git.sesse.net Git - ffmpeg/blob - libavcodec/mips/h264pred_mmi.c

   1 /*
   2  * Loongson SIMD optimized h264pred
   3  *
   4  * Copyright (c) 2015 Loongson Technology Corporation Limited
   5  * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
   6  *                    Zhang Shuangshuang <zhangshuangshuang@ict.ac.cn>
   7  *
   8  * This file is part of FFmpeg.
   9  *
  10  * FFmpeg is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU Lesser General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2.1 of the License, or (at your option) any later version.
  14  *
  15  * FFmpeg is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public
  21  * License along with FFmpeg; if not, write to the Free Software
  22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  23  */
  24
  25 #include "h264pred_mips.h"
  26 #include "constants.h"
  27
  28 void ff_pred16x16_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
  29 {
  30     __asm__ volatile (
  31         "dsubu $2, %0, %1                   \r\n"
  32         "daddu $3, %0, $0                   \r\n"
  33         "ldl $4, 7($2)                      \r\n"
  34         "ldr $4, 0($2)                      \r\n"
  35         "ldl $5, 15($2)                     \r\n"
  36         "ldr $5, 8($2)                      \r\n"
  37         "dli $6, 0x10                       \r\n"
  38         "1:                                 \r\n"
  39         "sdl $4, 7($3)                      \r\n"
  40         "sdr $4, 0($3)                      \r\n"
  41         "sdl $5, 15($3)                     \r\n"
  42         "sdr $5, 8($3)                      \r\n"
  43         "daddu $3, %1                       \r\n"
  44         "daddiu $6, -1                      \r\n"
  45         "bnez $6, 1b                        \r\n"
  46         ::"r"(src),"r"(stride)
  47         : "$2","$3","$4","$5","$6","memory"
  48     );
  49 }
  50
  51 void ff_pred16x16_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
  52 {
  53     __asm__ volatile (
  54         "daddiu $2, %0, -1                  \r\n"
  55         "daddu $3, %0, $0                   \r\n"
  56         "dli $6, 0x10                       \r\n"
  57         "1:                                 \r\n"
  58         "lbu $4, 0($2)                      \r\n"
  59         "dmul $5, $4, %2                    \r\n"
  60         "sdl $5, 7($3)                      \r\n"
  61         "sdr $5, 0($3)                      \r\n"
  62         "sdl $5, 15($3)                     \r\n"
  63         "sdr $5, 8($3)                      \r\n"
  64         "daddu $2, %1                       \r\n"
  65         "daddu $3, %1                       \r\n"
  66         "daddiu $6, -1                      \r\n"
  67         "bnez $6, 1b                        \r\n"
  68         ::"r"(src),"r"(stride),"r"(ff_pb_1)
  69         : "$2","$3","$4","$5","$6","memory"
  70     );
  71 }
  72
  73 void ff_pred16x16_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
  74 {
  75     __asm__ volatile (
  76         "daddiu $2, %0, -1                  \r\n"
  77         "dli $6, 0x10                       \r\n"
  78         "xor $8, $8, $8                     \r\n"
  79         "1:                                 \r\n"
  80         "lbu $4, 0($2)                      \r\n"
  81         "daddu $8, $8, $4                   \r\n"
  82         "daddu $2, $2, %1                   \r\n"
  83         "daddiu $6, $6, -1                  \r\n"
  84         "bnez $6, 1b                        \r\n"
  85         "dli $6, 0x10                       \r\n"
  86         "negu $3, %1                        \r\n"
  87         "daddu $2, %0, $3                   \r\n"
  88         "2:                                 \r\n"
  89         "lbu $4, 0($2)                      \r\n"
  90         "daddu $8, $8, $4                   \r\n"
  91         "daddiu $2, $2, 1                   \r\n"
  92         "daddiu $6, $6, -1                  \r\n"
  93         "bnez $6, 2b                        \r\n"
  94         "daddiu $8, $8, 0x10                \r\n"
  95         "dsra $8, 5                         \r\n"
  96         "dmul $5, $8, %2                    \r\n"
  97         "daddu $2, %0, $0                   \r\n"
  98         "dli $6, 0x10                       \r\n"
  99         "3:                                 \r\n"
 100         "sdl $5, 7($2)                      \r\n"
 101         "sdr $5, 0($2)                      \r\n"
 102         "sdl $5, 15($2)                     \r\n"
 103         "sdr $5, 8($2)                      \r\n"
 104         "daddu $2, $2, %1                   \r\n"
 105         "daddiu $6, $6, -1                  \r\n"
 106         "bnez $6, 3b                        \r\n"
 107         ::"r"(src),"r"(stride),"r"(ff_pb_1)
 108         : "$2","$3","$4","$5","$6","$8","memory"
 109     );
 110 }
 111
 112 void ff_pred8x8l_top_dc_8_mmi(uint8_t *src, int has_topleft,
 113         int has_topright, ptrdiff_t stride)
 114 {
 115     int y;
 116     uint32_t dc;
 117
 118     __asm__ volatile (
 119         "ldl $8, 7(%1)                      \r\n"
 120         "ldr $8, 0(%1)                      \r\n"
 121         "ldl $9, 7(%2)                      \r\n"
 122         "ldr $9, 0(%2)                      \r\n"
 123         "ldl $10, 7(%3)                     \r\n"
 124         "ldr $10, 0(%3)                     \r\n"
 125         "dmtc1 $8, $f2                      \r\n"
 126         "dmtc1 $9, $f4                      \r\n"
 127         "dmtc1 $10, $f6                     \r\n"
 128         "dmtc1 $0, $f0                      \r\n"
 129         "punpcklbh $f8, $f2, $f0            \r\n"
 130         "punpckhbh $f10, $f2, $f0           \r\n"
 131         "punpcklbh $f12, $f4, $f0           \r\n"
 132         "punpckhbh $f14, $f4, $f0           \r\n"
 133         "punpcklbh $f16, $f6, $f0           \r\n"
 134         "punpckhbh $f18, $f6, $f0           \r\n"
 135         "bnez %4, 1f                        \r\n"
 136         "pinsrh_0 $f8, $f8, $f12            \r\n"
 137         "1:                                 \r\n"
 138         "bnez %5, 2f                        \r\n"
 139         "pinsrh_3 $f18, $f18, $f14          \r\n"
 140         "2:                                 \r\n"
 141         "daddiu $8, $0, 2                   \r\n"
 142         "dmtc1 $8, $f20                     \r\n"
 143         "pshufh $f22, $f20, $f0             \r\n"
 144         "pmullh $f12, $f12, $f22            \r\n"
 145         "pmullh $f14, $f14, $f22            \r\n"
 146         "paddh $f8, $f8, $f12               \r\n"
 147         "paddh $f10, $f10, $f14             \r\n"
 148         "paddh $f8, $f8, $f16               \r\n"
 149         "paddh $f10, $f10, $f18             \r\n"
 150         "paddh $f8, $f8, $f22               \r\n"
 151         "paddh $f10, $f10, $f22             \r\n"
 152         "psrah $f8, $f8, $f20               \r\n"
 153         "psrah $f10, $f10, $f20             \r\n"
 154         "packushb $f4, $f8, $f10            \r\n"
 155         "biadd $f2, $f4                     \r\n"
 156         "mfc1 $9, $f2                       \r\n"
 157         "addiu $9, $9, 4                    \r\n"
 158         "dsrl $9, $9, 3                     \r\n"
 159         "li $8, 0x01010101                  \r\n"
 160         "mul %0, $9, $8                     \r\n"
 161         : "=r"(dc)
 162         : "r"(src-stride-1),"r"(src-stride),"r"(src-stride+1),
 163           "r"(has_topleft),"r"(has_topright)
 164         : "$8","$9","$10"
 165     );
 166
 167     for (y=0; y<8; y++) {
 168         AV_WN4PA(((uint32_t*)src)+0, dc);
 169         AV_WN4PA(((uint32_t*)src)+1, dc);
 170         src += stride;
 171     }
 172 }
 173
 174 void ff_pred8x8l_dc_8_mmi(uint8_t *src, int has_topleft,
 175         int has_topright, ptrdiff_t stride)
 176 {
 177     int y;
 178     uint32_t dc, dc1, dc2;
 179
 180     const int l0 = ((has_topleft ? src[-1+-1*stride] : src[-1+0*stride]) + 2*src[-1+0*stride] + src[-1+1*stride] + 2) >> 2;
 181     const int l1 = (src[-1+0*stride] + 2*src[-1+1*stride] + src[-1+2*stride] + 2) >> 2;
 182     const int l2 = (src[-1+1*stride] + 2*src[-1+2*stride] + src[-1+3*stride] + 2) >> 2;
 183     const int l3 = (src[-1+2*stride] + 2*src[-1+3*stride] + src[-1+4*stride] + 2) >> 2;
 184     const int l4 = (src[-1+3*stride] + 2*src[-1+4*stride] + src[-1+5*stride] + 2) >> 2;
 185     const int l5 = (src[-1+4*stride] + 2*src[-1+5*stride] + src[-1+6*stride] + 2) >> 2;
 186     const int l6 = (src[-1+5*stride] + 2*src[-1+6*stride] + src[-1+7*stride] + 2) >> 2;
 187     const int l7 = (src[-1+6*stride] + 2*src[-1+7*stride] + src[-1+7*stride] + 2) >> 2;
 188
 189     __asm__ volatile (
 190         "ldl $8, 7(%1)                      \r\n"
 191         "ldr $8, 0(%1)                      \r\n"
 192         "ldl $9, 7(%2)                      \r\n"
 193         "ldr $9, 0(%2)                      \r\n"
 194         "ldl $10, 7(%3)                     \r\n"
 195         "ldr $10, 0(%3)                     \r\n"
 196         "dmtc1 $8, $f2                      \r\n"
 197         "dmtc1 $9, $f4                      \r\n"
 198         "dmtc1 $10, $f6                     \r\n"
 199         "dmtc1 $0, $f0                      \r\n"
 200         "punpcklbh $f8, $f2, $f0            \r\n"
 201         "punpckhbh $f10, $f2, $f0           \r\n"
 202         "punpcklbh $f12, $f4, $f0           \r\n"
 203         "punpckhbh $f14, $f4, $f0           \r\n"
 204         "punpcklbh $f16, $f6, $f0           \r\n"
 205         "punpckhbh $f18, $f6, $f0           \r\n"
 206         "daddiu $8, $0, 3                   \r\n"
 207         "dmtc1 $8, $f20                     \r\n"
 208         "pshufh $f28, $f10, $f20            \r\n"
 209         "pshufh $f30, $f18, $f20            \r\n"
 210         "pinsrh_3 $f10, $f10, $f30          \r\n"
 211         "pinsrh_3 $f18, $f18, $f28          \r\n"
 212         "bnez %4, 1f                        \r\n"
 213         "pinsrh_0 $f8, $f8, $f12            \r\n"
 214         "1:                                 \r\n"
 215         "bnez %5, 2f                        \r\n"
 216         "pshufh $f30, $f14, $f20            \r\n"
 217         "pinsrh_3 $f10, $f10, $f30          \r\n"
 218         "2:                                 \r\n"
 219         "daddiu $8, $0, 2                   \r\n"
 220         "dmtc1 $8, $f20                     \r\n"
 221         "pshufh $f22, $f20, $f0             \r\n"
 222         "pmullh $f12, $f12, $f22            \r\n"
 223         "pmullh $f14, $f14, $f22            \r\n"
 224         "paddh $f8, $f8, $f12               \r\n"
 225         "paddh $f10, $f10, $f14             \r\n"
 226         "paddh $f8, $f8, $f16               \r\n"
 227         "paddh $f10, $f10, $f18             \r\n"
 228         "paddh $f8, $f8, $f22               \r\n"
 229         "paddh $f10, $f10, $f22             \r\n"
 230         "psrah $f8, $f8, $f20               \r\n"
 231         "psrah $f10, $f10, $f20             \r\n"
 232         "packushb $f4, $f8, $f10            \r\n"
 233         "biadd $f2, $f4                     \r\n"
 234         "mfc1 %0, $f2                       \r\n"
 235         : "=r"(dc2)
 236         : "r"(src-stride-1),"r"(src-stride),"r"(src-stride+1),
 237           "r"(has_topleft),"r"(has_topright)
 238         : "$8","$9","$10"
 239     );
 240
 241     dc1 = l0+l1+l2+l3+l4+l5+l6+l7;
 242     dc = PIXEL_SPLAT_X4((dc1+dc2+8)>>4);
 243
 244     for (y=0; y<8; y++) {
 245         AV_WN4PA(((uint32_t*)src)+0, dc);
 246         AV_WN4PA(((uint32_t*)src)+1, dc);
 247         src += stride;
 248     }
 249 }
 250
 251 void ff_pred8x8l_horizontal_8_mmi(uint8_t *src, int has_topleft,
 252         int has_topright, ptrdiff_t stride)
 253 {
 254     const int l0 = ((has_topleft ? src[-1+-1*stride] : src[-1+0*stride]) + 2*src[-1+0*stride] + src[-1+1*stride] + 2) >> 2;
 255     const int l1 = (src[-1+0*stride] + 2*src[-1+1*stride] + src[-1+2*stride] + 2) >> 2;
 256     const int l2 = (src[-1+1*stride] + 2*src[-1+2*stride] + src[-1+3*stride] + 2) >> 2;
 257     const int l3 = (src[-1+2*stride] + 2*src[-1+3*stride] + src[-1+4*stride] + 2) >> 2;
 258     const int l4 = (src[-1+3*stride] + 2*src[-1+4*stride] + src[-1+5*stride] + 2) >> 2;
 259     const int l5 = (src[-1+4*stride] + 2*src[-1+5*stride] + src[-1+6*stride] + 2) >> 2;
 260     const int l6 = (src[-1+5*stride] + 2*src[-1+6*stride] + src[-1+7*stride] + 2) >> 2;
 261     const int l7 = (src[-1+6*stride] + 2*src[-1+7*stride] + src[-1+7*stride] + 2) >> 2;
 262
 263     AV_WN4PA(src+0*stride, PIXEL_SPLAT_X4(l0));
 264     AV_WN4PA(src+0*stride+4, PIXEL_SPLAT_X4(l0));
 265     AV_WN4PA(src+1*stride, PIXEL_SPLAT_X4(l1));
 266     AV_WN4PA(src+1*stride+4, PIXEL_SPLAT_X4(l1));
 267     AV_WN4PA(src+2*stride, PIXEL_SPLAT_X4(l2));
 268     AV_WN4PA(src+2*stride+4, PIXEL_SPLAT_X4(l2));
 269     AV_WN4PA(src+3*stride, PIXEL_SPLAT_X4(l3));
 270     AV_WN4PA(src+3*stride+4, PIXEL_SPLAT_X4(l3));
 271     AV_WN4PA(src+4*stride, PIXEL_SPLAT_X4(l4));
 272     AV_WN4PA(src+4*stride+4, PIXEL_SPLAT_X4(l4));
 273     AV_WN4PA(src+5*stride, PIXEL_SPLAT_X4(l5));
 274     AV_WN4PA(src+5*stride+4, PIXEL_SPLAT_X4(l5));
 275     AV_WN4PA(src+6*stride, PIXEL_SPLAT_X4(l6));
 276     AV_WN4PA(src+6*stride+4, PIXEL_SPLAT_X4(l6));
 277     AV_WN4PA(src+7*stride, PIXEL_SPLAT_X4(l7));
 278     AV_WN4PA(src+7*stride+4, PIXEL_SPLAT_X4(l7));
 279 }
 280
 281 void ff_pred8x8l_vertical_8_mmi(uint8_t *src, int has_topleft,
 282         int has_topright, ptrdiff_t stride)
 283 {
 284     int y;
 285     uint32_t a, b;
 286
 287     __asm__ volatile (
 288         "ldl $8, 7(%1)                      \r\n"
 289         "ldr $8, 0(%1)                      \r\n"
 290         "ldl $9, 7(%2)                      \r\n"
 291         "ldr $9, 0(%2)                      \r\n"
 292         "ldl $10, 7(%3)                     \r\n"
 293         "ldr $10, 0(%3)                     \r\n"
 294         "dmtc1 $8, $f2                      \r\n"
 295         "dmtc1 $9, $f4                      \r\n"
 296         "dmtc1 $10, $f6                     \r\n"
 297         "dmtc1 $0, $f0                      \r\n"
 298         "punpcklbh $f8, $f2, $f0            \r\n"
 299         "punpckhbh $f10, $f2, $f0           \r\n"
 300         "punpcklbh $f12, $f4, $f0           \r\n"
 301         "punpckhbh $f14, $f4, $f0           \r\n"
 302         "punpcklbh $f16, $f6, $f0           \r\n"
 303         "punpckhbh $f18, $f6, $f0           \r\n"
 304         "bnez %4, 1f                        \r\n"
 305         "pinsrh_0 $f8, $f8, $f12            \r\n"
 306         "1:                                 \r\n"
 307         "bnez %5, 2f                        \r\n"
 308         "pinsrh_3 $f18, $f18, $f14          \r\n"
 309         "2:                                 \r\n"
 310         "daddiu $8, $0, 2                   \r\n"
 311         "dmtc1 $8, $f20                     \r\n"
 312         "pshufh $f22, $f20, $f0             \r\n"
 313         "pmullh $f12, $f12, $f22            \r\n"
 314         "pmullh $f14, $f14, $f22            \r\n"
 315         "paddh $f8, $f8, $f12               \r\n"
 316         "paddh $f10, $f10, $f14             \r\n"
 317         "paddh $f8, $f8, $f16               \r\n"
 318         "paddh $f10, $f10, $f18             \r\n"
 319         "paddh $f8, $f8, $f22               \r\n"
 320         "paddh $f10, $f10, $f22             \r\n"
 321         "psrah $f8, $f8, $f20               \r\n"
 322         "psrah $f10, $f10, $f20             \r\n"
 323         "packushb $f4, $f8, $f10            \r\n"
 324         "sdc1 $f4, %0                       \r\n"
 325         : "=m"(*src)
 326         : "r"(src-stride-1),"r"(src-stride),"r"(src-stride+1),
 327           "r"(has_topleft),"r"(has_topright)
 328         : "$8","$9","$10"
 329     );
 330
 331     a = AV_RN4PA(((uint32_t*)src)+0);
 332     b = AV_RN4PA(((uint32_t*)src)+1);
 333
 334     for (y=1; y<8; y++) {
 335         AV_WN4PA(((uint32_t*)(src+y*stride))+0, a);
 336         AV_WN4PA(((uint32_t*)(src+y*stride))+1, b);
 337     }
 338 }
 339
 340 void ff_pred4x4_dc_8_mmi(uint8_t *src, const uint8_t *topright,
 341         ptrdiff_t stride)
 342 {
 343     const int dc = (src[-stride] + src[1-stride] + src[2-stride]
 344                  + src[3-stride] + src[-1+0*stride] + src[-1+1*stride]
 345                  + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
 346
 347     __asm__ volatile (
 348         "daddu $2, %2, $0                   \r\n"
 349         "dmul $3, $2, %3                    \r\n"
 350         "xor $4, $4, $4                     \r\n"
 351         "gsswx $3, 0(%0,$4)                 \r\n"
 352         "daddu $4, %1                       \r\n"
 353         "gsswx $3, 0(%0,$4)                 \r\n"
 354         "daddu $4, %1                       \r\n"
 355         "gsswx $3, 0(%0,$4)                 \r\n"
 356         "daddu $4, %1                       \r\n"
 357         "gsswx $3, 0(%0,$4)                 \r\n"
 358         ::"r"(src),"r"(stride),"r"(dc),"r"(ff_pb_1)
 359         : "$2","$3","$4","memory"
 360     );
 361 }
 362
 363 void ff_pred8x8_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
 364 {
 365     __asm__ volatile (
 366         "dsubu $2, %0, %1                   \r\n"
 367         "daddu $3, %0, $0                   \r\n"
 368         "ldl $4, 7($2)                      \r\n"
 369         "ldr $4, 0($2)                      \r\n"
 370         "dli $5, 0x8                        \r\n"
 371         "1:                                 \r\n"
 372         "sdl $4, 7($3)                      \r\n"
 373         "sdr $4, 0($3)                      \r\n"
 374         "daddu $3, %1                       \r\n"
 375         "daddiu $5, -1                      \r\n"
 376         "bnez $5, 1b                        \r\n"
 377         ::"r"(src),"r"(stride)
 378         : "$2","$3","$4","$5","memory"
 379     );
 380 }
 381
 382 void ff_pred8x8_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
 383 {
 384     __asm__ volatile (
 385         "daddiu $2, %0, -1                  \r\n"
 386         "daddu $3, %0, $0                   \r\n"
 387         "dli $6, 0x8                        \r\n"
 388         "1:                                 \r\n"
 389         "lbu $4, 0($2)                      \r\n"
 390         "dmul $5, $4, %2                    \r\n"
 391         "sdl $5, 7($3)                      \r\n"
 392         "sdr $5, 0($3)                      \r\n"
 393         "daddu $2, %1                       \r\n"
 394         "daddu $3, %1                       \r\n"
 395         "daddiu $6, -1                      \r\n"
 396         "bnez $6, 1b                        \r\n"
 397         ::"r"(src),"r"(stride),"r"(ff_pb_1)
 398         : "$2","$3","$4","$5","$6","memory"
 399     );
 400 }
 401
 402 static void ff_pred16x16_plane_compat_8_mmi(uint8_t *src, ptrdiff_t stride,
 403         const int svq3, const int rv40)
 404 {
 405     __asm__ volatile (
 406         "negu $2, %1                        \r\n"
 407         "daddu $3, %0, $2                   \r\n"
 408         "xor $f8, $f8, $f8                  \r\n"
 409         "gslwlc1 $f0, 2($3)                 \r\n"
 410         "gslwrc1 $f0, -1($3)                \r\n"
 411         "gslwlc1 $f2, 6($3)                 \r\n"
 412         "gslwrc1 $f2, 3($3)                 \r\n"
 413         "gslwlc1 $f4, 11($3)                \r\n"
 414         "gslwrc1 $f4, 8($3)                 \r\n"
 415         "gslwlc1 $f6, 15($3)                \r\n"
 416         "gslwrc1 $f6, 12($3)                \r\n"
 417         "punpcklbh $f0, $f0, $f8            \r\n"
 418         "punpcklbh $f2, $f2, $f8            \r\n"
 419         "punpcklbh $f4, $f4, $f8            \r\n"
 420         "punpcklbh $f6, $f6, $f8            \r\n"
 421         "dmtc1 %4, $f20                     \r\n"
 422         "dmtc1 %5, $f22                     \r\n"
 423         "dmtc1 %6, $f24                     \r\n"
 424         "dmtc1 %7, $f26                     \r\n"
 425         "pmullh $f0, $f0, $f20              \r\n"
 426         "pmullh $f2, $f2, $f22              \r\n"
 427         "pmullh $f4, $f4, $f24              \r\n"
 428         "pmullh $f6, $f6, $f26              \r\n"
 429         "paddsh $f0, $f0, $f4               \r\n"
 430         "paddsh $f2, $f2, $f6               \r\n"
 431         "paddsh $f0, $f0, $f2               \r\n"
 432         "dli $4, 0xE                        \r\n"
 433         "dmtc1 $4, $f28                     \r\n"
 434         "pshufh $f2, $f0, $f28              \r\n"
 435         "paddsh $f0, $f0, $f2               \r\n"
 436         "dli $4, 0x1                        \r\n"
 437         "dmtc1 $4, $f30                     \r\n"
 438         "pshufh $f2, $f0, $f30              \r\n"
 439         "paddsh $f10, $f0, $f2              \r\n"
 440         "daddiu $3, %0, -1                  \r\n"
 441         "daddu $3, $2                       \r\n"
 442         "lbu $4, 0($3)                      \r\n"
 443         "lbu $8, 16($3)                     \r\n"
 444         "daddu $3, %1                       \r\n"
 445         "lbu $5, 0($3)                      \r\n"
 446         "daddu $3, %1                       \r\n"
 447         "lbu $6, 0($3)                      \r\n"
 448         "daddu $3, %1                       \r\n"
 449         "lbu $7, 0($3)                      \r\n"
 450         "dsll $5, 16                        \r\n"
 451         "dsll $6, 32                        \r\n"
 452         "dsll $7, 48                        \r\n"
 453         "or $6, $7                          \r\n"
 454         "or $4, $5                          \r\n"
 455         "or $4, $6                          \r\n"
 456         "dmtc1 $4, $f0                      \r\n"
 457         "daddu $3, %1                       \r\n"
 458         "lbu $4, 0($3)                      \r\n"
 459         "daddu $3, %1                       \r\n"
 460         "lbu $5, 0($3)                      \r\n"
 461         "daddu $3, %1                       \r\n"
 462         "lbu $6, 0($3)                      \r\n"
 463         "daddu $3, %1                       \r\n"
 464         "lbu $7, 0($3)                      \r\n"
 465         "dsll $5, 16                        \r\n"
 466         "dsll $6, 32                        \r\n"
 467         "dsll $7, 48                        \r\n"
 468         "or $6, $7                          \r\n"
 469         "or $4, $5                          \r\n"
 470         "or $4, $6                          \r\n"
 471         "dmtc1 $4, $f2                      \r\n"
 472         "daddu $3, %1                       \r\n"
 473         "daddu $3, %1                       \r\n"
 474         "lbu $4, 0($3)                      \r\n"
 475         "daddu $3, %1                       \r\n"
 476         "lbu $5, 0($3)                      \r\n"
 477         "daddu $3, %1                       \r\n"
 478         "lbu $6, 0($3)                      \r\n"
 479         "daddu $3, %1                       \r\n"
 480         "lbu $7, 0($3)                      \r\n"
 481         "dsll $5, 16                        \r\n"
 482         "dsll $6, 32                        \r\n"
 483         "dsll $7, 48                        \r\n"
 484         "or $6, $7                          \r\n"
 485         "or $4, $5                          \r\n"
 486         "or $4, $6                          \r\n"
 487         "dmtc1 $4, $f4                      \r\n"
 488         "daddu $3, %1                       \r\n"
 489         "lbu $4, 0($3)                      \r\n"
 490         "daddu $3, %1                       \r\n"
 491         "lbu $5, 0($3)                      \r\n"
 492         "daddu $3, %1                       \r\n"
 493         "lbu $6, 0($3)                      \r\n"
 494         "daddu $3, %1                       \r\n"
 495         "lbu $7, 0($3)                      \r\n"
 496         "daddu $8, $7                       \r\n"
 497         "daddiu $8, 1                       \r\n"
 498         "dsll $8, 4                         \r\n"
 499         "dsll $5, 16                        \r\n"
 500         "dsll $6, 32                        \r\n"
 501         "dsll $7, 48                        \r\n"
 502         "or $6, $7                          \r\n"
 503         "or $4, $5                          \r\n"
 504         "or $4, $6                          \r\n"
 505         "dmtc1 $4, $f6                      \r\n"
 506         "pmullh $f0, $f0, $f20              \r\n"
 507         "pmullh $f2, $f2, $f22              \r\n"
 508         "pmullh $f4, $f4, $f24              \r\n"
 509         "pmullh $f6, $f6, $f26              \r\n"
 510         "paddsh $f0, $f0, $f4               \r\n"
 511         "paddsh $f2, $f2, $f6               \r\n"
 512         "paddsh $f0, $f0, $f2               \r\n"
 513         "pshufh $f2, $f0, $f28              \r\n"
 514         "paddsh $f0, $f0, $f2               \r\n"
 515         "pshufh $f2, $f0, $f30              \r\n"
 516         "paddsh $f12, $f0, $f2              \r\n"
 517         "dmfc1 $2, $f10                     \r\n"
 518         "dsll $2, 48                        \r\n"
 519         "dsra $2, 48                        \r\n"
 520         "dmfc1 $3, $f12                     \r\n"
 521         "dsll $3, 48                        \r\n"
 522         "dsra $3, 48                        \r\n"
 523         "beqz %2, 1f                        \r\n"
 524         "dli $4, 4                          \r\n"
 525         "ddiv $2, $4                        \r\n"
 526         "ddiv $3, $4                        \r\n"
 527         "dli $4, 5                          \r\n"
 528         "dmul $2, $4                        \r\n"
 529         "dmul $3, $4                        \r\n"
 530         "dli $4, 16                         \r\n"
 531         "ddiv $2, $4                        \r\n"
 532         "ddiv $3, $4                        \r\n"
 533         "daddu $4, $2, $0                   \r\n"
 534         "daddu $2, $3, $0                   \r\n"
 535         "daddu $3, $4, $0                   \r\n"
 536         "b 2f                               \r\n"
 537         "1:                                 \r\n"
 538         "beqz %3, 1f                        \r\n"
 539         "dsra $4, $2, 2                     \r\n"
 540         "daddu $2, $4                       \r\n"
 541         "dsra $4, $3, 2                     \r\n"
 542         "daddu $3, $4                       \r\n"
 543         "dsra $2, 4                         \r\n"
 544         "dsra $3, 4                         \r\n"
 545         "b 2f                               \r\n"
 546         "1:                                 \r\n"
 547         "dli $4, 5                          \r\n"
 548         "dmul $2, $4                        \r\n"
 549         "dmul $3, $4                        \r\n"
 550         "daddiu $2, 32                      \r\n"
 551         "daddiu $3, 32                      \r\n"
 552         "dsra $2, 6                         \r\n"
 553         "dsra $3, 6                         \r\n"
 554         "2:                                 \r\n"
 555         "daddu $5, $2, $3                   \r\n"
 556         "dli $4, 7                          \r\n"
 557         "dmul $5, $4                        \r\n"
 558         "dsubu $8, $5                       \r\n"
 559         "dmtc1 $0, $f8                      \r\n"
 560         "dmtc1 $2, $f0                      \r\n"
 561         "pshufh $f0, $f0, $f8               \r\n"
 562         "dmtc1 $3, $f10                     \r\n"
 563         "pshufh $f10, $f10, $f8             \r\n"
 564         "dmtc1 $8, $f12                     \r\n"
 565         "pshufh $f12, $f12, $f8             \r\n"
 566         "dli $4, 5                          \r\n"
 567         "dmtc1 $4, $f14                     \r\n"
 568         "dmtc1 %8, $f2                      \r\n"
 569         "pmullh $f2, $f2, $f0               \r\n"
 570         "dmtc1 %9, $f4                      \r\n"
 571         "pmullh $f4, $f4, $f0               \r\n"
 572         "dmtc1 %10, $f6                      \r\n"
 573         "pmullh $f6, $f6, $f0               \r\n"
 574         "dmtc1 %11, $f8                      \r\n"
 575         "pmullh $f8, $f8, $f0               \r\n"
 576         "daddu $3, %0, $0                   \r\n"
 577         "dli $2, 16                         \r\n"
 578         "1:                                 \r\n"
 579         "paddsh $f16, $f2, $f12             \r\n"
 580         "psrah $f16, $f16, $f14             \r\n"
 581         "paddsh $f18, $f4, $f12             \r\n"
 582         "psrah $f18, $f18, $f14             \r\n"
 583         "packushb $f20, $f16, $f18          \r\n"
 584         "gssdlc1 $f20, 7($3)                \r\n"
 585         "gssdrc1 $f20, 0($3)                \r\n"
 586         "paddsh $f16, $f6, $f12             \r\n"
 587         "psrah $f16, $f16, $f14             \r\n"
 588         "paddsh $f18, $f8, $f12             \r\n"
 589         "psrah $f18, $f18, $f14             \r\n"
 590         "packushb $f20, $f16, $f18          \r\n"
 591         "gssdlc1 $f20, 15($3)               \r\n"
 592         "gssdrc1 $f20, 8($3)                \r\n"
 593         "paddsh $f12, $f12, $f10            \r\n"
 594         "daddu $3, %1                       \r\n"
 595         "daddiu $2, -1                      \r\n"
 596         "bnez $2, 1b                        \r\n"
 597         ::"r"(src),"r"(stride),"r"(svq3),"r"(rv40),
 598           "r"(ff_pw_m8tom5),"r"(ff_pw_m4tom1),"r"(ff_pw_1to4),"r"(ff_pw_5to8),
 599           "r"(ff_pw_0to3),"r"(ff_pw_4to7),"r"(ff_pw_8tob),"r"(ff_pw_ctof)
 600         : "$2","$3","$4","$5","$6","$7","$8","memory"
 601     );
 602 }
 603
 604 void ff_pred16x16_plane_svq3_8_mmi(uint8_t *src, ptrdiff_t stride)
 605 {
 606     ff_pred16x16_plane_compat_8_mmi(src, stride, 1, 0);
 607 }
 608
 609 void ff_pred16x16_plane_rv40_8_mmi(uint8_t *src, ptrdiff_t stride)
 610 {
 611     ff_pred16x16_plane_compat_8_mmi(src, stride, 0, 1);
 612 }
 613
 614 void ff_pred16x16_plane_h264_8_mmi(uint8_t *src, ptrdiff_t stride)
 615 {
 616     ff_pred16x16_plane_compat_8_mmi(src, stride, 0, 0);
 617 }
 618
 619 void ff_pred8x8_top_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
 620 {
 621     __asm__ volatile (
 622         "dli $2, 2                          \r\n"
 623         "xor $f0, $f0, $f0                  \r\n"
 624         "xor $f2, $f2, $f2                  \r\n"
 625         "xor $f30, $f30, $f30               \r\n"
 626         "negu $3, %1                        \r\n"
 627         "daddu $3, $3, %0                   \r\n"
 628         "gsldlc1 $f4, 7($3)                 \r\n"
 629         "gsldrc1 $f4, 0($3)                 \r\n"
 630         "punpcklbh $f0, $f4, $f30           \r\n"
 631         "punpckhbh $f2, $f4, $f30           \r\n"
 632         "biadd $f0, $f0                     \r\n"
 633         "biadd $f2, $f2                     \r\n"
 634         "pshufh $f0, $f0, $f30              \r\n"
 635         "pshufh $f2, $f2, $f30              \r\n"
 636         "dmtc1 $2, $f4                      \r\n"
 637         "pshufh $f4, $f4, $f30              \r\n"
 638         "paddush $f0, $f0, $f4              \r\n"
 639         "paddush $f2, $f2, $f4              \r\n"
 640         "dmtc1 $2, $f4                      \r\n"
 641         "psrlh $f0, $f0, $f4                \r\n"
 642         "psrlh $f2, $f2, $f4                \r\n"
 643         "packushb $f4, $f0, $f2             \r\n"
 644         "dli $2, 8                          \r\n"
 645         "1:                                 \r\n"
 646         "gssdlc1 $f4, 7(%0)                 \r\n"
 647         "gssdrc1 $f4, 0(%0)                 \r\n"
 648         "daddu %0, %0, %1                   \r\n"
 649         "daddiu $2, $2, -1                  \r\n"
 650         "bnez $2, 1b                        \r\n"
 651         ::"r"(src),"r"(stride)
 652         : "$2","$3","memory"
 653     );
 654 }
 655
 656 void ff_pred8x8_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
 657 {
 658     __asm__ volatile (
 659         "negu $2, %1                        \r\n"
 660         "daddu $2, $2, %0                   \r\n"
 661         "daddiu $5, $2, 4                   \r\n"
 662         "lbu $6, 0($2)                      \r\n"
 663         "daddu $3, $0, $6                   \r\n"
 664         "daddiu $2, 1                       \r\n"
 665         "lbu $6, 0($5)                      \r\n"
 666         "daddu $4, $0, $6                   \r\n"
 667         "daddiu $5, 1                       \r\n"
 668         "lbu $6, 0($2)                      \r\n"
 669         "daddu $3, $3, $6                   \r\n"
 670         "daddiu $2, 1                       \r\n"
 671         "lbu $6, 0($5)                      \r\n"
 672         "daddu $4, $4, $6                   \r\n"
 673         "daddiu $5, 1                       \r\n"
 674         "lbu $6, 0($2)                      \r\n"
 675         "daddu $3, $3, $6                   \r\n"
 676         "daddiu $2, 1                       \r\n"
 677         "lbu $6, 0($5)                      \r\n"
 678         "daddu $4, $4, $6                   \r\n"
 679         "daddiu $5, 1                       \r\n"
 680         "lbu $6, 0($2)                      \r\n"
 681         "daddu $3, $3, $6                   \r\n"
 682         "daddiu $2, 1                       \r\n"
 683         "lbu $6, 0($5)                      \r\n"
 684         "daddu $4, $4, $6                   \r\n"
 685         "daddiu $5, 1                       \r\n"
 686         "dli $6, -1                         \r\n"
 687         "daddu $6, $6, %0                   \r\n"
 688         "lbu $5, 0($6)                      \r\n"
 689         "daddu $7, $0, $5                   \r\n"
 690         "daddu $6, $6, %1                   \r\n"
 691         "lbu $5, 0($6)                      \r\n"
 692         "daddu $7, $7, $5                   \r\n"
 693         "daddu $6, $6, %1                   \r\n"
 694         "lbu $5, 0($6)                      \r\n"
 695         "daddu $7, $7, $5                   \r\n"
 696         "daddu $6, $6, %1                   \r\n"
 697         "lbu $5, 0($6)                      \r\n"
 698         "daddu $7, $7, $5                   \r\n"
 699         "daddu $6, $6, %1                   \r\n"
 700         "lbu $5, 0($6)                      \r\n"
 701         "daddu $8, $0, $5                   \r\n"
 702         "daddu $6, $6, %1                   \r\n"
 703         "lbu $5, 0($6)                      \r\n"
 704         "daddu $8, $8, $5                   \r\n"
 705         "daddu $6, $6, %1                   \r\n"
 706         "lbu $5, 0($6)                      \r\n"
 707         "daddu $8, $8, $5                   \r\n"
 708         "daddu $6, $6, %1                   \r\n"
 709         "lbu $5, 0($6)                      \r\n"
 710         "daddu $8, $8, $5                   \r\n"
 711         "daddu $3, $3, $7                   \r\n"
 712         "daddiu $3, $3, 4                   \r\n"
 713         "daddiu $4, $4, 2                   \r\n"
 714         "daddiu $5, $8, 2                   \r\n"
 715         "daddu $6, $4, $5                   \r\n"
 716         "dsrl $3, 3                         \r\n"
 717         "dsrl $4, 2                         \r\n"
 718         "dsrl $5, 2                         \r\n"
 719         "dsrl $6, 3                         \r\n"
 720         "xor $f30, $f30, $f30               \r\n"
 721         "dmtc1 $3, $f0                      \r\n"
 722         "pshufh $f0, $f0, $f30              \r\n"
 723         "dmtc1 $4, $f2                      \r\n"
 724         "pshufh $f2, $f2, $f30              \r\n"
 725         "dmtc1 $5, $f4                      \r\n"
 726         "pshufh $f4, $f4, $f30              \r\n"
 727         "dmtc1 $6, $f6                      \r\n"
 728         "pshufh $f6, $f6, $f30              \r\n"
 729         "packushb $f0, $f0, $f2             \r\n"
 730         "packushb $f2, $f4, $f6             \r\n"
 731         "daddu $2, $0, %0                   \r\n"
 732         "sdc1 $f0, 0($2)                    \r\n"
 733         "daddu $2, $2, %1                   \r\n"
 734         "sdc1 $f0, 0($2)                    \r\n"
 735         "daddu $2, $2, %1                   \r\n"
 736         "sdc1 $f0, 0($2)                    \r\n"
 737         "daddu $2, $2, %1                   \r\n"
 738         "sdc1 $f0, 0($2)                    \r\n"
 739         "daddu $2, $2, %1                   \r\n"
 740         "sdc1 $f2, 0($2)                    \r\n"
 741         "daddu $2, $2, %1                   \r\n"
 742         "sdc1 $f2, 0($2)                    \r\n"
 743         "daddu $2, $2, %1                   \r\n"
 744         "sdc1 $f2, 0($2)                    \r\n"
 745         "daddu $2, $2, %1                   \r\n"
 746         "sdc1 $f2, 0($2)                    \r\n"
 747         ::"r"(src),"r"(stride)
 748         :"$2","$3","$4","$5","$6","$7","$8","memory"
 749     );
 750 }
 751
 752 void ff_pred8x16_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
 753 {
 754     __asm__ volatile (
 755         "dsubu $2, %0, %1                   \r\n"
 756         "daddu $3, %0, $0                   \r\n"
 757         "ldl $4, 7($2)                      \r\n"
 758         "ldr $4, 0($2)                      \r\n"
 759         "dli $5, 0x10                       \r\n"
 760         "1:                                 \r\n"
 761         "sdl $4, 7($3)                      \r\n"
 762         "sdr $4, 0($3)                      \r\n"
 763         "daddu $3, %1                       \r\n"
 764         "daddiu $5, -1                      \r\n"
 765         "bnez $5, 1b                        \r\n"
 766         ::"r"(src),"r"(stride)
 767         : "$2","$3","$4","$5","memory"
 768     );
 769 }
 770
 771 void ff_pred8x16_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
 772 {
 773     __asm__ volatile (
 774         "daddiu $2, %0, -1                  \r\n"
 775         "daddu $3, %0, $0                   \r\n"
 776         "dli $6, 0x10                       \r\n"
 777         "1:                                 \r\n"
 778         "lbu $4, 0($2)                      \r\n"
 779         "dmul $5, $4, %2                    \r\n"
 780         "sdl $5, 7($3)                      \r\n"
 781         "sdr $5, 0($3)                      \r\n"
 782         "daddu $2, %1                       \r\n"
 783         "daddu $3, %1                       \r\n"
 784         "daddiu $6, -1                      \r\n"
 785         "bnez $6, 1b                        \r\n"
 786         ::"r"(src),"r"(stride),"r"(ff_pb_1)
 787         : "$2","$3","$4","$5","$6","memory"
 788     );
 789 }