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
  27 void ff_pred16x16_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
  28 {
  29     __asm__ volatile (
  30         "dli $8, 16                         \r\n"
  31         "gsldlc1 $f2, 7(%[srcA])            \r\n"
  32         "gsldrc1 $f2, 0(%[srcA])            \r\n"
  33         "gsldlc1 $f4, 15(%[srcA])           \r\n"
  34         "gsldrc1 $f4, 8(%[srcA])            \r\n"
  35         "1:                                 \r\n"
  36         "gssdlc1 $f2, 7(%[src])             \r\n"
  37         "gssdrc1 $f2, 0(%[src])             \r\n"
  38         "gssdlc1 $f4, 15(%[src])            \r\n"
  39         "gssdrc1 $f4, 8(%[src])             \r\n"
  40         "daddu %[src], %[src], %[stride]    \r\n"
  41         "daddi $8, $8, -1                   \r\n"
  42         "bnez $8, 1b                        \r\n"
  43         : [src]"+&r"(src)
  44         : [stride]"r"(stride),[srcA]"r"(src-stride)
  45         : "$8","$f2","$f4"
  46     );
  47 }
  48
  49 void ff_pred16x16_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
  50 {
  51     __asm__ volatile (
  52         "daddiu $2, %[src], -1              \r\n"
  53         "daddu $3, %[src], $0               \r\n"
  54         "dli $6, 0x10                       \r\n"
  55         "1:                                 \r\n"
  56         "lbu $4, 0($2)                      \r\n"
  57         "dmul $5, $4, %[ff_pb_1]            \r\n"
  58         "sdl $5, 7($3)                      \r\n"
  59         "sdr $5, 0($3)                      \r\n"
  60         "sdl $5, 15($3)                     \r\n"
  61         "sdr $5, 8($3)                      \r\n"
  62         "daddu $2, %[stride]                \r\n"
  63         "daddu $3, %[stride]                \r\n"
  64         "daddiu $6, -1                      \r\n"
  65         "bnez $6, 1b                        \r\n"
  66         ::[src]"r"(src),[stride]"r"(stride),[ff_pb_1]"r"(ff_pb_1)
  67         : "$2","$3","$4","$5","$6"
  68     );
  69 }
  70
  71 void ff_pred16x16_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
  72 {
  73     __asm__ volatile (
  74         "daddiu $2, %[src], -1              \r\n"
  75         "dli $6, 0x10                       \r\n"
  76         "xor $8, $8, $8                     \r\n"
  77         "1:                                 \r\n"
  78         "lbu $4, 0($2)                      \r\n"
  79         "daddu $8, $8, $4                   \r\n"
  80         "daddu $2, $2, %[stride]            \r\n"
  81         "daddiu $6, $6, -1                  \r\n"
  82         "bnez $6, 1b                        \r\n"
  83         "dli $6, 0x10                       \r\n"
  84         "negu $3, %[stride]                 \r\n"
  85         "daddu $2, %[src], $3               \r\n"
  86         "2:                                 \r\n"
  87         "lbu $4, 0($2)                      \r\n"
  88         "daddu $8, $8, $4                   \r\n"
  89         "daddiu $2, $2, 1                   \r\n"
  90         "daddiu $6, $6, -1                  \r\n"
  91         "bnez $6, 2b                        \r\n"
  92         "daddiu $8, $8, 0x10                \r\n"
  93         "dsra $8, 5                         \r\n"
  94         "dmul $5, $8, %[ff_pb_1]            \r\n"
  95         "daddu $2, %[src], $0               \r\n"
  96         "dli $6, 0x10                       \r\n"
  97         "3:                                 \r\n"
  98         "sdl $5, 7($2)                      \r\n"
  99         "sdr $5, 0($2)                      \r\n"
 100         "sdl $5, 15($2)                     \r\n"
 101         "sdr $5, 8($2)                      \r\n"
 102         "daddu $2, $2, %[stride]            \r\n"
 103         "daddiu $6, $6, -1                  \r\n"
 104         "bnez $6, 3b                        \r\n"
 105         ::[src]"r"(src),[stride]"r"(stride),[ff_pb_1]"r"(ff_pb_1)
 106         : "$2","$3","$4","$5","$6","$8"
 107     );
 108 }
 109
 110 void ff_pred8x8l_top_dc_8_mmi(uint8_t *src, int has_topleft,
 111         int has_topright, ptrdiff_t stride)
 112 {
 113     uint32_t dc;
 114
 115     __asm__ volatile (
 116         "ldl $8, 7(%[srcA])                 \r\n"
 117         "ldr $8, 0(%[srcA])                 \r\n"
 118         "ldl $9, 7(%[src0])                 \r\n"
 119         "ldr $9, 0(%[src0])                 \r\n"
 120         "ldl $10, 7(%[src1])                \r\n"
 121         "ldr $10, 0(%[src1])                \r\n"
 122         "dmtc1 $8, $f2                      \r\n"
 123         "dmtc1 $9, $f4                      \r\n"
 124         "dmtc1 $10, $f6                     \r\n"
 125         "dmtc1 $0, $f0                      \r\n"
 126         "punpcklbh $f8, $f2, $f0            \r\n"
 127         "punpckhbh $f10, $f2, $f0           \r\n"
 128         "punpcklbh $f12, $f4, $f0           \r\n"
 129         "punpckhbh $f14, $f4, $f0           \r\n"
 130         "punpcklbh $f16, $f6, $f0           \r\n"
 131         "punpckhbh $f18, $f6, $f0           \r\n"
 132         "bnez %[has_topleft], 1f            \r\n"
 133         "pinsrh_0 $f8, $f8, $f12            \r\n"
 134         "1:                                 \r\n"
 135         "bnez %[has_topright], 2f           \r\n"
 136         "pinsrh_3 $f18, $f18, $f14          \r\n"
 137         "2:                                 \r\n"
 138         "daddiu $8, $0, 2                   \r\n"
 139         "dmtc1 $8, $f20                     \r\n"
 140         "pshufh $f22, $f20, $f0             \r\n"
 141         "pmullh $f12, $f12, $f22            \r\n"
 142         "pmullh $f14, $f14, $f22            \r\n"
 143         "paddh $f8, $f8, $f12               \r\n"
 144         "paddh $f10, $f10, $f14             \r\n"
 145         "paddh $f8, $f8, $f16               \r\n"
 146         "paddh $f10, $f10, $f18             \r\n"
 147         "paddh $f8, $f8, $f22               \r\n"
 148         "paddh $f10, $f10, $f22             \r\n"
 149         "psrah $f8, $f8, $f20               \r\n"
 150         "psrah $f10, $f10, $f20             \r\n"
 151         "packushb $f4, $f8, $f10            \r\n"
 152         "biadd $f2, $f4                     \r\n"
 153         "mfc1 $9, $f2                       \r\n"
 154         "addiu $9, $9, 4                    \r\n"
 155         "dsrl $9, $9, 3                     \r\n"
 156         "mul %[dc], $9, %[ff_pb_1]          \r\n"
 157         : [dc]"=r"(dc)
 158         : [srcA]"r"(src-stride-1),[src0]"r"(src-stride),
 159           [src1]"r"(src-stride+1),[has_topleft]"r"(has_topleft),
 160           [has_topright]"r"(has_topright),[ff_pb_1]"r"(ff_pb_1)
 161         : "$8","$9","$10","$f2","$f4","$f6","$f8","$f10","$f12","$f14","$f16",
 162           "$f18","$f20","$f22"
 163     );
 164
 165     __asm__ volatile (
 166         "dli $8, 8                          \r\n"
 167         "1:                                 \r\n"
 168         "punpcklwd $f2, %[dc], %[dc]        \r\n"
 169         "gssdlc1 $f2, 7(%[src])             \r\n"
 170         "gssdrc1 $f2, 0(%[src])             \r\n"
 171         "daddu %[src], %[src], %[stride]    \r\n"
 172         "daddi $8, $8, -1                   \r\n"
 173         "bnez $8, 1b                        \r\n"
 174         : [src]"+&r"(src)
 175         : [dc]"f"(dc),[stride]"r"(stride)
 176         : "$8","$f2"
 177     );
 178 }
 179
 180 void ff_pred8x8l_dc_8_mmi(uint8_t *src, int has_topleft,
 181         int has_topright, ptrdiff_t stride)
 182 {
 183     uint32_t dc, dc1, dc2;
 184
 185     const int l0 = ((has_topleft ? src[-1+-1*stride] : src[-1+0*stride]) + 2*src[-1+0*stride] + src[-1+1*stride] + 2) >> 2;
 186     const int l1 = (src[-1+0*stride] + 2*src[-1+1*stride] + src[-1+2*stride] + 2) >> 2;
 187     const int l2 = (src[-1+1*stride] + 2*src[-1+2*stride] + src[-1+3*stride] + 2) >> 2;
 188     const int l3 = (src[-1+2*stride] + 2*src[-1+3*stride] + src[-1+4*stride] + 2) >> 2;
 189     const int l4 = (src[-1+3*stride] + 2*src[-1+4*stride] + src[-1+5*stride] + 2) >> 2;
 190     const int l5 = (src[-1+4*stride] + 2*src[-1+5*stride] + src[-1+6*stride] + 2) >> 2;
 191     const int l6 = (src[-1+5*stride] + 2*src[-1+6*stride] + src[-1+7*stride] + 2) >> 2;
 192     const int l7 = (src[-1+6*stride] + 2*src[-1+7*stride] + src[-1+7*stride] + 2) >> 2;
 193
 194     __asm__ volatile (
 195         "ldl $8, 7(%[srcA])                 \r\n"
 196         "ldr $8, 0(%[srcA])                 \r\n"
 197         "ldl $9, 7(%[src0])                 \r\n"
 198         "ldr $9, 0(%[src0])                 \r\n"
 199         "ldl $10, 7(%[src1])                \r\n"
 200         "ldr $10, 0(%[src1])                \r\n"
 201         "dmtc1 $8, $f2                      \r\n"
 202         "dmtc1 $9, $f4                      \r\n"
 203         "dmtc1 $10, $f6                     \r\n"
 204         "dmtc1 $0, $f0                      \r\n"
 205         "punpcklbh $f8, $f2, $f0            \r\n"
 206         "punpckhbh $f10, $f2, $f0           \r\n"
 207         "punpcklbh $f12, $f4, $f0           \r\n"
 208         "punpckhbh $f14, $f4, $f0           \r\n"
 209         "punpcklbh $f16, $f6, $f0           \r\n"
 210         "punpckhbh $f18, $f6, $f0           \r\n"
 211         "daddiu $8, $0, 3                   \r\n"
 212         "dmtc1 $8, $f20                     \r\n"
 213         "pshufh $f28, $f10, $f20            \r\n"
 214         "pshufh $f30, $f18, $f20            \r\n"
 215         "pinsrh_3 $f10, $f10, $f30          \r\n"
 216         "pinsrh_3 $f18, $f18, $f28          \r\n"
 217         "bnez %[has_topleft], 1f            \r\n"
 218         "pinsrh_0 $f8, $f8, $f12            \r\n"
 219         "1:                                 \r\n"
 220         "bnez %[has_topright], 2f           \r\n"
 221         "pshufh $f30, $f14, $f20            \r\n"
 222         "pinsrh_3 $f10, $f10, $f30          \r\n"
 223         "2:                                 \r\n"
 224         "daddiu $8, $0, 2                   \r\n"
 225         "dmtc1 $8, $f20                     \r\n"
 226         "pshufh $f22, $f20, $f0             \r\n"
 227         "pmullh $f12, $f12, $f22            \r\n"
 228         "pmullh $f14, $f14, $f22            \r\n"
 229         "paddh $f8, $f8, $f12               \r\n"
 230         "paddh $f10, $f10, $f14             \r\n"
 231         "paddh $f8, $f8, $f16               \r\n"
 232         "paddh $f10, $f10, $f18             \r\n"
 233         "paddh $f8, $f8, $f22               \r\n"
 234         "paddh $f10, $f10, $f22             \r\n"
 235         "psrah $f8, $f8, $f20               \r\n"
 236         "psrah $f10, $f10, $f20             \r\n"
 237         "packushb $f4, $f8, $f10            \r\n"
 238         "biadd $f2, $f4                     \r\n"
 239         "mfc1 %[dc2], $f2                   \r\n"
 240         : [dc2]"=r"(dc2)
 241         : [srcA]"r"(src-stride-1),[src0]"r"(src-stride),
 242           [src1]"r"(src-stride+1),[has_topleft]"r"(has_topleft),
 243           [has_topright]"r"(has_topright)
 244         : "$8","$9","$10","$f2","$f4","$f6","$f8","$f10","$f12","$f14","$f16",
 245           "$f18","$f20","$f22"
 246     );
 247
 248     dc1 = l0+l1+l2+l3+l4+l5+l6+l7;
 249     dc = ((dc1+dc2+8)>>4)*0x01010101U;
 250
 251     __asm__ volatile (
 252         "dli $8, 8                          \r\n"
 253         "1:                                 \r\n"
 254         "punpcklwd $f2, %[dc], %[dc]        \r\n"
 255         "gssdlc1 $f2, 7(%[src])             \r\n"
 256         "gssdrc1 $f2, 0(%[src])             \r\n"
 257         "daddu %[src], %[src], %[stride]    \r\n"
 258         "daddi $8, $8, -1                   \r\n"
 259         "bnez $8, 1b                        \r\n"
 260         : [src]"+&r"(src)
 261         : [dc]"f"(dc),[stride]"r"(stride)
 262         : "$8","$f2"
 263     );
 264 }
 265
 266 void ff_pred8x8l_vertical_8_mmi(uint8_t *src, int has_topleft,
 267         int has_topright, ptrdiff_t stride)
 268 {
 269     __asm__ volatile (
 270         "ldl $8, 7(%[srcA])                 \r\n"
 271         "ldr $8, 0(%[srcA])                 \r\n"
 272         "ldl $9, 7(%[src0])                 \r\n"
 273         "ldr $9, 0(%[src0])                 \r\n"
 274         "ldl $10, 7(%[src1])                \r\n"
 275         "ldr $10, 0(%[src1])                \r\n"
 276         "dmtc1 $8, $f2                      \r\n"
 277         "dmtc1 $9, $f4                      \r\n"
 278         "dmtc1 $10, $f6                     \r\n"
 279         "dmtc1 $0, $f0                      \r\n"
 280         "punpcklbh $f8, $f2, $f0            \r\n"
 281         "punpckhbh $f10, $f2, $f0           \r\n"
 282         "punpcklbh $f12, $f4, $f0           \r\n"
 283         "punpckhbh $f14, $f4, $f0           \r\n"
 284         "punpcklbh $f16, $f6, $f0           \r\n"
 285         "punpckhbh $f18, $f6, $f0           \r\n"
 286         "bnez %[has_topleft], 1f            \r\n"
 287         "pinsrh_0 $f8, $f8, $f12            \r\n"
 288         "1:                                 \r\n"
 289         "bnez %[has_topright], 2f           \r\n"
 290         "pinsrh_3 $f18, $f18, $f14          \r\n"
 291         "2:                                 \r\n"
 292         "daddiu $8, $0, 2                   \r\n"
 293         "dmtc1 $8, $f20                     \r\n"
 294         "pshufh $f22, $f20, $f0             \r\n"
 295         "pmullh $f12, $f12, $f22            \r\n"
 296         "pmullh $f14, $f14, $f22            \r\n"
 297         "paddh $f8, $f8, $f12               \r\n"
 298         "paddh $f10, $f10, $f14             \r\n"
 299         "paddh $f8, $f8, $f16               \r\n"
 300         "paddh $f10, $f10, $f18             \r\n"
 301         "paddh $f8, $f8, $f22               \r\n"
 302         "paddh $f10, $f10, $f22             \r\n"
 303         "psrah $f8, $f8, $f20               \r\n"
 304         "psrah $f10, $f10, $f20             \r\n"
 305         "packushb $f4, $f8, $f10            \r\n"
 306         "sdc1 $f4, 0(%[src])                \r\n"
 307         : [src]"=r"(src)
 308         : [srcA]"r"(src-stride-1),[src0]"r"(src-stride),
 309           [src1]"r"(src-stride+1),[has_topleft]"r"(has_topleft),
 310           [has_topright]"r"(has_topright)
 311         : "$8","$9","$10","$f2","$f4","$f6","$f8","$f10","$f12","$f14","$f16",
 312           "$f18","$f20","$f22"
 313     );
 314
 315     __asm__ volatile (
 316         "dli $8, 7                          \r\n"
 317         "gsldlc1 $f2, 7(%[src])             \r\n"
 318         "gsldrc1 $f2, 0(%[src])             \r\n"
 319         "dadd %[src], %[src], %[stride]     \r\n"
 320         "1:                                 \r\n"
 321         "gssdlc1 $f2, 7(%[src])             \r\n"
 322         "gssdrc1 $f2, 0(%[src])             \r\n"
 323         "daddu %[src], %[src], %[stride]    \r\n"
 324         "daddi $8, $8, -1                   \r\n"
 325         "bnez $8, 1b                        \r\n"
 326         : [src]"+&r"(src)
 327         : [stride]"r"(stride)
 328         : "$8","$f2"
 329     );
 330 }
 331
 332 void ff_pred4x4_dc_8_mmi(uint8_t *src, const uint8_t *topright,
 333         ptrdiff_t stride)
 334 {
 335     const int dc = (src[-stride] + src[1-stride] + src[2-stride]
 336                  + src[3-stride] + src[-1+0*stride] + src[-1+1*stride]
 337                  + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
 338
 339     __asm__ volatile (
 340         "daddu $2, %[dc], $0                \r\n"
 341         "dmul $3, $2, %[ff_pb_1]            \r\n"
 342         "xor $4, $4, $4                     \r\n"
 343         "gsswx $3, 0(%[src],$4)             \r\n"
 344         "daddu $4, %[stride]                \r\n"
 345         "gsswx $3, 0(%[src],$4)             \r\n"
 346         "daddu $4, %[stride]                \r\n"
 347         "gsswx $3, 0(%[src],$4)             \r\n"
 348         "daddu $4, %[stride]                \r\n"
 349         "gsswx $3, 0(%[src],$4)             \r\n"
 350         ::[src]"r"(src),[stride]"r"(stride),[dc]"r"(dc),[ff_pb_1]"r"(ff_pb_1)
 351         : "$2","$3","$4"
 352     );
 353 }
 354
 355 void ff_pred8x8_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
 356 {
 357     __asm__ volatile (
 358         "dsubu $2, %[src], %[stride]        \r\n"
 359         "daddu $3, %[src], $0               \r\n"
 360         "ldl $4, 7($2)                      \r\n"
 361         "ldr $4, 0($2)                      \r\n"
 362         "dli $5, 0x8                        \r\n"
 363         "1:                                 \r\n"
 364         "sdl $4, 7($3)                      \r\n"
 365         "sdr $4, 0($3)                      \r\n"
 366         "daddu $3, %[stride]                \r\n"
 367         "daddiu $5, -1                      \r\n"
 368         "bnez $5, 1b                        \r\n"
 369         ::[src]"r"(src),[stride]"r"(stride)
 370         : "$2","$3","$4","$5"
 371     );
 372 }
 373
 374 void ff_pred8x8_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
 375 {
 376     __asm__ volatile (
 377         "daddiu $2, %[src], -1              \r\n"
 378         "daddu $3, %[src], $0               \r\n"
 379         "dli $6, 0x8                        \r\n"
 380         "1:                                 \r\n"
 381         "lbu $4, 0($2)                      \r\n"
 382         "dmul $5, $4, %[ff_pb_1]            \r\n"
 383         "sdl $5, 7($3)                      \r\n"
 384         "sdr $5, 0($3)                      \r\n"
 385         "daddu $2, %[stride]                \r\n"
 386         "daddu $3, %[stride]                \r\n"
 387         "daddiu $6, -1                      \r\n"
 388         "bnez $6, 1b                        \r\n"
 389         ::[src]"r"(src),[stride]"r"(stride),[ff_pb_1]"r"(ff_pb_1)
 390         : "$2","$3","$4","$5","$6"
 391     );
 392 }
 393
 394 static void ff_pred16x16_plane_compat_8_mmi(uint8_t *src, ptrdiff_t stride,
 395         const int svq3, const int rv40)
 396 {
 397     __asm__ volatile (
 398         "negu $2, %[stride]                 \r\n"
 399         "daddu $3, %[src], $2               \r\n"
 400         "xor $f8, $f8, $f8                  \r\n"
 401         "gslwlc1 $f0, 2($3)                 \r\n"
 402         "gslwrc1 $f0, -1($3)                \r\n"
 403         "gslwlc1 $f2, 6($3)                 \r\n"
 404         "gslwrc1 $f2, 3($3)                 \r\n"
 405         "gslwlc1 $f4, 11($3)                \r\n"
 406         "gslwrc1 $f4, 8($3)                 \r\n"
 407         "gslwlc1 $f6, 15($3)                \r\n"
 408         "gslwrc1 $f6, 12($3)                \r\n"
 409         "punpcklbh $f0, $f0, $f8            \r\n"
 410         "punpcklbh $f2, $f2, $f8            \r\n"
 411         "punpcklbh $f4, $f4, $f8            \r\n"
 412         "punpcklbh $f6, $f6, $f8            \r\n"
 413         "dmtc1 %[ff_pw_m8tom5], $f20        \r\n"
 414         "dmtc1 %[ff_pw_m4tom1], $f22        \r\n"
 415         "dmtc1 %[ff_pw_1to4], $f24          \r\n"
 416         "dmtc1 %[ff_pw_5to8], $f26          \r\n"
 417         "pmullh $f0, $f0, $f20              \r\n"
 418         "pmullh $f2, $f2, $f22              \r\n"
 419         "pmullh $f4, $f4, $f24              \r\n"
 420         "pmullh $f6, $f6, $f26              \r\n"
 421         "paddsh $f0, $f0, $f4               \r\n"
 422         "paddsh $f2, $f2, $f6               \r\n"
 423         "paddsh $f0, $f0, $f2               \r\n"
 424         "dli $4, 0xE                        \r\n"
 425         "dmtc1 $4, $f28                     \r\n"
 426         "pshufh $f2, $f0, $f28              \r\n"
 427         "paddsh $f0, $f0, $f2               \r\n"
 428         "dli $4, 0x1                        \r\n"
 429         "dmtc1 $4, $f30                     \r\n"
 430         "pshufh $f2, $f0, $f30              \r\n"
 431         "paddsh $f10, $f0, $f2              \r\n"
 432         "daddiu $3, %[src], -1              \r\n"
 433         "daddu $3, $2                       \r\n"
 434         "lbu $4, 0($3)                      \r\n"
 435         "lbu $8, 16($3)                     \r\n"
 436         "daddu $3, %[stride]                \r\n"
 437         "lbu $5, 0($3)                      \r\n"
 438         "daddu $3, %[stride]                \r\n"
 439         "lbu $6, 0($3)                      \r\n"
 440         "daddu $3, %[stride]                \r\n"
 441         "lbu $7, 0($3)                      \r\n"
 442         "dsll $5, 16                        \r\n"
 443         "dsll $6, 32                        \r\n"
 444         "dsll $7, 48                        \r\n"
 445         "or $6, $7                          \r\n"
 446         "or $4, $5                          \r\n"
 447         "or $4, $6                          \r\n"
 448         "dmtc1 $4, $f0                      \r\n"
 449         "daddu $3, %[stride]                \r\n"
 450         "lbu $4, 0($3)                      \r\n"
 451         "daddu $3, %[stride]                \r\n"
 452         "lbu $5, 0($3)                      \r\n"
 453         "daddu $3, %[stride]                \r\n"
 454         "lbu $6, 0($3)                      \r\n"
 455         "daddu $3, %[stride]                \r\n"
 456         "lbu $7, 0($3)                      \r\n"
 457         "dsll $5, 16                        \r\n"
 458         "dsll $6, 32                        \r\n"
 459         "dsll $7, 48                        \r\n"
 460         "or $6, $7                          \r\n"
 461         "or $4, $5                          \r\n"
 462         "or $4, $6                          \r\n"
 463         "dmtc1 $4, $f2                      \r\n"
 464         "daddu $3, %[stride]                \r\n"
 465         "daddu $3, %[stride]                \r\n"
 466         "lbu $4, 0($3)                      \r\n"
 467         "daddu $3, %[stride]                \r\n"
 468         "lbu $5, 0($3)                      \r\n"
 469         "daddu $3, %[stride]                \r\n"
 470         "lbu $6, 0($3)                      \r\n"
 471         "daddu $3, %[stride]                \r\n"
 472         "lbu $7, 0($3)                      \r\n"
 473         "dsll $5, 16                        \r\n"
 474         "dsll $6, 32                        \r\n"
 475         "dsll $7, 48                        \r\n"
 476         "or $6, $7                          \r\n"
 477         "or $4, $5                          \r\n"
 478         "or $4, $6                          \r\n"
 479         "dmtc1 $4, $f4                      \r\n"
 480         "daddu $3, %[stride]                \r\n"
 481         "lbu $4, 0($3)                      \r\n"
 482         "daddu $3, %[stride]                \r\n"
 483         "lbu $5, 0($3)                      \r\n"
 484         "daddu $3, %[stride]                \r\n"
 485         "lbu $6, 0($3)                      \r\n"
 486         "daddu $3, %[stride]                \r\n"
 487         "lbu $7, 0($3)                      \r\n"
 488         "daddu $8, $7                       \r\n"
 489         "daddiu $8, 1                       \r\n"
 490         "dsll $8, 4                         \r\n"
 491         "dsll $5, 16                        \r\n"
 492         "dsll $6, 32                        \r\n"
 493         "dsll $7, 48                        \r\n"
 494         "or $6, $7                          \r\n"
 495         "or $4, $5                          \r\n"
 496         "or $4, $6                          \r\n"
 497         "dmtc1 $4, $f6                      \r\n"
 498         "pmullh $f0, $f0, $f20              \r\n"
 499         "pmullh $f2, $f2, $f22              \r\n"
 500         "pmullh $f4, $f4, $f24              \r\n"
 501         "pmullh $f6, $f6, $f26              \r\n"
 502         "paddsh $f0, $f0, $f4               \r\n"
 503         "paddsh $f2, $f2, $f6               \r\n"
 504         "paddsh $f0, $f0, $f2               \r\n"
 505         "pshufh $f2, $f0, $f28              \r\n"
 506         "paddsh $f0, $f0, $f2               \r\n"
 507         "pshufh $f2, $f0, $f30              \r\n"
 508         "paddsh $f12, $f0, $f2              \r\n"
 509         "dmfc1 $2, $f10                     \r\n"
 510         "dsll $2, 48                        \r\n"
 511         "dsra $2, 48                        \r\n"
 512         "dmfc1 $3, $f12                     \r\n"
 513         "dsll $3, 48                        \r\n"
 514         "dsra $3, 48                        \r\n"
 515         "beqz %[svq3], 1f                   \r\n"
 516         "dli $4, 4                          \r\n"
 517         "ddiv $2, $4                        \r\n"
 518         "ddiv $3, $4                        \r\n"
 519         "dli $4, 5                          \r\n"
 520         "dmul $2, $4                        \r\n"
 521         "dmul $3, $4                        \r\n"
 522         "dli $4, 16                         \r\n"
 523         "ddiv $2, $4                        \r\n"
 524         "ddiv $3, $4                        \r\n"
 525         "daddu $4, $2, $0                   \r\n"
 526         "daddu $2, $3, $0                   \r\n"
 527         "daddu $3, $4, $0                   \r\n"
 528         "b 2f                               \r\n"
 529         "1:                                 \r\n"
 530         "beqz %[rv40], 1f                   \r\n"
 531         "dsra $4, $2, 2                     \r\n"
 532         "daddu $2, $4                       \r\n"
 533         "dsra $4, $3, 2                     \r\n"
 534         "daddu $3, $4                       \r\n"
 535         "dsra $2, 4                         \r\n"
 536         "dsra $3, 4                         \r\n"
 537         "b 2f                               \r\n"
 538         "1:                                 \r\n"
 539         "dli $4, 5                          \r\n"
 540         "dmul $2, $4                        \r\n"
 541         "dmul $3, $4                        \r\n"
 542         "daddiu $2, 32                      \r\n"
 543         "daddiu $3, 32                      \r\n"
 544         "dsra $2, 6                         \r\n"
 545         "dsra $3, 6                         \r\n"
 546         "2:                                 \r\n"
 547         "daddu $5, $2, $3                   \r\n"
 548         "dli $4, 7                          \r\n"
 549         "dmul $5, $4                        \r\n"
 550         "dsubu $8, $5                       \r\n"
 551         "dmtc1 $0, $f8                      \r\n"
 552         "dmtc1 $2, $f0                      \r\n"
 553         "pshufh $f0, $f0, $f8               \r\n"
 554         "dmtc1 $3, $f10                     \r\n"
 555         "pshufh $f10, $f10, $f8             \r\n"
 556         "dmtc1 $8, $f12                     \r\n"
 557         "pshufh $f12, $f12, $f8             \r\n"
 558         "dli $4, 5                          \r\n"
 559         "dmtc1 $4, $f14                     \r\n"
 560         "pmullh $f2, %[ff_pw_0to3], $f0     \r\n"
 561         "pmullh $f4, %[ff_pw_4to7], $f0     \r\n"
 562         "pmullh $f6, %[ff_pw_8tob], $f0     \r\n"
 563         "pmullh $f8, %[ff_pw_ctof], $f0     \r\n"
 564         "daddu $3, %[src], $0               \r\n"
 565         "dli $2, 16                         \r\n"
 566         "1:                                 \r\n"
 567         "paddsh $f16, $f2, $f12             \r\n"
 568         "psrah $f16, $f16, $f14             \r\n"
 569         "paddsh $f18, $f4, $f12             \r\n"
 570         "psrah $f18, $f18, $f14             \r\n"
 571         "packushb $f20, $f16, $f18          \r\n"
 572         "gssdlc1 $f20, 7($3)                \r\n"
 573         "gssdrc1 $f20, 0($3)                \r\n"
 574         "paddsh $f16, $f6, $f12             \r\n"
 575         "psrah $f16, $f16, $f14             \r\n"
 576         "paddsh $f18, $f8, $f12             \r\n"
 577         "psrah $f18, $f18, $f14             \r\n"
 578         "packushb $f20, $f16, $f18          \r\n"
 579         "gssdlc1 $f20, 15($3)               \r\n"
 580         "gssdrc1 $f20, 8($3)                \r\n"
 581         "paddsh $f12, $f12, $f10            \r\n"
 582         "daddu $3, %[stride]                \r\n"
 583         "daddiu $2, -1                      \r\n"
 584         "bnez $2, 1b                        \r\n"
 585         ::[src]"r"(src),[stride]"r"(stride),[svq3]"r"(svq3),[rv40]"r"(rv40),
 586           [ff_pw_m8tom5]"r"(ff_pw_m8tom5),[ff_pw_m4tom1]"r"(ff_pw_m4tom1),
 587           [ff_pw_1to4]"r"(ff_pw_1to4),[ff_pw_5to8]"r"(ff_pw_5to8),
 588           [ff_pw_0to3]"f"(ff_pw_0to3),[ff_pw_4to7]"f"(ff_pw_4to7),
 589           [ff_pw_8tob]"f"(ff_pw_8tob),[ff_pw_ctof]"f"(ff_pw_ctof)
 590         : "$2","$3","$4","$5","$6","$7","$8","$f0","$f2","$f4","$f6","$f8",
 591           "$f10","$f12","$f14","$f16","$f18","$f20","$f22","$f24","$f26",
 592           "$f28","$f30"
 593     );
 594 }
 595
 596 void ff_pred16x16_plane_svq3_8_mmi(uint8_t *src, ptrdiff_t stride)
 597 {
 598     ff_pred16x16_plane_compat_8_mmi(src, stride, 1, 0);
 599 }
 600
 601 void ff_pred16x16_plane_rv40_8_mmi(uint8_t *src, ptrdiff_t stride)
 602 {
 603     ff_pred16x16_plane_compat_8_mmi(src, stride, 0, 1);
 604 }
 605
 606 void ff_pred16x16_plane_h264_8_mmi(uint8_t *src, ptrdiff_t stride)
 607 {
 608     ff_pred16x16_plane_compat_8_mmi(src, stride, 0, 0);
 609 }
 610
 611 void ff_pred8x8_top_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
 612 {
 613     __asm__ volatile (
 614         "dli $2, 2                          \r\n"
 615         "xor $f0, $f0, $f0                  \r\n"
 616         "xor $f2, $f2, $f2                  \r\n"
 617         "xor $f30, $f30, $f30               \r\n"
 618         "negu $3, %[stride]                 \r\n"
 619         "daddu $3, $3, %[src]               \r\n"
 620         "gsldlc1 $f4, 7($3)                 \r\n"
 621         "gsldrc1 $f4, 0($3)                 \r\n"
 622         "punpcklbh $f0, $f4, $f30           \r\n"
 623         "punpckhbh $f2, $f4, $f30           \r\n"
 624         "biadd $f0, $f0                     \r\n"
 625         "biadd $f2, $f2                     \r\n"
 626         "pshufh $f0, $f0, $f30              \r\n"
 627         "pshufh $f2, $f2, $f30              \r\n"
 628         "dmtc1 $2, $f4                      \r\n"
 629         "pshufh $f4, $f4, $f30              \r\n"
 630         "paddush $f0, $f0, $f4              \r\n"
 631         "paddush $f2, $f2, $f4              \r\n"
 632         "dmtc1 $2, $f4                      \r\n"
 633         "psrlh $f0, $f0, $f4                \r\n"
 634         "psrlh $f2, $f2, $f4                \r\n"
 635         "packushb $f4, $f0, $f2             \r\n"
 636         "dli $2, 8                          \r\n"
 637         "1:                                 \r\n"
 638         "gssdlc1 $f4, 7(%[src])             \r\n"
 639         "gssdrc1 $f4, 0(%[src])             \r\n"
 640         "daddu %[src], %0, %[stride]        \r\n"
 641         "daddiu $2, $2, -1                  \r\n"
 642         "bnez $2, 1b                        \r\n"
 643         ::[src]"r"(src),[stride]"r"(stride)
 644         : "$2","$3","$f0","$f2","$f4","$f30"
 645     );
 646 }
 647
 648 void ff_pred8x8_dc_8_mmi(uint8_t *src, ptrdiff_t stride)
 649 {
 650     __asm__ volatile (
 651         "negu $2, %[stride]                 \r\n"
 652         "daddu $2, $2, %[src]               \r\n"
 653         "daddiu $5, $2, 4                   \r\n"
 654         "lbu $6, 0($2)                      \r\n"
 655         "daddu $3, $0, $6                   \r\n"
 656         "daddiu $2, 1                       \r\n"
 657         "lbu $6, 0($5)                      \r\n"
 658         "daddu $4, $0, $6                   \r\n"
 659         "daddiu $5, 1                       \r\n"
 660         "lbu $6, 0($2)                      \r\n"
 661         "daddu $3, $3, $6                   \r\n"
 662         "daddiu $2, 1                       \r\n"
 663         "lbu $6, 0($5)                      \r\n"
 664         "daddu $4, $4, $6                   \r\n"
 665         "daddiu $5, 1                       \r\n"
 666         "lbu $6, 0($2)                      \r\n"
 667         "daddu $3, $3, $6                   \r\n"
 668         "daddiu $2, 1                       \r\n"
 669         "lbu $6, 0($5)                      \r\n"
 670         "daddu $4, $4, $6                   \r\n"
 671         "daddiu $5, 1                       \r\n"
 672         "lbu $6, 0($2)                      \r\n"
 673         "daddu $3, $3, $6                   \r\n"
 674         "daddiu $2, 1                       \r\n"
 675         "lbu $6, 0($5)                      \r\n"
 676         "daddu $4, $4, $6                   \r\n"
 677         "daddiu $5, 1                       \r\n"
 678         "dli $6, -1                         \r\n"
 679         "daddu $6, $6, %[src]               \r\n"
 680         "lbu $5, 0($6)                      \r\n"
 681         "daddu $7, $0, $5                   \r\n"
 682         "daddu $6, $6, %[stride]            \r\n"
 683         "lbu $5, 0($6)                      \r\n"
 684         "daddu $7, $7, $5                   \r\n"
 685         "daddu $6, $6, %[stride]            \r\n"
 686         "lbu $5, 0($6)                      \r\n"
 687         "daddu $7, $7, $5                   \r\n"
 688         "daddu $6, $6, %[stride]            \r\n"
 689         "lbu $5, 0($6)                      \r\n"
 690         "daddu $7, $7, $5                   \r\n"
 691         "daddu $6, $6, %[stride]            \r\n"
 692         "lbu $5, 0($6)                      \r\n"
 693         "daddu $8, $0, $5                   \r\n"
 694         "daddu $6, $6, %[stride]            \r\n"
 695         "lbu $5, 0($6)                      \r\n"
 696         "daddu $8, $8, $5                   \r\n"
 697         "daddu $6, $6, %[stride]            \r\n"
 698         "lbu $5, 0($6)                      \r\n"
 699         "daddu $8, $8, $5                   \r\n"
 700         "daddu $6, $6, %[stride]            \r\n"
 701         "lbu $5, 0($6)                      \r\n"
 702         "daddu $8, $8, $5                   \r\n"
 703         "daddu $3, $3, $7                   \r\n"
 704         "daddiu $3, $3, 4                   \r\n"
 705         "daddiu $4, $4, 2                   \r\n"
 706         "daddiu $5, $8, 2                   \r\n"
 707         "daddu $6, $4, $5                   \r\n"
 708         "dsrl $3, 3                         \r\n"
 709         "dsrl $4, 2                         \r\n"
 710         "dsrl $5, 2                         \r\n"
 711         "dsrl $6, 3                         \r\n"
 712         "xor $f30, $f30, $f30               \r\n"
 713         "dmtc1 $3, $f0                      \r\n"
 714         "pshufh $f0, $f0, $f30              \r\n"
 715         "dmtc1 $4, $f2                      \r\n"
 716         "pshufh $f2, $f2, $f30              \r\n"
 717         "dmtc1 $5, $f4                      \r\n"
 718         "pshufh $f4, $f4, $f30              \r\n"
 719         "dmtc1 $6, $f6                      \r\n"
 720         "pshufh $f6, $f6, $f30              \r\n"
 721         "packushb $f0, $f0, $f2             \r\n"
 722         "packushb $f2, $f4, $f6             \r\n"
 723         "daddu $2, $0, %[src]               \r\n"
 724         "sdc1 $f0, 0($2)                    \r\n"
 725         "daddu $2, $2, %[stride]            \r\n"
 726         "sdc1 $f0, 0($2)                    \r\n"
 727         "daddu $2, $2, %[stride]            \r\n"
 728         "sdc1 $f0, 0($2)                    \r\n"
 729         "daddu $2, $2, %[stride]            \r\n"
 730         "sdc1 $f0, 0($2)                    \r\n"
 731         "daddu $2, $2, %[stride]            \r\n"
 732         "sdc1 $f2, 0($2)                    \r\n"
 733         "daddu $2, $2, %[stride]            \r\n"
 734         "sdc1 $f2, 0($2)                    \r\n"
 735         "daddu $2, $2, %[stride]            \r\n"
 736         "sdc1 $f2, 0($2)                    \r\n"
 737         "daddu $2, $2, %[stride]            \r\n"
 738         "sdc1 $f2, 0($2)                    \r\n"
 739         ::[src]"r"(src),[stride]"r"(stride)
 740         : "$2","$3","$4","$5","$6","$7","$8","$f0","$f2","$f4","$f6","$f30"
 741     );
 742 }
 743
 744 void ff_pred8x16_vertical_8_mmi(uint8_t *src, ptrdiff_t stride)
 745 {
 746     __asm__ volatile (
 747         "gsldlc1 $f2, 7(%[srcA])            \r\n"
 748         "gsldrc1 $f2, 0(%[srcA])            \r\n"
 749         "dli $8, 16                         \r\n"
 750         "1:                                 \r\n"
 751         "gssdlc1 $f2, 7(%[src])             \r\n"
 752         "gssdrc1 $f2, 0(%[src])             \r\n"
 753         "daddu %[src], %[src], %[stride]    \r\n"
 754         "daddi $8, $8, -1                   \r\n"
 755         "bnez $8, 1b                        \r\n"
 756         : [src]"+&r"(src)
 757         : [stride]"r"(stride),[srcA]"r"(src-stride)
 758         : "$8","$f2"
 759     );
 760 }
 761
 762 void ff_pred8x16_horizontal_8_mmi(uint8_t *src, ptrdiff_t stride)
 763 {
 764     __asm__ volatile (
 765         "daddiu $2, %[src], -1              \r\n"
 766         "daddu $3, %[src], $0               \r\n"
 767         "dli $6, 0x10                       \r\n"
 768         "1:                                 \r\n"
 769         "lbu $4, 0($2)                      \r\n"
 770         "dmul $5, $4, %[ff_pb_1]            \r\n"
 771         "sdl $5, 7($3)                      \r\n"
 772         "sdr $5, 0($3)                      \r\n"
 773         "daddu $2, %[stride]                \r\n"
 774         "daddu $3, %[stride]                \r\n"
 775         "daddiu $6, -1                      \r\n"
 776         "bnez $6, 1b                        \r\n"
 777         ::[src]"r"(src),[stride]"r"(stride),[ff_pb_1]"r"(ff_pb_1)
 778         : "$2","$3","$4","$5","$6"
 779     );
 780 }