2342 |
|
| 2343 |
vis_ld64_2(ref, 8, TMP2);
|
| 2344 |
|
| 2345 |
vis_ld64_2(ref, 16, TMP4);
|
| 2346 |
|
| 2347 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 2348 |
|
| 2349 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 2350 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2351 |
|
| 2352 |
vis_ld64(constants128[0], CONST_128);
|
| 2353 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2354 |
|
| 2355 |
if (off != 0x7) { |
| 2356 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2357 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2358 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 2359 |
} else {
|
| 2360 |
vis_src1(TMP2, REF_2); |
| 2361 |
vis_src1(TMP4, REF_6); |
| 2362 |
} |
| 2363 |
|
| 2364 |
ref += stride; |
| 2365 |
height = (height >> 1) - 1; |
| 2366 |
|
| 2367 |
do { /* 34 cycles */ |
| 2368 |
vis_ld64(ref[0], TMP0);
|
| 2369 |
vis_xor(REF_0, REF_2, TMP6); |
| 2370 |
|
| 2371 |
vis_ld64_2(ref, 8, TMP2);
|
| 2372 |
vis_xor(REF_4, REF_6, TMP8); |
| 2373 |
|
| 2374 |
vis_ld64_2(ref, 16, TMP4);
|
| 2375 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2376 |
ref += stride; |
| 2377 |
|
| 2378 |
vis_ld64(ref[0], TMP14);
|
| 2379 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2380 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2381 |
|
| 2382 |
vis_ld64_2(ref, 8, TMP16);
|
| 2383 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2384 |
vis_and(REF_0, REF_2, TMP10); |
| 2385 |
|
| 2386 |
vis_ld64_2(ref, 16, TMP18);
|
| 2387 |
ref += stride; |
| 2388 |
vis_and(REF_4, REF_6, TMP12); |
| 2389 |
|
| 2390 |
vis_alignaddr_g0((void *)off);
|
| 2391 |
|
| 2392 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2393 |
|
| 2394 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2395 |
|
| 2396 |
if (off != 0x7) { |
| 2397 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2398 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2399 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 2400 |
} else {
|
| 2401 |
vis_src1(TMP2, REF_2); |
| 2402 |
vis_src1(TMP4, REF_6); |
| 2403 |
} |
| 2404 |
|
| 2405 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2406 |
|
| 2407 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2408 |
|
| 2409 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2410 |
vis_st64(TMP6, dest[0]);
|
| 2411 |
|
| 2412 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2413 |
vis_st64_2(TMP8, dest, 8);
|
| 2414 |
dest += stride; |
| 2415 |
|
| 2416 |
vis_xor(REF_0, REF_2, TMP6); |
| 2417 |
|
| 2418 |
vis_xor(REF_4, REF_6, TMP8); |
| 2419 |
|
| 2420 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2421 |
|
| 2422 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2423 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2424 |
|
| 2425 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2426 |
vis_and(REF_0, REF_2, TMP10); |
| 2427 |
|
| 2428 |
vis_and(REF_4, REF_6, TMP12); |
| 2429 |
|
| 2430 |
vis_alignaddr_g0((void *)off);
|
| 2431 |
|
| 2432 |
vis_faligndata(TMP14, TMP16, REF_0); |
| 2433 |
|
| 2434 |
vis_faligndata(TMP16, TMP18, REF_4); |
| 2435 |
|
| 2436 |
if (off != 0x7) { |
| 2437 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2438 |
vis_faligndata(TMP14, TMP16, REF_2); |
| 2439 |
vis_faligndata(TMP16, TMP18, REF_6); |
| 2440 |
} else {
|
| 2441 |
vis_src1(TMP16, REF_2); |
| 2442 |
vis_src1(TMP18, REF_6); |
| 2443 |
} |
| 2444 |
|
| 2445 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2446 |
|
| 2447 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2448 |
|
| 2449 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2450 |
vis_st64(TMP6, dest[0]);
|
| 2451 |
|
| 2452 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2453 |
vis_st64_2(TMP8, dest, 8);
|
| 2454 |
dest += stride; |
| 2455 |
} while (--height);
|
| 2456 |
|
| 2457 |
vis_ld64(ref[0], TMP0);
|
| 2458 |
vis_xor(REF_0, REF_2, TMP6); |
| 2459 |
|
| 2460 |
vis_ld64_2(ref, 8, TMP2);
|
| 2461 |
vis_xor(REF_4, REF_6, TMP8); |
| 2462 |
|
| 2463 |
vis_ld64_2(ref, 16, TMP4);
|
| 2464 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2465 |
|
| 2466 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2467 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2468 |
|
| 2469 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2470 |
vis_and(REF_0, REF_2, TMP10); |
| 2471 |
|
| 2472 |
vis_and(REF_4, REF_6, TMP12); |
| 2473 |
|
| 2474 |
vis_alignaddr_g0((void *)off);
|
| 2475 |
|
| 2476 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2477 |
|
| 2478 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2479 |
|
| 2480 |
if (off != 0x7) { |
| 2481 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2482 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2483 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 2484 |
} else {
|
| 2485 |
vis_src1(TMP2, REF_2); |
| 2486 |
vis_src1(TMP4, REF_6); |
| 2487 |
} |
| 2488 |
|
| 2489 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2490 |
|
| 2491 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2492 |
|
| 2493 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2494 |
vis_st64(TMP6, dest[0]);
|
| 2495 |
|
| 2496 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2497 |
vis_st64_2(TMP8, dest, 8);
|
| 2498 |
dest += stride; |
| 2499 |
|
| 2500 |
vis_xor(REF_0, REF_2, TMP6); |
| 2501 |
|
| 2502 |
vis_xor(REF_4, REF_6, TMP8); |
| 2503 |
|
| 2504 |
vis_and(TMP6, MASK_fe, TMP6); |
| 2505 |
|
| 2506 |
vis_mul8x16(CONST_128, TMP6, TMP6); |
| 2507 |
vis_and(TMP8, MASK_fe, TMP8); |
| 2508 |
|
| 2509 |
vis_mul8x16(CONST_128, TMP8, TMP8); |
| 2510 |
vis_and(REF_0, REF_2, TMP10); |
| 2511 |
|
| 2512 |
vis_and(REF_4, REF_6, TMP12); |
| 2513 |
|
| 2514 |
vis_and(TMP6, MASK_7f, TMP6); |
| 2515 |
|
| 2516 |
vis_and(TMP8, MASK_7f, TMP8); |
| 2517 |
|
| 2518 |
vis_padd16(TMP10, TMP6, TMP6); |
| 2519 |
vis_st64(TMP6, dest[0]);
|
| 2520 |
|
| 2521 |
vis_padd16(TMP12, TMP8, TMP8); |
| 2522 |
vis_st64_2(TMP8, dest, 8);
|
| 2523 |
} |
| 2524 |
|
| 2525 |
static void MC_put_no_round_x_8_vis (uint8_t * dest, const uint8_t * ref, |
| 2526 |
const int stride, int height) |
| 2527 |
{
|
| 2528 |
unsigned long off = (unsigned long) ref & 0x7; |
| 2529 |
unsigned long off_plus_1 = off + 1; |
| 2530 |
|
| 2531 |
ref = vis_alignaddr(ref); |
| 2532 |
|
| 2533 |
vis_ld64(ref[0], TMP0);
|
| 2534 |
|
| 2535 |
vis_ld64(ref[8], TMP2);
|
| 2536 |
|
| 2537 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 2538 |
|
| 2539 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 2540 |
|
| 2541 |
vis_ld64(constants128[0], CONST_128);
|
| 2542 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2543 |
|
| 2544 |
if (off != 0x7) { |
| 2545 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2546 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2547 |
} else {
|
| 2548 |
vis_src1(TMP2, REF_2); |
| 2549 |
} |
| 2550 |
|
| 2551 |
ref += stride; |
| 2552 |
height = (height >> 1) - 1; |
| 2553 |
|
| 2554 |
do { /* 20 cycles */ |
| 2555 |
vis_ld64(ref[0], TMP0);
|
| 2556 |
vis_xor(REF_0, REF_2, TMP4); |
| 2557 |
|
| 2558 |
vis_ld64_2(ref, 8, TMP2);
|
| 2559 |
vis_and(TMP4, MASK_fe, TMP4); |
| 2560 |
ref += stride; |
| 2561 |
|
| 2562 |
vis_ld64(ref[0], TMP8);
|
| 2563 |
vis_and(REF_0, REF_2, TMP6); |
| 2564 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 2565 |
|
| 2566 |
vis_alignaddr_g0((void *)off);
|
| 2567 |
|
| 2568 |
vis_ld64_2(ref, 8, TMP10);
|
| 2569 |
ref += stride; |
| 2570 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2571 |
|
| 2572 |
if (off != 0x7) { |
| 2573 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2574 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2575 |
} else {
|
| 2576 |
vis_src1(TMP2, REF_2); |
| 2577 |
} |
| 2578 |
|
| 2579 |
vis_and(TMP4, MASK_7f, TMP4); |
| 2580 |
|
| 2581 |
vis_padd16(TMP6, TMP4, DST_0); |
| 2582 |
vis_st64(DST_0, dest[0]);
|
| 2583 |
dest += stride; |
| 2584 |
|
| 2585 |
vis_xor(REF_0, REF_2, TMP12); |
| 2586 |
|
| 2587 |
vis_and(TMP12, MASK_fe, TMP12); |
| 2588 |
|
| 2589 |
vis_and(REF_0, REF_2, TMP14); |
| 2590 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 2591 |
|
| 2592 |
vis_alignaddr_g0((void *)off);
|
| 2593 |
vis_faligndata(TMP8, TMP10, REF_0); |
| 2594 |
if (off != 0x7) { |
| 2595 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2596 |
vis_faligndata(TMP8, TMP10, REF_2); |
| 2597 |
} else {
|
| 2598 |
vis_src1(TMP10, REF_2); |
| 2599 |
} |
| 2600 |
|
| 2601 |
vis_and(TMP12, MASK_7f, TMP12); |
| 2602 |
|
| 2603 |
vis_padd16(TMP14, TMP12, DST_0); |
| 2604 |
vis_st64(DST_0, dest[0]);
|
| 2605 |
dest += stride; |
| 2606 |
} while (--height);
|
| 2607 |
|
| 2608 |
vis_ld64(ref[0], TMP0);
|
| 2609 |
vis_xor(REF_0, REF_2, TMP4); |
| 2610 |
|
| 2611 |
vis_ld64_2(ref, 8, TMP2);
|
| 2612 |
vis_and(TMP4, MASK_fe, TMP4); |
| 2613 |
|
| 2614 |
vis_and(REF_0, REF_2, TMP6); |
| 2615 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 2616 |
|
| 2617 |
vis_alignaddr_g0((void *)off);
|
| 2618 |
|
| 2619 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2620 |
|
| 2621 |
if (off != 0x7) { |
| 2622 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2623 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2624 |
} else {
|
| 2625 |
vis_src1(TMP2, REF_2); |
| 2626 |
} |
| 2627 |
|
| 2628 |
vis_and(TMP4, MASK_7f, TMP4); |
| 2629 |
|
| 2630 |
vis_padd16(TMP6, TMP4, DST_0); |
| 2631 |
vis_st64(DST_0, dest[0]);
|
| 2632 |
dest += stride; |
| 2633 |
|
| 2634 |
vis_xor(REF_0, REF_2, TMP12); |
| 2635 |
|
| 2636 |
vis_and(TMP12, MASK_fe, TMP12); |
| 2637 |
|
| 2638 |
vis_and(REF_0, REF_2, TMP14); |
| 2639 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 2640 |
|
| 2641 |
vis_and(TMP12, MASK_7f, TMP12); |
| 2642 |
|
| 2643 |
vis_padd16(TMP14, TMP12, DST_0); |
| 2644 |
vis_st64(DST_0, dest[0]);
|
| 2645 |
dest += stride; |
| 2646 |
} |
| 2647 |
|
| 2648 |
static void MC_avg_no_round_x_16_vis (uint8_t * dest, const uint8_t * ref, |
| 2649 |
const int stride, int height) |
| 2650 |
{
|
| 2651 |
unsigned long off = (unsigned long) ref & 0x7; |
| 2652 |
unsigned long off_plus_1 = off + 1; |
| 2653 |
|
| 2654 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 2655 |
|
| 2656 |
vis_ld64(constants3[0], CONST_3);
|
| 2657 |
vis_fzero(ZERO); |
| 2658 |
vis_ld64(constants256_512[0], CONST_256);
|
| 2659 |
|
| 2660 |
ref = vis_alignaddr(ref); |
| 2661 |
do { /* 26 cycles */ |
| 2662 |
vis_ld64(ref[0], TMP0);
|
| 2663 |
|
| 2664 |
vis_ld64(ref[8], TMP2);
|
| 2665 |
|
| 2666 |
vis_alignaddr_g0((void *)off);
|
| 2667 |
|
| 2668 |
vis_ld64(ref[16], TMP4);
|
| 2669 |
|
| 2670 |
vis_ld64(dest[0], DST_0);
|
| 2671 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2672 |
|
| 2673 |
vis_ld64(dest[8], DST_2);
|
| 2674 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2675 |
|
| 2676 |
if (off != 0x7) { |
| 2677 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2678 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2679 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 2680 |
} else {
|
| 2681 |
vis_src1(TMP2, REF_2); |
| 2682 |
vis_src1(TMP4, REF_6); |
| 2683 |
} |
| 2684 |
|
| 2685 |
vis_mul8x16au(REF_0, CONST_256, TMP0); |
| 2686 |
|
| 2687 |
vis_pmerge(ZERO, REF_2, TMP4); |
| 2688 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 2689 |
|
| 2690 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 2691 |
|
| 2692 |
vis_padd16(TMP0, TMP4, TMP0); |
| 2693 |
|
| 2694 |
vis_mul8x16al(DST_0, CONST_512, TMP4); |
| 2695 |
vis_padd16(TMP2, TMP6, TMP2); |
| 2696 |
|
| 2697 |
vis_mul8x16al(DST_1, CONST_512, TMP6); |
| 2698 |
|
| 2699 |
vis_mul8x16au(REF_6, CONST_256, TMP12); |
| 2700 |
|
| 2701 |
vis_padd16(TMP0, TMP4, TMP0); |
| 2702 |
vis_mul8x16au(REF_6_1, CONST_256, TMP14); |
| 2703 |
|
| 2704 |
vis_padd16(TMP2, TMP6, TMP2); |
| 2705 |
vis_mul8x16au(REF_4, CONST_256, TMP16); |
| 2706 |
|
| 2707 |
vis_padd16(TMP0, CONST_3, TMP8); |
| 2708 |
vis_mul8x16au(REF_4_1, CONST_256, TMP18); |
| 2709 |
|
| 2710 |
vis_padd16(TMP2, CONST_3, TMP10); |
| 2711 |
vis_pack16(TMP8, DST_0); |
| 2712 |
|
| 2713 |
vis_pack16(TMP10, DST_1); |
| 2714 |
vis_padd16(TMP16, TMP12, TMP0); |
| 2715 |
|
| 2716 |
vis_st64(DST_0, dest[0]);
|
| 2717 |
vis_mul8x16al(DST_2, CONST_512, TMP4); |
| 2718 |
vis_padd16(TMP18, TMP14, TMP2); |
| 2719 |
|
| 2720 |
vis_mul8x16al(DST_3, CONST_512, TMP6); |
| 2721 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 2722 |
|
| 2723 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 2724 |
|
| 2725 |
vis_padd16(TMP0, TMP4, TMP0); |
| 2726 |
|
| 2727 |
vis_padd16(TMP2, TMP6, TMP2); |
| 2728 |
vis_pack16(TMP0, DST_2); |
| 2729 |
|
| 2730 |
vis_pack16(TMP2, DST_3); |
| 2731 |
vis_st64(DST_2, dest[8]);
|
| 2732 |
|
| 2733 |
ref += stride; |
| 2734 |
dest += stride; |
| 2735 |
} while (--height);
|
| 2736 |
} |
| 2737 |
|
| 2738 |
static void MC_avg_no_round_x_8_vis (uint8_t * dest, const uint8_t * ref, |
| 2739 |
const int stride, int height) |
| 2740 |
{
|
| 2741 |
unsigned long off = (unsigned long) ref & 0x7; |
| 2742 |
unsigned long off_plus_1 = off + 1; |
| 2743 |
int stride_times_2 = stride << 1; |
| 2744 |
|
| 2745 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 2746 |
|
| 2747 |
vis_ld64(constants3[0], CONST_3);
|
| 2748 |
vis_fzero(ZERO); |
| 2749 |
vis_ld64(constants256_512[0], CONST_256);
|
| 2750 |
|
| 2751 |
ref = vis_alignaddr(ref); |
| 2752 |
height >>= 2;
|
| 2753 |
do { /* 47 cycles */ |
| 2754 |
vis_ld64(ref[0], TMP0);
|
| 2755 |
|
| 2756 |
vis_ld64_2(ref, 8, TMP2);
|
| 2757 |
ref += stride; |
| 2758 |
|
| 2759 |
vis_alignaddr_g0((void *)off);
|
| 2760 |
|
| 2761 |
vis_ld64(ref[0], TMP4);
|
| 2762 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2763 |
|
| 2764 |
vis_ld64_2(ref, 8, TMP6);
|
| 2765 |
ref += stride; |
| 2766 |
|
| 2767 |
vis_ld64(ref[0], TMP8);
|
| 2768 |
|
| 2769 |
vis_ld64_2(ref, 8, TMP10);
|
| 2770 |
ref += stride; |
| 2771 |
vis_faligndata(TMP4, TMP6, REF_4); |
| 2772 |
|
| 2773 |
vis_ld64(ref[0], TMP12);
|
| 2774 |
|
| 2775 |
vis_ld64_2(ref, 8, TMP14);
|
| 2776 |
ref += stride; |
| 2777 |
vis_faligndata(TMP8, TMP10, REF_S0); |
| 2778 |
|
| 2779 |
vis_faligndata(TMP12, TMP14, REF_S4); |
| 2780 |
|
| 2781 |
if (off != 0x7) { |
| 2782 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 2783 |
|
| 2784 |
vis_ld64(dest[0], DST_0);
|
| 2785 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 2786 |
|
| 2787 |
vis_ld64_2(dest, stride, DST_2); |
| 2788 |
vis_faligndata(TMP4, TMP6, REF_6); |
| 2789 |
|
| 2790 |
vis_faligndata(TMP8, TMP10, REF_S2); |
| 2791 |
|
| 2792 |
vis_faligndata(TMP12, TMP14, REF_S6); |
| 2793 |
} else {
|
| 2794 |
vis_ld64(dest[0], DST_0);
|
| 2795 |
vis_src1(TMP2, REF_2); |
| 2796 |
|
| 2797 |
vis_ld64_2(dest, stride, DST_2); |
| 2798 |
vis_src1(TMP6, REF_6); |
| 2799 |
|
| 2800 |
vis_src1(TMP10, REF_S2); |
| 2801 |
|
| 2802 |
vis_src1(TMP14, REF_S6); |
| 2803 |
} |
| 2804 |
|
| 2805 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 2806 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 2807 |
|
| 2808 |
vis_pmerge(ZERO, REF_2, TMP4); |
| 2809 |
vis_mul8x16au(REF_2_1, CONST_256, TMP6); |
| 2810 |
|
| 2811 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 2812 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 2813 |
|
| 2814 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 2815 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 2816 |
|
| 2817 |
vis_padd16(TMP0, TMP4, TMP0); |
| 2818 |
vis_mul8x16au(REF_4, CONST_256, TMP8); |
| 2819 |
|
| 2820 |
vis_padd16(TMP2, TMP6, TMP2); |
| 2821 |
vis_mul8x16au(REF_4_1, CONST_256, TMP10); |
| 2822 |
|
| 2823 |
vis_padd16(TMP0, TMP16, TMP0); |
| 2824 |
vis_mul8x16au(REF_6, CONST_256, TMP12); |
| 2825 |
|
| 2826 |
vis_padd16(TMP2, TMP18, TMP2); |
| 2827 |
vis_mul8x16au(REF_6_1, CONST_256, TMP14); |
| 2828 |
|
| 2829 |
vis_padd16(TMP8, CONST_3, TMP8); |
| 2830 |
vis_mul8x16al(DST_2, CONST_512, TMP16); |
| 2831 |
|
| 2832 |
vis_padd16(TMP8, TMP12, TMP8); |
| 2833 |
vis_mul8x16al(DST_3, CONST_512, TMP18); |
| 2834 |
|
| 2835 |
vis_padd16(TMP10, TMP14, TMP10); |
| 2836 |
vis_pack16(TMP0, DST_0); |
| 2837 |
|
| 2838 |
vis_pack16(TMP2, DST_1); |
| 2839 |
vis_st64(DST_0, dest[0]);
|
| 2840 |
dest += stride; |
| 2841 |
vis_padd16(TMP10, CONST_3, TMP10); |
| 2842 |
|
| 2843 |
vis_ld64_2(dest, stride, DST_0); |
| 2844 |
vis_padd16(TMP8, TMP16, TMP8); |
| 2845 |
|
| 2846 |
vis_ld64_2(dest, stride_times_2, TMP4/*DST_2*/);
|
| 2847 |
vis_padd16(TMP10, TMP18, TMP10); |
| 2848 |
vis_pack16(TMP8, DST_2); |
| 2849 |
|
| 2850 |
vis_pack16(TMP10, DST_3); |
| 2851 |
vis_st64(DST_2, dest[0]);
|
| 2852 |
dest += stride; |
| 2853 |
|
| 2854 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP2); |
| 2855 |
vis_pmerge(ZERO, REF_S0, TMP0); |
| 2856 |
|
| 2857 |
vis_pmerge(ZERO, REF_S2, TMP24); |
| 2858 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP6); |
| 2859 |
|
| 2860 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 2861 |
vis_mul8x16au(REF_S4, CONST_256, TMP8); |
| 2862 |
|
| 2863 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 2864 |
vis_mul8x16au(REF_S4_1, CONST_256, TMP10); |
| 2865 |
|
| 2866 |
vis_padd16(TMP0, TMP24, TMP0); |
| 2867 |
vis_mul8x16au(REF_S6, CONST_256, TMP12); |
| 2868 |
|
| 2869 |
vis_padd16(TMP2, TMP6, TMP2); |
| 2870 |
vis_mul8x16au(REF_S6_1, CONST_256, TMP14); |
| 2871 |
|
| 2872 |
vis_padd16(TMP8, CONST_3, TMP8); |
| 2873 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 2874 |
|
| 2875 |
vis_padd16(TMP10, CONST_3, TMP10); |
| 2876 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 2877 |
|
| 2878 |
vis_padd16(TMP8, TMP12, TMP8); |
| 2879 |
vis_mul8x16al(TMP4/*DST_2*/, CONST_512, TMP20);
|
| 2880 |
|
| 2881 |
vis_mul8x16al(TMP5/*DST_3*/, CONST_512, TMP22);
|
| 2882 |
vis_padd16(TMP0, TMP16, TMP0); |
| 2883 |
|
| 2884 |
vis_padd16(TMP2, TMP18, TMP2); |
| 2885 |
vis_pack16(TMP0, DST_0); |
| 2886 |
|
| 2887 |
vis_padd16(TMP10, TMP14, TMP10); |
| 2888 |
vis_pack16(TMP2, DST_1); |
| 2889 |
vis_st64(DST_0, dest[0]);
|
| 2890 |
dest += stride; |
| 2891 |
|
| 2892 |
vis_padd16(TMP8, TMP20, TMP8); |
| 2893 |
|
| 2894 |
vis_padd16(TMP10, TMP22, TMP10); |
| 2895 |
vis_pack16(TMP8, DST_2); |
| 2896 |
|
| 2897 |
vis_pack16(TMP10, DST_3); |
| 2898 |
vis_st64(DST_2, dest[0]);
|
| 2899 |
dest += stride; |
| 2900 |
} while (--height);
|
| 2901 |
} |
| 2902 |
|
| 2903 |
static void MC_put_no_round_y_16_vis (uint8_t * dest, const uint8_t * ref, |
| 2904 |
const int stride, int height) |
| 2905 |
{
|
| 2906 |
ref = vis_alignaddr(ref); |
| 2907 |
vis_ld64(ref[0], TMP0);
|
| 2908 |
|
| 2909 |
vis_ld64_2(ref, 8, TMP2);
|
| 2910 |
|
| 2911 |
vis_ld64_2(ref, 16, TMP4);
|
| 2912 |
ref += stride; |
| 2913 |
|
| 2914 |
vis_ld64(ref[0], TMP6);
|
| 2915 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2916 |
|
| 2917 |
vis_ld64_2(ref, 8, TMP8);
|
| 2918 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2919 |
|
| 2920 |
vis_ld64_2(ref, 16, TMP10);
|
| 2921 |
ref += stride; |
| 2922 |
|
| 2923 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 2924 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 2925 |
|
| 2926 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 2927 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 2928 |
|
| 2929 |
vis_ld64(constants128[0], CONST_128);
|
| 2930 |
height = (height >> 1) - 1; |
| 2931 |
do { /* 24 cycles */ |
| 2932 |
vis_ld64(ref[0], TMP0);
|
| 2933 |
vis_xor(REF_0, REF_2, TMP12); |
| 2934 |
|
| 2935 |
vis_ld64_2(ref, 8, TMP2);
|
| 2936 |
vis_xor(REF_4, REF_6, TMP16); |
| 2937 |
|
| 2938 |
vis_ld64_2(ref, 16, TMP4);
|
| 2939 |
ref += stride; |
| 2940 |
vis_and(REF_0, REF_2, TMP14); |
| 2941 |
|
| 2942 |
vis_ld64(ref[0], TMP6);
|
| 2943 |
vis_and(REF_4, REF_6, TMP18); |
| 2944 |
|
| 2945 |
vis_ld64_2(ref, 8, TMP8);
|
| 2946 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 2947 |
|
| 2948 |
vis_ld64_2(ref, 16, TMP10);
|
| 2949 |
ref += stride; |
| 2950 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 2951 |
|
| 2952 |
vis_and(TMP12, MASK_fe, TMP12); |
| 2953 |
|
| 2954 |
vis_and(TMP16, MASK_fe, TMP16); |
| 2955 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 2956 |
|
| 2957 |
vis_mul8x16(CONST_128, TMP16, TMP16); |
| 2958 |
vis_xor(REF_0, REF_2, TMP0); |
| 2959 |
|
| 2960 |
vis_xor(REF_4, REF_6, TMP2); |
| 2961 |
|
| 2962 |
vis_and(REF_0, REF_2, TMP20); |
| 2963 |
|
| 2964 |
vis_and(TMP12, MASK_7f, TMP12); |
| 2965 |
|
| 2966 |
vis_and(TMP16, MASK_7f, TMP16); |
| 2967 |
|
| 2968 |
vis_padd16(TMP14, TMP12, TMP12); |
| 2969 |
vis_st64(TMP12, dest[0]);
|
| 2970 |
|
| 2971 |
vis_padd16(TMP18, TMP16, TMP16); |
| 2972 |
vis_st64_2(TMP16, dest, 8);
|
| 2973 |
dest += stride; |
| 2974 |
|
| 2975 |
vis_and(REF_4, REF_6, TMP18); |
| 2976 |
|
| 2977 |
vis_and(TMP0, MASK_fe, TMP0); |
| 2978 |
|
| 2979 |
vis_and(TMP2, MASK_fe, TMP2); |
| 2980 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 2981 |
|
| 2982 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 2983 |
vis_mul8x16(CONST_128, TMP2, TMP2); |
| 2984 |
|
| 2985 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 2986 |
|
| 2987 |
vis_and(TMP0, MASK_7f, TMP0); |
| 2988 |
|
| 2989 |
vis_and(TMP2, MASK_7f, TMP2); |
| 2990 |
|
| 2991 |
vis_padd16(TMP20, TMP0, TMP0); |
| 2992 |
vis_st64(TMP0, dest[0]);
|
| 2993 |
|
| 2994 |
vis_padd16(TMP18, TMP2, TMP2); |
| 2995 |
vis_st64_2(TMP2, dest, 8);
|
| 2996 |
dest += stride; |
| 2997 |
} while (--height);
|
| 2998 |
|
| 2999 |
vis_ld64(ref[0], TMP0);
|
| 3000 |
vis_xor(REF_0, REF_2, TMP12); |
| 3001 |
|
| 3002 |
vis_ld64_2(ref, 8, TMP2);
|
| 3003 |
vis_xor(REF_4, REF_6, TMP16); |
| 3004 |
|
| 3005 |
vis_ld64_2(ref, 16, TMP4);
|
| 3006 |
vis_and(REF_0, REF_2, TMP14); |
| 3007 |
|
| 3008 |
vis_and(REF_4, REF_6, TMP18); |
| 3009 |
|
| 3010 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3011 |
|
| 3012 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 3013 |
|
| 3014 |
vis_and(TMP12, MASK_fe, TMP12); |
| 3015 |
|
| 3016 |
vis_and(TMP16, MASK_fe, TMP16); |
| 3017 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 3018 |
|
| 3019 |
vis_mul8x16(CONST_128, TMP16, TMP16); |
| 3020 |
vis_xor(REF_0, REF_2, TMP0); |
| 3021 |
|
| 3022 |
vis_xor(REF_4, REF_6, TMP2); |
| 3023 |
|
| 3024 |
vis_and(REF_0, REF_2, TMP20); |
| 3025 |
|
| 3026 |
vis_and(TMP12, MASK_7f, TMP12); |
| 3027 |
|
| 3028 |
vis_and(TMP16, MASK_7f, TMP16); |
| 3029 |
|
| 3030 |
vis_padd16(TMP14, TMP12, TMP12); |
| 3031 |
vis_st64(TMP12, dest[0]);
|
| 3032 |
|
| 3033 |
vis_padd16(TMP18, TMP16, TMP16); |
| 3034 |
vis_st64_2(TMP16, dest, 8);
|
| 3035 |
dest += stride; |
| 3036 |
|
| 3037 |
vis_and(REF_4, REF_6, TMP18); |
| 3038 |
|
| 3039 |
vis_and(TMP0, MASK_fe, TMP0); |
| 3040 |
|
| 3041 |
vis_and(TMP2, MASK_fe, TMP2); |
| 3042 |
vis_mul8x16(CONST_128, TMP0, TMP0); |
| 3043 |
|
| 3044 |
vis_mul8x16(CONST_128, TMP2, TMP2); |
| 3045 |
|
| 3046 |
vis_and(TMP0, MASK_7f, TMP0); |
| 3047 |
|
| 3048 |
vis_and(TMP2, MASK_7f, TMP2); |
| 3049 |
|
| 3050 |
vis_padd16(TMP20, TMP0, TMP0); |
| 3051 |
vis_st64(TMP0, dest[0]);
|
| 3052 |
|
| 3053 |
vis_padd16(TMP18, TMP2, TMP2); |
| 3054 |
vis_st64_2(TMP2, dest, 8);
|
| 3055 |
} |
| 3056 |
|
| 3057 |
static void MC_put_no_round_y_8_vis (uint8_t * dest, const uint8_t * ref, |
| 3058 |
const int stride, int height) |
| 3059 |
{
|
| 3060 |
ref = vis_alignaddr(ref); |
| 3061 |
vis_ld64(ref[0], TMP0);
|
| 3062 |
|
| 3063 |
vis_ld64_2(ref, 8, TMP2);
|
| 3064 |
ref += stride; |
| 3065 |
|
| 3066 |
vis_ld64(ref[0], TMP4);
|
| 3067 |
|
| 3068 |
vis_ld64_2(ref, 8, TMP6);
|
| 3069 |
ref += stride; |
| 3070 |
|
| 3071 |
vis_ld64(constants_fe[0], MASK_fe);
|
| 3072 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3073 |
|
| 3074 |
vis_ld64(constants_7f[0], MASK_7f);
|
| 3075 |
vis_faligndata(TMP4, TMP6, REF_2); |
| 3076 |
|
| 3077 |
vis_ld64(constants128[0], CONST_128);
|
| 3078 |
height = (height >> 1) - 1; |
| 3079 |
do { /* 12 cycles */ |
| 3080 |
vis_ld64(ref[0], TMP0);
|
| 3081 |
vis_xor(REF_0, REF_2, TMP4); |
| 3082 |
|
| 3083 |
vis_ld64_2(ref, 8, TMP2);
|
| 3084 |
ref += stride; |
| 3085 |
vis_and(TMP4, MASK_fe, TMP4); |
| 3086 |
|
| 3087 |
vis_and(REF_0, REF_2, TMP6); |
| 3088 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 3089 |
|
| 3090 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3091 |
vis_ld64(ref[0], TMP0);
|
| 3092 |
|
| 3093 |
vis_ld64_2(ref, 8, TMP2);
|
| 3094 |
ref += stride; |
| 3095 |
vis_xor(REF_0, REF_2, TMP12); |
| 3096 |
|
| 3097 |
vis_and(TMP4, MASK_7f, TMP4); |
| 3098 |
|
| 3099 |
vis_and(TMP12, MASK_fe, TMP12); |
| 3100 |
|
| 3101 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 3102 |
vis_and(REF_0, REF_2, TMP14); |
| 3103 |
|
| 3104 |
vis_padd16(TMP6, TMP4, DST_0); |
| 3105 |
vis_st64(DST_0, dest[0]);
|
| 3106 |
dest += stride; |
| 3107 |
|
| 3108 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 3109 |
|
| 3110 |
vis_and(TMP12, MASK_7f, TMP12); |
| 3111 |
|
| 3112 |
vis_padd16(TMP14, TMP12, DST_0); |
| 3113 |
vis_st64(DST_0, dest[0]);
|
| 3114 |
dest += stride; |
| 3115 |
} while (--height);
|
| 3116 |
|
| 3117 |
vis_ld64(ref[0], TMP0);
|
| 3118 |
vis_xor(REF_0, REF_2, TMP4); |
| 3119 |
|
| 3120 |
vis_ld64_2(ref, 8, TMP2);
|
| 3121 |
vis_and(TMP4, MASK_fe, TMP4); |
| 3122 |
|
| 3123 |
vis_and(REF_0, REF_2, TMP6); |
| 3124 |
vis_mul8x16(CONST_128, TMP4, TMP4); |
| 3125 |
|
| 3126 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3127 |
|
| 3128 |
vis_xor(REF_0, REF_2, TMP12); |
| 3129 |
|
| 3130 |
vis_and(TMP4, MASK_7f, TMP4); |
| 3131 |
|
| 3132 |
vis_and(TMP12, MASK_fe, TMP12); |
| 3133 |
|
| 3134 |
vis_mul8x16(CONST_128, TMP12, TMP12); |
| 3135 |
vis_and(REF_0, REF_2, TMP14); |
| 3136 |
|
| 3137 |
vis_padd16(TMP6, TMP4, DST_0); |
| 3138 |
vis_st64(DST_0, dest[0]);
|
| 3139 |
dest += stride; |
| 3140 |
|
| 3141 |
vis_and(TMP12, MASK_7f, TMP12); |
| 3142 |
|
| 3143 |
vis_padd16(TMP14, TMP12, DST_0); |
| 3144 |
vis_st64(DST_0, dest[0]);
|
| 3145 |
} |
| 3146 |
|
| 3147 |
static void MC_avg_no_round_y_16_vis (uint8_t * dest, const uint8_t * ref, |
| 3148 |
const int stride, int height) |
| 3149 |
{
|
| 3150 |
int stride_8 = stride + 8; |
| 3151 |
int stride_16 = stride + 16; |
| 3152 |
|
| 3153 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3154 |
|
| 3155 |
ref = vis_alignaddr(ref); |
| 3156 |
|
| 3157 |
vis_ld64(ref[ 0], TMP0);
|
| 3158 |
vis_fzero(ZERO); |
| 3159 |
|
| 3160 |
vis_ld64(ref[ 8], TMP2);
|
| 3161 |
|
| 3162 |
vis_ld64(ref[16], TMP4);
|
| 3163 |
|
| 3164 |
vis_ld64(constants3[0], CONST_3);
|
| 3165 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 3166 |
|
| 3167 |
vis_ld64(constants256_512[0], CONST_256);
|
| 3168 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 3169 |
height >>= 1;
|
| 3170 |
|
| 3171 |
do { /* 31 cycles */ |
| 3172 |
vis_ld64_2(ref, stride, TMP0); |
| 3173 |
vis_pmerge(ZERO, REF_2, TMP12); |
| 3174 |
vis_mul8x16au(REF_2_1, CONST_256, TMP14); |
| 3175 |
|
| 3176 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3177 |
vis_pmerge(ZERO, REF_6, TMP16); |
| 3178 |
vis_mul8x16au(REF_6_1, CONST_256, TMP18); |
| 3179 |
|
| 3180 |
vis_ld64_2(ref, stride_16, TMP4); |
| 3181 |
ref += stride; |
| 3182 |
|
| 3183 |
vis_ld64(dest[0], DST_0);
|
| 3184 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3185 |
|
| 3186 |
vis_ld64_2(dest, 8, DST_2);
|
| 3187 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 3188 |
|
| 3189 |
vis_ld64_2(ref, stride, TMP6); |
| 3190 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 3191 |
vis_mul8x16au(REF_0_1, CONST_256, TMP2); |
| 3192 |
|
| 3193 |
vis_ld64_2(ref, stride_8, TMP8); |
| 3194 |
vis_pmerge(ZERO, REF_4, TMP4); |
| 3195 |
|
| 3196 |
vis_ld64_2(ref, stride_16, TMP10); |
| 3197 |
ref += stride; |
| 3198 |
|
| 3199 |
vis_ld64_2(dest, stride, REF_S0/*DST_4*/);
|
| 3200 |
vis_faligndata(TMP6, TMP8, REF_2); |
| 3201 |
vis_mul8x16au(REF_4_1, CONST_256, TMP6); |
| 3202 |
|
| 3203 |
vis_ld64_2(dest, stride_8, REF_S2/*DST_6*/);
|
| 3204 |
vis_faligndata(TMP8, TMP10, REF_6); |
| 3205 |
vis_mul8x16al(DST_0, CONST_512, TMP20); |
| 3206 |
|
| 3207 |
vis_padd16(TMP0, CONST_3, TMP0); |
| 3208 |
vis_mul8x16al(DST_1, CONST_512, TMP22); |
| 3209 |
|
| 3210 |
vis_padd16(TMP2, CONST_3, TMP2); |
| 3211 |
vis_mul8x16al(DST_2, CONST_512, TMP24); |
| 3212 |
|
| 3213 |
vis_padd16(TMP4, CONST_3, TMP4); |
| 3214 |
vis_mul8x16al(DST_3, CONST_512, TMP26); |
| 3215 |
|
| 3216 |
vis_padd16(TMP6, CONST_3, TMP6); |
| 3217 |
|
| 3218 |
vis_padd16(TMP12, TMP20, TMP12); |
| 3219 |
vis_mul8x16al(REF_S0, CONST_512, TMP20); |
| 3220 |
|
| 3221 |
vis_padd16(TMP14, TMP22, TMP14); |
| 3222 |
vis_mul8x16al(REF_S0_1, CONST_512, TMP22); |
| 3223 |
|
| 3224 |
vis_padd16(TMP16, TMP24, TMP16); |
| 3225 |
vis_mul8x16al(REF_S2, CONST_512, TMP24); |
| 3226 |
|
| 3227 |
vis_padd16(TMP18, TMP26, TMP18); |
| 3228 |
vis_mul8x16al(REF_S2_1, CONST_512, TMP26); |
| 3229 |
|
| 3230 |
vis_padd16(TMP12, TMP0, TMP12); |
| 3231 |
vis_mul8x16au(REF_2, CONST_256, TMP28); |
| 3232 |
|
| 3233 |
vis_padd16(TMP14, TMP2, TMP14); |
| 3234 |
vis_mul8x16au(REF_2_1, CONST_256, TMP30); |
| 3235 |
|
| 3236 |
vis_padd16(TMP16, TMP4, TMP16); |
| 3237 |
vis_mul8x16au(REF_6, CONST_256, REF_S4); |
| 3238 |
|
| 3239 |
vis_padd16(TMP18, TMP6, TMP18); |
| 3240 |
vis_mul8x16au(REF_6_1, CONST_256, REF_S6); |
| 3241 |
|
| 3242 |
vis_pack16(TMP12, DST_0); |
| 3243 |
vis_padd16(TMP28, TMP0, TMP12); |
| 3244 |
|
| 3245 |
vis_pack16(TMP14, DST_1); |
| 3246 |
vis_st64(DST_0, dest[0]);
|
| 3247 |
vis_padd16(TMP30, TMP2, TMP14); |
| 3248 |
|
| 3249 |
vis_pack16(TMP16, DST_2); |
| 3250 |
vis_padd16(REF_S4, TMP4, TMP16); |
| 3251 |
|
| 3252 |
vis_pack16(TMP18, DST_3); |
| 3253 |
vis_st64_2(DST_2, dest, 8);
|
| 3254 |
dest += stride; |
| 3255 |
vis_padd16(REF_S6, TMP6, TMP18); |
| 3256 |
|
| 3257 |
vis_padd16(TMP12, TMP20, TMP12); |
| 3258 |
|
| 3259 |
vis_padd16(TMP14, TMP22, TMP14); |
| 3260 |
vis_pack16(TMP12, DST_0); |
| 3261 |
|
| 3262 |
vis_padd16(TMP16, TMP24, TMP16); |
| 3263 |
vis_pack16(TMP14, DST_1); |
| 3264 |
vis_st64(DST_0, dest[0]);
|
| 3265 |
|
| 3266 |
vis_padd16(TMP18, TMP26, TMP18); |
| 3267 |
vis_pack16(TMP16, DST_2); |
| 3268 |
|
| 3269 |
vis_pack16(TMP18, DST_3); |
| 3270 |
vis_st64_2(DST_2, dest, 8);
|
| 3271 |
dest += stride; |
| 3272 |
} while (--height);
|
| 3273 |
} |
| 3274 |
|
| 3275 |
static void MC_avg_no_round_y_8_vis (uint8_t * dest, const uint8_t * ref, |
| 3276 |
const int stride, int height) |
| 3277 |
{
|
| 3278 |
int stride_8 = stride + 8; |
| 3279 |
|
| 3280 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3281 |
|
| 3282 |
ref = vis_alignaddr(ref); |
| 3283 |
|
| 3284 |
vis_ld64(ref[ 0], TMP0);
|
| 3285 |
vis_fzero(ZERO); |
| 3286 |
|
| 3287 |
vis_ld64(ref[ 8], TMP2);
|
| 3288 |
|
| 3289 |
vis_ld64(constants3[0], CONST_3);
|
| 3290 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 3291 |
|
| 3292 |
vis_ld64(constants256_512[0], CONST_256);
|
| 3293 |
|
| 3294 |
height >>= 1;
|
| 3295 |
do { /* 20 cycles */ |
| 3296 |
vis_ld64_2(ref, stride, TMP0); |
| 3297 |
vis_pmerge(ZERO, REF_2, TMP8); |
| 3298 |
vis_mul8x16au(REF_2_1, CONST_256, TMP10); |
| 3299 |
|
| 3300 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3301 |
ref += stride; |
| 3302 |
|
| 3303 |
vis_ld64(dest[0], DST_0);
|
| 3304 |
|
| 3305 |
vis_ld64_2(dest, stride, DST_2); |
| 3306 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3307 |
|
| 3308 |
vis_ld64_2(ref, stride, TMP4); |
| 3309 |
vis_mul8x16al(DST_0, CONST_512, TMP16); |
| 3310 |
vis_pmerge(ZERO, REF_0, TMP12); |
| 3311 |
|
| 3312 |
vis_ld64_2(ref, stride_8, TMP6); |
| 3313 |
ref += stride; |
| 3314 |
vis_mul8x16al(DST_1, CONST_512, TMP18); |
| 3315 |
vis_pmerge(ZERO, REF_0_1, TMP14); |
| 3316 |
|
| 3317 |
vis_padd16(TMP12, CONST_3, TMP12); |
| 3318 |
vis_mul8x16al(DST_2, CONST_512, TMP24); |
| 3319 |
|
| 3320 |
vis_padd16(TMP14, CONST_3, TMP14); |
| 3321 |
vis_mul8x16al(DST_3, CONST_512, TMP26); |
| 3322 |
|
| 3323 |
vis_faligndata(TMP4, TMP6, REF_2); |
| 3324 |
|
| 3325 |
vis_padd16(TMP8, TMP12, TMP8); |
| 3326 |
|
| 3327 |
vis_padd16(TMP10, TMP14, TMP10); |
| 3328 |
vis_mul8x16au(REF_2, CONST_256, TMP20); |
| 3329 |
|
| 3330 |
vis_padd16(TMP8, TMP16, TMP0); |
| 3331 |
vis_mul8x16au(REF_2_1, CONST_256, TMP22); |
| 3332 |
|
| 3333 |
vis_padd16(TMP10, TMP18, TMP2); |
| 3334 |
vis_pack16(TMP0, DST_0); |
| 3335 |
|
| 3336 |
vis_pack16(TMP2, DST_1); |
| 3337 |
vis_st64(DST_0, dest[0]);
|
| 3338 |
dest += stride; |
| 3339 |
vis_padd16(TMP12, TMP20, TMP12); |
| 3340 |
|
| 3341 |
vis_padd16(TMP14, TMP22, TMP14); |
| 3342 |
|
| 3343 |
vis_padd16(TMP12, TMP24, TMP0); |
| 3344 |
|
| 3345 |
vis_padd16(TMP14, TMP26, TMP2); |
| 3346 |
vis_pack16(TMP0, DST_2); |
| 3347 |
|
| 3348 |
vis_pack16(TMP2, DST_3); |
| 3349 |
vis_st64(DST_2, dest[0]);
|
| 3350 |
dest += stride; |
| 3351 |
} while (--height);
|
| 3352 |
} |
| 3353 |
|
| 3354 |
static void MC_put_no_round_xy_16_vis (uint8_t * dest, const uint8_t * ref, |
| 3355 |
const int stride, int height) |
| 3356 |
{
|
| 3357 |
unsigned long off = (unsigned long) ref & 0x7; |
| 3358 |
unsigned long off_plus_1 = off + 1; |
| 3359 |
int stride_8 = stride + 8; |
| 3360 |
int stride_16 = stride + 16; |
| 3361 |
|
| 3362 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3363 |
|
| 3364 |
ref = vis_alignaddr(ref); |
| 3365 |
|
| 3366 |
vis_ld64(ref[ 0], TMP0);
|
| 3367 |
vis_fzero(ZERO); |
| 3368 |
|
| 3369 |
vis_ld64(ref[ 8], TMP2);
|
| 3370 |
|
| 3371 |
vis_ld64(ref[16], TMP4);
|
| 3372 |
|
| 3373 |
vis_ld64(constants1[0], CONST_1);
|
| 3374 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 3375 |
|
| 3376 |
vis_ld64(constants256_512[0], CONST_256);
|
| 3377 |
vis_faligndata(TMP2, TMP4, REF_S4); |
| 3378 |
|
| 3379 |
if (off != 0x7) { |
| 3380 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3381 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 3382 |
vis_faligndata(TMP2, TMP4, REF_S6); |
| 3383 |
} else {
|
| 3384 |
vis_src1(TMP2, REF_S2); |
| 3385 |
vis_src1(TMP4, REF_S6); |
| 3386 |
} |
| 3387 |
|
| 3388 |
height >>= 1;
|
| 3389 |
do {
|
| 3390 |
vis_ld64_2(ref, stride, TMP0); |
| 3391 |
vis_mul8x16au(REF_S0, CONST_256, TMP12); |
| 3392 |
vis_pmerge(ZERO, REF_S0_1, TMP14); |
| 3393 |
|
| 3394 |
vis_alignaddr_g0((void *)off);
|
| 3395 |
|
| 3396 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3397 |
vis_mul8x16au(REF_S2, CONST_256, TMP16); |
| 3398 |
vis_pmerge(ZERO, REF_S2_1, TMP18); |
| 3399 |
|
| 3400 |
vis_ld64_2(ref, stride_16, TMP4); |
| 3401 |
ref += stride; |
| 3402 |
vis_mul8x16au(REF_S4, CONST_256, TMP20); |
| 3403 |
vis_pmerge(ZERO, REF_S4_1, TMP22); |
| 3404 |
|
| 3405 |
vis_ld64_2(ref, stride, TMP6); |
| 3406 |
vis_mul8x16au(REF_S6, CONST_256, TMP24); |
| 3407 |
vis_pmerge(ZERO, REF_S6_1, TMP26); |
| 3408 |
|
| 3409 |
vis_ld64_2(ref, stride_8, TMP8); |
| 3410 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3411 |
|
| 3412 |
vis_ld64_2(ref, stride_16, TMP10); |
| 3413 |
ref += stride; |
| 3414 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 3415 |
|
| 3416 |
vis_faligndata(TMP6, TMP8, REF_S0); |
| 3417 |
|
| 3418 |
vis_faligndata(TMP8, TMP10, REF_S4); |
| 3419 |
|
| 3420 |
if (off != 0x7) { |
| 3421 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3422 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 3423 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 3424 |
vis_faligndata(TMP6, TMP8, REF_S2); |
| 3425 |
vis_faligndata(TMP8, TMP10, REF_S6); |
| 3426 |
} else {
|
| 3427 |
vis_src1(TMP2, REF_2); |
| 3428 |
vis_src1(TMP4, REF_6); |
| 3429 |
vis_src1(TMP8, REF_S2); |
| 3430 |
vis_src1(TMP10, REF_S6); |
| 3431 |
} |
| 3432 |
|
| 3433 |
vis_mul8x16au(REF_0, CONST_256, TMP0); |
| 3434 |
vis_pmerge(ZERO, REF_0_1, TMP2); |
| 3435 |
|
| 3436 |
vis_mul8x16au(REF_2, CONST_256, TMP4); |
| 3437 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 3438 |
|
| 3439 |
vis_padd16(TMP0, CONST_2, TMP8); |
| 3440 |
vis_mul8x16au(REF_4, CONST_256, TMP0); |
| 3441 |
|
| 3442 |
vis_padd16(TMP2, CONST_1, TMP10); |
| 3443 |
vis_mul8x16au(REF_4_1, CONST_256, TMP2); |
| 3444 |
|
| 3445 |
vis_padd16(TMP8, TMP4, TMP8); |
| 3446 |
vis_mul8x16au(REF_6, CONST_256, TMP4); |
| 3447 |
|
| 3448 |
vis_padd16(TMP10, TMP6, TMP10); |
| 3449 |
vis_mul8x16au(REF_6_1, CONST_256, TMP6); |
| 3450 |
|
| 3451 |
vis_padd16(TMP12, TMP8, TMP12); |
| 3452 |
|
| 3453 |
vis_padd16(TMP14, TMP10, TMP14); |
| 3454 |
|
| 3455 |
vis_padd16(TMP12, TMP16, TMP12); |
| 3456 |
|
| 3457 |
vis_padd16(TMP14, TMP18, TMP14); |
| 3458 |
vis_pack16(TMP12, DST_0); |
| 3459 |
|
| 3460 |
vis_pack16(TMP14, DST_1); |
| 3461 |
vis_st64(DST_0, dest[0]);
|
| 3462 |
vis_padd16(TMP0, CONST_1, TMP12); |
| 3463 |
|
| 3464 |
vis_mul8x16au(REF_S0, CONST_256, TMP0); |
| 3465 |
vis_padd16(TMP2, CONST_1, TMP14); |
| 3466 |
|
| 3467 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP2); |
| 3468 |
vis_padd16(TMP12, TMP4, TMP12); |
| 3469 |
|
| 3470 |
vis_mul8x16au(REF_S2, CONST_256, TMP4); |
| 3471 |
vis_padd16(TMP14, TMP6, TMP14); |
| 3472 |
|
| 3473 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP6); |
| 3474 |
vis_padd16(TMP20, TMP12, TMP20); |
| 3475 |
|
| 3476 |
vis_padd16(TMP22, TMP14, TMP22); |
| 3477 |
|
| 3478 |
vis_padd16(TMP20, TMP24, TMP20); |
| 3479 |
|
| 3480 |
vis_padd16(TMP22, TMP26, TMP22); |
| 3481 |
vis_pack16(TMP20, DST_2); |
| 3482 |
|
| 3483 |
vis_pack16(TMP22, DST_3); |
| 3484 |
vis_st64_2(DST_2, dest, 8);
|
| 3485 |
dest += stride; |
| 3486 |
vis_padd16(TMP0, TMP4, TMP24); |
| 3487 |
|
| 3488 |
vis_mul8x16au(REF_S4, CONST_256, TMP0); |
| 3489 |
vis_padd16(TMP2, TMP6, TMP26); |
| 3490 |
|
| 3491 |
vis_mul8x16au(REF_S4_1, CONST_256, TMP2); |
| 3492 |
vis_padd16(TMP24, TMP8, TMP24); |
| 3493 |
|
| 3494 |
vis_padd16(TMP26, TMP10, TMP26); |
| 3495 |
vis_pack16(TMP24, DST_0); |
| 3496 |
|
| 3497 |
vis_pack16(TMP26, DST_1); |
| 3498 |
vis_st64(DST_0, dest[0]);
|
| 3499 |
vis_pmerge(ZERO, REF_S6, TMP4); |
| 3500 |
|
| 3501 |
vis_pmerge(ZERO, REF_S6_1, TMP6); |
| 3502 |
|
| 3503 |
vis_padd16(TMP0, TMP4, TMP0); |
| 3504 |
|
| 3505 |
vis_padd16(TMP2, TMP6, TMP2); |
| 3506 |
|
| 3507 |
vis_padd16(TMP0, TMP12, TMP0); |
| 3508 |
|
| 3509 |
vis_padd16(TMP2, TMP14, TMP2); |
| 3510 |
vis_pack16(TMP0, DST_2); |
| 3511 |
|
| 3512 |
vis_pack16(TMP2, DST_3); |
| 3513 |
vis_st64_2(DST_2, dest, 8);
|
| 3514 |
dest += stride; |
| 3515 |
} while (--height);
|
| 3516 |
} |
| 3517 |
|
| 3518 |
static void MC_put_no_round_xy_8_vis (uint8_t * dest, const uint8_t * ref, |
| 3519 |
const int stride, int height) |
| 3520 |
{
|
| 3521 |
unsigned long off = (unsigned long) ref & 0x7; |
| 3522 |
unsigned long off_plus_1 = off + 1; |
| 3523 |
int stride_8 = stride + 8; |
| 3524 |
|
| 3525 |
vis_set_gsr(5 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3526 |
|
| 3527 |
ref = vis_alignaddr(ref); |
| 3528 |
|
| 3529 |
vis_ld64(ref[ 0], TMP0);
|
| 3530 |
vis_fzero(ZERO); |
| 3531 |
|
| 3532 |
vis_ld64(ref[ 8], TMP2);
|
| 3533 |
|
| 3534 |
vis_ld64(constants1[0], CONST_1);
|
| 3535 |
|
| 3536 |
vis_ld64(constants256_512[0], CONST_256);
|
| 3537 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 3538 |
|
| 3539 |
if (off != 0x7) { |
| 3540 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3541 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 3542 |
} else {
|
| 3543 |
vis_src1(TMP2, REF_S2); |
| 3544 |
} |
| 3545 |
|
| 3546 |
height >>= 1;
|
| 3547 |
do { /* 26 cycles */ |
| 3548 |
vis_ld64_2(ref, stride, TMP0); |
| 3549 |
vis_mul8x16au(REF_S0, CONST_256, TMP8); |
| 3550 |
vis_pmerge(ZERO, REF_S2, TMP12); |
| 3551 |
|
| 3552 |
vis_alignaddr_g0((void *)off);
|
| 3553 |
|
| 3554 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3555 |
ref += stride; |
| 3556 |
vis_mul8x16au(REF_S0_1, CONST_256, TMP10); |
| 3557 |
vis_pmerge(ZERO, REF_S2_1, TMP14); |
| 3558 |
|
| 3559 |
vis_ld64_2(ref, stride, TMP4); |
| 3560 |
|
| 3561 |
vis_ld64_2(ref, stride_8, TMP6); |
| 3562 |
ref += stride; |
| 3563 |
vis_faligndata(TMP0, TMP2, REF_S4); |
| 3564 |
|
| 3565 |
vis_pmerge(ZERO, REF_S4, TMP18); |
| 3566 |
|
| 3567 |
vis_pmerge(ZERO, REF_S4_1, TMP20); |
| 3568 |
|
| 3569 |
vis_faligndata(TMP4, TMP6, REF_S0); |
| 3570 |
|
| 3571 |
if (off != 0x7) { |
| 3572 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3573 |
vis_faligndata(TMP0, TMP2, REF_S6); |
| 3574 |
vis_faligndata(TMP4, TMP6, REF_S2); |
| 3575 |
} else {
|
| 3576 |
vis_src1(TMP2, REF_S6); |
| 3577 |
vis_src1(TMP6, REF_S2); |
| 3578 |
} |
| 3579 |
|
| 3580 |
vis_padd16(TMP18, CONST_1, TMP18); |
| 3581 |
vis_mul8x16au(REF_S6, CONST_256, TMP22); |
| 3582 |
|
| 3583 |
vis_padd16(TMP20, CONST_1, TMP20); |
| 3584 |
vis_mul8x16au(REF_S6_1, CONST_256, TMP24); |
| 3585 |
|
| 3586 |
vis_mul8x16au(REF_S0, CONST_256, TMP26); |
| 3587 |
vis_pmerge(ZERO, REF_S0_1, TMP28); |
| 3588 |
|
| 3589 |
vis_mul8x16au(REF_S2, CONST_256, TMP30); |
| 3590 |
vis_padd16(TMP18, TMP22, TMP18); |
| 3591 |
|
| 3592 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP32); |
| 3593 |
vis_padd16(TMP20, TMP24, TMP20); |
| 3594 |
|
| 3595 |
vis_padd16(TMP8, TMP18, TMP8); |
| 3596 |
|
| 3597 |
vis_padd16(TMP10, TMP20, TMP10); |
| 3598 |
|
| 3599 |
vis_padd16(TMP8, TMP12, TMP8); |
| 3600 |
|
| 3601 |
vis_padd16(TMP10, TMP14, TMP10); |
| 3602 |
vis_pack16(TMP8, DST_0); |
| 3603 |
|
| 3604 |
vis_pack16(TMP10, DST_1); |
| 3605 |
vis_st64(DST_0, dest[0]);
|
| 3606 |
dest += stride; |
| 3607 |
vis_padd16(TMP18, TMP26, TMP18); |
| 3608 |
|
| 3609 |
vis_padd16(TMP20, TMP28, TMP20); |
| 3610 |
|
| 3611 |
vis_padd16(TMP18, TMP30, TMP18); |
| 3612 |
|
| 3613 |
vis_padd16(TMP20, TMP32, TMP20); |
| 3614 |
vis_pack16(TMP18, DST_2); |
| 3615 |
|
| 3616 |
vis_pack16(TMP20, DST_3); |
| 3617 |
vis_st64(DST_2, dest[0]);
|
| 3618 |
dest += stride; |
| 3619 |
} while (--height);
|
| 3620 |
} |
| 3621 |
|
| 3622 |
static void MC_avg_no_round_xy_16_vis (uint8_t * dest, const uint8_t * ref, |
| 3623 |
const int stride, int height) |
| 3624 |
{
|
| 3625 |
unsigned long off = (unsigned long) ref & 0x7; |
| 3626 |
unsigned long off_plus_1 = off + 1; |
| 3627 |
int stride_8 = stride + 8; |
| 3628 |
int stride_16 = stride + 16; |
| 3629 |
|
| 3630 |
vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3631 |
|
| 3632 |
ref = vis_alignaddr(ref); |
| 3633 |
|
| 3634 |
vis_ld64(ref[ 0], TMP0);
|
| 3635 |
vis_fzero(ZERO); |
| 3636 |
|
| 3637 |
vis_ld64(ref[ 8], TMP2);
|
| 3638 |
|
| 3639 |
vis_ld64(ref[16], TMP4);
|
| 3640 |
|
| 3641 |
vis_ld64(constants6[0], CONST_6);
|
| 3642 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 3643 |
|
| 3644 |
vis_ld64(constants256_1024[0], CONST_256);
|
| 3645 |
vis_faligndata(TMP2, TMP4, REF_S4); |
| 3646 |
|
| 3647 |
if (off != 0x7) { |
| 3648 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3649 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 3650 |
vis_faligndata(TMP2, TMP4, REF_S6); |
| 3651 |
} else {
|
| 3652 |
vis_src1(TMP2, REF_S2); |
| 3653 |
vis_src1(TMP4, REF_S6); |
| 3654 |
} |
| 3655 |
|
| 3656 |
height >>= 1;
|
| 3657 |
do { /* 55 cycles */ |
| 3658 |
vis_ld64_2(ref, stride, TMP0); |
| 3659 |
vis_mul8x16au(REF_S0, CONST_256, TMP12); |
| 3660 |
vis_pmerge(ZERO, REF_S0_1, TMP14); |
| 3661 |
|
| 3662 |
vis_alignaddr_g0((void *)off);
|
| 3663 |
|
| 3664 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3665 |
vis_mul8x16au(REF_S2, CONST_256, TMP16); |
| 3666 |
vis_pmerge(ZERO, REF_S2_1, TMP18); |
| 3667 |
|
| 3668 |
vis_ld64_2(ref, stride_16, TMP4); |
| 3669 |
ref += stride; |
| 3670 |
vis_mul8x16au(REF_S4, CONST_256, TMP20); |
| 3671 |
vis_pmerge(ZERO, REF_S4_1, TMP22); |
| 3672 |
|
| 3673 |
vis_ld64_2(ref, stride, TMP6); |
| 3674 |
vis_mul8x16au(REF_S6, CONST_256, TMP24); |
| 3675 |
vis_pmerge(ZERO, REF_S6_1, TMP26); |
| 3676 |
|
| 3677 |
vis_ld64_2(ref, stride_8, TMP8); |
| 3678 |
vis_faligndata(TMP0, TMP2, REF_0); |
| 3679 |
|
| 3680 |
vis_ld64_2(ref, stride_16, TMP10); |
| 3681 |
ref += stride; |
| 3682 |
vis_faligndata(TMP2, TMP4, REF_4); |
| 3683 |
|
| 3684 |
vis_ld64(dest[0], DST_0);
|
| 3685 |
vis_faligndata(TMP6, TMP8, REF_S0); |
| 3686 |
|
| 3687 |
vis_ld64_2(dest, 8, DST_2);
|
| 3688 |
vis_faligndata(TMP8, TMP10, REF_S4); |
| 3689 |
|
| 3690 |
if (off != 0x7) { |
| 3691 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3692 |
vis_faligndata(TMP0, TMP2, REF_2); |
| 3693 |
vis_faligndata(TMP2, TMP4, REF_6); |
| 3694 |
vis_faligndata(TMP6, TMP8, REF_S2); |
| 3695 |
vis_faligndata(TMP8, TMP10, REF_S6); |
| 3696 |
} else {
|
| 3697 |
vis_src1(TMP2, REF_2); |
| 3698 |
vis_src1(TMP4, REF_6); |
| 3699 |
vis_src1(TMP8, REF_S2); |
| 3700 |
vis_src1(TMP10, REF_S6); |
| 3701 |
} |
| 3702 |
|
| 3703 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 3704 |
vis_pmerge(ZERO, REF_0, TMP0); |
| 3705 |
|
| 3706 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 3707 |
vis_pmerge(ZERO, REF_0_1, TMP2); |
| 3708 |
|
| 3709 |
vis_mul8x16au(REF_2, CONST_256, TMP4); |
| 3710 |
vis_pmerge(ZERO, REF_2_1, TMP6); |
| 3711 |
|
| 3712 |
vis_mul8x16al(DST_2, CONST_1024, REF_0); |
| 3713 |
vis_padd16(TMP0, CONST_6, TMP0); |
| 3714 |
|
| 3715 |
vis_mul8x16al(DST_3, CONST_1024, REF_2); |
| 3716 |
vis_padd16(TMP2, CONST_6, TMP2); |
| 3717 |
|
| 3718 |
vis_padd16(TMP0, TMP4, TMP0); |
| 3719 |
vis_mul8x16au(REF_4, CONST_256, TMP4); |
| 3720 |
|
| 3721 |
vis_padd16(TMP2, TMP6, TMP2); |
| 3722 |
vis_mul8x16au(REF_4_1, CONST_256, TMP6); |
| 3723 |
|
| 3724 |
vis_padd16(TMP12, TMP0, TMP12); |
| 3725 |
vis_mul8x16au(REF_6, CONST_256, TMP8); |
| 3726 |
|
| 3727 |
vis_padd16(TMP14, TMP2, TMP14); |
| 3728 |
vis_mul8x16au(REF_6_1, CONST_256, TMP10); |
| 3729 |
|
| 3730 |
vis_padd16(TMP12, TMP16, TMP12); |
| 3731 |
vis_mul8x16au(REF_S0, CONST_256, REF_4); |
| 3732 |
|
| 3733 |
vis_padd16(TMP14, TMP18, TMP14); |
| 3734 |
vis_mul8x16au(REF_S0_1, CONST_256, REF_6); |
| 3735 |
|
| 3736 |
vis_padd16(TMP12, TMP30, TMP12); |
| 3737 |
|
| 3738 |
vis_padd16(TMP14, TMP32, TMP14); |
| 3739 |
vis_pack16(TMP12, DST_0); |
| 3740 |
|
| 3741 |
vis_pack16(TMP14, DST_1); |
| 3742 |
vis_st64(DST_0, dest[0]);
|
| 3743 |
vis_padd16(TMP4, CONST_6, TMP4); |
| 3744 |
|
| 3745 |
vis_ld64_2(dest, stride, DST_0); |
| 3746 |
vis_padd16(TMP6, CONST_6, TMP6); |
| 3747 |
vis_mul8x16au(REF_S2, CONST_256, TMP12); |
| 3748 |
|
| 3749 |
vis_padd16(TMP4, TMP8, TMP4); |
| 3750 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP14); |
| 3751 |
|
| 3752 |
vis_padd16(TMP6, TMP10, TMP6); |
| 3753 |
|
| 3754 |
vis_padd16(TMP20, TMP4, TMP20); |
| 3755 |
|
| 3756 |
vis_padd16(TMP22, TMP6, TMP22); |
| 3757 |
|
| 3758 |
vis_padd16(TMP20, TMP24, TMP20); |
| 3759 |
|
| 3760 |
vis_padd16(TMP22, TMP26, TMP22); |
| 3761 |
|
| 3762 |
vis_padd16(TMP20, REF_0, TMP20); |
| 3763 |
vis_mul8x16au(REF_S4, CONST_256, REF_0); |
| 3764 |
|
| 3765 |
vis_padd16(TMP22, REF_2, TMP22); |
| 3766 |
vis_pack16(TMP20, DST_2); |
| 3767 |
|
| 3768 |
vis_pack16(TMP22, DST_3); |
| 3769 |
vis_st64_2(DST_2, dest, 8);
|
| 3770 |
dest += stride; |
| 3771 |
|
| 3772 |
vis_ld64_2(dest, 8, DST_2);
|
| 3773 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 3774 |
vis_pmerge(ZERO, REF_S4_1, REF_2); |
| 3775 |
|
| 3776 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 3777 |
vis_padd16(REF_4, TMP0, TMP8); |
| 3778 |
|
| 3779 |
vis_mul8x16au(REF_S6, CONST_256, REF_4); |
| 3780 |
vis_padd16(REF_6, TMP2, TMP10); |
| 3781 |
|
| 3782 |
vis_mul8x16au(REF_S6_1, CONST_256, REF_6); |
| 3783 |
vis_padd16(TMP8, TMP12, TMP8); |
| 3784 |
|
| 3785 |
vis_padd16(TMP10, TMP14, TMP10); |
| 3786 |
|
| 3787 |
vis_padd16(TMP8, TMP30, TMP8); |
| 3788 |
|
| 3789 |
vis_padd16(TMP10, TMP32, TMP10); |
| 3790 |
vis_pack16(TMP8, DST_0); |
| 3791 |
|
| 3792 |
vis_pack16(TMP10, DST_1); |
| 3793 |
vis_st64(DST_0, dest[0]);
|
| 3794 |
|
| 3795 |
vis_padd16(REF_0, TMP4, REF_0); |
| 3796 |
|
| 3797 |
vis_mul8x16al(DST_2, CONST_1024, TMP30); |
| 3798 |
vis_padd16(REF_2, TMP6, REF_2); |
| 3799 |
|
| 3800 |
vis_mul8x16al(DST_3, CONST_1024, TMP32); |
| 3801 |
vis_padd16(REF_0, REF_4, REF_0); |
| 3802 |
|
| 3803 |
vis_padd16(REF_2, REF_6, REF_2); |
| 3804 |
|
| 3805 |
vis_padd16(REF_0, TMP30, REF_0); |
| 3806 |
|
| 3807 |
/* stall */
|
| 3808 |
|
| 3809 |
vis_padd16(REF_2, TMP32, REF_2); |
| 3810 |
vis_pack16(REF_0, DST_2); |
| 3811 |
|
| 3812 |
vis_pack16(REF_2, DST_3); |
| 3813 |
vis_st64_2(DST_2, dest, 8);
|
| 3814 |
dest += stride; |
| 3815 |
} while (--height);
|
| 3816 |
} |
| 3817 |
|
| 3818 |
static void MC_avg_no_round_xy_8_vis (uint8_t * dest, const uint8_t * ref, |
| 3819 |
const int stride, int height) |
| 3820 |
{
|
| 3821 |
unsigned long off = (unsigned long) ref & 0x7; |
| 3822 |
unsigned long off_plus_1 = off + 1; |
| 3823 |
int stride_8 = stride + 8; |
| 3824 |
|
| 3825 |
vis_set_gsr(4 << VIS_GSR_SCALEFACT_SHIFT);
|
| 3826 |
|
| 3827 |
ref = vis_alignaddr(ref); |
| 3828 |
|
| 3829 |
vis_ld64(ref[0], TMP0);
|
| 3830 |
vis_fzero(ZERO); |
| 3831 |
|
| 3832 |
vis_ld64_2(ref, 8, TMP2);
|
| 3833 |
|
| 3834 |
vis_ld64(constants6[0], CONST_6);
|
| 3835 |
|
| 3836 |
vis_ld64(constants256_1024[0], CONST_256);
|
| 3837 |
vis_faligndata(TMP0, TMP2, REF_S0); |
| 3838 |
|
| 3839 |
if (off != 0x7) { |
| 3840 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3841 |
vis_faligndata(TMP0, TMP2, REF_S2); |
| 3842 |
} else {
|
| 3843 |
vis_src1(TMP2, REF_S2); |
| 3844 |
} |
| 3845 |
|
| 3846 |
height >>= 1;
|
| 3847 |
do { /* 31 cycles */ |
| 3848 |
vis_ld64_2(ref, stride, TMP0); |
| 3849 |
vis_mul8x16au(REF_S0, CONST_256, TMP8); |
| 3850 |
vis_pmerge(ZERO, REF_S0_1, TMP10); |
| 3851 |
|
| 3852 |
vis_ld64_2(ref, stride_8, TMP2); |
| 3853 |
ref += stride; |
| 3854 |
vis_mul8x16au(REF_S2, CONST_256, TMP12); |
| 3855 |
vis_pmerge(ZERO, REF_S2_1, TMP14); |
| 3856 |
|
| 3857 |
vis_alignaddr_g0((void *)off);
|
| 3858 |
|
| 3859 |
vis_ld64_2(ref, stride, TMP4); |
| 3860 |
vis_faligndata(TMP0, TMP2, REF_S4); |
| 3861 |
|
| 3862 |
vis_ld64_2(ref, stride_8, TMP6); |
| 3863 |
ref += stride; |
| 3864 |
|
| 3865 |
vis_ld64(dest[0], DST_0);
|
| 3866 |
vis_faligndata(TMP4, TMP6, REF_S0); |
| 3867 |
|
| 3868 |
vis_ld64_2(dest, stride, DST_2); |
| 3869 |
|
| 3870 |
if (off != 0x7) { |
| 3871 |
vis_alignaddr_g0((void *)off_plus_1);
|
| 3872 |
vis_faligndata(TMP0, TMP2, REF_S6); |
| 3873 |
vis_faligndata(TMP4, TMP6, REF_S2); |
| 3874 |
} else {
|
| 3875 |
vis_src1(TMP2, REF_S6); |
| 3876 |
vis_src1(TMP6, REF_S2); |
| 3877 |
} |
| 3878 |
|
| 3879 |
vis_mul8x16al(DST_0, CONST_1024, TMP30); |
| 3880 |
vis_pmerge(ZERO, REF_S4, TMP22); |
| 3881 |
|
| 3882 |
vis_mul8x16al(DST_1, CONST_1024, TMP32); |
| 3883 |
vis_pmerge(ZERO, REF_S4_1, TMP24); |
| 3884 |
|
| 3885 |
vis_mul8x16au(REF_S6, CONST_256, TMP26); |
| 3886 |
vis_pmerge(ZERO, REF_S6_1, TMP28); |
| 3887 |
|
| 3888 |
vis_mul8x16au(REF_S0, CONST_256, REF_S4); |
| 3889 |
vis_padd16(TMP22, CONST_6, TMP22); |
| 3890 |
|
| 3891 |
vis_mul8x16au(REF_S0_1, CONST_256, REF_S6); |
| 3892 |
vis_padd16(TMP24, CONST_6, TMP24); |
| 3893 |
|
| 3894 |
vis_mul8x16al(DST_2, CONST_1024, REF_0); |
| 3895 |
vis_padd16(TMP22, TMP26, TMP22); |
| 3896 |
|
| 3897 |
vis_mul8x16al(DST_3, CONST_1024, REF_2); |
| 3898 |
vis_padd16(TMP24, TMP28, TMP24); |
| 3899 |
|
| 3900 |
vis_mul8x16au(REF_S2, CONST_256, TMP26); |
| 3901 |
vis_padd16(TMP8, TMP22, TMP8); |
| 3902 |
|
| 3903 |
vis_mul8x16au(REF_S2_1, CONST_256, TMP28); |
| 3904 |
vis_padd16(TMP10, TMP24, TMP10); |
| 3905 |
|
| 3906 |
vis_padd16(TMP8, TMP12, TMP8); |
| 3907 |
|
| 3908 |
vis_padd16(TMP10, TMP14, TMP10); |
| 3909 |
|
| 3910 |
vis_padd16(TMP8, TMP30, TMP8); |
| 3911 |
|
| 3912 |
vis_padd16(TMP10, TMP32, TMP10); |
| 3913 |
vis_pack16(TMP8, DST_0); |
| 3914 |
|
| 3915 |
vis_pack16(TMP10, DST_1); |
| 3916 |
vis_st64(DST_0, dest[0]);
|
| 3917 |
dest += stride; |
| 3918 |
|
| 3919 |
vis_padd16(REF_S4, TMP22, TMP12); |
| 3920 |
|
| 3921 |
vis_padd16(REF_S6, TMP24, TMP14); |
| 3922 |
|
| 3923 |
vis_padd16(TMP12, TMP26, TMP12); |
| 3924 |
|
| 3925 |
vis_padd16(TMP14, TMP28, TMP14); |
| 3926 |
|
| 3927 |
vis_padd16(TMP12, REF_0, TMP12); |
| 3928 |
|
| 3929 |
vis_padd16(TMP14, REF_2, TMP14); |
| 3930 |
vis_pack16(TMP12, DST_2); |
| 3931 |
|
| 3932 |
vis_pack16(TMP14, DST_3); |
| 3933 |
vis_st64(DST_2, dest[0]);
|
| 3934 |
dest += stride; |
| 3935 |
} while (--height);
|
| 3936 |
} |
| 3937 |
|
| 3938 |
/* End of no rounding code */
|
| 3939 |
|
| 3940 |
#define ACCEL_SPARC_VIS 1 |
| 3941 |
#define ACCEL_SPARC_VIS2 2 |
| 3942 |
|
| 3943 |
static int vis_level(void) |
| 3944 |
{
|
| 3945 |
int accel = 0; |
| 3946 |
accel |= ACCEL_SPARC_VIS; |
| 3947 |
accel |= ACCEL_SPARC_VIS2; |
| 3948 |
return accel;
|
| 3949 |
} |
| 3950 |
|
| 3951 |
/* libavcodec initialization code */
|
| 3952 |
void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx)
|
| 3953 |
{
|
| 3954 |
/* VIS-specific optimizations */
|
| 3955 |
int accel = vis_level ();
|
| 3956 |
const int h264_high_depth = avctx->codec_id == CODEC_ID_H264 && avctx->bits_per_raw_sample > 8; |
| 3957 |
|
| 3958 |
if (accel & ACCEL_SPARC_VIS) {
|
| 3959 |
if(avctx->idct_algo==FF_IDCT_SIMPLEVIS){
|
| 3960 |
c->idct_put = ff_simple_idct_put_vis; |
| 3961 |
c->idct_add = ff_simple_idct_add_vis; |
| 3962 |
c->idct = ff_simple_idct_vis; |
| 3963 |
c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM; |
| 3964 |
} |
| 3965 |
|
| 3966 |
if (!h264_high_depth) {
|
| 3967 |
c->put_pixels_tab[0][0] = MC_put_o_16_vis; |
| 3968 |
c->put_pixels_tab[0][1] = MC_put_x_16_vis; |
| 3969 |
c->put_pixels_tab[0][2] = MC_put_y_16_vis; |
| 3970 |
c->put_pixels_tab[0][3] = MC_put_xy_16_vis; |
| 3971 |
|
| 3972 |
c->put_pixels_tab[1][0] = MC_put_o_8_vis; |
| 3973 |
c->put_pixels_tab[1][1] = MC_put_x_8_vis; |
| 3974 |
c->put_pixels_tab[1][2] = MC_put_y_8_vis; |
| 3975 |
c->put_pixels_tab[1][3] = MC_put_xy_8_vis; |
| 3976 |
|
| 3977 |
c->avg_pixels_tab[0][0] = MC_avg_o_16_vis; |
| 3978 |
c->avg_pixels_tab[0][1] = MC_avg_x_16_vis; |
| 3979 |
c->avg_pixels_tab[0][2] = MC_avg_y_16_vis; |
| 3980 |
c->avg_pixels_tab[0][3] = MC_avg_xy_16_vis; |
| 3981 |
|
| 3982 |
c->avg_pixels_tab[1][0] = MC_avg_o_8_vis; |
| 3983 |
c->avg_pixels_tab[1][1] = MC_avg_x_8_vis; |
| 3984 |
c->avg_pixels_tab[1][2] = MC_avg_y_8_vis; |
| 3985 |
c->avg_pixels_tab[1][3] = MC_avg_xy_8_vis; |
| 3986 |
|
| 3987 |
c->put_no_rnd_pixels_tab[0][0] = MC_put_no_round_o_16_vis; |
| 3988 |
c->put_no_rnd_pixels_tab[0][1] = MC_put_no_round_x_16_vis; |
| 3989 |
c->put_no_rnd_pixels_tab[0][2] = MC_put_no_round_y_16_vis; |
| 3990 |
c->put_no_rnd_pixels_tab[0][3] = MC_put_no_round_xy_16_vis; |
| 3991 |
|
| 3992 |
c->put_no_rnd_pixels_tab[1][0] = MC_put_no_round_o_8_vis; |
| 3993 |
c->put_no_rnd_pixels_tab[1][1] = MC_put_no_round_x_8_vis; |
| 3994 |
c->put_no_rnd_pixels_tab[1][2] = MC_put_no_round_y_8_vis; |
| 3995 |
c->put_no_rnd_pixels_tab[1][3] = MC_put_no_round_xy_8_vis; |
| 3996 |
|
| 3997 |
c->avg_no_rnd_pixels_tab[0][0] = MC_avg_no_round_o_16_vis; |
| 3998 |
c->avg_no_rnd_pixels_tab[0][1] = MC_avg_no_round_x_16_vis; |
| 3999 |
c->avg_no_rnd_pixels_tab[0][2] = MC_avg_no_round_y_16_vis; |
| 4000 |
c->avg_no_rnd_pixels_tab[0][3] = MC_avg_no_round_xy_16_vis; |
| 4001 |
|
| 4002 |
c->avg_no_rnd_pixels_tab[1][0] = MC_avg_no_round_o_8_vis; |
| 4003 |
c->avg_no_rnd_pixels_tab[1][1] = MC_avg_no_round_x_8_vis; |
| 4004 |
c->avg_no_rnd_pixels_tab[1][2] = MC_avg_no_round_y_8_vis; |
| 4005 |
c->avg_no_rnd_pixels_tab[1][3] = MC_avg_no_round_xy_8_vis; |
| 4006 |
} |
| 4007 |
} |
| 4008 |
} |