plugins/gpu_neon/psx_gpu/vector_ops.h

   1 /*
   2  * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public License as
   6  * published by the Free Software Foundation; either version 2 of
   7  * the License, or (at your option) any later version.
   8  *
   9  * This program is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12  * General Public License for more details.
  13  */
  14
  15 #ifndef VECTOR_OPS
  16 #define VECTOR_OPS
  17
  18 #include "vector_types.h"
  19
  20
  21 #define foreach_element(iterations, operation)                                 \
  22 {                                                                              \
  23   u32 _i;                                                                      \
  24   for(_i = 0; _i < iterations; _i++)                                           \
  25   {                                                                            \
  26     operation;                                                                 \
  27   }                                                                            \
  28 }                                                                              \
  29
  30 #define load_64b(dest, source)                                                 \
  31  *((u64 *)(dest).e) = *((u64 *)(source))                                       \
  32
  33 #define load_128b(dest, source)                                                \
  34  *((u64 *)(dest).e) = *((u64 *)(source));                                      \
  35  *((u64 *)(dest).e + 1) = *(((u64 *)(source)) + 1)                             \
  36
  37 #define load_8x16b(dest, source)                                               \
  38   foreach_element(8, (dest).e[_i] = ((u16 *)(source))[_i])                     \
  39
  40 #define store_64b(source, dest)                                                \
  41  *((u64 *)(dest)) = *((u64 *)(source).e)                                       \
  42
  43 #define store_128b(source, dest)                                               \
  44  *((u64 *)(dest)) = *((u64 *)(source).e);                                      \
  45  *(((u64 *)(dest)) + 1) = *((u64 *)(source).e + 1)                             \
  46
  47 #define store_8x16b(source, dest)                                              \
  48   foreach_element(8, ((u16 *)dest)[_i] = (source).e[_i])                       \
  49
  50
  51 #define split_8x16b(dest, source)                                              \
  52   foreach_element(8,                                                           \
  53   {                                                                            \
  54     (dest).e[_i * 2] = (source).e[_i];                                         \
  55     (dest).e[(_i * 2) + 1] = (source).e[_i] >> 8;                              \
  56   })                                                                           \
  57
  58 #define merge_16x8b(dest, source)                                              \
  59   foreach_element(8,                                                           \
  60     (dest).e[_i] = (source).e[_i * 2] | ((source).e[(_i * 2) + 1] << 8))       \
  61
  62 #define vector_cast(vec_to, source)                                            \
  63   (*((volatile vec_to *)(&(source))))                                          \
  64
  65 #define vector_cast_high(vec_to, source)                                       \
  66   (*((volatile vec_to *)((u8 *)source.e + (sizeof(source.e) / 2))))            \
  67
  68
  69 #define dup_8x8b(dest, value)                                                  \
  70   foreach_element(8, (dest).e[_i] = value)                                     \
  71
  72 #define dup_16x8b(dest, value)                                                 \
  73   foreach_element(16, (dest).e[_i] = value)                                    \
  74
  75 #define dup_4x16b(dest, value)                                                 \
  76   foreach_element(4, (dest).e[_i] = value)                                     \
  77
  78 #define dup_8x16b(dest, value)                                                 \
  79   foreach_element(8, (dest).e[_i] = value)                                     \
  80
  81 #define dup_2x32b(dest, value)                                                 \
  82   foreach_element(2, (dest).e[_i] = value)                                     \
  83
  84 #define dup_4x32b(dest, value)                                                 \
  85   foreach_element(4, (dest).e[_i] = value)                                     \
  86
  87 #define shr_narrow_8x16b(dest, source, shift)                                  \
  88   foreach_element(8, (dest).e[_i] = (u16)(source).e[_i] >> (shift))            \
  89
  90 #define shr_narrow_2x64b(dest, source, shift)                                  \
  91   foreach_element(2, (dest).e[_i] = (source).e[_i] >> (shift))                 \
  92
  93 #define shr_8x8b(dest, source, shift)                                          \
  94   foreach_element(8, (dest).e[_i] = (u8)(source).e[_i] >> (shift))             \
  95
  96 #define shl_8x8b(dest, source, shift)                                          \
  97   foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
  98
  99 #define shr_8x16b(dest, source, shift)                                         \
 100   foreach_element(8, (dest).e[_i] = (u16)(source).e[_i] >> (shift))            \
 101
 102 #define shr_2x32b(dest, source, shift)                                         \
 103   foreach_element(2, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
 104
 105 #define shr_4x16b(dest, source, shift)                                         \
 106   foreach_element(4, (dest).e[_i] = (u16)(source).e[_i] >> (shift))            \
 107
 108 #define shl_4x16b(dest, source, shift)                                         \
 109   foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] << (shift))            \
 110
 111 #define shr_4x32b(dest, source, shift)                                         \
 112   foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
 113
 114 #define shr_narrow_4x32b(dest, source, shift)                                  \
 115   foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
 116
 117 #define shl_8x16b(dest, source, shift)                                         \
 118   foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
 119
 120 #define shl_4x32b(dest, source, shift)                                         \
 121   foreach_element(4, (dest).e[_i] = (source).e[_i] << (shift))                 \
 122
 123 #define shl_2x32b(dest, source, shift)                                         \
 124   foreach_element(2, (dest).e[_i] = (source).e[_i] << (shift))                 \
 125
 126 #define shl_1x64b(dest, source, shift)                                         \
 127   ((dest).e[0] = (source).e[0] << (shift))                                     \
 128
 129 #define shl_2x64b(dest, source, shift)                                         \
 130   foreach_element(2, (dest).e[_i] = (source).e[_i] << (shift))                 \
 131
 132 #define shl_variable_2x64b(dest, source_a, source_b)                           \
 133   foreach_element(2,                                                           \
 134    (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
 135
 136 #define shl_variable_8x16b(dest, source_a, source_b)                           \
 137   foreach_element(8,                                                           \
 138    (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
 139
 140 #define shl_variable_4x16b(dest, source_a, source_b)                           \
 141   foreach_element(4,                                                           \
 142    (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
 143
 144 #define shr_1x64b(dest, source, shift)                                         \
 145   ((dest).e[0] = (source).e[0] >> (shift))                                     \
 146
 147 #define shl_long_8x8b(dest, source, shift)                                     \
 148   foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
 149
 150 #define shl_long_4x16b(dest, source, shift)                                    \
 151   foreach_element(4, (dest).e[_i] = (source).e[_i] << (shift))                 \
 152
 153 #define shrq_narrow_signed_8x16b(dest, source, shift)                          \
 154   foreach_element(8,                                                           \
 155   {                                                                            \
 156     s32 result = ((s16)(source).e[_i]) >> shift;                               \
 157     if(result < 0)                                                             \
 158       result = 0;                                                              \
 159     if(result > 0xFF)                                                          \
 160       result = 0xFF;                                                           \
 161     (dest).e[_i] = result;                                                     \
 162   })                                                                           \
 163
 164 #define shl_reg_4x32b(dest, source_a, source_b)                                \
 165   foreach_element(4,                                                           \
 166   {                                                                            \
 167     s8 shift  = (source_b).e[_i];                                              \
 168     if(shift < 0)                                                              \
 169       dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
 170     else                                                                       \
 171       dest.e[_i] = (source_a).e[_i] << shift;                                  \
 172   })                                                                           \
 173
 174 #define shl_reg_2x32b(dest, source_a, source_b)                                \
 175   foreach_element(2,                                                           \
 176   {                                                                            \
 177     s8 shift  = (source_b).e[_i];                                              \
 178     if(shift < 0)                                                              \
 179       dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
 180     else                                                                       \
 181       dest.e[_i] = (source_a).e[_i] << shift;                                  \
 182   })                                                                           \
 183
 184 #define shl_reg_2x64b(dest, source_a, source_b)                                \
 185   foreach_element(2,                                                           \
 186   {                                                                            \
 187     s8 shift  = (source_b).e[_i];                                              \
 188     if(shift < 0)                                                              \
 189       dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
 190     else                                                                       \
 191       dest.e[_i] = (source_a).e[_i] << shift;                                  \
 192   })                                                                           \
 193
 194
 195 #define sri_8x8b(dest, source, shift)                                          \
 196   foreach_element(8, (dest).e[_i] = ((dest).e[_i] & ~(0xFF >> (shift))) |      \
 197    ((u8)(source).e[_i] >> (shift)))                                            \
 198
 199 #define sli_8x8b(dest, source, shift)                                          \
 200   foreach_element(8, (dest).e[_i] = ((dest).e[_i] & ~(0xFF << (shift))) |      \
 201    ((source).e[_i] << (shift)))                                                \
 202
 203
 204
 205 #define mov_narrow_8x16b(dest, source)                                         \
 206   foreach_element(8, (dest).e[_i] = (source).e[_i])                            \
 207
 208 #define mov_narrow_4x32b(dest, source)                                         \
 209   foreach_element(4, (dest).e[_i] = (source).e[_i])                            \
 210
 211 #define mov_narrow_2x64b(dest, source)                                         \
 212   foreach_element(2, (dest).e[_i] = (source).e[_i])                            \
 213
 214 #define mov_wide_8x8b(dest, source)                                            \
 215   foreach_element(8, (dest).e[_i] = (source).e[_i])                            \
 216
 217 #define mov_wide_2x32b(dest, source)                                           \
 218   foreach_element(2, (dest).e[_i] = (source).e[_i])                            \
 219
 220 #define mvn_4x16b(dest, source)                                                \
 221   foreach_element(4, (dest).e[_i] = ~((source).e[_i]))                         \
 222
 223 #define add_4x16b(dest, source_a, source_b)                                    \
 224   foreach_element(4, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 225
 226 #define add_4x32b(dest, source_a, source_b)                                    \
 227   foreach_element(4, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 228
 229 #define add_2x32b(dest, source_a, source_b)                                    \
 230   foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 231
 232 #define add_8x16b(dest, source_a, source_b)                                    \
 233   foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 234
 235 #define add_16x8b(dest, source_a, source_b)                                    \
 236   foreach_element(16, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])      \
 237
 238 #define add_8x8b(dest, source_a, source_b)                                     \
 239   foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 240
 241 #define add_1x64b(dest, source_a, source_b)                                    \
 242   (dest).e[0] = (source_a).e[0] + (source_b).e[0]                              \
 243
 244 #define add_2x64b(dest, source_a, source_b)                                    \
 245   foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 246
 247 #define add_high_narrow_2x64b(dest, source_a, source_b)                        \
 248   foreach_element(2,                                                           \
 249    ((dest).e[_i] = (source_a).e[_i] + (source_b).e[_i]) >> 32)                 \
 250
 251 #define add_high_narrow_4x32b(dest, source_a, source_b)                        \
 252   foreach_element(4,                                                           \
 253    ((dest).e[_i] = ((source_a).e[_i] + (source_b).e[_i]) >> 16))               \
 254
 255 #define sub_4x16b(dest, source_a, source_b)                                    \
 256   foreach_element(4, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
 257
 258 #define sub_4x32b(dest, source_a, source_b)                                    \
 259   foreach_element(4, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
 260
 261 #define sub_2x32b(dest, source_a, source_b)                                    \
 262   foreach_element(2, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
 263
 264 #define sub_wide_8x8b(dest, source_a, source_b)                                \
 265   foreach_element(8, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
 266
 267 #define add_wide_8x8b(dest, source_a, source_b)                                \
 268   foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 269
 270 #define add_wide_2x32b(dest, source_a, source_b)                               \
 271   foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
 272
 273 #define addq_8x8b(dest, source_a, source_b)                                    \
 274   foreach_element(8,                                                           \
 275   {                                                                            \
 276     u32 result = (source_a).e[_i] + (source_b).e[_i];                          \
 277     if(result > 0xFF)                                                          \
 278       result = 0xFF;                                                           \
 279     (dest).e[_i] = result;                                                     \
 280   })                                                                           \
 281
 282 #define subq_8x8b(dest, source_a, source_b)                                    \
 283   foreach_element(8,                                                           \
 284   {                                                                            \
 285     u32 result = (source_a).e[_i] - (source_b).e[_i];                          \
 286     if(result > 0xFF)                                                          \
 287       result = 0;                                                              \
 288     (dest).e[_i] = result;                                                     \
 289   })                                                                           \
 290
 291 #define subs_long_8x8b(dest, source_a, source_b)                               \
 292   subs_8x8b(dest, source_a, source_b)                                          \
 293
 294 #define subs_16x8b(dest, source_a, source_b)                                   \
 295   foreach_element(16,                                                          \
 296   {                                                                            \
 297     u32 result = (source_a).e[_i] - (source_b).e[_i];                          \
 298     if(result > 0xFF)                                                          \
 299       result = 0;                                                              \
 300     (dest).e[_i] = result;                                                     \
 301   })                                                                           \
 302
 303 #define subs_8x16b(dest, source_a, source_b)                                   \
 304   foreach_element(8,                                                           \
 305   {                                                                            \
 306     s32 result = (source_a).e[_i] - (source_b).e[_i];                          \
 307     if(result < 0)                                                             \
 308       result = 0;                                                              \
 309                                                                                \
 310     (dest).e[_i] = result;                                                     \
 311   })                                                                           \
 312
 313 #define sub_8x16b(dest, source_a, source_b)                                    \
 314   foreach_element(8, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
 315
 316 #define sub_16x8b(dest, source_a, source_b)                                    \
 317   foreach_element(16, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])      \
 318
 319 #define orn_8x16b(dest, source_a, source_b)                                    \
 320   foreach_element(8, (dest).e[_i] = (source_a).e[_i] | ~((source_b).e[_i]))    \
 321
 322 #define and_4x16b(dest, source_a, source_b)                                    \
 323   foreach_element(4, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
 324
 325 #define and_8x16b(dest, source_a, source_b)                                    \
 326   foreach_element(8, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
 327
 328 #define and_4x32b(dest, source_a, source_b)                                    \
 329   foreach_element(4, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
 330
 331 #define and_16x8b(dest, source_a, source_b)                                    \
 332   foreach_element(16, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])      \
 333
 334 #define and_8x8b(dest, source_a, source_b)                                     \
 335   foreach_element(8, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
 336
 337 #define and_2x32b(dest, source_a, source_b)                                    \
 338   foreach_element(2, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
 339
 340 #define bic_8x8b(dest, source_a, source_b)                                     \
 341   foreach_element(8, (dest).e[_i] = (source_a).e[_i] & ~((source_b).e[_i]))    \
 342
 343 #define bic_8x16b(dest, source_a, source_b)                                    \
 344   foreach_element(8, (dest).e[_i] = (source_a).e[_i] & ~((source_b).e[_i]))    \
 345
 346 #define bic_immediate_4x16b(dest, value)                                       \
 347   foreach_element(4, (dest).e[_i] = (dest).e[_i] & ~(value))                   \
 348
 349 #define bic_immediate_8x16b(dest, value)                                       \
 350   foreach_element(8, (dest).e[_i] = (dest).e[_i] & ~(value))                   \
 351
 352 #define or_8x16b(dest, source_a, source_b)                                     \
 353   foreach_element(8, (dest).e[_i] = (source_a).e[_i] | (source_b).e[_i])       \
 354
 355 #define or_immediate_8x16b(dest, source_a, value)                              \
 356   foreach_element(8, (dest).e[_i] = (source_a).e[_i] | (value))                \
 357
 358 #define eor_8x16b(dest, source_a, source_b)                                    \
 359   foreach_element(8, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
 360
 361 #define eor_4x32b(dest, source_a, source_b)                                    \
 362   foreach_element(4, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
 363
 364 #define eor_2x32b(dest, source_a, source_b)                                    \
 365   foreach_element(2, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
 366
 367 #define zip_8x16b(dest, source_a, source_b)                                    \
 368   foreach_element(8, (dest).e[_i] =                                            \
 369    (u8)(source_a).e[_i] | ((u8)(source_b).e[_i] << 8))                         \
 370
 371 #define zip_4x32b(dest, source_a, source_b)                                    \
 372   foreach_element(4, (dest).e[_i] =                                            \
 373    (u16)(source_a).e[_i] | ((u16)(source_b).e[_i] << 16))                      \
 374
 375 #define zip_2x64b(dest, source_a, source_b)                                    \
 376   foreach_element(2, (dest).e[_i] =                                            \
 377    (u64)(source_a).e[_i] | ((u64)(source_b).e[_i] << 32))                      \
 378
 379 #define unzip_8x8b(dest_a, dest_b, source)                                     \
 380   foreach_element(8,                                                           \
 381   {                                                                            \
 382     (dest_a).e[_i] = (source).e[_i];                                           \
 383     (dest_b).e[_i] = ((source).e[_i]) >> 8;                                    \
 384   })                                                                           \
 385
 386 #define unzip_16x8b(dest_a, dest_b, source_a, source_b)                        \
 387   foreach_element(8,                                                           \
 388   {                                                                            \
 389     (dest_a).e[_i] = (source_a).e[_i];                                         \
 390     (dest_b).e[_i] = (source_a).e[_i] >> 8;                                    \
 391   });                                                                          \
 392   foreach_element(8,                                                           \
 393   {                                                                            \
 394     (dest_a).e[_i + 8] = (source_b).e[_i];                                     \
 395     (dest_b).e[_i + 8] = (source_b).e[_i] >> 8;                                \
 396   })                                                                           \
 397
 398 #define tbl_16(dest, indexes, table)                                           \
 399   foreach_element(8,                                                           \
 400   {                                                                            \
 401     u32 index = indexes.e[_i];                                                 \
 402     if(index < 16)                                                             \
 403       (dest).e[_i] = table.e[index];                                           \
 404     else                                                                       \
 405       (dest).e[_i] = 0;                                                        \
 406   })                                                                           \
 407
 408 #define cmpeqz_8x16b(dest, source)                                             \
 409   foreach_element(8, (dest).e[_i] = ~(((source).e[_i] == 0) - 1))              \
 410
 411 #define cmpltz_8x16b(dest, source)                                             \
 412   foreach_element(8, (dest).e[_i] = ((s16)(source).e[_i] >> 15))               \
 413
 414 #define cmpltz_4x32b(dest, source)                                             \
 415   foreach_element(4, (dest).e[_i] = ((s32)(source).e[_i] >> 31))               \
 416
 417 #define cmpltz_2x32b(dest, source)                                             \
 418   foreach_element(2, (dest).e[_i] = ((s32)(source).e[_i] >> 31))               \
 419
 420 #define cmplte_4x16b(dest, source_a, source_b)                                 \
 421   foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] <= source_b.e[_i]) - 1)) \
 422
 423 #define cmplt_4x16b(dest, source_a, source_b)                                  \
 424   foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] < source_b.e[_i]) - 1))  \
 425
 426 #define cmpgt_4x16b(dest, source_a, source_b)                                  \
 427   foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] > source_b.e[_i]) - 1))  \
 428
 429 #define tst_8x16b(dest, source_a, source_b)                                    \
 430   foreach_element(8,                                                           \
 431    (dest).e[_i] = ~(((source_a.e[_i] & source_b.e[_i]) != 0) - 1))             \
 432
 433 #define andi_8x8b(dest, source_a, value)                                       \
 434   foreach_element(8, (dest).e[_i] = (source_a).e[_i] & value)                  \
 435
 436 #define average_8x16b(dest, source_a, source_b)                                \
 437   foreach_element(8,                                                           \
 438    (dest).e[_i] = ((source_a).e[_i] + (source_b).e[_i]) >> 1)                  \
 439
 440
 441 #define mul_8x8b(dest, source_a, source_b)                                     \
 442   foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 443
 444 #define mul_8x16b(dest, source_a, source_b)                                    \
 445   foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 446
 447 #define mul_2x32b(dest, source_a, source_b)                                    \
 448   foreach_element(2, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 449
 450 #define mul_4x32b(dest, source_a, source_b)                                    \
 451   foreach_element(4, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 452
 453 #define mul_long_8x8b(dest, source_a, source_b)                                \
 454   foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 455
 456 #define mul_long_4x16b(dest, source_a, source_b)                               \
 457   foreach_element(4, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
 458
 459 #define mul_long_2x32b(dest, source_a, source_b)                               \
 460   foreach_element(2,                                                           \
 461    (dest).e[_i] = (source_a).e[_i] * (s64)((source_b).e[_i]))                  \
 462
 463 #define mul_scalar_2x32b(dest, source, value)                                  \
 464   foreach_element(2, (dest).e[_i] = (source).e[_i] * value)                    \
 465
 466 #define mul_scalar_long_8x16b(dest, source, value)                             \
 467   foreach_element(8, (dest).e[_i] = (source).e[_i] * value)                    \
 468
 469 #define mul_scalar_long_2x32b(dest, source, value)                             \
 470   foreach_element(2, (dest).e[_i] = (source).e[_i] * value)                    \
 471
 472 #define mla_2x32b(dest, source_a, source_b)                                    \
 473   foreach_element(2, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
 474
 475 #define mla_4x32b(dest, source_a, source_b)                                    \
 476   foreach_element(4, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
 477
 478 #define mla_scalar_long_2x32b(dest, source, value)                             \
 479   foreach_element(2, (dest).e[_i] += (source).e[_i] * value)                   \
 480
 481 #define mla_long_8x8b(dest, source_a, source_b)                                \
 482   foreach_element(8, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
 483
 484 #define mla_long_2x32b(dest, source_a, source_b)                               \
 485   foreach_element(2, (dest).e[_i] += (source_a).e[_i] * (s64)(source_b).e[_i]) \
 486
 487 #define mla_scalar_4x32b(dest, source, value)                                  \
 488   foreach_element(4, (dest).e[_i] += (source).e[_i] * value)                   \
 489
 490 #define mla_scalar_2x32b(dest, source, value)                                  \
 491   foreach_element(2, (dest).e[_i] += (source).e[_i] * value)                   \
 492
 493 #define mls_scalar_4x32b(dest, source, value)                                  \
 494   foreach_element(4, (dest).e[_i] -= (source).e[_i] * value)                   \
 495
 496 #define mls_scalar_2x32b(dest, source, value)                                  \
 497   foreach_element(2, (dest).e[_i] -= (source).e[_i] * value)                   \
 498
 499 #define mls_scalar_long_2x32b(dest, source, value)                             \
 500   foreach_element(2, (dest).e[_i] -= (source).e[_i] * value)                   \
 501
 502 #define rev_2x32b(dest, source)                                                \
 503 {                                                                              \
 504   u32 tmp = source.e[1];                                                       \
 505   (dest).e[1] = source.e[0];                                                   \
 506   (dest).e[0] = tmp;                                                           \
 507 }                                                                              \
 508
 509 #define abs_4x32b(dest, source)                                                \
 510   foreach_element(4, (dest).e[_i] = abs(source.e[_i]))                         \
 511
 512 #define abs_2x32b(dest, source)                                                \
 513   foreach_element(2, (dest).e[_i] = abs(source.e[_i]))                         \
 514
 515 #define neg_2x32b(dest, source)                                                \
 516   foreach_element(2, (dest).e[_i] = -((source).e[_i]))                         \
 517
 518
 519 #define shrq_narrow_8x16b(dest, source, shift)                                 \
 520   foreach_element(8,                                                           \
 521   {                                                                            \
 522     u32 result = ((source).e[_i]) >> shift;                                    \
 523     if(result > 0xFF)                                                          \
 524       result = 0xFF;                                                           \
 525     (dest).e[_i] = result;                                                     \
 526   })                                                                           \
 527
 528 #define min_4x16b(dest, source_a, source_b)                                    \
 529   foreach_element(4,                                                           \
 530   {                                                                            \
 531     s32 result = (source_a).e[_i];                                             \
 532     if((source_b).e[_i] < result)                                              \
 533       result = (source_b).e[_i];                                               \
 534     (dest).e[_i] = result;                                                     \
 535   })                                                                           \
 536
 537 #define min_8x16b(dest, source_a, source_b)                                    \
 538   foreach_element(8,                                                           \
 539   {                                                                            \
 540     s32 result = (source_a).e[_i];                                             \
 541     if((source_b).e[_i] < result)                                              \
 542       result = (source_b).e[_i];                                               \
 543     (dest).e[_i] = result;                                                     \
 544   })                                                                           \
 545
 546 #define min_8x8b(dest, source_a, source_b)                                     \
 547   foreach_element(8,                                                           \
 548   {                                                                            \
 549     u32 result = (source_a).e[_i];                                             \
 550     if((source_b).e[_i] < result)                                              \
 551       result = (source_b).e[_i];                                               \
 552     (dest).e[_i] = result;                                                     \
 553   })                                                                           \
 554
 555 #define min_16x8b(dest, source_a, source_b)                                    \
 556   foreach_element(16,                                                          \
 557   {                                                                            \
 558     u32 result = (source_a).e[_i];                                             \
 559     if((source_b).e[_i] < result)                                              \
 560       result = (source_b).e[_i];                                               \
 561     (dest).e[_i] = result;                                                     \
 562   })                                                                           \
 563
 564 #define max_8x16b(dest, source_a, source_b)                                    \
 565   foreach_element(8,                                                           \
 566   {                                                                            \
 567     s32 result = (source_a).e[_i];                                             \
 568     if((source_b).e[_i] > result)                                              \
 569       result = (source_b).e[_i];                                               \
 570     (dest).e[_i] = result;                                                     \
 571   })                                                                           \
 572
 573 #define bsl_8x16b(dest_mask, source_a, source_b)                               \
 574   foreach_element(8, dest_mask.e[_i] = ((source_a).e[_i] & dest_mask.e[_i]) |  \
 575    ((source_b).e[_i] & ~(dest_mask.e[_i])))                                    \
 576
 577 #define bif_8x16b(dest, source, mask)                                          \
 578   foreach_element(8, dest.e[_i] = ((source).e[_i] & ~(mask.e[_i])) |           \
 579    ((dest).e[_i] & mask.e[_i]))                                                \
 580
 581 #define bsl_4x32b(dest_mask, source_a, source_b)                               \
 582   foreach_element(4, dest_mask.e[_i] = ((source_a).e[_i] & dest_mask.e[_i]) |  \
 583    ((source_b).e[_i] & ~(dest_mask.e[_i])))                                    \
 584
 585 #define bit_4x16b(dest, source, mask)                                          \
 586   foreach_element(4, dest.e[_i] = ((source).e[_i] & mask.e[_i]) |              \
 587    ((dest).e[_i] & ~(mask.e[_i])))                                             \
 588
 589 #endif