gcc/doc/gcc/extensions-to-the-c-language-family/target-builtins/x86-built-in-functions.rst

   1 ..
   2   Copyright 1988-2022 Free Software Foundation, Inc.
   3   This is part of the GCC manual.
   4   For copying conditions, see the copyright.rst file.
   5
   6 .. _x86-built-in-functions:
   7
   8 x86 Built-in Functions
   9 ^^^^^^^^^^^^^^^^^^^^^^
  10
  11 These built-in functions are available for the x86-32 and x86-64 family
  12 of computers, depending on the command-line switches used.
  13
  14 If you specify command-line switches such as :option:`-msse`,
  15 the compiler could use the extended instruction sets even if the built-ins
  16 are not used explicitly in the program.  For this reason, applications
  17 that perform run-time CPU detection must compile separate files for each
  18 supported architecture, using the appropriate flags.  In particular,
  19 the file containing the CPU detection code should be compiled without
  20 these options.
  21
  22 The following machine modes are available for use with MMX built-in functions
  23 (see :ref:`vector-extensions`): ``V2SI`` for a vector of two 32-bit integers,
  24 ``V4HI`` for a vector of four 16-bit integers, and ``V8QI`` for a
  25 vector of eight 8-bit integers.  Some of the built-in functions operate on
  26 MMX registers as a whole 64-bit entity, these use ``V1DI`` as their mode.
  27
  28 If 3DNow! extensions are enabled, ``V2SF`` is used as a mode for a vector
  29 of two 32-bit floating-point values.
  30
  31 If SSE extensions are enabled, ``V4SF`` is used for a vector of four 32-bit
  32 floating-point values.  Some instructions use a vector of four 32-bit
  33 integers, these use ``V4SI``.  Finally, some instructions operate on an
  34 entire vector register, interpreting it as a 128-bit integer, these use mode
  35 ``TI``.
  36
  37 The x86-32 and x86-64 family of processors use additional built-in
  38 functions for efficient use of ``TF`` (``__float128``) 128-bit
  39 floating point and ``TC`` 128-bit complex floating-point values.
  40
  41 The following floating-point built-in functions are always available.  All
  42 of them implement the function that is part of the name.
  43
  44 .. code-block:: c++
  45
  46   __float128 __builtin_fabsq (__float128)
  47   __float128 __builtin_copysignq (__float128, __float128)
  48
  49 The following built-in functions are always available.
  50
  51 .. function:: __float128 __builtin_infq (void)
  52
  53   Similar to ``__builtin_inf``, except the return type is ``__float128``.
  54
  55   .. index:: __builtin_infq
  56
  57 .. function:: __float128 __builtin_huge_valq (void)
  58
  59   Similar to ``__builtin_huge_val``, except the return type is ``__float128``.
  60
  61   .. index:: __builtin_huge_valq
  62
  63 .. function:: __float128 __builtin_nanq (void)
  64
  65   Similar to ``__builtin_nan``, except the return type is ``__float128``.
  66
  67   .. index:: __builtin_nanq
  68
  69 .. function:: __float128 __builtin_nansq (void)
  70
  71   Similar to ``__builtin_nans``, except the return type is ``__float128``.
  72
  73   .. index:: __builtin_nansq
  74
  75   The following built-in function is always available.
  76
  77 .. function:: void __builtin_ia32_pause (void)
  78
  79   Generates the ``pause`` machine instruction with a compiler memory
  80   barrier.
  81
  82 The following built-in functions are always available and can be used to
  83 check the target platform type.
  84
  85 .. function:: void __builtin_cpu_init (void)
  86
  87   This function runs the CPU detection code to check the type of CPU and the
  88   features supported.  This built-in function needs to be invoked along with the built-in functions
  89   to check CPU type and features, ``__builtin_cpu_is`` and
  90   ``__builtin_cpu_supports``, only when used in a function that is
  91   executed before any constructors are called.  The CPU detection code is
  92   automatically executed in a very high priority constructor.
  93
  94   For example, this function has to be used in ``ifunc`` resolvers that
  95   check for CPU type using the built-in functions ``__builtin_cpu_is``
  96   and ``__builtin_cpu_supports``, or in constructors on targets that
  97   don't support constructor priority.
  98
  99   .. code-block:: c++
 100
 101     static void (*resolve_memcpy (void)) (void)
 102     {
 103       // ifunc resolvers fire before constructors, explicitly call the init
 104       // function.
 105       __builtin_cpu_init ();
 106       if (__builtin_cpu_supports ("ssse3"))
 107         return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
 108       else
 109         return default_memcpy;
 110     }
 111
 112     void *memcpy (void *, const void *, size_t)
 113          __attribute__ ((ifunc ("resolve_memcpy")));
 114
 115 .. function:: int __builtin_cpu_is (const char *cpuname)
 116
 117   This function returns a positive integer if the run-time CPU
 118   is of type :samp:`{cpuname}`
 119   and returns ``0`` otherwise. The following CPU names can be detected:
 120
 121   :samp:`amd`
 122     AMD CPU.
 123
 124   :samp:`intel`
 125     Intel CPU.
 126
 127   :samp:`atom`
 128     Intel Atom CPU.
 129
 130   :samp:`slm`
 131     Intel Silvermont CPU.
 132
 133   :samp:`core2`
 134     Intel Core 2 CPU.
 135
 136   :samp:`corei7`
 137     Intel Core i7 CPU.
 138
 139   :samp:`nehalem`
 140     Intel Core i7 Nehalem CPU.
 141
 142   :samp:`westmere`
 143     Intel Core i7 Westmere CPU.
 144
 145   :samp:`sandybridge`
 146     Intel Core i7 Sandy Bridge CPU.
 147
 148   :samp:`ivybridge`
 149     Intel Core i7 Ivy Bridge CPU.
 150
 151   :samp:`haswell`
 152     Intel Core i7 Haswell CPU.
 153
 154   :samp:`broadwell`
 155     Intel Core i7 Broadwell CPU.
 156
 157   :samp:`skylake`
 158     Intel Core i7 Skylake CPU.
 159
 160   :samp:`skylake-avx512`
 161     Intel Core i7 Skylake AVX512 CPU.
 162
 163   :samp:`cannonlake`
 164     Intel Core i7 Cannon Lake CPU.
 165
 166   :samp:`icelake-client`
 167     Intel Core i7 Ice Lake Client CPU.
 168
 169   :samp:`icelake-server`
 170     Intel Core i7 Ice Lake Server CPU.
 171
 172   :samp:`cascadelake`
 173     Intel Core i7 Cascadelake CPU.
 174
 175   :samp:`tigerlake`
 176     Intel Core i7 Tigerlake CPU.
 177
 178   :samp:`cooperlake`
 179     Intel Core i7 Cooperlake CPU.
 180
 181   :samp:`sapphirerapids`
 182     Intel Core i7 sapphirerapids CPU.
 183
 184   :samp:`alderlake`
 185     Intel Core i7 Alderlake CPU.
 186
 187   :samp:`rocketlake`
 188     Intel Core i7 Rocketlake CPU.
 189
 190   :samp:`graniterapids`
 191     Intel Core i7 graniterapids CPU.
 192
 193   :samp:`bonnell`
 194     Intel Atom Bonnell CPU.
 195
 196   :samp:`silvermont`
 197     Intel Atom Silvermont CPU.
 198
 199   :samp:`goldmont`
 200     Intel Atom Goldmont CPU.
 201
 202   :samp:`goldmont-plus`
 203     Intel Atom Goldmont Plus CPU.
 204
 205   :samp:`tremont`
 206     Intel Atom Tremont CPU.
 207
 208   :samp:`sierraforest`
 209     Intel Atom Sierra Forest CPU.
 210
 211   :samp:`grandridge`
 212     Intel Atom Grand Ridge CPU.
 213
 214   :samp:`knl`
 215     Intel Knights Landing CPU.
 216
 217   :samp:`knm`
 218     Intel Knights Mill CPU.
 219
 220   :samp:`lujiazui`
 221     ZHAOXIN lujiazui CPU.
 222
 223   :samp:`amdfam10h`
 224     AMD Family 10h CPU.
 225
 226   :samp:`barcelona`
 227     AMD Family 10h Barcelona CPU.
 228
 229   :samp:`shanghai`
 230     AMD Family 10h Shanghai CPU.
 231
 232   :samp:`istanbul`
 233     AMD Family 10h Istanbul CPU.
 234
 235   :samp:`btver1`
 236     AMD Family 14h CPU.
 237
 238   :samp:`amdfam15h`
 239     AMD Family 15h CPU.
 240
 241   :samp:`bdver1`
 242     AMD Family 15h Bulldozer version 1.
 243
 244   :samp:`bdver2`
 245     AMD Family 15h Bulldozer version 2.
 246
 247   :samp:`bdver3`
 248     AMD Family 15h Bulldozer version 3.
 249
 250   :samp:`bdver4`
 251     AMD Family 15h Bulldozer version 4.
 252
 253   :samp:`btver2`
 254     AMD Family 16h CPU.
 255
 256   :samp:`amdfam17h`
 257     AMD Family 17h CPU.
 258
 259   :samp:`znver1`
 260     AMD Family 17h Zen version 1.
 261
 262   :samp:`znver2`
 263     AMD Family 17h Zen version 2.
 264
 265   :samp:`amdfam19h`
 266     AMD Family 19h CPU.
 267
 268   :samp:`znver3`
 269     AMD Family 19h Zen version 3.
 270
 271   :samp:`znver4`
 272     AMD Family 19h Zen version 4.
 273
 274   :samp:`x86-64`
 275     Baseline x86-64 microarchitecture level (as defined in x86-64 psABI).
 276
 277   :samp:`x86-64-v2`
 278     x86-64-v2 microarchitecture level.
 279
 280   :samp:`x86-64-v3`
 281     x86-64-v3 microarchitecture level.
 282
 283   :samp:`x86-64-v4`
 284     x86-64-v4 microarchitecture level.
 285
 286     Here is an example:
 287
 288   .. code-block:: c++
 289
 290     if (__builtin_cpu_is ("corei7"))
 291       {
 292          do_corei7 (); // Core i7 specific implementation.
 293       }
 294     else
 295       {
 296          do_generic (); // Generic implementation.
 297       }
 298
 299 .. function:: int __builtin_cpu_supports (const char *feature)
 300
 301   This function returns a positive integer if the run-time CPU
 302   supports :samp:`{feature}`
 303   and returns ``0`` otherwise. The following features can be detected:
 304
 305   :samp:`cmov`
 306     CMOV instruction.
 307
 308   :samp:`mmx`
 309     MMX instructions.
 310
 311   :samp:`popcnt`
 312     POPCNT instruction.
 313
 314   :samp:`sse`
 315     SSE instructions.
 316
 317   :samp:`sse2`
 318     SSE2 instructions.
 319
 320   :samp:`sse3`
 321     SSE3 instructions.
 322
 323   :samp:`ssse3`
 324     SSSE3 instructions.
 325
 326   :samp:`sse4.1`
 327     SSE4.1 instructions.
 328
 329   :samp:`sse4.2`
 330     SSE4.2 instructions.
 331
 332   :samp:`avx`
 333     AVX instructions.
 334
 335   :samp:`avx2`
 336     AVX2 instructions.
 337
 338   :samp:`sse4a`
 339     SSE4A instructions.
 340
 341   :samp:`fma4`
 342     FMA4 instructions.
 343
 344   :samp:`xop`
 345     XOP instructions.
 346
 347   :samp:`fma`
 348     FMA instructions.
 349
 350   :samp:`avx512f`
 351     AVX512F instructions.
 352
 353   :samp:`bmi`
 354     BMI instructions.
 355
 356   :samp:`bmi2`
 357     BMI2 instructions.
 358
 359   :samp:`aes`
 360     AES instructions.
 361
 362   :samp:`pclmul`
 363     PCLMUL instructions.
 364
 365   :samp:`avx512vl`
 366     AVX512VL instructions.
 367
 368   :samp:`avx512bw`
 369     AVX512BW instructions.
 370
 371   :samp:`avx512dq`
 372     AVX512DQ instructions.
 373
 374   :samp:`avx512cd`
 375     AVX512CD instructions.
 376
 377   :samp:`avx512er`
 378     AVX512ER instructions.
 379
 380   :samp:`avx512pf`
 381     AVX512PF instructions.
 382
 383   :samp:`avx512vbmi`
 384     AVX512VBMI instructions.
 385
 386   :samp:`avx512ifma`
 387     AVX512IFMA instructions.
 388
 389   :samp:`avx5124vnniw`
 390     AVX5124VNNIW instructions.
 391
 392   :samp:`avx5124fmaps`
 393     AVX5124FMAPS instructions.
 394
 395   :samp:`avx512vpopcntdq`
 396     AVX512VPOPCNTDQ instructions.
 397
 398   :samp:`avx512vbmi2`
 399     AVX512VBMI2 instructions.
 400
 401   :samp:`gfni`
 402     GFNI instructions.
 403
 404   :samp:`vpclmulqdq`
 405     VPCLMULQDQ instructions.
 406
 407   :samp:`avx512vnni`
 408     AVX512VNNI instructions.
 409
 410   :samp:`avx512bitalg`
 411     AVX512BITALG instructions.
 412
 413   Here is an example:
 414
 415   .. code-block:: c++
 416
 417     if (__builtin_cpu_supports ("popcnt"))
 418       {
 419          asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
 420       }
 421     else
 422       {
 423          count = generic_countbits (n); //generic implementation.
 424       }
 425
 426 The following built-in functions are made available by :option:`-mmmx`.
 427 All of them generate the machine instruction that is part of the name.
 428
 429 .. code-block:: c++
 430
 431   v8qi __builtin_ia32_paddb (v8qi, v8qi);
 432   v4hi __builtin_ia32_paddw (v4hi, v4hi);
 433   v2si __builtin_ia32_paddd (v2si, v2si);
 434   v8qi __builtin_ia32_psubb (v8qi, v8qi);
 435   v4hi __builtin_ia32_psubw (v4hi, v4hi);
 436   v2si __builtin_ia32_psubd (v2si, v2si);
 437   v8qi __builtin_ia32_paddsb (v8qi, v8qi);
 438   v4hi __builtin_ia32_paddsw (v4hi, v4hi);
 439   v8qi __builtin_ia32_psubsb (v8qi, v8qi);
 440   v4hi __builtin_ia32_psubsw (v4hi, v4hi);
 441   v8qi __builtin_ia32_paddusb (v8qi, v8qi);
 442   v4hi __builtin_ia32_paddusw (v4hi, v4hi);
 443   v8qi __builtin_ia32_psubusb (v8qi, v8qi);
 444   v4hi __builtin_ia32_psubusw (v4hi, v4hi);
 445   v4hi __builtin_ia32_pmullw (v4hi, v4hi);
 446   v4hi __builtin_ia32_pmulhw (v4hi, v4hi);
 447   di __builtin_ia32_pand (di, di);
 448   di __builtin_ia32_pandn (di,di);
 449   di __builtin_ia32_por (di, di);
 450   di __builtin_ia32_pxor (di, di);
 451   v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi);
 452   v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi);
 453   v2si __builtin_ia32_pcmpeqd (v2si, v2si);
 454   v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi);
 455   v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi);
 456   v2si __builtin_ia32_pcmpgtd (v2si, v2si);
 457   v8qi __builtin_ia32_punpckhbw (v8qi, v8qi);
 458   v4hi __builtin_ia32_punpckhwd (v4hi, v4hi);
 459   v2si __builtin_ia32_punpckhdq (v2si, v2si);
 460   v8qi __builtin_ia32_punpcklbw (v8qi, v8qi);
 461   v4hi __builtin_ia32_punpcklwd (v4hi, v4hi);
 462   v2si __builtin_ia32_punpckldq (v2si, v2si);
 463   v8qi __builtin_ia32_packsswb (v4hi, v4hi);
 464   v4hi __builtin_ia32_packssdw (v2si, v2si);
 465   v8qi __builtin_ia32_packuswb (v4hi, v4hi);
 466
 467   v4hi __builtin_ia32_psllw (v4hi, v4hi);
 468   v2si __builtin_ia32_pslld (v2si, v2si);
 469   v1di __builtin_ia32_psllq (v1di, v1di);
 470   v4hi __builtin_ia32_psrlw (v4hi, v4hi);
 471   v2si __builtin_ia32_psrld (v2si, v2si);
 472   v1di __builtin_ia32_psrlq (v1di, v1di);
 473   v4hi __builtin_ia32_psraw (v4hi, v4hi);
 474   v2si __builtin_ia32_psrad (v2si, v2si);
 475   v4hi __builtin_ia32_psllwi (v4hi, int);
 476   v2si __builtin_ia32_pslldi (v2si, int);
 477   v1di __builtin_ia32_psllqi (v1di, int);
 478   v4hi __builtin_ia32_psrlwi (v4hi, int);
 479   v2si __builtin_ia32_psrldi (v2si, int);
 480   v1di __builtin_ia32_psrlqi (v1di, int);
 481   v4hi __builtin_ia32_psrawi (v4hi, int);
 482   v2si __builtin_ia32_psradi (v2si, int);
 483
 484 The following built-in functions are made available either with
 485 :option:`-msse`, or with :option:`-m3dnowa`.  All of them generate
 486 the machine instruction that is part of the name.
 487
 488 .. code-block:: c++
 489
 490   v4hi __builtin_ia32_pmulhuw (v4hi, v4hi);
 491   v8qi __builtin_ia32_pavgb (v8qi, v8qi);
 492   v4hi __builtin_ia32_pavgw (v4hi, v4hi);
 493   v1di __builtin_ia32_psadbw (v8qi, v8qi);
 494   v8qi __builtin_ia32_pmaxub (v8qi, v8qi);
 495   v4hi __builtin_ia32_pmaxsw (v4hi, v4hi);
 496   v8qi __builtin_ia32_pminub (v8qi, v8qi);
 497   v4hi __builtin_ia32_pminsw (v4hi, v4hi);
 498   int __builtin_ia32_pmovmskb (v8qi);
 499   void __builtin_ia32_maskmovq (v8qi, v8qi, char *);
 500   void __builtin_ia32_movntq (di *, di);
 501   void __builtin_ia32_sfence (void);
 502
 503 The following built-in functions are available when :option:`-msse` is used.
 504 All of them generate the machine instruction that is part of the name.
 505
 506 .. code-block:: c++
 507
 508   int __builtin_ia32_comieq (v4sf, v4sf);
 509   int __builtin_ia32_comineq (v4sf, v4sf);
 510   int __builtin_ia32_comilt (v4sf, v4sf);
 511   int __builtin_ia32_comile (v4sf, v4sf);
 512   int __builtin_ia32_comigt (v4sf, v4sf);
 513   int __builtin_ia32_comige (v4sf, v4sf);
 514   int __builtin_ia32_ucomieq (v4sf, v4sf);
 515   int __builtin_ia32_ucomineq (v4sf, v4sf);
 516   int __builtin_ia32_ucomilt (v4sf, v4sf);
 517   int __builtin_ia32_ucomile (v4sf, v4sf);
 518   int __builtin_ia32_ucomigt (v4sf, v4sf);
 519   int __builtin_ia32_ucomige (v4sf, v4sf);
 520   v4sf __builtin_ia32_addps (v4sf, v4sf);
 521   v4sf __builtin_ia32_subps (v4sf, v4sf);
 522   v4sf __builtin_ia32_mulps (v4sf, v4sf);
 523   v4sf __builtin_ia32_divps (v4sf, v4sf);
 524   v4sf __builtin_ia32_addss (v4sf, v4sf);
 525   v4sf __builtin_ia32_subss (v4sf, v4sf);
 526   v4sf __builtin_ia32_mulss (v4sf, v4sf);
 527   v4sf __builtin_ia32_divss (v4sf, v4sf);
 528   v4sf __builtin_ia32_cmpeqps (v4sf, v4sf);
 529   v4sf __builtin_ia32_cmpltps (v4sf, v4sf);
 530   v4sf __builtin_ia32_cmpleps (v4sf, v4sf);
 531   v4sf __builtin_ia32_cmpgtps (v4sf, v4sf);
 532   v4sf __builtin_ia32_cmpgeps (v4sf, v4sf);
 533   v4sf __builtin_ia32_cmpunordps (v4sf, v4sf);
 534   v4sf __builtin_ia32_cmpneqps (v4sf, v4sf);
 535   v4sf __builtin_ia32_cmpnltps (v4sf, v4sf);
 536   v4sf __builtin_ia32_cmpnleps (v4sf, v4sf);
 537   v4sf __builtin_ia32_cmpngtps (v4sf, v4sf);
 538   v4sf __builtin_ia32_cmpngeps (v4sf, v4sf);
 539   v4sf __builtin_ia32_cmpordps (v4sf, v4sf);
 540   v4sf __builtin_ia32_cmpeqss (v4sf, v4sf);
 541   v4sf __builtin_ia32_cmpltss (v4sf, v4sf);
 542   v4sf __builtin_ia32_cmpless (v4sf, v4sf);
 543   v4sf __builtin_ia32_cmpunordss (v4sf, v4sf);
 544   v4sf __builtin_ia32_cmpneqss (v4sf, v4sf);
 545   v4sf __builtin_ia32_cmpnltss (v4sf, v4sf);
 546   v4sf __builtin_ia32_cmpnless (v4sf, v4sf);
 547   v4sf __builtin_ia32_cmpordss (v4sf, v4sf);
 548   v4sf __builtin_ia32_maxps (v4sf, v4sf);
 549   v4sf __builtin_ia32_maxss (v4sf, v4sf);
 550   v4sf __builtin_ia32_minps (v4sf, v4sf);
 551   v4sf __builtin_ia32_minss (v4sf, v4sf);
 552   v4sf __builtin_ia32_andps (v4sf, v4sf);
 553   v4sf __builtin_ia32_andnps (v4sf, v4sf);
 554   v4sf __builtin_ia32_orps (v4sf, v4sf);
 555   v4sf __builtin_ia32_xorps (v4sf, v4sf);
 556   v4sf __builtin_ia32_movss (v4sf, v4sf);
 557   v4sf __builtin_ia32_movhlps (v4sf, v4sf);
 558   v4sf __builtin_ia32_movlhps (v4sf, v4sf);
 559   v4sf __builtin_ia32_unpckhps (v4sf, v4sf);
 560   v4sf __builtin_ia32_unpcklps (v4sf, v4sf);
 561   v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si);
 562   v4sf __builtin_ia32_cvtsi2ss (v4sf, int);
 563   v2si __builtin_ia32_cvtps2pi (v4sf);
 564   int __builtin_ia32_cvtss2si (v4sf);
 565   v2si __builtin_ia32_cvttps2pi (v4sf);
 566   int __builtin_ia32_cvttss2si (v4sf);
 567   v4sf __builtin_ia32_rcpps (v4sf);
 568   v4sf __builtin_ia32_rsqrtps (v4sf);
 569   v4sf __builtin_ia32_sqrtps (v4sf);
 570   v4sf __builtin_ia32_rcpss (v4sf);
 571   v4sf __builtin_ia32_rsqrtss (v4sf);
 572   v4sf __builtin_ia32_sqrtss (v4sf);
 573   v4sf __builtin_ia32_shufps (v4sf, v4sf, int);
 574   void __builtin_ia32_movntps (float *, v4sf);
 575   int __builtin_ia32_movmskps (v4sf);
 576
 577 The following built-in functions are available when :option:`-msse` is used.
 578
 579 .. function:: v4sf __builtin_ia32_loadups (float *)
 580
 581   Generates the ``movups`` machine instruction as a load from memory.
 582
 583 .. function:: void __builtin_ia32_storeups (float *, v4sf)
 584
 585   Generates the ``movups`` machine instruction as a store to memory.
 586
 587 .. function:: v4sf __builtin_ia32_loadss (float *)
 588
 589   Generates the ``movss`` machine instruction as a load from memory.
 590
 591 .. function:: v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
 592
 593   Generates the ``movhps`` machine instruction as a load from memory.
 594
 595 .. function:: v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
 596
 597   Generates the ``movlps`` machine instruction as a load from memory
 598
 599 .. function:: void __builtin_ia32_storehps (v2sf *, v4sf)
 600
 601   Generates the ``movhps`` machine instruction as a store to memory.
 602
 603 .. function:: void __builtin_ia32_storelps (v2sf *, v4sf)
 604
 605   Generates the ``movlps`` machine instruction as a store to memory.
 606
 607   The following built-in functions are available when :option:`-msse2` is used.
 608
 609 All of them generate the machine instruction that is part of the name.
 610
 611 .. code-block:: c++
 612
 613   int __builtin_ia32_comisdeq (v2df, v2df);
 614   int __builtin_ia32_comisdlt (v2df, v2df);
 615   int __builtin_ia32_comisdle (v2df, v2df);
 616   int __builtin_ia32_comisdgt (v2df, v2df);
 617   int __builtin_ia32_comisdge (v2df, v2df);
 618   int __builtin_ia32_comisdneq (v2df, v2df);
 619   int __builtin_ia32_ucomisdeq (v2df, v2df);
 620   int __builtin_ia32_ucomisdlt (v2df, v2df);
 621   int __builtin_ia32_ucomisdle (v2df, v2df);
 622   int __builtin_ia32_ucomisdgt (v2df, v2df);
 623   int __builtin_ia32_ucomisdge (v2df, v2df);
 624   int __builtin_ia32_ucomisdneq (v2df, v2df);
 625   v2df __builtin_ia32_cmpeqpd (v2df, v2df);
 626   v2df __builtin_ia32_cmpltpd (v2df, v2df);
 627   v2df __builtin_ia32_cmplepd (v2df, v2df);
 628   v2df __builtin_ia32_cmpgtpd (v2df, v2df);
 629   v2df __builtin_ia32_cmpgepd (v2df, v2df);
 630   v2df __builtin_ia32_cmpunordpd (v2df, v2df);
 631   v2df __builtin_ia32_cmpneqpd (v2df, v2df);
 632   v2df __builtin_ia32_cmpnltpd (v2df, v2df);
 633   v2df __builtin_ia32_cmpnlepd (v2df, v2df);
 634   v2df __builtin_ia32_cmpngtpd (v2df, v2df);
 635   v2df __builtin_ia32_cmpngepd (v2df, v2df);
 636   v2df __builtin_ia32_cmpordpd (v2df, v2df);
 637   v2df __builtin_ia32_cmpeqsd (v2df, v2df);
 638   v2df __builtin_ia32_cmpltsd (v2df, v2df);
 639   v2df __builtin_ia32_cmplesd (v2df, v2df);
 640   v2df __builtin_ia32_cmpunordsd (v2df, v2df);
 641   v2df __builtin_ia32_cmpneqsd (v2df, v2df);
 642   v2df __builtin_ia32_cmpnltsd (v2df, v2df);
 643   v2df __builtin_ia32_cmpnlesd (v2df, v2df);
 644   v2df __builtin_ia32_cmpordsd (v2df, v2df);
 645   v2di __builtin_ia32_paddq (v2di, v2di);
 646   v2di __builtin_ia32_psubq (v2di, v2di);
 647   v2df __builtin_ia32_addpd (v2df, v2df);
 648   v2df __builtin_ia32_subpd (v2df, v2df);
 649   v2df __builtin_ia32_mulpd (v2df, v2df);
 650   v2df __builtin_ia32_divpd (v2df, v2df);
 651   v2df __builtin_ia32_addsd (v2df, v2df);
 652   v2df __builtin_ia32_subsd (v2df, v2df);
 653   v2df __builtin_ia32_mulsd (v2df, v2df);
 654   v2df __builtin_ia32_divsd (v2df, v2df);
 655   v2df __builtin_ia32_minpd (v2df, v2df);
 656   v2df __builtin_ia32_maxpd (v2df, v2df);
 657   v2df __builtin_ia32_minsd (v2df, v2df);
 658   v2df __builtin_ia32_maxsd (v2df, v2df);
 659   v2df __builtin_ia32_andpd (v2df, v2df);
 660   v2df __builtin_ia32_andnpd (v2df, v2df);
 661   v2df __builtin_ia32_orpd (v2df, v2df);
 662   v2df __builtin_ia32_xorpd (v2df, v2df);
 663   v2df __builtin_ia32_movsd (v2df, v2df);
 664   v2df __builtin_ia32_unpckhpd (v2df, v2df);
 665   v2df __builtin_ia32_unpcklpd (v2df, v2df);
 666   v16qi __builtin_ia32_paddb128 (v16qi, v16qi);
 667   v8hi __builtin_ia32_paddw128 (v8hi, v8hi);
 668   v4si __builtin_ia32_paddd128 (v4si, v4si);
 669   v2di __builtin_ia32_paddq128 (v2di, v2di);
 670   v16qi __builtin_ia32_psubb128 (v16qi, v16qi);
 671   v8hi __builtin_ia32_psubw128 (v8hi, v8hi);
 672   v4si __builtin_ia32_psubd128 (v4si, v4si);
 673   v2di __builtin_ia32_psubq128 (v2di, v2di);
 674   v8hi __builtin_ia32_pmullw128 (v8hi, v8hi);
 675   v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi);
 676   v2di __builtin_ia32_pand128 (v2di, v2di);
 677   v2di __builtin_ia32_pandn128 (v2di, v2di);
 678   v2di __builtin_ia32_por128 (v2di, v2di);
 679   v2di __builtin_ia32_pxor128 (v2di, v2di);
 680   v16qi __builtin_ia32_pavgb128 (v16qi, v16qi);
 681   v8hi __builtin_ia32_pavgw128 (v8hi, v8hi);
 682   v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi);
 683   v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi);
 684   v4si __builtin_ia32_pcmpeqd128 (v4si, v4si);
 685   v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi);
 686   v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi);
 687   v4si __builtin_ia32_pcmpgtd128 (v4si, v4si);
 688   v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi);
 689   v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi);
 690   v16qi __builtin_ia32_pminub128 (v16qi, v16qi);
 691   v8hi __builtin_ia32_pminsw128 (v8hi, v8hi);
 692   v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi);
 693   v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi);
 694   v4si __builtin_ia32_punpckhdq128 (v4si, v4si);
 695   v2di __builtin_ia32_punpckhqdq128 (v2di, v2di);
 696   v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi);
 697   v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi);
 698   v4si __builtin_ia32_punpckldq128 (v4si, v4si);
 699   v2di __builtin_ia32_punpcklqdq128 (v2di, v2di);
 700   v16qi __builtin_ia32_packsswb128 (v8hi, v8hi);
 701   v8hi __builtin_ia32_packssdw128 (v4si, v4si);
 702   v16qi __builtin_ia32_packuswb128 (v8hi, v8hi);
 703   v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi);
 704   void __builtin_ia32_maskmovdqu (v16qi, v16qi);
 705   v2df __builtin_ia32_loadupd (double *);
 706   void __builtin_ia32_storeupd (double *, v2df);
 707   v2df __builtin_ia32_loadhpd (v2df, double const *);
 708   v2df __builtin_ia32_loadlpd (v2df, double const *);
 709   int __builtin_ia32_movmskpd (v2df);
 710   int __builtin_ia32_pmovmskb128 (v16qi);
 711   void __builtin_ia32_movnti (int *, int);
 712   void __builtin_ia32_movnti64 (long long int *, long long int);
 713   void __builtin_ia32_movntpd (double *, v2df);
 714   void __builtin_ia32_movntdq (v2df *, v2df);
 715   v4si __builtin_ia32_pshufd (v4si, int);
 716   v8hi __builtin_ia32_pshuflw (v8hi, int);
 717   v8hi __builtin_ia32_pshufhw (v8hi, int);
 718   v2di __builtin_ia32_psadbw128 (v16qi, v16qi);
 719   v2df __builtin_ia32_sqrtpd (v2df);
 720   v2df __builtin_ia32_sqrtsd (v2df);
 721   v2df __builtin_ia32_shufpd (v2df, v2df, int);
 722   v2df __builtin_ia32_cvtdq2pd (v4si);
 723   v4sf __builtin_ia32_cvtdq2ps (v4si);
 724   v4si __builtin_ia32_cvtpd2dq (v2df);
 725   v2si __builtin_ia32_cvtpd2pi (v2df);
 726   v4sf __builtin_ia32_cvtpd2ps (v2df);
 727   v4si __builtin_ia32_cvttpd2dq (v2df);
 728   v2si __builtin_ia32_cvttpd2pi (v2df);
 729   v2df __builtin_ia32_cvtpi2pd (v2si);
 730   int __builtin_ia32_cvtsd2si (v2df);
 731   int __builtin_ia32_cvttsd2si (v2df);
 732   long long __builtin_ia32_cvtsd2si64 (v2df);
 733   long long __builtin_ia32_cvttsd2si64 (v2df);
 734   v4si __builtin_ia32_cvtps2dq (v4sf);
 735   v2df __builtin_ia32_cvtps2pd (v4sf);
 736   v4si __builtin_ia32_cvttps2dq (v4sf);
 737   v2df __builtin_ia32_cvtsi2sd (v2df, int);
 738   v2df __builtin_ia32_cvtsi642sd (v2df, long long);
 739   v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df);
 740   v2df __builtin_ia32_cvtss2sd (v2df, v4sf);
 741   void __builtin_ia32_clflush (const void *);
 742   void __builtin_ia32_lfence (void);
 743   void __builtin_ia32_mfence (void);
 744   v16qi __builtin_ia32_loaddqu (const char *);
 745   void __builtin_ia32_storedqu (char *, v16qi);
 746   v1di __builtin_ia32_pmuludq (v2si, v2si);
 747   v2di __builtin_ia32_pmuludq128 (v4si, v4si);
 748   v8hi __builtin_ia32_psllw128 (v8hi, v8hi);
 749   v4si __builtin_ia32_pslld128 (v4si, v4si);
 750   v2di __builtin_ia32_psllq128 (v2di, v2di);
 751   v8hi __builtin_ia32_psrlw128 (v8hi, v8hi);
 752   v4si __builtin_ia32_psrld128 (v4si, v4si);
 753   v2di __builtin_ia32_psrlq128 (v2di, v2di);
 754   v8hi __builtin_ia32_psraw128 (v8hi, v8hi);
 755   v4si __builtin_ia32_psrad128 (v4si, v4si);
 756   v2di __builtin_ia32_pslldqi128 (v2di, int);
 757   v8hi __builtin_ia32_psllwi128 (v8hi, int);
 758   v4si __builtin_ia32_pslldi128 (v4si, int);
 759   v2di __builtin_ia32_psllqi128 (v2di, int);
 760   v2di __builtin_ia32_psrldqi128 (v2di, int);
 761   v8hi __builtin_ia32_psrlwi128 (v8hi, int);
 762   v4si __builtin_ia32_psrldi128 (v4si, int);
 763   v2di __builtin_ia32_psrlqi128 (v2di, int);
 764   v8hi __builtin_ia32_psrawi128 (v8hi, int);
 765   v4si __builtin_ia32_psradi128 (v4si, int);
 766   v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi);
 767   v2di __builtin_ia32_movq128 (v2di);
 768
 769 The following built-in functions are available when :option:`-msse3` is used.
 770 All of them generate the machine instruction that is part of the name.
 771
 772 .. code-block:: c++
 773
 774   v2df __builtin_ia32_addsubpd (v2df, v2df);
 775   v4sf __builtin_ia32_addsubps (v4sf, v4sf);
 776   v2df __builtin_ia32_haddpd (v2df, v2df);
 777   v4sf __builtin_ia32_haddps (v4sf, v4sf);
 778   v2df __builtin_ia32_hsubpd (v2df, v2df);
 779   v4sf __builtin_ia32_hsubps (v4sf, v4sf);
 780   v16qi __builtin_ia32_lddqu (char const *);
 781   void __builtin_ia32_monitor (void *, unsigned int, unsigned int);
 782   v4sf __builtin_ia32_movshdup (v4sf);
 783   v4sf __builtin_ia32_movsldup (v4sf);
 784   void __builtin_ia32_mwait (unsigned int, unsigned int);
 785
 786 The following built-in functions are available when :option:`-mssse3` is used.
 787 All of them generate the machine instruction that is part of the name.
 788
 789 .. code-block:: c++
 790
 791   v2si __builtin_ia32_phaddd (v2si, v2si);
 792   v4hi __builtin_ia32_phaddw (v4hi, v4hi);
 793   v4hi __builtin_ia32_phaddsw (v4hi, v4hi);
 794   v2si __builtin_ia32_phsubd (v2si, v2si);
 795   v4hi __builtin_ia32_phsubw (v4hi, v4hi);
 796   v4hi __builtin_ia32_phsubsw (v4hi, v4hi);
 797   v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi);
 798   v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi);
 799   v8qi __builtin_ia32_pshufb (v8qi, v8qi);
 800   v8qi __builtin_ia32_psignb (v8qi, v8qi);
 801   v2si __builtin_ia32_psignd (v2si, v2si);
 802   v4hi __builtin_ia32_psignw (v4hi, v4hi);
 803   v1di __builtin_ia32_palignr (v1di, v1di, int);
 804   v8qi __builtin_ia32_pabsb (v8qi);
 805   v2si __builtin_ia32_pabsd (v2si);
 806   v4hi __builtin_ia32_pabsw (v4hi);
 807
 808 The following built-in functions are available when :option:`-mssse3` is used.
 809 All of them generate the machine instruction that is part of the name.
 810
 811 .. code-block:: c++
 812
 813   v4si __builtin_ia32_phaddd128 (v4si, v4si);
 814   v8hi __builtin_ia32_phaddw128 (v8hi, v8hi);
 815   v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi);
 816   v4si __builtin_ia32_phsubd128 (v4si, v4si);
 817   v8hi __builtin_ia32_phsubw128 (v8hi, v8hi);
 818   v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi);
 819   v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi);
 820   v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi);
 821   v16qi __builtin_ia32_pshufb128 (v16qi, v16qi);
 822   v16qi __builtin_ia32_psignb128 (v16qi, v16qi);
 823   v4si __builtin_ia32_psignd128 (v4si, v4si);
 824   v8hi __builtin_ia32_psignw128 (v8hi, v8hi);
 825   v2di __builtin_ia32_palignr128 (v2di, v2di, int);
 826   v16qi __builtin_ia32_pabsb128 (v16qi);
 827   v4si __builtin_ia32_pabsd128 (v4si);
 828   v8hi __builtin_ia32_pabsw128 (v8hi);
 829
 830 The following built-in functions are available when :option:`-msse4.1` is
 831 used.  All of them generate the machine instruction that is part of the
 832 name.
 833
 834 .. code-block:: c++
 835
 836   v2df __builtin_ia32_blendpd (v2df, v2df, const int);
 837   v4sf __builtin_ia32_blendps (v4sf, v4sf, const int);
 838   v2df __builtin_ia32_blendvpd (v2df, v2df, v2df);
 839   v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf);
 840   v2df __builtin_ia32_dppd (v2df, v2df, const int);
 841   v4sf __builtin_ia32_dpps (v4sf, v4sf, const int);
 842   v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int);
 843   v2di __builtin_ia32_movntdqa (v2di *);
 844   v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int);
 845   v8hi __builtin_ia32_packusdw128 (v4si, v4si);
 846   v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi);
 847   v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int);
 848   v2di __builtin_ia32_pcmpeqq (v2di, v2di);
 849   v8hi __builtin_ia32_phminposuw128 (v8hi);
 850   v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi);
 851   v4si __builtin_ia32_pmaxsd128 (v4si, v4si);
 852   v4si __builtin_ia32_pmaxud128 (v4si, v4si);
 853   v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi);
 854   v16qi __builtin_ia32_pminsb128 (v16qi, v16qi);
 855   v4si __builtin_ia32_pminsd128 (v4si, v4si);
 856   v4si __builtin_ia32_pminud128 (v4si, v4si);
 857   v8hi __builtin_ia32_pminuw128 (v8hi, v8hi);
 858   v4si __builtin_ia32_pmovsxbd128 (v16qi);
 859   v2di __builtin_ia32_pmovsxbq128 (v16qi);
 860   v8hi __builtin_ia32_pmovsxbw128 (v16qi);
 861   v2di __builtin_ia32_pmovsxdq128 (v4si);
 862   v4si __builtin_ia32_pmovsxwd128 (v8hi);
 863   v2di __builtin_ia32_pmovsxwq128 (v8hi);
 864   v4si __builtin_ia32_pmovzxbd128 (v16qi);
 865   v2di __builtin_ia32_pmovzxbq128 (v16qi);
 866   v8hi __builtin_ia32_pmovzxbw128 (v16qi);
 867   v2di __builtin_ia32_pmovzxdq128 (v4si);
 868   v4si __builtin_ia32_pmovzxwd128 (v8hi);
 869   v2di __builtin_ia32_pmovzxwq128 (v8hi);
 870   v2di __builtin_ia32_pmuldq128 (v4si, v4si);
 871   v4si __builtin_ia32_pmulld128 (v4si, v4si);
 872   int __builtin_ia32_ptestc128 (v2di, v2di);
 873   int __builtin_ia32_ptestnzc128 (v2di, v2di);
 874   int __builtin_ia32_ptestz128 (v2di, v2di);
 875   v2df __builtin_ia32_roundpd (v2df, const int);
 876   v4sf __builtin_ia32_roundps (v4sf, const int);
 877   v2df __builtin_ia32_roundsd (v2df, v2df, const int);
 878   v4sf __builtin_ia32_roundss (v4sf, v4sf, const int);
 879
 880 The following built-in functions are available when :option:`-msse4.1` is
 881 used.
 882
 883 .. function:: v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
 884
 885   Generates the ``insertps`` machine instruction.
 886
 887 .. function:: int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
 888
 889   Generates the ``pextrb`` machine instruction.
 890
 891 .. function:: v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
 892
 893   Generates the ``pinsrb`` machine instruction.
 894
 895 .. function:: v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
 896
 897   Generates the ``pinsrd`` machine instruction.
 898
 899 .. function:: v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
 900
 901   Generates the ``pinsrq`` machine instruction in 64bit mode.
 902
 903 The following built-in functions are changed to generate new SSE4.1
 904 instructions when :option:`-msse4.1` is used.
 905
 906 .. function:: float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
 907
 908   Generates the ``extractps`` machine instruction.
 909
 910 .. function:: int __builtin_ia32_vec_ext_v4si (v4si, const int)
 911
 912   Generates the ``pextrd`` machine instruction.
 913
 914 .. function:: long long __builtin_ia32_vec_ext_v2di (v2di, const int)
 915
 916   Generates the ``pextrq`` machine instruction in 64bit mode.
 917
 918 The following built-in functions are available when :option:`-msse4.2` is
 919 used.  All of them generate the machine instruction that is part of the
 920 name.
 921
 922 .. code-block:: c++
 923
 924   v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int);
 925   int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int);
 926   int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int);
 927   int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int);
 928   int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int);
 929   int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int);
 930   int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int);
 931   v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int);
 932   int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int);
 933   int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int);
 934   int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int);
 935   int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int);
 936   int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int);
 937   int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int);
 938   v2di __builtin_ia32_pcmpgtq (v2di, v2di);
 939
 940 The following built-in functions are available when :option:`-msse4.2` is
 941 used.
 942
 943 .. function:: unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
 944
 945   Generates the ``crc32b`` machine instruction.
 946
 947 .. function:: unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
 948
 949   Generates the ``crc32w`` machine instruction.
 950
 951 .. function:: unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
 952
 953   Generates the ``crc32l`` machine instruction.
 954
 955 .. function:: unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
 956
 957   Generates the ``crc32q`` machine instruction.
 958
 959 The following built-in functions are changed to generate new SSE4.2
 960 instructions when :option:`-msse4.2` is used.
 961
 962 .. function:: int __builtin_popcount (unsigned int)
 963
 964   Generates the ``popcntl`` machine instruction.
 965
 966 .. function:: int __builtin_popcountl (unsigned long)
 967
 968   Generates the ``popcntl`` or ``popcntq`` machine instruction,
 969   depending on the size of ``unsigned long``.
 970
 971 .. function:: int __builtin_popcountll (unsigned long long)
 972
 973   Generates the ``popcntq`` machine instruction.
 974
 975 The following built-in functions are available when :option:`-mavx` is
 976 used. All of them generate the machine instruction that is part of the
 977 name.
 978
 979 .. code-block:: c++
 980
 981   v4df __builtin_ia32_addpd256 (v4df,v4df);
 982   v8sf __builtin_ia32_addps256 (v8sf,v8sf);
 983   v4df __builtin_ia32_addsubpd256 (v4df,v4df);
 984   v8sf __builtin_ia32_addsubps256 (v8sf,v8sf);
 985   v4df __builtin_ia32_andnpd256 (v4df,v4df);
 986   v8sf __builtin_ia32_andnps256 (v8sf,v8sf);
 987   v4df __builtin_ia32_andpd256 (v4df,v4df);
 988   v8sf __builtin_ia32_andps256 (v8sf,v8sf);
 989   v4df __builtin_ia32_blendpd256 (v4df,v4df,int);
 990   v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int);
 991   v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df);
 992   v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf);
 993   v2df __builtin_ia32_cmppd (v2df,v2df,int);
 994   v4df __builtin_ia32_cmppd256 (v4df,v4df,int);
 995   v4sf __builtin_ia32_cmpps (v4sf,v4sf,int);
 996   v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int);
 997   v2df __builtin_ia32_cmpsd (v2df,v2df,int);
 998   v4sf __builtin_ia32_cmpss (v4sf,v4sf,int);
 999   v4df __builtin_ia32_cvtdq2pd256 (v4si);
1000   v8sf __builtin_ia32_cvtdq2ps256 (v8si);
1001   v4si __builtin_ia32_cvtpd2dq256 (v4df);
1002   v4sf __builtin_ia32_cvtpd2ps256 (v4df);
1003   v8si __builtin_ia32_cvtps2dq256 (v8sf);
1004   v4df __builtin_ia32_cvtps2pd256 (v4sf);
1005   v4si __builtin_ia32_cvttpd2dq256 (v4df);
1006   v8si __builtin_ia32_cvttps2dq256 (v8sf);
1007   v4df __builtin_ia32_divpd256 (v4df,v4df);
1008   v8sf __builtin_ia32_divps256 (v8sf,v8sf);
1009   v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int);
1010   v4df __builtin_ia32_haddpd256 (v4df,v4df);
1011   v8sf __builtin_ia32_haddps256 (v8sf,v8sf);
1012   v4df __builtin_ia32_hsubpd256 (v4df,v4df);
1013   v8sf __builtin_ia32_hsubps256 (v8sf,v8sf);
1014   v32qi __builtin_ia32_lddqu256 (pcchar);
1015   v32qi __builtin_ia32_loaddqu256 (pcchar);
1016   v4df __builtin_ia32_loadupd256 (pcdouble);
1017   v8sf __builtin_ia32_loadups256 (pcfloat);
1018   v2df __builtin_ia32_maskloadpd (pcv2df,v2df);
1019   v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df);
1020   v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf);
1021   v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf);
1022   void __builtin_ia32_maskstorepd (pv2df,v2df,v2df);
1023   void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df);
1024   void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf);
1025   void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf);
1026   v4df __builtin_ia32_maxpd256 (v4df,v4df);
1027   v8sf __builtin_ia32_maxps256 (v8sf,v8sf);
1028   v4df __builtin_ia32_minpd256 (v4df,v4df);
1029   v8sf __builtin_ia32_minps256 (v8sf,v8sf);
1030   v4df __builtin_ia32_movddup256 (v4df);
1031   int __builtin_ia32_movmskpd256 (v4df);
1032   int __builtin_ia32_movmskps256 (v8sf);
1033   v8sf __builtin_ia32_movshdup256 (v8sf);
1034   v8sf __builtin_ia32_movsldup256 (v8sf);
1035   v4df __builtin_ia32_mulpd256 (v4df,v4df);
1036   v8sf __builtin_ia32_mulps256 (v8sf,v8sf);
1037   v4df __builtin_ia32_orpd256 (v4df,v4df);
1038   v8sf __builtin_ia32_orps256 (v8sf,v8sf);
1039   v2df __builtin_ia32_pd_pd256 (v4df);
1040   v4df __builtin_ia32_pd256_pd (v2df);
1041   v4sf __builtin_ia32_ps_ps256 (v8sf);
1042   v8sf __builtin_ia32_ps256_ps (v4sf);
1043   int __builtin_ia32_ptestc256 (v4di,v4di,ptest);
1044   int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest);
1045   int __builtin_ia32_ptestz256 (v4di,v4di,ptest);
1046   v8sf __builtin_ia32_rcpps256 (v8sf);
1047   v4df __builtin_ia32_roundpd256 (v4df,int);
1048   v8sf __builtin_ia32_roundps256 (v8sf,int);
1049   v8sf __builtin_ia32_rsqrtps_nr256 (v8sf);
1050   v8sf __builtin_ia32_rsqrtps256 (v8sf);
1051   v4df __builtin_ia32_shufpd256 (v4df,v4df,int);
1052   v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int);
1053   v4si __builtin_ia32_si_si256 (v8si);
1054   v8si __builtin_ia32_si256_si (v4si);
1055   v4df __builtin_ia32_sqrtpd256 (v4df);
1056   v8sf __builtin_ia32_sqrtps_nr256 (v8sf);
1057   v8sf __builtin_ia32_sqrtps256 (v8sf);
1058   void __builtin_ia32_storedqu256 (pchar,v32qi);
1059   void __builtin_ia32_storeupd256 (pdouble,v4df);
1060   void __builtin_ia32_storeups256 (pfloat,v8sf);
1061   v4df __builtin_ia32_subpd256 (v4df,v4df);
1062   v8sf __builtin_ia32_subps256 (v8sf,v8sf);
1063   v4df __builtin_ia32_unpckhpd256 (v4df,v4df);
1064   v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf);
1065   v4df __builtin_ia32_unpcklpd256 (v4df,v4df);
1066   v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf);
1067   v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df);
1068   v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf);
1069   v4df __builtin_ia32_vbroadcastsd256 (pcdouble);
1070   v4sf __builtin_ia32_vbroadcastss (pcfloat);
1071   v8sf __builtin_ia32_vbroadcastss256 (pcfloat);
1072   v2df __builtin_ia32_vextractf128_pd256 (v4df,int);
1073   v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int);
1074   v4si __builtin_ia32_vextractf128_si256 (v8si,int);
1075   v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int);
1076   v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int);
1077   v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int);
1078   v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int);
1079   v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int);
1080   v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int);
1081   v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int);
1082   v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int);
1083   v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int);
1084   v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int);
1085   v2df __builtin_ia32_vpermilpd (v2df,int);
1086   v4df __builtin_ia32_vpermilpd256 (v4df,int);
1087   v4sf __builtin_ia32_vpermilps (v4sf,int);
1088   v8sf __builtin_ia32_vpermilps256 (v8sf,int);
1089   v2df __builtin_ia32_vpermilvarpd (v2df,v2di);
1090   v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di);
1091   v4sf __builtin_ia32_vpermilvarps (v4sf,v4si);
1092   v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si);
1093   int __builtin_ia32_vtestcpd (v2df,v2df,ptest);
1094   int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest);
1095   int __builtin_ia32_vtestcps (v4sf,v4sf,ptest);
1096   int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest);
1097   int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest);
1098   int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest);
1099   int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest);
1100   int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest);
1101   int __builtin_ia32_vtestzpd (v2df,v2df,ptest);
1102   int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest);
1103   int __builtin_ia32_vtestzps (v4sf,v4sf,ptest);
1104   int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest);
1105   void __builtin_ia32_vzeroall (void);
1106   void __builtin_ia32_vzeroupper (void);
1107   v4df __builtin_ia32_xorpd256 (v4df,v4df);
1108   v8sf __builtin_ia32_xorps256 (v8sf,v8sf);
1109
1110 The following built-in functions are available when :option:`-mavx2` is
1111 used. All of them generate the machine instruction that is part of the
1112 name.
1113
1114 .. code-block:: c++
1115
1116   v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int);
1117   v32qi __builtin_ia32_pabsb256 (v32qi);
1118   v16hi __builtin_ia32_pabsw256 (v16hi);
1119   v8si __builtin_ia32_pabsd256 (v8si);
1120   v16hi __builtin_ia32_packssdw256 (v8si,v8si);
1121   v32qi __builtin_ia32_packsswb256 (v16hi,v16hi);
1122   v16hi __builtin_ia32_packusdw256 (v8si,v8si);
1123   v32qi __builtin_ia32_packuswb256 (v16hi,v16hi);
1124   v32qi __builtin_ia32_paddb256 (v32qi,v32qi);
1125   v16hi __builtin_ia32_paddw256 (v16hi,v16hi);
1126   v8si __builtin_ia32_paddd256 (v8si,v8si);
1127   v4di __builtin_ia32_paddq256 (v4di,v4di);
1128   v32qi __builtin_ia32_paddsb256 (v32qi,v32qi);
1129   v16hi __builtin_ia32_paddsw256 (v16hi,v16hi);
1130   v32qi __builtin_ia32_paddusb256 (v32qi,v32qi);
1131   v16hi __builtin_ia32_paddusw256 (v16hi,v16hi);
1132   v4di __builtin_ia32_palignr256 (v4di,v4di,int);
1133   v4di __builtin_ia32_andsi256 (v4di,v4di);
1134   v4di __builtin_ia32_andnotsi256 (v4di,v4di);
1135   v32qi __builtin_ia32_pavgb256 (v32qi,v32qi);
1136   v16hi __builtin_ia32_pavgw256 (v16hi,v16hi);
1137   v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi);
1138   v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int);
1139   v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi);
1140   v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi);
1141   v8si __builtin_ia32_pcmpeqd256 (c8si,v8si);
1142   v4di __builtin_ia32_pcmpeqq256 (v4di,v4di);
1143   v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi);
1144   v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi);
1145   v8si __builtin_ia32_pcmpgtd256 (v8si,v8si);
1146   v4di __builtin_ia32_pcmpgtq256 (v4di,v4di);
1147   v16hi __builtin_ia32_phaddw256 (v16hi,v16hi);
1148   v8si __builtin_ia32_phaddd256 (v8si,v8si);
1149   v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi);
1150   v16hi __builtin_ia32_phsubw256 (v16hi,v16hi);
1151   v8si __builtin_ia32_phsubd256 (v8si,v8si);
1152   v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi);
1153   v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi);
1154   v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi);
1155   v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi);
1156   v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi);
1157   v8si __builtin_ia32_pmaxsd256 (v8si,v8si);
1158   v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi);
1159   v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi);
1160   v8si __builtin_ia32_pmaxud256 (v8si,v8si);
1161   v32qi __builtin_ia32_pminsb256 (v32qi,v32qi);
1162   v16hi __builtin_ia32_pminsw256 (v16hi,v16hi);
1163   v8si __builtin_ia32_pminsd256 (v8si,v8si);
1164   v32qi __builtin_ia32_pminub256 (v32qi,v32qi);
1165   v16hi __builtin_ia32_pminuw256 (v16hi,v16hi);
1166   v8si __builtin_ia32_pminud256 (v8si,v8si);
1167   int __builtin_ia32_pmovmskb256 (v32qi);
1168   v16hi __builtin_ia32_pmovsxbw256 (v16qi);
1169   v8si __builtin_ia32_pmovsxbd256 (v16qi);
1170   v4di __builtin_ia32_pmovsxbq256 (v16qi);
1171   v8si __builtin_ia32_pmovsxwd256 (v8hi);
1172   v4di __builtin_ia32_pmovsxwq256 (v8hi);
1173   v4di __builtin_ia32_pmovsxdq256 (v4si);
1174   v16hi __builtin_ia32_pmovzxbw256 (v16qi);
1175   v8si __builtin_ia32_pmovzxbd256 (v16qi);
1176   v4di __builtin_ia32_pmovzxbq256 (v16qi);
1177   v8si __builtin_ia32_pmovzxwd256 (v8hi);
1178   v4di __builtin_ia32_pmovzxwq256 (v8hi);
1179   v4di __builtin_ia32_pmovzxdq256 (v4si);
1180   v4di __builtin_ia32_pmuldq256 (v8si,v8si);
1181   v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi);
1182   v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi);
1183   v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi);
1184   v16hi __builtin_ia32_pmullw256 (v16hi,v16hi);
1185   v8si __builtin_ia32_pmulld256 (v8si,v8si);
1186   v4di __builtin_ia32_pmuludq256 (v8si,v8si);
1187   v4di __builtin_ia32_por256 (v4di,v4di);
1188   v16hi __builtin_ia32_psadbw256 (v32qi,v32qi);
1189   v32qi __builtin_ia32_pshufb256 (v32qi,v32qi);
1190   v8si __builtin_ia32_pshufd256 (v8si,int);
1191   v16hi __builtin_ia32_pshufhw256 (v16hi,int);
1192   v16hi __builtin_ia32_pshuflw256 (v16hi,int);
1193   v32qi __builtin_ia32_psignb256 (v32qi,v32qi);
1194   v16hi __builtin_ia32_psignw256 (v16hi,v16hi);
1195   v8si __builtin_ia32_psignd256 (v8si,v8si);
1196   v4di __builtin_ia32_pslldqi256 (v4di,int);
1197   v16hi __builtin_ia32_psllwi256 (16hi,int);
1198   v16hi __builtin_ia32_psllw256(v16hi,v8hi);
1199   v8si __builtin_ia32_pslldi256 (v8si,int);
1200   v8si __builtin_ia32_pslld256(v8si,v4si);
1201   v4di __builtin_ia32_psllqi256 (v4di,int);
1202   v4di __builtin_ia32_psllq256(v4di,v2di);
1203   v16hi __builtin_ia32_psrawi256 (v16hi,int);
1204   v16hi __builtin_ia32_psraw256 (v16hi,v8hi);
1205   v8si __builtin_ia32_psradi256 (v8si,int);
1206   v8si __builtin_ia32_psrad256 (v8si,v4si);
1207   v4di __builtin_ia32_psrldqi256 (v4di, int);
1208   v16hi __builtin_ia32_psrlwi256 (v16hi,int);
1209   v16hi __builtin_ia32_psrlw256 (v16hi,v8hi);
1210   v8si __builtin_ia32_psrldi256 (v8si,int);
1211   v8si __builtin_ia32_psrld256 (v8si,v4si);
1212   v4di __builtin_ia32_psrlqi256 (v4di,int);
1213   v4di __builtin_ia32_psrlq256(v4di,v2di);
1214   v32qi __builtin_ia32_psubb256 (v32qi,v32qi);
1215   v32hi __builtin_ia32_psubw256 (v16hi,v16hi);
1216   v8si __builtin_ia32_psubd256 (v8si,v8si);
1217   v4di __builtin_ia32_psubq256 (v4di,v4di);
1218   v32qi __builtin_ia32_psubsb256 (v32qi,v32qi);
1219   v16hi __builtin_ia32_psubsw256 (v16hi,v16hi);
1220   v32qi __builtin_ia32_psubusb256 (v32qi,v32qi);
1221   v16hi __builtin_ia32_psubusw256 (v16hi,v16hi);
1222   v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi);
1223   v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi);
1224   v8si __builtin_ia32_punpckhdq256 (v8si,v8si);
1225   v4di __builtin_ia32_punpckhqdq256 (v4di,v4di);
1226   v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi);
1227   v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi);
1228   v8si __builtin_ia32_punpckldq256 (v8si,v8si);
1229   v4di __builtin_ia32_punpcklqdq256 (v4di,v4di);
1230   v4di __builtin_ia32_pxor256 (v4di,v4di);
1231   v4di __builtin_ia32_movntdqa256 (pv4di);
1232   v4sf __builtin_ia32_vbroadcastss_ps (v4sf);
1233   v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf);
1234   v4df __builtin_ia32_vbroadcastsd_pd256 (v2df);
1235   v4di __builtin_ia32_vbroadcastsi256 (v2di);
1236   v4si __builtin_ia32_pblendd128 (v4si,v4si);
1237   v8si __builtin_ia32_pblendd256 (v8si,v8si);
1238   v32qi __builtin_ia32_pbroadcastb256 (v16qi);
1239   v16hi __builtin_ia32_pbroadcastw256 (v8hi);
1240   v8si __builtin_ia32_pbroadcastd256 (v4si);
1241   v4di __builtin_ia32_pbroadcastq256 (v2di);
1242   v16qi __builtin_ia32_pbroadcastb128 (v16qi);
1243   v8hi __builtin_ia32_pbroadcastw128 (v8hi);
1244   v4si __builtin_ia32_pbroadcastd128 (v4si);
1245   v2di __builtin_ia32_pbroadcastq128 (v2di);
1246   v8si __builtin_ia32_permvarsi256 (v8si,v8si);
1247   v4df __builtin_ia32_permdf256 (v4df,int);
1248   v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf);
1249   v4di __builtin_ia32_permdi256 (v4di,int);
1250   v4di __builtin_ia32_permti256 (v4di,v4di,int);
1251   v4di __builtin_ia32_extract128i256 (v4di,int);
1252   v4di __builtin_ia32_insert128i256 (v4di,v2di,int);
1253   v8si __builtin_ia32_maskloadd256 (pcv8si,v8si);
1254   v4di __builtin_ia32_maskloadq256 (pcv4di,v4di);
1255   v4si __builtin_ia32_maskloadd (pcv4si,v4si);
1256   v2di __builtin_ia32_maskloadq (pcv2di,v2di);
1257   void __builtin_ia32_maskstored256 (pv8si,v8si,v8si);
1258   void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di);
1259   void __builtin_ia32_maskstored (pv4si,v4si,v4si);
1260   void __builtin_ia32_maskstoreq (pv2di,v2di,v2di);
1261   v8si __builtin_ia32_psllv8si (v8si,v8si);
1262   v4si __builtin_ia32_psllv4si (v4si,v4si);
1263   v4di __builtin_ia32_psllv4di (v4di,v4di);
1264   v2di __builtin_ia32_psllv2di (v2di,v2di);
1265   v8si __builtin_ia32_psrav8si (v8si,v8si);
1266   v4si __builtin_ia32_psrav4si (v4si,v4si);
1267   v8si __builtin_ia32_psrlv8si (v8si,v8si);
1268   v4si __builtin_ia32_psrlv4si (v4si,v4si);
1269   v4di __builtin_ia32_psrlv4di (v4di,v4di);
1270   v2di __builtin_ia32_psrlv2di (v2di,v2di);
1271   v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int);
1272   v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int);
1273   v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int);
1274   v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int);
1275   v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int);
1276   v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int);
1277   v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int);
1278   v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int);
1279   v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int);
1280   v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int);
1281   v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int);
1282   v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int);
1283   v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int);
1284   v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int);
1285   v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int);
1286   v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int);
1287
1288 The following built-in functions are available when :option:`-maes` is
1289 used.  All of them generate the machine instruction that is part of the
1290 name.
1291
1292 .. code-block:: c++
1293
1294   v2di __builtin_ia32_aesenc128 (v2di, v2di);
1295   v2di __builtin_ia32_aesenclast128 (v2di, v2di);
1296   v2di __builtin_ia32_aesdec128 (v2di, v2di);
1297   v2di __builtin_ia32_aesdeclast128 (v2di, v2di);
1298   v2di __builtin_ia32_aeskeygenassist128 (v2di, const int);
1299   v2di __builtin_ia32_aesimc128 (v2di);
1300
1301 The following built-in function is available when :option:`-mpclmul` is
1302 used.
1303
1304 .. function:: v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
1305
1306   Generates the ``pclmulqdq`` machine instruction.
1307
1308 The following built-in function is available when :option:`-mfsgsbase` is
1309 used.  All of them generate the machine instruction that is part of the
1310 name.
1311
1312 .. code-block:: c++
1313
1314   unsigned int __builtin_ia32_rdfsbase32 (void);
1315   unsigned long long __builtin_ia32_rdfsbase64 (void);
1316   unsigned int __builtin_ia32_rdgsbase32 (void);
1317   unsigned long long __builtin_ia32_rdgsbase64 (void);
1318   void _writefsbase_u32 (unsigned int);
1319   void _writefsbase_u64 (unsigned long long);
1320   void _writegsbase_u32 (unsigned int);
1321   void _writegsbase_u64 (unsigned long long);
1322
1323 The following built-in function is available when :option:`-mrdrnd` is
1324 used.  All of them generate the machine instruction that is part of the
1325 name.
1326
1327 .. code-block:: c++
1328
1329   unsigned int __builtin_ia32_rdrand16_step (unsigned short *);
1330   unsigned int __builtin_ia32_rdrand32_step (unsigned int *);
1331   unsigned int __builtin_ia32_rdrand64_step (unsigned long long *);
1332
1333 The following built-in function is available when :option:`-mptwrite` is
1334 used.  All of them generate the machine instruction that is part of the
1335 name.
1336
1337 .. code-block:: c++
1338
1339   void __builtin_ia32_ptwrite32 (unsigned);
1340   void __builtin_ia32_ptwrite64 (unsigned long long);
1341
1342 The following built-in functions are available when :option:`-msse4a` is used.
1343 All of them generate the machine instruction that is part of the name.
1344
1345 .. code-block:: c++
1346
1347   void __builtin_ia32_movntsd (double *, v2df);
1348   void __builtin_ia32_movntss (float *, v4sf);
1349   v2di __builtin_ia32_extrq  (v2di, v16qi);
1350   v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int);
1351   v2di __builtin_ia32_insertq (v2di, v2di);
1352   v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int);
1353
1354 The following built-in functions are available when :option:`-mxop` is used.
1355
1356 .. code-block:: c++
1357
1358   v2df __builtin_ia32_vfrczpd (v2df);
1359   v4sf __builtin_ia32_vfrczps (v4sf);
1360   v2df __builtin_ia32_vfrczsd (v2df);
1361   v4sf __builtin_ia32_vfrczss (v4sf);
1362   v4df __builtin_ia32_vfrczpd256 (v4df);
1363   v8sf __builtin_ia32_vfrczps256 (v8sf);
1364   v2di __builtin_ia32_vpcmov (v2di, v2di, v2di);
1365   v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di);
1366   v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si);
1367   v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi);
1368   v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi);
1369   v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df);
1370   v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf);
1371   v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di);
1372   v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si);
1373   v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi);
1374   v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi);
1375   v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df);
1376   v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf);
1377   v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi);
1378   v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi);
1379   v4si __builtin_ia32_vpcomeqd (v4si, v4si);
1380   v2di __builtin_ia32_vpcomeqq (v2di, v2di);
1381   v16qi __builtin_ia32_vpcomequb (v16qi, v16qi);
1382   v4si __builtin_ia32_vpcomequd (v4si, v4si);
1383   v2di __builtin_ia32_vpcomequq (v2di, v2di);
1384   v8hi __builtin_ia32_vpcomequw (v8hi, v8hi);
1385   v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi);
1386   v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi);
1387   v4si __builtin_ia32_vpcomfalsed (v4si, v4si);
1388   v2di __builtin_ia32_vpcomfalseq (v2di, v2di);
1389   v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi);
1390   v4si __builtin_ia32_vpcomfalseud (v4si, v4si);
1391   v2di __builtin_ia32_vpcomfalseuq (v2di, v2di);
1392   v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi);
1393   v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi);
1394   v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi);
1395   v4si __builtin_ia32_vpcomged (v4si, v4si);
1396   v2di __builtin_ia32_vpcomgeq (v2di, v2di);
1397   v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi);
1398   v4si __builtin_ia32_vpcomgeud (v4si, v4si);
1399   v2di __builtin_ia32_vpcomgeuq (v2di, v2di);
1400   v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi);
1401   v8hi __builtin_ia32_vpcomgew (v8hi, v8hi);
1402   v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi);
1403   v4si __builtin_ia32_vpcomgtd (v4si, v4si);
1404   v2di __builtin_ia32_vpcomgtq (v2di, v2di);
1405   v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi);
1406   v4si __builtin_ia32_vpcomgtud (v4si, v4si);
1407   v2di __builtin_ia32_vpcomgtuq (v2di, v2di);
1408   v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi);
1409   v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi);
1410   v16qi __builtin_ia32_vpcomleb (v16qi, v16qi);
1411   v4si __builtin_ia32_vpcomled (v4si, v4si);
1412   v2di __builtin_ia32_vpcomleq (v2di, v2di);
1413   v16qi __builtin_ia32_vpcomleub (v16qi, v16qi);
1414   v4si __builtin_ia32_vpcomleud (v4si, v4si);
1415   v2di __builtin_ia32_vpcomleuq (v2di, v2di);
1416   v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi);
1417   v8hi __builtin_ia32_vpcomlew (v8hi, v8hi);
1418   v16qi __builtin_ia32_vpcomltb (v16qi, v16qi);
1419   v4si __builtin_ia32_vpcomltd (v4si, v4si);
1420   v2di __builtin_ia32_vpcomltq (v2di, v2di);
1421   v16qi __builtin_ia32_vpcomltub (v16qi, v16qi);
1422   v4si __builtin_ia32_vpcomltud (v4si, v4si);
1423   v2di __builtin_ia32_vpcomltuq (v2di, v2di);
1424   v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi);
1425   v8hi __builtin_ia32_vpcomltw (v8hi, v8hi);
1426   v16qi __builtin_ia32_vpcomneb (v16qi, v16qi);
1427   v4si __builtin_ia32_vpcomned (v4si, v4si);
1428   v2di __builtin_ia32_vpcomneq (v2di, v2di);
1429   v16qi __builtin_ia32_vpcomneub (v16qi, v16qi);
1430   v4si __builtin_ia32_vpcomneud (v4si, v4si);
1431   v2di __builtin_ia32_vpcomneuq (v2di, v2di);
1432   v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi);
1433   v8hi __builtin_ia32_vpcomnew (v8hi, v8hi);
1434   v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi);
1435   v4si __builtin_ia32_vpcomtrued (v4si, v4si);
1436   v2di __builtin_ia32_vpcomtrueq (v2di, v2di);
1437   v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi);
1438   v4si __builtin_ia32_vpcomtrueud (v4si, v4si);
1439   v2di __builtin_ia32_vpcomtrueuq (v2di, v2di);
1440   v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi);
1441   v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi);
1442   v4si __builtin_ia32_vphaddbd (v16qi);
1443   v2di __builtin_ia32_vphaddbq (v16qi);
1444   v8hi __builtin_ia32_vphaddbw (v16qi);
1445   v2di __builtin_ia32_vphadddq (v4si);
1446   v4si __builtin_ia32_vphaddubd (v16qi);
1447   v2di __builtin_ia32_vphaddubq (v16qi);
1448   v8hi __builtin_ia32_vphaddubw (v16qi);
1449   v2di __builtin_ia32_vphaddudq (v4si);
1450   v4si __builtin_ia32_vphadduwd (v8hi);
1451   v2di __builtin_ia32_vphadduwq (v8hi);
1452   v4si __builtin_ia32_vphaddwd (v8hi);
1453   v2di __builtin_ia32_vphaddwq (v8hi);
1454   v8hi __builtin_ia32_vphsubbw (v16qi);
1455   v2di __builtin_ia32_vphsubdq (v4si);
1456   v4si __builtin_ia32_vphsubwd (v8hi);
1457   v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si);
1458   v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di);
1459   v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di);
1460   v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si);
1461   v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di);
1462   v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di);
1463   v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si);
1464   v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi);
1465   v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si);
1466   v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi);
1467   v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si);
1468   v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si);
1469   v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi);
1470   v16qi __builtin_ia32_vprotb (v16qi, v16qi);
1471   v4si __builtin_ia32_vprotd (v4si, v4si);
1472   v2di __builtin_ia32_vprotq (v2di, v2di);
1473   v8hi __builtin_ia32_vprotw (v8hi, v8hi);
1474   v16qi __builtin_ia32_vpshab (v16qi, v16qi);
1475   v4si __builtin_ia32_vpshad (v4si, v4si);
1476   v2di __builtin_ia32_vpshaq (v2di, v2di);
1477   v8hi __builtin_ia32_vpshaw (v8hi, v8hi);
1478   v16qi __builtin_ia32_vpshlb (v16qi, v16qi);
1479   v4si __builtin_ia32_vpshld (v4si, v4si);
1480   v2di __builtin_ia32_vpshlq (v2di, v2di);
1481   v8hi __builtin_ia32_vpshlw (v8hi, v8hi);
1482
1483 The following built-in functions are available when :option:`-mfma4` is used.
1484 All of them generate the machine instruction that is part of the name.
1485
1486 .. code-block:: c++
1487
1488   v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df);
1489   v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf);
1490   v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df);
1491   v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf);
1492   v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df);
1493   v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf);
1494   v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df);
1495   v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf);
1496   v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df);
1497   v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf);
1498   v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df);
1499   v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf);
1500   v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df);
1501   v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf);
1502   v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df);
1503   v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf);
1504   v2df __builtin_ia32_vfmaddsubpd  (v2df, v2df, v2df);
1505   v4sf __builtin_ia32_vfmaddsubps  (v4sf, v4sf, v4sf);
1506   v2df __builtin_ia32_vfmsubaddpd  (v2df, v2df, v2df);
1507   v4sf __builtin_ia32_vfmsubaddps  (v4sf, v4sf, v4sf);
1508   v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df);
1509   v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf);
1510   v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df);
1511   v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf);
1512   v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df);
1513   v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf);
1514   v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df);
1515   v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf);
1516   v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df);
1517   v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf);
1518   v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df);
1519   v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf);
1520
1521 The following built-in functions are available when :option:`-mlwp` is used.
1522
1523 .. code-block:: c++
1524
1525   void __builtin_ia32_llwpcb16 (void *);
1526   void __builtin_ia32_llwpcb32 (void *);
1527   void __builtin_ia32_llwpcb64 (void *);
1528   void * __builtin_ia32_llwpcb16 (void);
1529   void * __builtin_ia32_llwpcb32 (void);
1530   void * __builtin_ia32_llwpcb64 (void);
1531   void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short);
1532   void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int);
1533   void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int);
1534   unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short);
1535   unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int);
1536   unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int);
1537
1538 The following built-in functions are available when :option:`-mbmi` is used.
1539 All of them generate the machine instruction that is part of the name.
1540
1541 .. code-block:: c++
1542
1543   unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
1544   unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
1545
1546 The following built-in functions are available when :option:`-mbmi2` is used.
1547 All of them generate the machine instruction that is part of the name.
1548
1549 .. code-block:: c++
1550
1551   unsigned int _bzhi_u32 (unsigned int, unsigned int);
1552   unsigned int _pdep_u32 (unsigned int, unsigned int);
1553   unsigned int _pext_u32 (unsigned int, unsigned int);
1554   unsigned long long _bzhi_u64 (unsigned long long, unsigned long long);
1555   unsigned long long _pdep_u64 (unsigned long long, unsigned long long);
1556   unsigned long long _pext_u64 (unsigned long long, unsigned long long);
1557
1558 The following built-in functions are available when :option:`-mlzcnt` is used.
1559 All of them generate the machine instruction that is part of the name.
1560
1561 .. code-block:: c++
1562
1563   unsigned short __builtin_ia32_lzcnt_u16(unsigned short);
1564   unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
1565   unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
1566
1567 The following built-in functions are available when :option:`-mfxsr` is used.
1568 All of them generate the machine instruction that is part of the name.
1569
1570 .. code-block:: c++
1571
1572   void __builtin_ia32_fxsave (void *);
1573   void __builtin_ia32_fxrstor (void *);
1574   void __builtin_ia32_fxsave64 (void *);
1575   void __builtin_ia32_fxrstor64 (void *);
1576
1577 The following built-in functions are available when :option:`-mxsave` is used.
1578 All of them generate the machine instruction that is part of the name.
1579
1580 .. code-block:: c++
1581
1582   void __builtin_ia32_xsave (void *, long long);
1583   void __builtin_ia32_xrstor (void *, long long);
1584   void __builtin_ia32_xsave64 (void *, long long);
1585   void __builtin_ia32_xrstor64 (void *, long long);
1586
1587 The following built-in functions are available when :option:`-mxsaveopt` is used.
1588 All of them generate the machine instruction that is part of the name.
1589
1590 .. code-block:: c++
1591
1592   void __builtin_ia32_xsaveopt (void *, long long);
1593   void __builtin_ia32_xsaveopt64 (void *, long long);
1594
1595 The following built-in functions are available when :option:`-mtbm` is used.
1596 Both of them generate the immediate form of the bextr machine instruction.
1597
1598 .. code-block:: c++
1599
1600   unsigned int __builtin_ia32_bextri_u32 (unsigned int,
1601                                           const unsigned int);
1602   unsigned long long __builtin_ia32_bextri_u64 (unsigned long long,
1603                                                 const unsigned long long);
1604
1605 The following built-in functions are available when :option:`-m3dnow` is used.
1606 All of them generate the machine instruction that is part of the name.
1607
1608 .. code-block:: c++
1609
1610   void __builtin_ia32_femms (void);
1611   v8qi __builtin_ia32_pavgusb (v8qi, v8qi);
1612   v2si __builtin_ia32_pf2id (v2sf);
1613   v2sf __builtin_ia32_pfacc (v2sf, v2sf);
1614   v2sf __builtin_ia32_pfadd (v2sf, v2sf);
1615   v2si __builtin_ia32_pfcmpeq (v2sf, v2sf);
1616   v2si __builtin_ia32_pfcmpge (v2sf, v2sf);
1617   v2si __builtin_ia32_pfcmpgt (v2sf, v2sf);
1618   v2sf __builtin_ia32_pfmax (v2sf, v2sf);
1619   v2sf __builtin_ia32_pfmin (v2sf, v2sf);
1620   v2sf __builtin_ia32_pfmul (v2sf, v2sf);
1621   v2sf __builtin_ia32_pfrcp (v2sf);
1622   v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf);
1623   v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf);
1624   v2sf __builtin_ia32_pfrsqrt (v2sf);
1625   v2sf __builtin_ia32_pfsub (v2sf, v2sf);
1626   v2sf __builtin_ia32_pfsubr (v2sf, v2sf);
1627   v2sf __builtin_ia32_pi2fd (v2si);
1628   v4hi __builtin_ia32_pmulhrw (v4hi, v4hi);
1629
1630 The following built-in functions are available when :option:`-m3dnowa` is used.
1631 All of them generate the machine instruction that is part of the name.
1632
1633 .. code-block:: c++
1634
1635   v2si __builtin_ia32_pf2iw (v2sf);
1636   v2sf __builtin_ia32_pfnacc (v2sf, v2sf);
1637   v2sf __builtin_ia32_pfpnacc (v2sf, v2sf);
1638   v2sf __builtin_ia32_pi2fw (v2si);
1639   v2sf __builtin_ia32_pswapdsf (v2sf);
1640   v2si __builtin_ia32_pswapdsi (v2si);
1641
1642 The following built-in functions are available when :option:`-mrtm` is used
1643 They are used for restricted transactional memory. These are the internal
1644 low level functions. Normally the functions in
1645 :ref:`x86-transactional-memory-intrinsics` should be used instead.
1646
1647 .. code-block:: c++
1648
1649   int __builtin_ia32_xbegin ();
1650   void __builtin_ia32_xend ();
1651   void __builtin_ia32_xabort (status);
1652   int __builtin_ia32_xtest ();
1653
1654 The following built-in functions are available when :option:`-mmwaitx` is used.
1655 All of them generate the machine instruction that is part of the name.
1656
1657 .. code-block:: c++
1658
1659   void __builtin_ia32_monitorx (void *, unsigned int, unsigned int);
1660   void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int);
1661
1662 The following built-in functions are available when :option:`-mclzero` is used.
1663 All of them generate the machine instruction that is part of the name.
1664
1665 .. code-block:: c++
1666
1667   void __builtin_i32_clzero (void *);
1668
1669 The following built-in functions are available when :option:`-mpku` is used.
1670 They generate reads and writes to PKRU.
1671
1672 .. code-block:: c++
1673
1674   void __builtin_ia32_wrpkru (unsigned int);
1675   unsigned int __builtin_ia32_rdpkru ();
1676
1677 The following built-in functions are available when
1678 :option:`-mshstk` option is used.  They support shadow stack
1679 machine instructions from Intel Control-flow Enforcement Technology (CET).
1680 Each built-in function generates the  machine instruction that is part
1681 of the function's name.  These are the internal low-level functions.
1682 Normally the functions in :ref:`x86-control-flow-protection-intrinsics`
1683 should be used instead.
1684
1685 .. code-block:: c++
1686
1687   unsigned int __builtin_ia32_rdsspd (void);
1688   unsigned long long __builtin_ia32_rdsspq (void);
1689   void __builtin_ia32_incsspd (unsigned int);
1690   void __builtin_ia32_incsspq (unsigned long long);
1691   void __builtin_ia32_saveprevssp(void);
1692   void __builtin_ia32_rstorssp(void *);
1693   void __builtin_ia32_wrssd(unsigned int, void *);
1694   void __builtin_ia32_wrssq(unsigned long long, void *);
1695   void __builtin_ia32_wrussd(unsigned int, void *);
1696   void __builtin_ia32_wrussq(unsigned long long, void *);
1697   void __builtin_ia32_setssbsy(void);
1698   void __builtin_ia32_clrssbsy(void *);