clang  8.0.0svn
emmintrin.h
Go to the documentation of this file.
1 /*===---- emmintrin.h - SSE2 intrinsics ------------------------------------===
2  *
3  * Permission is hereby granted, free of charge, to any person obtaining a copy
4  * of this software and associated documentation files (the "Software"), to deal
5  * in the Software without restriction, including without limitation the rights
6  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7  * copies of the Software, and to permit persons to whom the Software is
8  * furnished to do so, subject to the following conditions:
9  *
10  * The above copyright notice and this permission notice shall be included in
11  * all copies or substantial portions of the Software.
12  *
13  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
18  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
19  * THE SOFTWARE.
20  *
21  *===-----------------------------------------------------------------------===
22  */
23 
24 #ifndef __EMMINTRIN_H
25 #define __EMMINTRIN_H
26 
27 #include <xmmintrin.h>
28 
29 typedef double __m128d __attribute__((__vector_size__(16)));
30 typedef long long __m128i __attribute__((__vector_size__(16)));
31 
32 /* Type defines. */
33 typedef double __v2df __attribute__ ((__vector_size__ (16)));
34 typedef long long __v2di __attribute__ ((__vector_size__ (16)));
35 typedef short __v8hi __attribute__((__vector_size__(16)));
36 typedef char __v16qi __attribute__((__vector_size__(16)));
37 
38 /* Unsigned types */
39 typedef unsigned long long __v2du __attribute__ ((__vector_size__ (16)));
40 typedef unsigned short __v8hu __attribute__((__vector_size__(16)));
41 typedef unsigned char __v16qu __attribute__((__vector_size__(16)));
42 
43 /* We need an explicitly signed variant for char. Note that this shouldn't
44  * appear in the interface though. */
45 typedef signed char __v16qs __attribute__((__vector_size__(16)));
46 
47 /* Define the default attributes for the functions in this file. */
48 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("sse2"), __min_vector_width__(128)))
49 #define __DEFAULT_FN_ATTRS_MMX __attribute__((__always_inline__, __nodebug__, __target__("mmx,sse2"), __min_vector_width__(64)))
50 
51 /// Adds lower double-precision values in both operands and returns the
52 /// sum in the lower 64 bits of the result. The upper 64 bits of the result
53 /// are copied from the upper double-precision value of the first operand.
54 ///
55 /// \headerfile <x86intrin.h>
56 ///
57 /// This intrinsic corresponds to the <c> VADDSD / ADDSD </c> instruction.
58 ///
59 /// \param __a
60 /// A 128-bit vector of [2 x double] containing one of the source operands.
61 /// \param __b
62 /// A 128-bit vector of [2 x double] containing one of the source operands.
63 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
64 /// sum of the lower 64 bits of both operands. The upper 64 bits are copied
65 /// from the upper 64 bits of the first source operand.
66 static __inline__ __m128d __DEFAULT_FN_ATTRS
67 _mm_add_sd(__m128d __a, __m128d __b)
68 {
69  __a[0] += __b[0];
70  return __a;
71 }
72 
73 /// Adds two 128-bit vectors of [2 x double].
74 ///
75 /// \headerfile <x86intrin.h>
76 ///
77 /// This intrinsic corresponds to the <c> VADDPD / ADDPD </c> instruction.
78 ///
79 /// \param __a
80 /// A 128-bit vector of [2 x double] containing one of the source operands.
81 /// \param __b
82 /// A 128-bit vector of [2 x double] containing one of the source operands.
83 /// \returns A 128-bit vector of [2 x double] containing the sums of both
84 /// operands.
85 static __inline__ __m128d __DEFAULT_FN_ATTRS
86 _mm_add_pd(__m128d __a, __m128d __b)
87 {
88  return (__m128d)((__v2df)__a + (__v2df)__b);
89 }
90 
91 /// Subtracts the lower double-precision value of the second operand
92 /// from the lower double-precision value of the first operand and returns
93 /// the difference in the lower 64 bits of the result. The upper 64 bits of
94 /// the result are copied from the upper double-precision value of the first
95 /// operand.
96 ///
97 /// \headerfile <x86intrin.h>
98 ///
99 /// This intrinsic corresponds to the <c> VSUBSD / SUBSD </c> instruction.
100 ///
101 /// \param __a
102 /// A 128-bit vector of [2 x double] containing the minuend.
103 /// \param __b
104 /// A 128-bit vector of [2 x double] containing the subtrahend.
105 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
106 /// difference of the lower 64 bits of both operands. The upper 64 bits are
107 /// copied from the upper 64 bits of the first source operand.
108 static __inline__ __m128d __DEFAULT_FN_ATTRS
109 _mm_sub_sd(__m128d __a, __m128d __b)
110 {
111  __a[0] -= __b[0];
112  return __a;
113 }
114 
115 /// Subtracts two 128-bit vectors of [2 x double].
116 ///
117 /// \headerfile <x86intrin.h>
118 ///
119 /// This intrinsic corresponds to the <c> VSUBPD / SUBPD </c> instruction.
120 ///
121 /// \param __a
122 /// A 128-bit vector of [2 x double] containing the minuend.
123 /// \param __b
124 /// A 128-bit vector of [2 x double] containing the subtrahend.
125 /// \returns A 128-bit vector of [2 x double] containing the differences between
126 /// both operands.
127 static __inline__ __m128d __DEFAULT_FN_ATTRS
128 _mm_sub_pd(__m128d __a, __m128d __b)
129 {
130  return (__m128d)((__v2df)__a - (__v2df)__b);
131 }
132 
133 /// Multiplies lower double-precision values in both operands and returns
134 /// the product in the lower 64 bits of the result. The upper 64 bits of the
135 /// result are copied from the upper double-precision value of the first
136 /// operand.
137 ///
138 /// \headerfile <x86intrin.h>
139 ///
140 /// This intrinsic corresponds to the <c> VMULSD / MULSD </c> instruction.
141 ///
142 /// \param __a
143 /// A 128-bit vector of [2 x double] containing one of the source operands.
144 /// \param __b
145 /// A 128-bit vector of [2 x double] containing one of the source operands.
146 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
147 /// product of the lower 64 bits of both operands. The upper 64 bits are
148 /// copied from the upper 64 bits of the first source operand.
149 static __inline__ __m128d __DEFAULT_FN_ATTRS
150 _mm_mul_sd(__m128d __a, __m128d __b)
151 {
152  __a[0] *= __b[0];
153  return __a;
154 }
155 
156 /// Multiplies two 128-bit vectors of [2 x double].
157 ///
158 /// \headerfile <x86intrin.h>
159 ///
160 /// This intrinsic corresponds to the <c> VMULPD / MULPD </c> instruction.
161 ///
162 /// \param __a
163 /// A 128-bit vector of [2 x double] containing one of the operands.
164 /// \param __b
165 /// A 128-bit vector of [2 x double] containing one of the operands.
166 /// \returns A 128-bit vector of [2 x double] containing the products of both
167 /// operands.
168 static __inline__ __m128d __DEFAULT_FN_ATTRS
169 _mm_mul_pd(__m128d __a, __m128d __b)
170 {
171  return (__m128d)((__v2df)__a * (__v2df)__b);
172 }
173 
174 /// Divides the lower double-precision value of the first operand by the
175 /// lower double-precision value of the second operand and returns the
176 /// quotient in the lower 64 bits of the result. The upper 64 bits of the
177 /// result are copied from the upper double-precision value of the first
178 /// operand.
179 ///
180 /// \headerfile <x86intrin.h>
181 ///
182 /// This intrinsic corresponds to the <c> VDIVSD / DIVSD </c> instruction.
183 ///
184 /// \param __a
185 /// A 128-bit vector of [2 x double] containing the dividend.
186 /// \param __b
187 /// A 128-bit vector of [2 x double] containing divisor.
188 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
189 /// quotient of the lower 64 bits of both operands. The upper 64 bits are
190 /// copied from the upper 64 bits of the first source operand.
191 static __inline__ __m128d __DEFAULT_FN_ATTRS
192 _mm_div_sd(__m128d __a, __m128d __b)
193 {
194  __a[0] /= __b[0];
195  return __a;
196 }
197 
198 /// Performs an element-by-element division of two 128-bit vectors of
199 /// [2 x double].
200 ///
201 /// \headerfile <x86intrin.h>
202 ///
203 /// This intrinsic corresponds to the <c> VDIVPD / DIVPD </c> instruction.
204 ///
205 /// \param __a
206 /// A 128-bit vector of [2 x double] containing the dividend.
207 /// \param __b
208 /// A 128-bit vector of [2 x double] containing the divisor.
209 /// \returns A 128-bit vector of [2 x double] containing the quotients of both
210 /// operands.
211 static __inline__ __m128d __DEFAULT_FN_ATTRS
212 _mm_div_pd(__m128d __a, __m128d __b)
213 {
214  return (__m128d)((__v2df)__a / (__v2df)__b);
215 }
216 
217 /// Calculates the square root of the lower double-precision value of
218 /// the second operand and returns it in the lower 64 bits of the result.
219 /// The upper 64 bits of the result are copied from the upper
220 /// double-precision value of the first operand.
221 ///
222 /// \headerfile <x86intrin.h>
223 ///
224 /// This intrinsic corresponds to the <c> VSQRTSD / SQRTSD </c> instruction.
225 ///
226 /// \param __a
227 /// A 128-bit vector of [2 x double] containing one of the operands. The
228 /// upper 64 bits of this operand are copied to the upper 64 bits of the
229 /// result.
230 /// \param __b
231 /// A 128-bit vector of [2 x double] containing one of the operands. The
232 /// square root is calculated using the lower 64 bits of this operand.
233 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
234 /// square root of the lower 64 bits of operand \a __b, and whose upper 64
235 /// bits are copied from the upper 64 bits of operand \a __a.
236 static __inline__ __m128d __DEFAULT_FN_ATTRS
237 _mm_sqrt_sd(__m128d __a, __m128d __b)
238 {
239  __m128d __c = __builtin_ia32_sqrtsd((__v2df)__b);
240  return __extension__ (__m128d) { __c[0], __a[1] };
241 }
242 
243 /// Calculates the square root of the each of two values stored in a
244 /// 128-bit vector of [2 x double].
245 ///
246 /// \headerfile <x86intrin.h>
247 ///
248 /// This intrinsic corresponds to the <c> VSQRTPD / SQRTPD </c> instruction.
249 ///
250 /// \param __a
251 /// A 128-bit vector of [2 x double].
252 /// \returns A 128-bit vector of [2 x double] containing the square roots of the
253 /// values in the operand.
254 static __inline__ __m128d __DEFAULT_FN_ATTRS
255 _mm_sqrt_pd(__m128d __a)
256 {
257  return __builtin_ia32_sqrtpd((__v2df)__a);
258 }
259 
260 /// Compares lower 64-bit double-precision values of both operands, and
261 /// returns the lesser of the pair of values in the lower 64-bits of the
262 /// result. The upper 64 bits of the result are copied from the upper
263 /// double-precision value of the first operand.
264 ///
265 /// \headerfile <x86intrin.h>
266 ///
267 /// This intrinsic corresponds to the <c> VMINSD / MINSD </c> instruction.
268 ///
269 /// \param __a
270 /// A 128-bit vector of [2 x double] containing one of the operands. The
271 /// lower 64 bits of this operand are used in the comparison.
272 /// \param __b
273 /// A 128-bit vector of [2 x double] containing one of the operands. The
274 /// lower 64 bits of this operand are used in the comparison.
275 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
276 /// minimum value between both operands. The upper 64 bits are copied from
277 /// the upper 64 bits of the first source operand.
278 static __inline__ __m128d __DEFAULT_FN_ATTRS
279 _mm_min_sd(__m128d __a, __m128d __b)
280 {
281  return __builtin_ia32_minsd((__v2df)__a, (__v2df)__b);
282 }
283 
284 /// Performs element-by-element comparison of the two 128-bit vectors of
285 /// [2 x double] and returns the vector containing the lesser of each pair of
286 /// values.
287 ///
288 /// \headerfile <x86intrin.h>
289 ///
290 /// This intrinsic corresponds to the <c> VMINPD / MINPD </c> instruction.
291 ///
292 /// \param __a
293 /// A 128-bit vector of [2 x double] containing one of the operands.
294 /// \param __b
295 /// A 128-bit vector of [2 x double] containing one of the operands.
296 /// \returns A 128-bit vector of [2 x double] containing the minimum values
297 /// between both operands.
298 static __inline__ __m128d __DEFAULT_FN_ATTRS
299 _mm_min_pd(__m128d __a, __m128d __b)
300 {
301  return __builtin_ia32_minpd((__v2df)__a, (__v2df)__b);
302 }
303 
304 /// Compares lower 64-bit double-precision values of both operands, and
305 /// returns the greater of the pair of values in the lower 64-bits of the
306 /// result. The upper 64 bits of the result are copied from the upper
307 /// double-precision value of the first operand.
308 ///
309 /// \headerfile <x86intrin.h>
310 ///
311 /// This intrinsic corresponds to the <c> VMAXSD / MAXSD </c> instruction.
312 ///
313 /// \param __a
314 /// A 128-bit vector of [2 x double] containing one of the operands. The
315 /// lower 64 bits of this operand are used in the comparison.
316 /// \param __b
317 /// A 128-bit vector of [2 x double] containing one of the operands. The
318 /// lower 64 bits of this operand are used in the comparison.
319 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
320 /// maximum value between both operands. The upper 64 bits are copied from
321 /// the upper 64 bits of the first source operand.
322 static __inline__ __m128d __DEFAULT_FN_ATTRS
323 _mm_max_sd(__m128d __a, __m128d __b)
324 {
325  return __builtin_ia32_maxsd((__v2df)__a, (__v2df)__b);
326 }
327 
328 /// Performs element-by-element comparison of the two 128-bit vectors of
329 /// [2 x double] and returns the vector containing the greater of each pair
330 /// of values.
331 ///
332 /// \headerfile <x86intrin.h>
333 ///
334 /// This intrinsic corresponds to the <c> VMAXPD / MAXPD </c> instruction.
335 ///
336 /// \param __a
337 /// A 128-bit vector of [2 x double] containing one of the operands.
338 /// \param __b
339 /// A 128-bit vector of [2 x double] containing one of the operands.
340 /// \returns A 128-bit vector of [2 x double] containing the maximum values
341 /// between both operands.
342 static __inline__ __m128d __DEFAULT_FN_ATTRS
343 _mm_max_pd(__m128d __a, __m128d __b)
344 {
345  return __builtin_ia32_maxpd((__v2df)__a, (__v2df)__b);
346 }
347 
348 /// Performs a bitwise AND of two 128-bit vectors of [2 x double].
349 ///
350 /// \headerfile <x86intrin.h>
351 ///
352 /// This intrinsic corresponds to the <c> VPAND / PAND </c> instruction.
353 ///
354 /// \param __a
355 /// A 128-bit vector of [2 x double] containing one of the source operands.
356 /// \param __b
357 /// A 128-bit vector of [2 x double] containing one of the source operands.
358 /// \returns A 128-bit vector of [2 x double] containing the bitwise AND of the
359 /// values between both operands.
360 static __inline__ __m128d __DEFAULT_FN_ATTRS
361 _mm_and_pd(__m128d __a, __m128d __b)
362 {
363  return (__m128d)((__v2du)__a & (__v2du)__b);
364 }
365 
366 /// Performs a bitwise AND of two 128-bit vectors of [2 x double], using
367 /// the one's complement of the values contained in the first source operand.
368 ///
369 /// \headerfile <x86intrin.h>
370 ///
371 /// This intrinsic corresponds to the <c> VPANDN / PANDN </c> instruction.
372 ///
373 /// \param __a
374 /// A 128-bit vector of [2 x double] containing the left source operand. The
375 /// one's complement of this value is used in the bitwise AND.
376 /// \param __b
377 /// A 128-bit vector of [2 x double] containing the right source operand.
378 /// \returns A 128-bit vector of [2 x double] containing the bitwise AND of the
379 /// values in the second operand and the one's complement of the first
380 /// operand.
381 static __inline__ __m128d __DEFAULT_FN_ATTRS
382 _mm_andnot_pd(__m128d __a, __m128d __b)
383 {
384  return (__m128d)(~(__v2du)__a & (__v2du)__b);
385 }
386 
387 /// Performs a bitwise OR of two 128-bit vectors of [2 x double].
388 ///
389 /// \headerfile <x86intrin.h>
390 ///
391 /// This intrinsic corresponds to the <c> VPOR / POR </c> instruction.
392 ///
393 /// \param __a
394 /// A 128-bit vector of [2 x double] containing one of the source operands.
395 /// \param __b
396 /// A 128-bit vector of [2 x double] containing one of the source operands.
397 /// \returns A 128-bit vector of [2 x double] containing the bitwise OR of the
398 /// values between both operands.
399 static __inline__ __m128d __DEFAULT_FN_ATTRS
400 _mm_or_pd(__m128d __a, __m128d __b)
401 {
402  return (__m128d)((__v2du)__a | (__v2du)__b);
403 }
404 
405 /// Performs a bitwise XOR of two 128-bit vectors of [2 x double].
406 ///
407 /// \headerfile <x86intrin.h>
408 ///
409 /// This intrinsic corresponds to the <c> VPXOR / PXOR </c> instruction.
410 ///
411 /// \param __a
412 /// A 128-bit vector of [2 x double] containing one of the source operands.
413 /// \param __b
414 /// A 128-bit vector of [2 x double] containing one of the source operands.
415 /// \returns A 128-bit vector of [2 x double] containing the bitwise XOR of the
416 /// values between both operands.
417 static __inline__ __m128d __DEFAULT_FN_ATTRS
418 _mm_xor_pd(__m128d __a, __m128d __b)
419 {
420  return (__m128d)((__v2du)__a ^ (__v2du)__b);
421 }
422 
423 /// Compares each of the corresponding double-precision values of the
424 /// 128-bit vectors of [2 x double] for equality. Each comparison yields 0x0
425 /// for false, 0xFFFFFFFFFFFFFFFF for true.
426 ///
427 /// \headerfile <x86intrin.h>
428 ///
429 /// This intrinsic corresponds to the <c> VCMPEQPD / CMPEQPD </c> instruction.
430 ///
431 /// \param __a
432 /// A 128-bit vector of [2 x double].
433 /// \param __b
434 /// A 128-bit vector of [2 x double].
435 /// \returns A 128-bit vector containing the comparison results.
436 static __inline__ __m128d __DEFAULT_FN_ATTRS
437 _mm_cmpeq_pd(__m128d __a, __m128d __b)
438 {
439  return (__m128d)__builtin_ia32_cmpeqpd((__v2df)__a, (__v2df)__b);
440 }
441 
442 /// Compares each of the corresponding double-precision values of the
443 /// 128-bit vectors of [2 x double] to determine if the values in the first
444 /// operand are less than those in the second operand. Each comparison
445 /// yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
446 ///
447 /// \headerfile <x86intrin.h>
448 ///
449 /// This intrinsic corresponds to the <c> VCMPLTPD / CMPLTPD </c> instruction.
450 ///
451 /// \param __a
452 /// A 128-bit vector of [2 x double].
453 /// \param __b
454 /// A 128-bit vector of [2 x double].
455 /// \returns A 128-bit vector containing the comparison results.
456 static __inline__ __m128d __DEFAULT_FN_ATTRS
457 _mm_cmplt_pd(__m128d __a, __m128d __b)
458 {
459  return (__m128d)__builtin_ia32_cmpltpd((__v2df)__a, (__v2df)__b);
460 }
461 
462 /// Compares each of the corresponding double-precision values of the
463 /// 128-bit vectors of [2 x double] to determine if the values in the first
464 /// operand are less than or equal to those in the second operand.
465 ///
466 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
467 ///
468 /// \headerfile <x86intrin.h>
469 ///
470 /// This intrinsic corresponds to the <c> VCMPLEPD / CMPLEPD </c> instruction.
471 ///
472 /// \param __a
473 /// A 128-bit vector of [2 x double].
474 /// \param __b
475 /// A 128-bit vector of [2 x double].
476 /// \returns A 128-bit vector containing the comparison results.
477 static __inline__ __m128d __DEFAULT_FN_ATTRS
478 _mm_cmple_pd(__m128d __a, __m128d __b)
479 {
480  return (__m128d)__builtin_ia32_cmplepd((__v2df)__a, (__v2df)__b);
481 }
482 
483 /// Compares each of the corresponding double-precision values of the
484 /// 128-bit vectors of [2 x double] to determine if the values in the first
485 /// operand are greater than those in the second operand.
486 ///
487 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
488 ///
489 /// \headerfile <x86intrin.h>
490 ///
491 /// This intrinsic corresponds to the <c> VCMPLTPD / CMPLTPD </c> instruction.
492 ///
493 /// \param __a
494 /// A 128-bit vector of [2 x double].
495 /// \param __b
496 /// A 128-bit vector of [2 x double].
497 /// \returns A 128-bit vector containing the comparison results.
498 static __inline__ __m128d __DEFAULT_FN_ATTRS
499 _mm_cmpgt_pd(__m128d __a, __m128d __b)
500 {
501  return (__m128d)__builtin_ia32_cmpltpd((__v2df)__b, (__v2df)__a);
502 }
503 
504 /// Compares each of the corresponding double-precision values of the
505 /// 128-bit vectors of [2 x double] to determine if the values in the first
506 /// operand are greater than or equal to those in the second operand.
507 ///
508 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
509 ///
510 /// \headerfile <x86intrin.h>
511 ///
512 /// This intrinsic corresponds to the <c> VCMPLEPD / CMPLEPD </c> instruction.
513 ///
514 /// \param __a
515 /// A 128-bit vector of [2 x double].
516 /// \param __b
517 /// A 128-bit vector of [2 x double].
518 /// \returns A 128-bit vector containing the comparison results.
519 static __inline__ __m128d __DEFAULT_FN_ATTRS
520 _mm_cmpge_pd(__m128d __a, __m128d __b)
521 {
522  return (__m128d)__builtin_ia32_cmplepd((__v2df)__b, (__v2df)__a);
523 }
524 
525 /// Compares each of the corresponding double-precision values of the
526 /// 128-bit vectors of [2 x double] to determine if the values in the first
527 /// operand are ordered with respect to those in the second operand.
528 ///
529 /// A pair of double-precision values are "ordered" with respect to each
530 /// other if neither value is a NaN. Each comparison yields 0x0 for false,
531 /// 0xFFFFFFFFFFFFFFFF for true.
532 ///
533 /// \headerfile <x86intrin.h>
534 ///
535 /// This intrinsic corresponds to the <c> VCMPORDPD / CMPORDPD </c> instruction.
536 ///
537 /// \param __a
538 /// A 128-bit vector of [2 x double].
539 /// \param __b
540 /// A 128-bit vector of [2 x double].
541 /// \returns A 128-bit vector containing the comparison results.
542 static __inline__ __m128d __DEFAULT_FN_ATTRS
543 _mm_cmpord_pd(__m128d __a, __m128d __b)
544 {
545  return (__m128d)__builtin_ia32_cmpordpd((__v2df)__a, (__v2df)__b);
546 }
547 
548 /// Compares each of the corresponding double-precision values of the
549 /// 128-bit vectors of [2 x double] to determine if the values in the first
550 /// operand are unordered with respect to those in the second operand.
551 ///
552 /// A pair of double-precision values are "unordered" with respect to each
553 /// other if one or both values are NaN. Each comparison yields 0x0 for
554 /// false, 0xFFFFFFFFFFFFFFFF for true.
555 ///
556 /// \headerfile <x86intrin.h>
557 ///
558 /// This intrinsic corresponds to the <c> VCMPUNORDPD / CMPUNORDPD </c>
559 /// instruction.
560 ///
561 /// \param __a
562 /// A 128-bit vector of [2 x double].
563 /// \param __b
564 /// A 128-bit vector of [2 x double].
565 /// \returns A 128-bit vector containing the comparison results.
566 static __inline__ __m128d __DEFAULT_FN_ATTRS
567 _mm_cmpunord_pd(__m128d __a, __m128d __b)
568 {
569  return (__m128d)__builtin_ia32_cmpunordpd((__v2df)__a, (__v2df)__b);
570 }
571 
572 /// Compares each of the corresponding double-precision values of the
573 /// 128-bit vectors of [2 x double] to determine if the values in the first
574 /// operand are unequal to those in the second operand.
575 ///
576 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
577 ///
578 /// \headerfile <x86intrin.h>
579 ///
580 /// This intrinsic corresponds to the <c> VCMPNEQPD / CMPNEQPD </c> instruction.
581 ///
582 /// \param __a
583 /// A 128-bit vector of [2 x double].
584 /// \param __b
585 /// A 128-bit vector of [2 x double].
586 /// \returns A 128-bit vector containing the comparison results.
587 static __inline__ __m128d __DEFAULT_FN_ATTRS
588 _mm_cmpneq_pd(__m128d __a, __m128d __b)
589 {
590  return (__m128d)__builtin_ia32_cmpneqpd((__v2df)__a, (__v2df)__b);
591 }
592 
593 /// Compares each of the corresponding double-precision values of the
594 /// 128-bit vectors of [2 x double] to determine if the values in the first
595 /// operand are not less than those in the second operand.
596 ///
597 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
598 ///
599 /// \headerfile <x86intrin.h>
600 ///
601 /// This intrinsic corresponds to the <c> VCMPNLTPD / CMPNLTPD </c> instruction.
602 ///
603 /// \param __a
604 /// A 128-bit vector of [2 x double].
605 /// \param __b
606 /// A 128-bit vector of [2 x double].
607 /// \returns A 128-bit vector containing the comparison results.
608 static __inline__ __m128d __DEFAULT_FN_ATTRS
609 _mm_cmpnlt_pd(__m128d __a, __m128d __b)
610 {
611  return (__m128d)__builtin_ia32_cmpnltpd((__v2df)__a, (__v2df)__b);
612 }
613 
614 /// Compares each of the corresponding double-precision values of the
615 /// 128-bit vectors of [2 x double] to determine if the values in the first
616 /// operand are not less than or equal to those in the second operand.
617 ///
618 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
619 ///
620 /// \headerfile <x86intrin.h>
621 ///
622 /// This intrinsic corresponds to the <c> VCMPNLEPD / CMPNLEPD </c> instruction.
623 ///
624 /// \param __a
625 /// A 128-bit vector of [2 x double].
626 /// \param __b
627 /// A 128-bit vector of [2 x double].
628 /// \returns A 128-bit vector containing the comparison results.
629 static __inline__ __m128d __DEFAULT_FN_ATTRS
630 _mm_cmpnle_pd(__m128d __a, __m128d __b)
631 {
632  return (__m128d)__builtin_ia32_cmpnlepd((__v2df)__a, (__v2df)__b);
633 }
634 
635 /// Compares each of the corresponding double-precision values of the
636 /// 128-bit vectors of [2 x double] to determine if the values in the first
637 /// operand are not greater than those in the second operand.
638 ///
639 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
640 ///
641 /// \headerfile <x86intrin.h>
642 ///
643 /// This intrinsic corresponds to the <c> VCMPNLTPD / CMPNLTPD </c> instruction.
644 ///
645 /// \param __a
646 /// A 128-bit vector of [2 x double].
647 /// \param __b
648 /// A 128-bit vector of [2 x double].
649 /// \returns A 128-bit vector containing the comparison results.
650 static __inline__ __m128d __DEFAULT_FN_ATTRS
651 _mm_cmpngt_pd(__m128d __a, __m128d __b)
652 {
653  return (__m128d)__builtin_ia32_cmpnltpd((__v2df)__b, (__v2df)__a);
654 }
655 
656 /// Compares each of the corresponding double-precision values of the
657 /// 128-bit vectors of [2 x double] to determine if the values in the first
658 /// operand are not greater than or equal to those in the second operand.
659 ///
660 /// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
661 ///
662 /// \headerfile <x86intrin.h>
663 ///
664 /// This intrinsic corresponds to the <c> VCMPNLEPD / CMPNLEPD </c> instruction.
665 ///
666 /// \param __a
667 /// A 128-bit vector of [2 x double].
668 /// \param __b
669 /// A 128-bit vector of [2 x double].
670 /// \returns A 128-bit vector containing the comparison results.
671 static __inline__ __m128d __DEFAULT_FN_ATTRS
672 _mm_cmpnge_pd(__m128d __a, __m128d __b)
673 {
674  return (__m128d)__builtin_ia32_cmpnlepd((__v2df)__b, (__v2df)__a);
675 }
676 
677 /// Compares the lower double-precision floating-point values in each of
678 /// the two 128-bit floating-point vectors of [2 x double] for equality.
679 ///
680 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
681 ///
682 /// \headerfile <x86intrin.h>
683 ///
684 /// This intrinsic corresponds to the <c> VCMPEQSD / CMPEQSD </c> instruction.
685 ///
686 /// \param __a
687 /// A 128-bit vector of [2 x double]. The lower double-precision value is
688 /// compared to the lower double-precision value of \a __b.
689 /// \param __b
690 /// A 128-bit vector of [2 x double]. The lower double-precision value is
691 /// compared to the lower double-precision value of \a __a.
692 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
693 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
694 static __inline__ __m128d __DEFAULT_FN_ATTRS
695 _mm_cmpeq_sd(__m128d __a, __m128d __b)
696 {
697  return (__m128d)__builtin_ia32_cmpeqsd((__v2df)__a, (__v2df)__b);
698 }
699 
700 /// Compares the lower double-precision floating-point values in each of
701 /// the two 128-bit floating-point vectors of [2 x double] to determine if
702 /// the value in the first parameter is less than the corresponding value in
703 /// the second parameter.
704 ///
705 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
706 ///
707 /// \headerfile <x86intrin.h>
708 ///
709 /// This intrinsic corresponds to the <c> VCMPLTSD / CMPLTSD </c> instruction.
710 ///
711 /// \param __a
712 /// A 128-bit vector of [2 x double]. The lower double-precision value is
713 /// compared to the lower double-precision value of \a __b.
714 /// \param __b
715 /// A 128-bit vector of [2 x double]. The lower double-precision value is
716 /// compared to the lower double-precision value of \a __a.
717 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
718 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
719 static __inline__ __m128d __DEFAULT_FN_ATTRS
720 _mm_cmplt_sd(__m128d __a, __m128d __b)
721 {
722  return (__m128d)__builtin_ia32_cmpltsd((__v2df)__a, (__v2df)__b);
723 }
724 
725 /// Compares the lower double-precision floating-point values in each of
726 /// the two 128-bit floating-point vectors of [2 x double] to determine if
727 /// the value in the first parameter is less than or equal to the
728 /// corresponding value in the second parameter.
729 ///
730 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
731 ///
732 /// \headerfile <x86intrin.h>
733 ///
734 /// This intrinsic corresponds to the <c> VCMPLESD / CMPLESD </c> instruction.
735 ///
736 /// \param __a
737 /// A 128-bit vector of [2 x double]. The lower double-precision value is
738 /// compared to the lower double-precision value of \a __b.
739 /// \param __b
740 /// A 128-bit vector of [2 x double]. The lower double-precision value is
741 /// compared to the lower double-precision value of \a __a.
742 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
743 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
744 static __inline__ __m128d __DEFAULT_FN_ATTRS
745 _mm_cmple_sd(__m128d __a, __m128d __b)
746 {
747  return (__m128d)__builtin_ia32_cmplesd((__v2df)__a, (__v2df)__b);
748 }
749 
750 /// Compares the lower double-precision floating-point values in each of
751 /// the two 128-bit floating-point vectors of [2 x double] to determine if
752 /// the value in the first parameter is greater than the corresponding value
753 /// in the second parameter.
754 ///
755 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
756 ///
757 /// \headerfile <x86intrin.h>
758 ///
759 /// This intrinsic corresponds to the <c> VCMPLTSD / CMPLTSD </c> instruction.
760 ///
761 /// \param __a
762 /// A 128-bit vector of [2 x double]. The lower double-precision value is
763 /// compared to the lower double-precision value of \a __b.
764 /// \param __b
765 /// A 128-bit vector of [2 x double]. The lower double-precision value is
766 /// compared to the lower double-precision value of \a __a.
767 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
768 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
769 static __inline__ __m128d __DEFAULT_FN_ATTRS
770 _mm_cmpgt_sd(__m128d __a, __m128d __b)
771 {
772  __m128d __c = __builtin_ia32_cmpltsd((__v2df)__b, (__v2df)__a);
773  return __extension__ (__m128d) { __c[0], __a[1] };
774 }
775 
776 /// Compares the lower double-precision floating-point values in each of
777 /// the two 128-bit floating-point vectors of [2 x double] to determine if
778 /// the value in the first parameter is greater than or equal to the
779 /// corresponding value in the second parameter.
780 ///
781 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
782 ///
783 /// \headerfile <x86intrin.h>
784 ///
785 /// This intrinsic corresponds to the <c> VCMPLESD / CMPLESD </c> instruction.
786 ///
787 /// \param __a
788 /// A 128-bit vector of [2 x double]. The lower double-precision value is
789 /// compared to the lower double-precision value of \a __b.
790 /// \param __b
791 /// A 128-bit vector of [2 x double]. The lower double-precision value is
792 /// compared to the lower double-precision value of \a __a.
793 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
794 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
795 static __inline__ __m128d __DEFAULT_FN_ATTRS
796 _mm_cmpge_sd(__m128d __a, __m128d __b)
797 {
798  __m128d __c = __builtin_ia32_cmplesd((__v2df)__b, (__v2df)__a);
799  return __extension__ (__m128d) { __c[0], __a[1] };
800 }
801 
802 /// Compares the lower double-precision floating-point values in each of
803 /// the two 128-bit floating-point vectors of [2 x double] to determine if
804 /// the value in the first parameter is "ordered" with respect to the
805 /// corresponding value in the second parameter.
806 ///
807 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true. A pair
808 /// of double-precision values are "ordered" with respect to each other if
809 /// neither value is a NaN.
810 ///
811 /// \headerfile <x86intrin.h>
812 ///
813 /// This intrinsic corresponds to the <c> VCMPORDSD / CMPORDSD </c> instruction.
814 ///
815 /// \param __a
816 /// A 128-bit vector of [2 x double]. The lower double-precision value is
817 /// compared to the lower double-precision value of \a __b.
818 /// \param __b
819 /// A 128-bit vector of [2 x double]. The lower double-precision value is
820 /// compared to the lower double-precision value of \a __a.
821 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
822 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
823 static __inline__ __m128d __DEFAULT_FN_ATTRS
824 _mm_cmpord_sd(__m128d __a, __m128d __b)
825 {
826  return (__m128d)__builtin_ia32_cmpordsd((__v2df)__a, (__v2df)__b);
827 }
828 
829 /// Compares the lower double-precision floating-point values in each of
830 /// the two 128-bit floating-point vectors of [2 x double] to determine if
831 /// the value in the first parameter is "unordered" with respect to the
832 /// corresponding value in the second parameter.
833 ///
834 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true. A pair
835 /// of double-precision values are "unordered" with respect to each other if
836 /// one or both values are NaN.
837 ///
838 /// \headerfile <x86intrin.h>
839 ///
840 /// This intrinsic corresponds to the <c> VCMPUNORDSD / CMPUNORDSD </c>
841 /// instruction.
842 ///
843 /// \param __a
844 /// A 128-bit vector of [2 x double]. The lower double-precision value is
845 /// compared to the lower double-precision value of \a __b.
846 /// \param __b
847 /// A 128-bit vector of [2 x double]. The lower double-precision value is
848 /// compared to the lower double-precision value of \a __a.
849 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
850 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
851 static __inline__ __m128d __DEFAULT_FN_ATTRS
852 _mm_cmpunord_sd(__m128d __a, __m128d __b)
853 {
854  return (__m128d)__builtin_ia32_cmpunordsd((__v2df)__a, (__v2df)__b);
855 }
856 
857 /// Compares the lower double-precision floating-point values in each of
858 /// the two 128-bit floating-point vectors of [2 x double] to determine if
859 /// the value in the first parameter is unequal to the corresponding value in
860 /// the second parameter.
861 ///
862 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
863 ///
864 /// \headerfile <x86intrin.h>
865 ///
866 /// This intrinsic corresponds to the <c> VCMPNEQSD / CMPNEQSD </c> instruction.
867 ///
868 /// \param __a
869 /// A 128-bit vector of [2 x double]. The lower double-precision value is
870 /// compared to the lower double-precision value of \a __b.
871 /// \param __b
872 /// A 128-bit vector of [2 x double]. The lower double-precision value is
873 /// compared to the lower double-precision value of \a __a.
874 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
875 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
876 static __inline__ __m128d __DEFAULT_FN_ATTRS
877 _mm_cmpneq_sd(__m128d __a, __m128d __b)
878 {
879  return (__m128d)__builtin_ia32_cmpneqsd((__v2df)__a, (__v2df)__b);
880 }
881 
882 /// Compares the lower double-precision floating-point values in each of
883 /// the two 128-bit floating-point vectors of [2 x double] to determine if
884 /// the value in the first parameter is not less than the corresponding
885 /// value in the second parameter.
886 ///
887 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
888 ///
889 /// \headerfile <x86intrin.h>
890 ///
891 /// This intrinsic corresponds to the <c> VCMPNLTSD / CMPNLTSD </c> instruction.
892 ///
893 /// \param __a
894 /// A 128-bit vector of [2 x double]. The lower double-precision value is
895 /// compared to the lower double-precision value of \a __b.
896 /// \param __b
897 /// A 128-bit vector of [2 x double]. The lower double-precision value is
898 /// compared to the lower double-precision value of \a __a.
899 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
900 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
901 static __inline__ __m128d __DEFAULT_FN_ATTRS
902 _mm_cmpnlt_sd(__m128d __a, __m128d __b)
903 {
904  return (__m128d)__builtin_ia32_cmpnltsd((__v2df)__a, (__v2df)__b);
905 }
906 
907 /// Compares the lower double-precision floating-point values in each of
908 /// the two 128-bit floating-point vectors of [2 x double] to determine if
909 /// the value in the first parameter is not less than or equal to the
910 /// corresponding value in the second parameter.
911 ///
912 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
913 ///
914 /// \headerfile <x86intrin.h>
915 ///
916 /// This intrinsic corresponds to the <c> VCMPNLESD / CMPNLESD </c> instruction.
917 ///
918 /// \param __a
919 /// A 128-bit vector of [2 x double]. The lower double-precision value is
920 /// compared to the lower double-precision value of \a __b.
921 /// \param __b
922 /// A 128-bit vector of [2 x double]. The lower double-precision value is
923 /// compared to the lower double-precision value of \a __a.
924 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
925 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
926 static __inline__ __m128d __DEFAULT_FN_ATTRS
927 _mm_cmpnle_sd(__m128d __a, __m128d __b)
928 {
929  return (__m128d)__builtin_ia32_cmpnlesd((__v2df)__a, (__v2df)__b);
930 }
931 
932 /// Compares the lower double-precision floating-point values in each of
933 /// the two 128-bit floating-point vectors of [2 x double] to determine if
934 /// the value in the first parameter is not greater than the corresponding
935 /// value in the second parameter.
936 ///
937 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
938 ///
939 /// \headerfile <x86intrin.h>
940 ///
941 /// This intrinsic corresponds to the <c> VCMPNLTSD / CMPNLTSD </c> instruction.
942 ///
943 /// \param __a
944 /// A 128-bit vector of [2 x double]. The lower double-precision value is
945 /// compared to the lower double-precision value of \a __b.
946 /// \param __b
947 /// A 128-bit vector of [2 x double]. The lower double-precision value is
948 /// compared to the lower double-precision value of \a __a.
949 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
950 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
951 static __inline__ __m128d __DEFAULT_FN_ATTRS
952 _mm_cmpngt_sd(__m128d __a, __m128d __b)
953 {
954  __m128d __c = __builtin_ia32_cmpnltsd((__v2df)__b, (__v2df)__a);
955  return __extension__ (__m128d) { __c[0], __a[1] };
956 }
957 
958 /// Compares the lower double-precision floating-point values in each of
959 /// the two 128-bit floating-point vectors of [2 x double] to determine if
960 /// the value in the first parameter is not greater than or equal to the
961 /// corresponding value in the second parameter.
962 ///
963 /// The comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
964 ///
965 /// \headerfile <x86intrin.h>
966 ///
967 /// This intrinsic corresponds to the <c> VCMPNLESD / CMPNLESD </c> instruction.
968 ///
969 /// \param __a
970 /// A 128-bit vector of [2 x double]. The lower double-precision value is
971 /// compared to the lower double-precision value of \a __b.
972 /// \param __b
973 /// A 128-bit vector of [2 x double]. The lower double-precision value is
974 /// compared to the lower double-precision value of \a __a.
975 /// \returns A 128-bit vector. The lower 64 bits contains the comparison
976 /// results. The upper 64 bits are copied from the upper 64 bits of \a __a.
977 static __inline__ __m128d __DEFAULT_FN_ATTRS
978 _mm_cmpnge_sd(__m128d __a, __m128d __b)
979 {
980  __m128d __c = __builtin_ia32_cmpnlesd((__v2df)__b, (__v2df)__a);
981  return __extension__ (__m128d) { __c[0], __a[1] };
982 }
983 
984 /// Compares the lower double-precision floating-point values in each of
985 /// the two 128-bit floating-point vectors of [2 x double] for equality.
986 ///
987 /// The comparison yields 0 for false, 1 for true. If either of the two
988 /// lower double-precision values is NaN, 0 is returned.
989 ///
990 /// \headerfile <x86intrin.h>
991 ///
992 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
993 ///
994 /// \param __a
995 /// A 128-bit vector of [2 x double]. The lower double-precision value is
996 /// compared to the lower double-precision value of \a __b.
997 /// \param __b
998 /// A 128-bit vector of [2 x double]. The lower double-precision value is
999 /// compared to the lower double-precision value of \a __a.
1000 /// \returns An integer containing the comparison results. If either of the two
1001 /// lower double-precision values is NaN, 0 is returned.
1002 static __inline__ int __DEFAULT_FN_ATTRS
1003 _mm_comieq_sd(__m128d __a, __m128d __b)
1004 {
1005  return __builtin_ia32_comisdeq((__v2df)__a, (__v2df)__b);
1006 }
1007 
1008 /// Compares the lower double-precision floating-point values in each of
1009 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1010 /// the value in the first parameter is less than the corresponding value in
1011 /// the second parameter.
1012 ///
1013 /// The comparison yields 0 for false, 1 for true. If either of the two
1014 /// lower double-precision values is NaN, 0 is returned.
1015 ///
1016 /// \headerfile <x86intrin.h>
1017 ///
1018 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
1019 ///
1020 /// \param __a
1021 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1022 /// compared to the lower double-precision value of \a __b.
1023 /// \param __b
1024 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1025 /// compared to the lower double-precision value of \a __a.
1026 /// \returns An integer containing the comparison results. If either of the two
1027 /// lower double-precision values is NaN, 0 is returned.
1028 static __inline__ int __DEFAULT_FN_ATTRS
1029 _mm_comilt_sd(__m128d __a, __m128d __b)
1030 {
1031  return __builtin_ia32_comisdlt((__v2df)__a, (__v2df)__b);
1032 }
1033 
1034 /// Compares the lower double-precision floating-point values in each of
1035 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1036 /// the value in the first parameter is less than or equal to the
1037 /// corresponding value in the second parameter.
1038 ///
1039 /// The comparison yields 0 for false, 1 for true. If either of the two
1040 /// lower double-precision values is NaN, 0 is returned.
1041 ///
1042 /// \headerfile <x86intrin.h>
1043 ///
1044 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
1045 ///
1046 /// \param __a
1047 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1048 /// compared to the lower double-precision value of \a __b.
1049 /// \param __b
1050 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1051 /// compared to the lower double-precision value of \a __a.
1052 /// \returns An integer containing the comparison results. If either of the two
1053 /// lower double-precision values is NaN, 0 is returned.
1054 static __inline__ int __DEFAULT_FN_ATTRS
1055 _mm_comile_sd(__m128d __a, __m128d __b)
1056 {
1057  return __builtin_ia32_comisdle((__v2df)__a, (__v2df)__b);
1058 }
1059 
1060 /// Compares the lower double-precision floating-point values in each of
1061 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1062 /// the value in the first parameter is greater than the corresponding value
1063 /// in the second parameter.
1064 ///
1065 /// The comparison yields 0 for false, 1 for true. If either of the two
1066 /// lower double-precision values is NaN, 0 is returned.
1067 ///
1068 /// \headerfile <x86intrin.h>
1069 ///
1070 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
1071 ///
1072 /// \param __a
1073 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1074 /// compared to the lower double-precision value of \a __b.
1075 /// \param __b
1076 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1077 /// compared to the lower double-precision value of \a __a.
1078 /// \returns An integer containing the comparison results. If either of the two
1079 /// lower double-precision values is NaN, 0 is returned.
1080 static __inline__ int __DEFAULT_FN_ATTRS
1081 _mm_comigt_sd(__m128d __a, __m128d __b)
1082 {
1083  return __builtin_ia32_comisdgt((__v2df)__a, (__v2df)__b);
1084 }
1085 
1086 /// Compares the lower double-precision floating-point values in each of
1087 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1088 /// the value in the first parameter is greater than or equal to the
1089 /// corresponding value in the second parameter.
1090 ///
1091 /// The comparison yields 0 for false, 1 for true. If either of the two
1092 /// lower double-precision values is NaN, 0 is returned.
1093 ///
1094 /// \headerfile <x86intrin.h>
1095 ///
1096 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
1097 ///
1098 /// \param __a
1099 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1100 /// compared to the lower double-precision value of \a __b.
1101 /// \param __b
1102 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1103 /// compared to the lower double-precision value of \a __a.
1104 /// \returns An integer containing the comparison results. If either of the two
1105 /// lower double-precision values is NaN, 0 is returned.
1106 static __inline__ int __DEFAULT_FN_ATTRS
1107 _mm_comige_sd(__m128d __a, __m128d __b)
1108 {
1109  return __builtin_ia32_comisdge((__v2df)__a, (__v2df)__b);
1110 }
1111 
1112 /// Compares the lower double-precision floating-point values in each of
1113 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1114 /// the value in the first parameter is unequal to the corresponding value in
1115 /// the second parameter.
1116 ///
1117 /// The comparison yields 0 for false, 1 for true. If either of the two
1118 /// lower double-precision values is NaN, 1 is returned.
1119 ///
1120 /// \headerfile <x86intrin.h>
1121 ///
1122 /// This intrinsic corresponds to the <c> VCOMISD / COMISD </c> instruction.
1123 ///
1124 /// \param __a
1125 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1126 /// compared to the lower double-precision value of \a __b.
1127 /// \param __b
1128 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1129 /// compared to the lower double-precision value of \a __a.
1130 /// \returns An integer containing the comparison results. If either of the two
1131 /// lower double-precision values is NaN, 1 is returned.
1132 static __inline__ int __DEFAULT_FN_ATTRS
1133 _mm_comineq_sd(__m128d __a, __m128d __b)
1134 {
1135  return __builtin_ia32_comisdneq((__v2df)__a, (__v2df)__b);
1136 }
1137 
1138 /// Compares the lower double-precision floating-point values in each of
1139 /// the two 128-bit floating-point vectors of [2 x double] for equality. The
1140 /// comparison yields 0 for false, 1 for true.
1141 ///
1142 /// If either of the two lower double-precision values is NaN, 0 is returned.
1143 ///
1144 /// \headerfile <x86intrin.h>
1145 ///
1146 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1147 ///
1148 /// \param __a
1149 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1150 /// compared to the lower double-precision value of \a __b.
1151 /// \param __b
1152 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1153 /// compared to the lower double-precision value of \a __a.
1154 /// \returns An integer containing the comparison results. If either of the two
1155 /// lower double-precision values is NaN, 0 is returned.
1156 static __inline__ int __DEFAULT_FN_ATTRS
1157 _mm_ucomieq_sd(__m128d __a, __m128d __b)
1158 {
1159  return __builtin_ia32_ucomisdeq((__v2df)__a, (__v2df)__b);
1160 }
1161 
1162 /// Compares the lower double-precision floating-point values in each of
1163 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1164 /// the value in the first parameter is less than the corresponding value in
1165 /// the second parameter.
1166 ///
1167 /// The comparison yields 0 for false, 1 for true. If either of the two lower
1168 /// double-precision values is NaN, 0 is returned.
1169 ///
1170 /// \headerfile <x86intrin.h>
1171 ///
1172 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1173 ///
1174 /// \param __a
1175 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1176 /// compared to the lower double-precision value of \a __b.
1177 /// \param __b
1178 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1179 /// compared to the lower double-precision value of \a __a.
1180 /// \returns An integer containing the comparison results. If either of the two
1181 /// lower double-precision values is NaN, 0 is returned.
1182 static __inline__ int __DEFAULT_FN_ATTRS
1183 _mm_ucomilt_sd(__m128d __a, __m128d __b)
1184 {
1185  return __builtin_ia32_ucomisdlt((__v2df)__a, (__v2df)__b);
1186 }
1187 
1188 /// Compares the lower double-precision floating-point values in each of
1189 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1190 /// the value in the first parameter is less than or equal to the
1191 /// corresponding value in the second parameter.
1192 ///
1193 /// The comparison yields 0 for false, 1 for true. If either of the two lower
1194 /// double-precision values is NaN, 0 is returned.
1195 ///
1196 /// \headerfile <x86intrin.h>
1197 ///
1198 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1199 ///
1200 /// \param __a
1201 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1202 /// compared to the lower double-precision value of \a __b.
1203 /// \param __b
1204 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1205 /// compared to the lower double-precision value of \a __a.
1206 /// \returns An integer containing the comparison results. If either of the two
1207 /// lower double-precision values is NaN, 0 is returned.
1208 static __inline__ int __DEFAULT_FN_ATTRS
1209 _mm_ucomile_sd(__m128d __a, __m128d __b)
1210 {
1211  return __builtin_ia32_ucomisdle((__v2df)__a, (__v2df)__b);
1212 }
1213 
1214 /// Compares the lower double-precision floating-point values in each of
1215 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1216 /// the value in the first parameter is greater than the corresponding value
1217 /// in the second parameter.
1218 ///
1219 /// The comparison yields 0 for false, 1 for true. If either of the two lower
1220 /// double-precision values is NaN, 0 is returned.
1221 ///
1222 /// \headerfile <x86intrin.h>
1223 ///
1224 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1225 ///
1226 /// \param __a
1227 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1228 /// compared to the lower double-precision value of \a __b.
1229 /// \param __b
1230 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1231 /// compared to the lower double-precision value of \a __a.
1232 /// \returns An integer containing the comparison results. If either of the two
1233 /// lower double-precision values is NaN, 0 is returned.
1234 static __inline__ int __DEFAULT_FN_ATTRS
1235 _mm_ucomigt_sd(__m128d __a, __m128d __b)
1236 {
1237  return __builtin_ia32_ucomisdgt((__v2df)__a, (__v2df)__b);
1238 }
1239 
1240 /// Compares the lower double-precision floating-point values in each of
1241 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1242 /// the value in the first parameter is greater than or equal to the
1243 /// corresponding value in the second parameter.
1244 ///
1245 /// The comparison yields 0 for false, 1 for true. If either of the two
1246 /// lower double-precision values is NaN, 0 is returned.
1247 ///
1248 /// \headerfile <x86intrin.h>
1249 ///
1250 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1251 ///
1252 /// \param __a
1253 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1254 /// compared to the lower double-precision value of \a __b.
1255 /// \param __b
1256 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1257 /// compared to the lower double-precision value of \a __a.
1258 /// \returns An integer containing the comparison results. If either of the two
1259 /// lower double-precision values is NaN, 0 is returned.
1260 static __inline__ int __DEFAULT_FN_ATTRS
1261 _mm_ucomige_sd(__m128d __a, __m128d __b)
1262 {
1263  return __builtin_ia32_ucomisdge((__v2df)__a, (__v2df)__b);
1264 }
1265 
1266 /// Compares the lower double-precision floating-point values in each of
1267 /// the two 128-bit floating-point vectors of [2 x double] to determine if
1268 /// the value in the first parameter is unequal to the corresponding value in
1269 /// the second parameter.
1270 ///
1271 /// The comparison yields 0 for false, 1 for true. If either of the two lower
1272 /// double-precision values is NaN, 1 is returned.
1273 ///
1274 /// \headerfile <x86intrin.h>
1275 ///
1276 /// This intrinsic corresponds to the <c> VUCOMISD / UCOMISD </c> instruction.
1277 ///
1278 /// \param __a
1279 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1280 /// compared to the lower double-precision value of \a __b.
1281 /// \param __b
1282 /// A 128-bit vector of [2 x double]. The lower double-precision value is
1283 /// compared to the lower double-precision value of \a __a.
1284 /// \returns An integer containing the comparison result. If either of the two
1285 /// lower double-precision values is NaN, 1 is returned.
1286 static __inline__ int __DEFAULT_FN_ATTRS
1287 _mm_ucomineq_sd(__m128d __a, __m128d __b)
1288 {
1289  return __builtin_ia32_ucomisdneq((__v2df)__a, (__v2df)__b);
1290 }
1291 
1292 /// Converts the two double-precision floating-point elements of a
1293 /// 128-bit vector of [2 x double] into two single-precision floating-point
1294 /// values, returned in the lower 64 bits of a 128-bit vector of [4 x float].
1295 /// The upper 64 bits of the result vector are set to zero.
1296 ///
1297 /// \headerfile <x86intrin.h>
1298 ///
1299 /// This intrinsic corresponds to the <c> VCVTPD2PS / CVTPD2PS </c> instruction.
1300 ///
1301 /// \param __a
1302 /// A 128-bit vector of [2 x double].
1303 /// \returns A 128-bit vector of [4 x float] whose lower 64 bits contain the
1304 /// converted values. The upper 64 bits are set to zero.
1305 static __inline__ __m128 __DEFAULT_FN_ATTRS
1307 {
1308  return __builtin_ia32_cvtpd2ps((__v2df)__a);
1309 }
1310 
1311 /// Converts the lower two single-precision floating-point elements of a
1312 /// 128-bit vector of [4 x float] into two double-precision floating-point
1313 /// values, returned in a 128-bit vector of [2 x double]. The upper two
1314 /// elements of the input vector are unused.
1315 ///
1316 /// \headerfile <x86intrin.h>
1317 ///
1318 /// This intrinsic corresponds to the <c> VCVTPS2PD / CVTPS2PD </c> instruction.
1319 ///
1320 /// \param __a
1321 /// A 128-bit vector of [4 x float]. The lower two single-precision
1322 /// floating-point elements are converted to double-precision values. The
1323 /// upper two elements are unused.
1324 /// \returns A 128-bit vector of [2 x double] containing the converted values.
1325 static __inline__ __m128d __DEFAULT_FN_ATTRS
1327 {
1328  return (__m128d) __builtin_convertvector(
1329  __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 0, 1), __v2df);
1330 }
1331 
1332 /// Converts the lower two integer elements of a 128-bit vector of
1333 /// [4 x i32] into two double-precision floating-point values, returned in a
1334 /// 128-bit vector of [2 x double].
1335 ///
1336 /// The upper two elements of the input vector are unused.
1337 ///
1338 /// \headerfile <x86intrin.h>
1339 ///
1340 /// This intrinsic corresponds to the <c> VCVTDQ2PD / CVTDQ2PD </c> instruction.
1341 ///
1342 /// \param __a
1343 /// A 128-bit integer vector of [4 x i32]. The lower two integer elements are
1344 /// converted to double-precision values.
1345 ///
1346 /// The upper two elements are unused.
1347 /// \returns A 128-bit vector of [2 x double] containing the converted values.
1348 static __inline__ __m128d __DEFAULT_FN_ATTRS
1350 {
1351  return (__m128d) __builtin_convertvector(
1352  __builtin_shufflevector((__v4si)__a, (__v4si)__a, 0, 1), __v2df);
1353 }
1354 
1355 /// Converts the two double-precision floating-point elements of a
1356 /// 128-bit vector of [2 x double] into two signed 32-bit integer values,
1357 /// returned in the lower 64 bits of a 128-bit vector of [4 x i32]. The upper
1358 /// 64 bits of the result vector are set to zero.
1359 ///
1360 /// \headerfile <x86intrin.h>
1361 ///
1362 /// This intrinsic corresponds to the <c> VCVTPD2DQ / CVTPD2DQ </c> instruction.
1363 ///
1364 /// \param __a
1365 /// A 128-bit vector of [2 x double].
1366 /// \returns A 128-bit vector of [4 x i32] whose lower 64 bits contain the
1367 /// converted values. The upper 64 bits are set to zero.
1368 static __inline__ __m128i __DEFAULT_FN_ATTRS
1370 {
1371  return __builtin_ia32_cvtpd2dq((__v2df)__a);
1372 }
1373 
1374 /// Converts the low-order element of a 128-bit vector of [2 x double]
1375 /// into a 32-bit signed integer value.
1376 ///
1377 /// \headerfile <x86intrin.h>
1378 ///
1379 /// This intrinsic corresponds to the <c> VCVTSD2SI / CVTSD2SI </c> instruction.
1380 ///
1381 /// \param __a
1382 /// A 128-bit vector of [2 x double]. The lower 64 bits are used in the
1383 /// conversion.
1384 /// \returns A 32-bit signed integer containing the converted value.
1385 static __inline__ int __DEFAULT_FN_ATTRS
1387 {
1388  return __builtin_ia32_cvtsd2si((__v2df)__a);
1389 }
1390 
1391 /// Converts the lower double-precision floating-point element of a
1392 /// 128-bit vector of [2 x double], in the second parameter, into a
1393 /// single-precision floating-point value, returned in the lower 32 bits of a
1394 /// 128-bit vector of [4 x float]. The upper 96 bits of the result vector are
1395 /// copied from the upper 96 bits of the first parameter.
1396 ///
1397 /// \headerfile <x86intrin.h>
1398 ///
1399 /// This intrinsic corresponds to the <c> VCVTSD2SS / CVTSD2SS </c> instruction.
1400 ///
1401 /// \param __a
1402 /// A 128-bit vector of [4 x float]. The upper 96 bits of this parameter are
1403 /// copied to the upper 96 bits of the result.
1404 /// \param __b
1405 /// A 128-bit vector of [2 x double]. The lower double-precision
1406 /// floating-point element is used in the conversion.
1407 /// \returns A 128-bit vector of [4 x float]. The lower 32 bits contain the
1408 /// converted value from the second parameter. The upper 96 bits are copied
1409 /// from the upper 96 bits of the first parameter.
1410 static __inline__ __m128 __DEFAULT_FN_ATTRS
1411 _mm_cvtsd_ss(__m128 __a, __m128d __b)
1412 {
1413  return (__m128)__builtin_ia32_cvtsd2ss((__v4sf)__a, (__v2df)__b);
1414 }
1415 
1416 /// Converts a 32-bit signed integer value, in the second parameter, into
1417 /// a double-precision floating-point value, returned in the lower 64 bits of
1418 /// a 128-bit vector of [2 x double]. The upper 64 bits of the result vector
1419 /// are copied from the upper 64 bits of the first parameter.
1420 ///
1421 /// \headerfile <x86intrin.h>
1422 ///
1423 /// This intrinsic corresponds to the <c> VCVTSI2SD / CVTSI2SD </c> instruction.
1424 ///
1425 /// \param __a
1426 /// A 128-bit vector of [2 x double]. The upper 64 bits of this parameter are
1427 /// copied to the upper 64 bits of the result.
1428 /// \param __b
1429 /// A 32-bit signed integer containing the value to be converted.
1430 /// \returns A 128-bit vector of [2 x double]. The lower 64 bits contain the
1431 /// converted value from the second parameter. The upper 64 bits are copied
1432 /// from the upper 64 bits of the first parameter.
1433 static __inline__ __m128d __DEFAULT_FN_ATTRS
1434 _mm_cvtsi32_sd(__m128d __a, int __b)
1435 {
1436  __a[0] = __b;
1437  return __a;
1438 }
1439 
1440 /// Converts the lower single-precision floating-point element of a
1441 /// 128-bit vector of [4 x float], in the second parameter, into a
1442 /// double-precision floating-point value, returned in the lower 64 bits of
1443 /// a 128-bit vector of [2 x double]. The upper 64 bits of the result vector
1444 /// are copied from the upper 64 bits of the first parameter.
1445 ///
1446 /// \headerfile <x86intrin.h>
1447 ///
1448 /// This intrinsic corresponds to the <c> VCVTSS2SD / CVTSS2SD </c> instruction.
1449 ///
1450 /// \param __a
1451 /// A 128-bit vector of [2 x double]. The upper 64 bits of this parameter are
1452 /// copied to the upper 64 bits of the result.
1453 /// \param __b
1454 /// A 128-bit vector of [4 x float]. The lower single-precision
1455 /// floating-point element is used in the conversion.
1456 /// \returns A 128-bit vector of [2 x double]. The lower 64 bits contain the
1457 /// converted value from the second parameter. The upper 64 bits are copied
1458 /// from the upper 64 bits of the first parameter.
1459 static __inline__ __m128d __DEFAULT_FN_ATTRS
1460 _mm_cvtss_sd(__m128d __a, __m128 __b)
1461 {
1462  __a[0] = __b[0];
1463  return __a;
1464 }
1465 
1466 /// Converts the two double-precision floating-point elements of a
1467 /// 128-bit vector of [2 x double] into two signed 32-bit integer values,
1468 /// returned in the lower 64 bits of a 128-bit vector of [4 x i32].
1469 ///
1470 /// If the result of either conversion is inexact, the result is truncated
1471 /// (rounded towards zero) regardless of the current MXCSR setting. The upper
1472 /// 64 bits of the result vector are set to zero.
1473 ///
1474 /// \headerfile <x86intrin.h>
1475 ///
1476 /// This intrinsic corresponds to the <c> VCVTTPD2DQ / CVTTPD2DQ </c>
1477 /// instruction.
1478 ///
1479 /// \param __a
1480 /// A 128-bit vector of [2 x double].
1481 /// \returns A 128-bit vector of [4 x i32] whose lower 64 bits contain the
1482 /// converted values. The upper 64 bits are set to zero.
1483 static __inline__ __m128i __DEFAULT_FN_ATTRS
1485 {
1486  return (__m128i)__builtin_ia32_cvttpd2dq((__v2df)__a);
1487 }
1488 
1489 /// Converts the low-order element of a [2 x double] vector into a 32-bit
1490 /// signed integer value, truncating the result when it is inexact.
1491 ///
1492 /// \headerfile <x86intrin.h>
1493 ///
1494 /// This intrinsic corresponds to the <c> VCVTTSD2SI / CVTTSD2SI </c>
1495 /// instruction.
1496 ///
1497 /// \param __a
1498 /// A 128-bit vector of [2 x double]. The lower 64 bits are used in the
1499 /// conversion.
1500 /// \returns A 32-bit signed integer containing the converted value.
1501 static __inline__ int __DEFAULT_FN_ATTRS
1503 {
1504  return __builtin_ia32_cvttsd2si((__v2df)__a);
1505 }
1506 
1507 /// Converts the two double-precision floating-point elements of a
1508 /// 128-bit vector of [2 x double] into two signed 32-bit integer values,
1509 /// returned in a 64-bit vector of [2 x i32].
1510 ///
1511 /// \headerfile <x86intrin.h>
1512 ///
1513 /// This intrinsic corresponds to the <c> CVTPD2PI </c> instruction.
1514 ///
1515 /// \param __a
1516 /// A 128-bit vector of [2 x double].
1517 /// \returns A 64-bit vector of [2 x i32] containing the converted values.
1518 static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
1520 {
1521  return (__m64)__builtin_ia32_cvtpd2pi((__v2df)__a);
1522 }
1523 
1524 /// Converts the two double-precision floating-point elements of a
1525 /// 128-bit vector of [2 x double] into two signed 32-bit integer values,
1526 /// returned in a 64-bit vector of [2 x i32].
1527 ///
1528 /// If the result of either conversion is inexact, the result is truncated
1529 /// (rounded towards zero) regardless of the current MXCSR setting.
1530 ///
1531 /// \headerfile <x86intrin.h>
1532 ///
1533 /// This intrinsic corresponds to the <c> CVTTPD2PI </c> instruction.
1534 ///
1535 /// \param __a
1536 /// A 128-bit vector of [2 x double].
1537 /// \returns A 64-bit vector of [2 x i32] containing the converted values.
1538 static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
1540 {
1541  return (__m64)__builtin_ia32_cvttpd2pi((__v2df)__a);
1542 }
1543 
1544 /// Converts the two signed 32-bit integer elements of a 64-bit vector of
1545 /// [2 x i32] into two double-precision floating-point values, returned in a
1546 /// 128-bit vector of [2 x double].
1547 ///
1548 /// \headerfile <x86intrin.h>
1549 ///
1550 /// This intrinsic corresponds to the <c> CVTPI2PD </c> instruction.
1551 ///
1552 /// \param __a
1553 /// A 64-bit vector of [2 x i32].
1554 /// \returns A 128-bit vector of [2 x double] containing the converted values.
1555 static __inline__ __m128d __DEFAULT_FN_ATTRS_MMX
1557 {
1558  return __builtin_ia32_cvtpi2pd((__v2si)__a);
1559 }
1560 
1561 /// Returns the low-order element of a 128-bit vector of [2 x double] as
1562 /// a double-precision floating-point value.
1563 ///
1564 /// \headerfile <x86intrin.h>
1565 ///
1566 /// This intrinsic has no corresponding instruction.
1567 ///
1568 /// \param __a
1569 /// A 128-bit vector of [2 x double]. The lower 64 bits are returned.
1570 /// \returns A double-precision floating-point value copied from the lower 64
1571 /// bits of \a __a.
1572 static __inline__ double __DEFAULT_FN_ATTRS
1574 {
1575  return __a[0];
1576 }
1577 
1578 /// Loads a 128-bit floating-point vector of [2 x double] from an aligned
1579 /// memory location.
1580 ///
1581 /// \headerfile <x86intrin.h>
1582 ///
1583 /// This intrinsic corresponds to the <c> VMOVAPD / MOVAPD </c> instruction.
1584 ///
1585 /// \param __dp
1586 /// A pointer to a 128-bit memory location. The address of the memory
1587 /// location has to be 16-byte aligned.
1588 /// \returns A 128-bit vector of [2 x double] containing the loaded values.
1589 static __inline__ __m128d __DEFAULT_FN_ATTRS
1590 _mm_load_pd(double const *__dp)
1591 {
1592  return *(__m128d*)__dp;
1593 }
1594 
1595 /// Loads a double-precision floating-point value from a specified memory
1596 /// location and duplicates it to both vector elements of a 128-bit vector of
1597 /// [2 x double].
1598 ///
1599 /// \headerfile <x86intrin.h>
1600 ///
1601 /// This intrinsic corresponds to the <c> VMOVDDUP / MOVDDUP </c> instruction.
1602 ///
1603 /// \param __dp
1604 /// A pointer to a memory location containing a double-precision value.
1605 /// \returns A 128-bit vector of [2 x double] containing the loaded and
1606 /// duplicated values.
1607 static __inline__ __m128d __DEFAULT_FN_ATTRS
1608 _mm_load1_pd(double const *__dp)
1609 {
1610  struct __mm_load1_pd_struct {
1611  double __u;
1612  } __attribute__((__packed__, __may_alias__));
1613  double __u = ((struct __mm_load1_pd_struct*)__dp)->__u;
1614  return __extension__ (__m128d){ __u, __u };
1615 }
1616 
1617 #define _mm_load_pd1(dp) _mm_load1_pd(dp)
1618 
1619 /// Loads two double-precision values, in reverse order, from an aligned
1620 /// memory location into a 128-bit vector of [2 x double].
1621 ///
1622 /// \headerfile <x86intrin.h>
1623 ///
1624 /// This intrinsic corresponds to the <c> VMOVAPD / MOVAPD </c> instruction +
1625 /// needed shuffling instructions. In AVX mode, the shuffling may be combined
1626 /// with the \c VMOVAPD, resulting in only a \c VPERMILPD instruction.
1627 ///
1628 /// \param __dp
1629 /// A 16-byte aligned pointer to an array of double-precision values to be
1630 /// loaded in reverse order.
1631 /// \returns A 128-bit vector of [2 x double] containing the reversed loaded
1632 /// values.
1633 static __inline__ __m128d __DEFAULT_FN_ATTRS
1634 _mm_loadr_pd(double const *__dp)
1635 {
1636  __m128d __u = *(__m128d*)__dp;
1637  return __builtin_shufflevector((__v2df)__u, (__v2df)__u, 1, 0);
1638 }
1639 
1640 /// Loads a 128-bit floating-point vector of [2 x double] from an
1641 /// unaligned memory location.
1642 ///
1643 /// \headerfile <x86intrin.h>
1644 ///
1645 /// This intrinsic corresponds to the <c> VMOVUPD / MOVUPD </c> instruction.
1646 ///
1647 /// \param __dp
1648 /// A pointer to a 128-bit memory location. The address of the memory
1649 /// location does not have to be aligned.
1650 /// \returns A 128-bit vector of [2 x double] containing the loaded values.
1651 static __inline__ __m128d __DEFAULT_FN_ATTRS
1652 _mm_loadu_pd(double const *__dp)
1653 {
1654  struct __loadu_pd {
1655  __m128d __v;
1656  } __attribute__((__packed__, __may_alias__));
1657  return ((struct __loadu_pd*)__dp)->__v;
1658 }
1659 
1660 /// Loads a 64-bit integer value to the low element of a 128-bit integer
1661 /// vector and clears the upper element.
1662 ///
1663 /// \headerfile <x86intrin.h>
1664 ///
1665 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
1666 ///
1667 /// \param __a
1668 /// A pointer to a 64-bit memory location. The address of the memory
1669 /// location does not have to be aligned.
1670 /// \returns A 128-bit vector of [2 x i64] containing the loaded value.
1671 static __inline__ __m128i __DEFAULT_FN_ATTRS
1672 _mm_loadu_si64(void const *__a)
1673 {
1674  struct __loadu_si64 {
1675  long long __v;
1676  } __attribute__((__packed__, __may_alias__));
1677  long long __u = ((struct __loadu_si64*)__a)->__v;
1678  return __extension__ (__m128i)(__v2di){__u, 0L};
1679 }
1680 
1681 /// Loads a 64-bit double-precision value to the low element of a
1682 /// 128-bit integer vector and clears the upper element.
1683 ///
1684 /// \headerfile <x86intrin.h>
1685 ///
1686 /// This intrinsic corresponds to the <c> VMOVSD / MOVSD </c> instruction.
1687 ///
1688 /// \param __dp
1689 /// A pointer to a memory location containing a double-precision value.
1690 /// The address of the memory location does not have to be aligned.
1691 /// \returns A 128-bit vector of [2 x double] containing the loaded value.
1692 static __inline__ __m128d __DEFAULT_FN_ATTRS
1693 _mm_load_sd(double const *__dp)
1694 {
1695  struct __mm_load_sd_struct {
1696  double __u;
1697  } __attribute__((__packed__, __may_alias__));
1698  double __u = ((struct __mm_load_sd_struct*)__dp)->__u;
1699  return __extension__ (__m128d){ __u, 0 };
1700 }
1701 
1702 /// Loads a double-precision value into the high-order bits of a 128-bit
1703 /// vector of [2 x double]. The low-order bits are copied from the low-order
1704 /// bits of the first operand.
1705 ///
1706 /// \headerfile <x86intrin.h>
1707 ///
1708 /// This intrinsic corresponds to the <c> VMOVHPD / MOVHPD </c> instruction.
1709 ///
1710 /// \param __a
1711 /// A 128-bit vector of [2 x double]. \n
1712 /// Bits [63:0] are written to bits [63:0] of the result.
1713 /// \param __dp
1714 /// A pointer to a 64-bit memory location containing a double-precision
1715 /// floating-point value that is loaded. The loaded value is written to bits
1716 /// [127:64] of the result. The address of the memory location does not have
1717 /// to be aligned.
1718 /// \returns A 128-bit vector of [2 x double] containing the moved values.
1719 static __inline__ __m128d __DEFAULT_FN_ATTRS
1720 _mm_loadh_pd(__m128d __a, double const *__dp)
1721 {
1722  struct __mm_loadh_pd_struct {
1723  double __u;
1724  } __attribute__((__packed__, __may_alias__));
1725  double __u = ((struct __mm_loadh_pd_struct*)__dp)->__u;
1726  return __extension__ (__m128d){ __a[0], __u };
1727 }
1728 
1729 /// Loads a double-precision value into the low-order bits of a 128-bit
1730 /// vector of [2 x double]. The high-order bits are copied from the
1731 /// high-order bits of the first operand.
1732 ///
1733 /// \headerfile <x86intrin.h>
1734 ///
1735 /// This intrinsic corresponds to the <c> VMOVLPD / MOVLPD </c> instruction.
1736 ///
1737 /// \param __a
1738 /// A 128-bit vector of [2 x double]. \n
1739 /// Bits [127:64] are written to bits [127:64] of the result.
1740 /// \param __dp
1741 /// A pointer to a 64-bit memory location containing a double-precision
1742 /// floating-point value that is loaded. The loaded value is written to bits
1743 /// [63:0] of the result. The address of the memory location does not have to
1744 /// be aligned.
1745 /// \returns A 128-bit vector of [2 x double] containing the moved values.
1746 static __inline__ __m128d __DEFAULT_FN_ATTRS
1747 _mm_loadl_pd(__m128d __a, double const *__dp)
1748 {
1749  struct __mm_loadl_pd_struct {
1750  double __u;
1751  } __attribute__((__packed__, __may_alias__));
1752  double __u = ((struct __mm_loadl_pd_struct*)__dp)->__u;
1753  return __extension__ (__m128d){ __u, __a[1] };
1754 }
1755 
1756 /// Constructs a 128-bit floating-point vector of [2 x double] with
1757 /// unspecified content. This could be used as an argument to another
1758 /// intrinsic function where the argument is required but the value is not
1759 /// actually used.
1760 ///
1761 /// \headerfile <x86intrin.h>
1762 ///
1763 /// This intrinsic has no corresponding instruction.
1764 ///
1765 /// \returns A 128-bit floating-point vector of [2 x double] with unspecified
1766 /// content.
1767 static __inline__ __m128d __DEFAULT_FN_ATTRS
1769 {
1770  return (__m128d)__builtin_ia32_undef128();
1771 }
1772 
1773 /// Constructs a 128-bit floating-point vector of [2 x double]. The lower
1774 /// 64 bits of the vector are initialized with the specified double-precision
1775 /// floating-point value. The upper 64 bits are set to zero.
1776 ///
1777 /// \headerfile <x86intrin.h>
1778 ///
1779 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
1780 ///
1781 /// \param __w
1782 /// A double-precision floating-point value used to initialize the lower 64
1783 /// bits of the result.
1784 /// \returns An initialized 128-bit floating-point vector of [2 x double]. The
1785 /// lower 64 bits contain the value of the parameter. The upper 64 bits are
1786 /// set to zero.
1787 static __inline__ __m128d __DEFAULT_FN_ATTRS
1788 _mm_set_sd(double __w)
1789 {
1790  return __extension__ (__m128d){ __w, 0 };
1791 }
1792 
1793 /// Constructs a 128-bit floating-point vector of [2 x double], with each
1794 /// of the two double-precision floating-point vector elements set to the
1795 /// specified double-precision floating-point value.
1796 ///
1797 /// \headerfile <x86intrin.h>
1798 ///
1799 /// This intrinsic corresponds to the <c> VMOVDDUP / MOVLHPS </c> instruction.
1800 ///
1801 /// \param __w
1802 /// A double-precision floating-point value used to initialize each vector
1803 /// element of the result.
1804 /// \returns An initialized 128-bit floating-point vector of [2 x double].
1805 static __inline__ __m128d __DEFAULT_FN_ATTRS
1806 _mm_set1_pd(double __w)
1807 {
1808  return __extension__ (__m128d){ __w, __w };
1809 }
1810 
1811 /// Constructs a 128-bit floating-point vector of [2 x double], with each
1812 /// of the two double-precision floating-point vector elements set to the
1813 /// specified double-precision floating-point value.
1814 ///
1815 /// \headerfile <x86intrin.h>
1816 ///
1817 /// This intrinsic corresponds to the <c> VMOVDDUP / MOVLHPS </c> instruction.
1818 ///
1819 /// \param __w
1820 /// A double-precision floating-point value used to initialize each vector
1821 /// element of the result.
1822 /// \returns An initialized 128-bit floating-point vector of [2 x double].
1823 static __inline__ __m128d __DEFAULT_FN_ATTRS
1824 _mm_set_pd1(double __w)
1825 {
1826  return _mm_set1_pd(__w);
1827 }
1828 
1829 /// Constructs a 128-bit floating-point vector of [2 x double]
1830 /// initialized with the specified double-precision floating-point values.
1831 ///
1832 /// \headerfile <x86intrin.h>
1833 ///
1834 /// This intrinsic corresponds to the <c> VUNPCKLPD / UNPCKLPD </c> instruction.
1835 ///
1836 /// \param __w
1837 /// A double-precision floating-point value used to initialize the upper 64
1838 /// bits of the result.
1839 /// \param __x
1840 /// A double-precision floating-point value used to initialize the lower 64
1841 /// bits of the result.
1842 /// \returns An initialized 128-bit floating-point vector of [2 x double].
1843 static __inline__ __m128d __DEFAULT_FN_ATTRS
1844 _mm_set_pd(double __w, double __x)
1845 {
1846  return __extension__ (__m128d){ __x, __w };
1847 }
1848 
1849 /// Constructs a 128-bit floating-point vector of [2 x double],
1850 /// initialized in reverse order with the specified double-precision
1851 /// floating-point values.
1852 ///
1853 /// \headerfile <x86intrin.h>
1854 ///
1855 /// This intrinsic corresponds to the <c> VUNPCKLPD / UNPCKLPD </c> instruction.
1856 ///
1857 /// \param __w
1858 /// A double-precision floating-point value used to initialize the lower 64
1859 /// bits of the result.
1860 /// \param __x
1861 /// A double-precision floating-point value used to initialize the upper 64
1862 /// bits of the result.
1863 /// \returns An initialized 128-bit floating-point vector of [2 x double].
1864 static __inline__ __m128d __DEFAULT_FN_ATTRS
1865 _mm_setr_pd(double __w, double __x)
1866 {
1867  return __extension__ (__m128d){ __w, __x };
1868 }
1869 
1870 /// Constructs a 128-bit floating-point vector of [2 x double]
1871 /// initialized to zero.
1872 ///
1873 /// \headerfile <x86intrin.h>
1874 ///
1875 /// This intrinsic corresponds to the <c> VXORPS / XORPS </c> instruction.
1876 ///
1877 /// \returns An initialized 128-bit floating-point vector of [2 x double] with
1878 /// all elements set to zero.
1879 static __inline__ __m128d __DEFAULT_FN_ATTRS
1881 {
1882  return __extension__ (__m128d){ 0, 0 };
1883 }
1884 
1885 /// Constructs a 128-bit floating-point vector of [2 x double]. The lower
1886 /// 64 bits are set to the lower 64 bits of the second parameter. The upper
1887 /// 64 bits are set to the upper 64 bits of the first parameter.
1888 ///
1889 /// \headerfile <x86intrin.h>
1890 ///
1891 /// This intrinsic corresponds to the <c> VBLENDPD / BLENDPD </c> instruction.
1892 ///
1893 /// \param __a
1894 /// A 128-bit vector of [2 x double]. The upper 64 bits are written to the
1895 /// upper 64 bits of the result.
1896 /// \param __b
1897 /// A 128-bit vector of [2 x double]. The lower 64 bits are written to the
1898 /// lower 64 bits of the result.
1899 /// \returns A 128-bit vector of [2 x double] containing the moved values.
1900 static __inline__ __m128d __DEFAULT_FN_ATTRS
1901 _mm_move_sd(__m128d __a, __m128d __b)
1902 {
1903  __a[0] = __b[0];
1904  return __a;
1905 }
1906 
1907 /// Stores the lower 64 bits of a 128-bit vector of [2 x double] to a
1908 /// memory location.
1909 ///
1910 /// \headerfile <x86intrin.h>
1911 ///
1912 /// This intrinsic corresponds to the <c> VMOVSD / MOVSD </c> instruction.
1913 ///
1914 /// \param __dp
1915 /// A pointer to a 64-bit memory location.
1916 /// \param __a
1917 /// A 128-bit vector of [2 x double] containing the value to be stored.
1918 static __inline__ void __DEFAULT_FN_ATTRS
1919 _mm_store_sd(double *__dp, __m128d __a)
1920 {
1921  struct __mm_store_sd_struct {
1922  double __u;
1923  } __attribute__((__packed__, __may_alias__));
1924  ((struct __mm_store_sd_struct*)__dp)->__u = __a[0];
1925 }
1926 
1927 /// Moves packed double-precision values from a 128-bit vector of
1928 /// [2 x double] to a memory location.
1929 ///
1930 /// \headerfile <x86intrin.h>
1931 ///
1932 /// This intrinsic corresponds to the <c>VMOVAPD / MOVAPS</c> instruction.
1933 ///
1934 /// \param __dp
1935 /// A pointer to an aligned memory location that can store two
1936 /// double-precision values.
1937 /// \param __a
1938 /// A packed 128-bit vector of [2 x double] containing the values to be
1939 /// moved.
1940 static __inline__ void __DEFAULT_FN_ATTRS
1941 _mm_store_pd(double *__dp, __m128d __a)
1942 {
1943  *(__m128d*)__dp = __a;
1944 }
1945 
1946 /// Moves the lower 64 bits of a 128-bit vector of [2 x double] twice to
1947 /// the upper and lower 64 bits of a memory location.
1948 ///
1949 /// \headerfile <x86intrin.h>
1950 ///
1951 /// This intrinsic corresponds to the
1952 /// <c> VMOVDDUP + VMOVAPD / MOVLHPS + MOVAPS </c> instruction.
1953 ///
1954 /// \param __dp
1955 /// A pointer to a memory location that can store two double-precision
1956 /// values.
1957 /// \param __a
1958 /// A 128-bit vector of [2 x double] whose lower 64 bits are copied to each
1959 /// of the values in \a __dp.
1960 static __inline__ void __DEFAULT_FN_ATTRS
1961 _mm_store1_pd(double *__dp, __m128d __a)
1962 {
1963  __a = __builtin_shufflevector((__v2df)__a, (__v2df)__a, 0, 0);
1964  _mm_store_pd(__dp, __a);
1965 }
1966 
1967 /// Moves the lower 64 bits of a 128-bit vector of [2 x double] twice to
1968 /// the upper and lower 64 bits of a memory location.
1969 ///
1970 /// \headerfile <x86intrin.h>
1971 ///
1972 /// This intrinsic corresponds to the
1973 /// <c> VMOVDDUP + VMOVAPD / MOVLHPS + MOVAPS </c> instruction.
1974 ///
1975 /// \param __dp
1976 /// A pointer to a memory location that can store two double-precision
1977 /// values.
1978 /// \param __a
1979 /// A 128-bit vector of [2 x double] whose lower 64 bits are copied to each
1980 /// of the values in \a __dp.
1981 static __inline__ void __DEFAULT_FN_ATTRS
1982 _mm_store_pd1(double *__dp, __m128d __a)
1983 {
1984  _mm_store1_pd(__dp, __a);
1985 }
1986 
1987 /// Stores a 128-bit vector of [2 x double] into an unaligned memory
1988 /// location.
1989 ///
1990 /// \headerfile <x86intrin.h>
1991 ///
1992 /// This intrinsic corresponds to the <c> VMOVUPD / MOVUPD </c> instruction.
1993 ///
1994 /// \param __dp
1995 /// A pointer to a 128-bit memory location. The address of the memory
1996 /// location does not have to be aligned.
1997 /// \param __a
1998 /// A 128-bit vector of [2 x double] containing the values to be stored.
1999 static __inline__ void __DEFAULT_FN_ATTRS
2000 _mm_storeu_pd(double *__dp, __m128d __a)
2001 {
2002  struct __storeu_pd {
2003  __m128d __v;
2004  } __attribute__((__packed__, __may_alias__));
2005  ((struct __storeu_pd*)__dp)->__v = __a;
2006 }
2007 
2008 /// Stores two double-precision values, in reverse order, from a 128-bit
2009 /// vector of [2 x double] to a 16-byte aligned memory location.
2010 ///
2011 /// \headerfile <x86intrin.h>
2012 ///
2013 /// This intrinsic corresponds to a shuffling instruction followed by a
2014 /// <c> VMOVAPD / MOVAPD </c> instruction.
2015 ///
2016 /// \param __dp
2017 /// A pointer to a 16-byte aligned memory location that can store two
2018 /// double-precision values.
2019 /// \param __a
2020 /// A 128-bit vector of [2 x double] containing the values to be reversed and
2021 /// stored.
2022 static __inline__ void __DEFAULT_FN_ATTRS
2023 _mm_storer_pd(double *__dp, __m128d __a)
2024 {
2025  __a = __builtin_shufflevector((__v2df)__a, (__v2df)__a, 1, 0);
2026  *(__m128d *)__dp = __a;
2027 }
2028 
2029 /// Stores the upper 64 bits of a 128-bit vector of [2 x double] to a
2030 /// memory location.
2031 ///
2032 /// \headerfile <x86intrin.h>
2033 ///
2034 /// This intrinsic corresponds to the <c> VMOVHPD / MOVHPD </c> instruction.
2035 ///
2036 /// \param __dp
2037 /// A pointer to a 64-bit memory location.
2038 /// \param __a
2039 /// A 128-bit vector of [2 x double] containing the value to be stored.
2040 static __inline__ void __DEFAULT_FN_ATTRS
2041 _mm_storeh_pd(double *__dp, __m128d __a)
2042 {
2043  struct __mm_storeh_pd_struct {
2044  double __u;
2045  } __attribute__((__packed__, __may_alias__));
2046  ((struct __mm_storeh_pd_struct*)__dp)->__u = __a[1];
2047 }
2048 
2049 /// Stores the lower 64 bits of a 128-bit vector of [2 x double] to a
2050 /// memory location.
2051 ///
2052 /// \headerfile <x86intrin.h>
2053 ///
2054 /// This intrinsic corresponds to the <c> VMOVLPD / MOVLPD </c> instruction.
2055 ///
2056 /// \param __dp
2057 /// A pointer to a 64-bit memory location.
2058 /// \param __a
2059 /// A 128-bit vector of [2 x double] containing the value to be stored.
2060 static __inline__ void __DEFAULT_FN_ATTRS
2061 _mm_storel_pd(double *__dp, __m128d __a)
2062 {
2063  struct __mm_storeh_pd_struct {
2064  double __u;
2065  } __attribute__((__packed__, __may_alias__));
2066  ((struct __mm_storeh_pd_struct*)__dp)->__u = __a[0];
2067 }
2068 
2069 /// Adds the corresponding elements of two 128-bit vectors of [16 x i8],
2070 /// saving the lower 8 bits of each sum in the corresponding element of a
2071 /// 128-bit result vector of [16 x i8].
2072 ///
2073 /// The integer elements of both parameters can be either signed or unsigned.
2074 ///
2075 /// \headerfile <x86intrin.h>
2076 ///
2077 /// This intrinsic corresponds to the <c> VPADDB / PADDB </c> instruction.
2078 ///
2079 /// \param __a
2080 /// A 128-bit vector of [16 x i8].
2081 /// \param __b
2082 /// A 128-bit vector of [16 x i8].
2083 /// \returns A 128-bit vector of [16 x i8] containing the sums of both
2084 /// parameters.
2085 static __inline__ __m128i __DEFAULT_FN_ATTRS
2086 _mm_add_epi8(__m128i __a, __m128i __b)
2087 {
2088  return (__m128i)((__v16qu)__a + (__v16qu)__b);
2089 }
2090 
2091 /// Adds the corresponding elements of two 128-bit vectors of [8 x i16],
2092 /// saving the lower 16 bits of each sum in the corresponding element of a
2093 /// 128-bit result vector of [8 x i16].
2094 ///
2095 /// The integer elements of both parameters can be either signed or unsigned.
2096 ///
2097 /// \headerfile <x86intrin.h>
2098 ///
2099 /// This intrinsic corresponds to the <c> VPADDW / PADDW </c> instruction.
2100 ///
2101 /// \param __a
2102 /// A 128-bit vector of [8 x i16].
2103 /// \param __b
2104 /// A 128-bit vector of [8 x i16].
2105 /// \returns A 128-bit vector of [8 x i16] containing the sums of both
2106 /// parameters.
2107 static __inline__ __m128i __DEFAULT_FN_ATTRS
2108 _mm_add_epi16(__m128i __a, __m128i __b)
2109 {
2110  return (__m128i)((__v8hu)__a + (__v8hu)__b);
2111 }
2112 
2113 /// Adds the corresponding elements of two 128-bit vectors of [4 x i32],
2114 /// saving the lower 32 bits of each sum in the corresponding element of a
2115 /// 128-bit result vector of [4 x i32].
2116 ///
2117 /// The integer elements of both parameters can be either signed or unsigned.
2118 ///
2119 /// \headerfile <x86intrin.h>
2120 ///
2121 /// This intrinsic corresponds to the <c> VPADDD / PADDD </c> instruction.
2122 ///
2123 /// \param __a
2124 /// A 128-bit vector of [4 x i32].
2125 /// \param __b
2126 /// A 128-bit vector of [4 x i32].
2127 /// \returns A 128-bit vector of [4 x i32] containing the sums of both
2128 /// parameters.
2129 static __inline__ __m128i __DEFAULT_FN_ATTRS
2130 _mm_add_epi32(__m128i __a, __m128i __b)
2131 {
2132  return (__m128i)((__v4su)__a + (__v4su)__b);
2133 }
2134 
2135 /// Adds two signed or unsigned 64-bit integer values, returning the
2136 /// lower 64 bits of the sum.
2137 ///
2138 /// \headerfile <x86intrin.h>
2139 ///
2140 /// This intrinsic corresponds to the <c> PADDQ </c> instruction.
2141 ///
2142 /// \param __a
2143 /// A 64-bit integer.
2144 /// \param __b
2145 /// A 64-bit integer.
2146 /// \returns A 64-bit integer containing the sum of both parameters.
2147 static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
2148 _mm_add_si64(__m64 __a, __m64 __b)
2149 {
2150  return (__m64)__builtin_ia32_paddq((__v1di)__a, (__v1di)__b);
2151 }
2152 
2153 /// Adds the corresponding elements of two 128-bit vectors of [2 x i64],
2154 /// saving the lower 64 bits of each sum in the corresponding element of a
2155 /// 128-bit result vector of [2 x i64].
2156 ///
2157 /// The integer elements of both parameters can be either signed or unsigned.
2158 ///
2159 /// \headerfile <x86intrin.h>
2160 ///
2161 /// This intrinsic corresponds to the <c> VPADDQ / PADDQ </c> instruction.
2162 ///
2163 /// \param __a
2164 /// A 128-bit vector of [2 x i64].
2165 /// \param __b
2166 /// A 128-bit vector of [2 x i64].
2167 /// \returns A 128-bit vector of [2 x i64] containing the sums of both
2168 /// parameters.
2169 static __inline__ __m128i __DEFAULT_FN_ATTRS
2170 _mm_add_epi64(__m128i __a, __m128i __b)
2171 {
2172  return (__m128i)((__v2du)__a + (__v2du)__b);
2173 }
2174 
2175 /// Adds, with saturation, the corresponding elements of two 128-bit
2176 /// signed [16 x i8] vectors, saving each sum in the corresponding element of
2177 /// a 128-bit result vector of [16 x i8]. Positive sums greater than 0x7F are
2178 /// saturated to 0x7F. Negative sums less than 0x80 are saturated to 0x80.
2179 ///
2180 /// \headerfile <x86intrin.h>
2181 ///
2182 /// This intrinsic corresponds to the <c> VPADDSB / PADDSB </c> instruction.
2183 ///
2184 /// \param __a
2185 /// A 128-bit signed [16 x i8] vector.
2186 /// \param __b
2187 /// A 128-bit signed [16 x i8] vector.
2188 /// \returns A 128-bit signed [16 x i8] vector containing the saturated sums of
2189 /// both parameters.
2190 static __inline__ __m128i __DEFAULT_FN_ATTRS
2191 _mm_adds_epi8(__m128i __a, __m128i __b)
2192 {
2193  return (__m128i)__builtin_ia32_paddsb128((__v16qi)__a, (__v16qi)__b);
2194 }
2195 
2196 /// Adds, with saturation, the corresponding elements of two 128-bit
2197 /// signed [8 x i16] vectors, saving each sum in the corresponding element of
2198 /// a 128-bit result vector of [8 x i16]. Positive sums greater than 0x7FFF
2199 /// are saturated to 0x7FFF. Negative sums less than 0x8000 are saturated to
2200 /// 0x8000.
2201 ///
2202 /// \headerfile <x86intrin.h>
2203 ///
2204 /// This intrinsic corresponds to the <c> VPADDSW / PADDSW </c> instruction.
2205 ///
2206 /// \param __a
2207 /// A 128-bit signed [8 x i16] vector.
2208 /// \param __b
2209 /// A 128-bit signed [8 x i16] vector.
2210 /// \returns A 128-bit signed [8 x i16] vector containing the saturated sums of
2211 /// both parameters.
2212 static __inline__ __m128i __DEFAULT_FN_ATTRS
2213 _mm_adds_epi16(__m128i __a, __m128i __b)
2214 {
2215  return (__m128i)__builtin_ia32_paddsw128((__v8hi)__a, (__v8hi)__b);
2216 }
2217 
2218 /// Adds, with saturation, the corresponding elements of two 128-bit
2219 /// unsigned [16 x i8] vectors, saving each sum in the corresponding element
2220 /// of a 128-bit result vector of [16 x i8]. Positive sums greater than 0xFF
2221 /// are saturated to 0xFF. Negative sums are saturated to 0x00.
2222 ///
2223 /// \headerfile <x86intrin.h>
2224 ///
2225 /// This intrinsic corresponds to the <c> VPADDUSB / PADDUSB </c> instruction.
2226 ///
2227 /// \param __a
2228 /// A 128-bit unsigned [16 x i8] vector.
2229 /// \param __b
2230 /// A 128-bit unsigned [16 x i8] vector.
2231 /// \returns A 128-bit unsigned [16 x i8] vector containing the saturated sums
2232 /// of both parameters.
2233 static __inline__ __m128i __DEFAULT_FN_ATTRS
2234 _mm_adds_epu8(__m128i __a, __m128i __b)
2235 {
2236  return (__m128i)__builtin_ia32_paddusb128((__v16qi)__a, (__v16qi)__b);
2237 }
2238 
2239 /// Adds, with saturation, the corresponding elements of two 128-bit
2240 /// unsigned [8 x i16] vectors, saving each sum in the corresponding element
2241 /// of a 128-bit result vector of [8 x i16]. Positive sums greater than
2242 /// 0xFFFF are saturated to 0xFFFF. Negative sums are saturated to 0x0000.
2243 ///
2244 /// \headerfile <x86intrin.h>
2245 ///
2246 /// This intrinsic corresponds to the <c> VPADDUSB / PADDUSB </c> instruction.
2247 ///
2248 /// \param __a
2249 /// A 128-bit unsigned [8 x i16] vector.
2250 /// \param __b
2251 /// A 128-bit unsigned [8 x i16] vector.
2252 /// \returns A 128-bit unsigned [8 x i16] vector containing the saturated sums
2253 /// of both parameters.
2254 static __inline__ __m128i __DEFAULT_FN_ATTRS
2255 _mm_adds_epu16(__m128i __a, __m128i __b)
2256 {
2257  return (__m128i)__builtin_ia32_paddusw128((__v8hi)__a, (__v8hi)__b);
2258 }
2259 
2260 /// Computes the rounded avarages of corresponding elements of two
2261 /// 128-bit unsigned [16 x i8] vectors, saving each result in the
2262 /// corresponding element of a 128-bit result vector of [16 x i8].
2263 ///
2264 /// \headerfile <x86intrin.h>
2265 ///
2266 /// This intrinsic corresponds to the <c> VPAVGB / PAVGB </c> instruction.
2267 ///
2268 /// \param __a
2269 /// A 128-bit unsigned [16 x i8] vector.
2270 /// \param __b
2271 /// A 128-bit unsigned [16 x i8] vector.
2272 /// \returns A 128-bit unsigned [16 x i8] vector containing the rounded
2273 /// averages of both parameters.
2274 static __inline__ __m128i __DEFAULT_FN_ATTRS
2275 _mm_avg_epu8(__m128i __a, __m128i __b)
2276 {
2277  typedef unsigned short __v16hu __attribute__ ((__vector_size__ (32)));
2278  return (__m128i)__builtin_convertvector(
2279  ((__builtin_convertvector((__v16qu)__a, __v16hu) +
2280  __builtin_convertvector((__v16qu)__b, __v16hu)) + 1)
2281  >> 1, __v16qu);
2282 }
2283 
2284 /// Computes the rounded avarages of corresponding elements of two
2285 /// 128-bit unsigned [8 x i16] vectors, saving each result in the
2286 /// corresponding element of a 128-bit result vector of [8 x i16].
2287 ///
2288 /// \headerfile <x86intrin.h>
2289 ///
2290 /// This intrinsic corresponds to the <c> VPAVGW / PAVGW </c> instruction.
2291 ///
2292 /// \param __a
2293 /// A 128-bit unsigned [8 x i16] vector.
2294 /// \param __b
2295 /// A 128-bit unsigned [8 x i16] vector.
2296 /// \returns A 128-bit unsigned [8 x i16] vector containing the rounded
2297 /// averages of both parameters.
2298 static __inline__ __m128i __DEFAULT_FN_ATTRS
2299 _mm_avg_epu16(__m128i __a, __m128i __b)
2300 {
2301  typedef unsigned int __v8su __attribute__ ((__vector_size__ (32)));
2302  return (__m128i)__builtin_convertvector(
2303  ((__builtin_convertvector((__v8hu)__a, __v8su) +
2304  __builtin_convertvector((__v8hu)__b, __v8su)) + 1)
2305  >> 1, __v8hu);
2306 }
2307 
2308 /// Multiplies the corresponding elements of two 128-bit signed [8 x i16]
2309 /// vectors, producing eight intermediate 32-bit signed integer products, and
2310 /// adds the consecutive pairs of 32-bit products to form a 128-bit signed
2311 /// [4 x i32] vector.
2312 ///
2313 /// For example, bits [15:0] of both parameters are multiplied producing a
2314 /// 32-bit product, bits [31:16] of both parameters are multiplied producing
2315 /// a 32-bit product, and the sum of those two products becomes bits [31:0]
2316 /// of the result.
2317 ///
2318 /// \headerfile <x86intrin.h>
2319 ///
2320 /// This intrinsic corresponds to the <c> VPMADDWD / PMADDWD </c> instruction.
2321 ///
2322 /// \param __a
2323 /// A 128-bit signed [8 x i16] vector.
2324 /// \param __b
2325 /// A 128-bit signed [8 x i16] vector.
2326 /// \returns A 128-bit signed [4 x i32] vector containing the sums of products
2327 /// of both parameters.
2328 static __inline__ __m128i __DEFAULT_FN_ATTRS
2329 _mm_madd_epi16(__m128i __a, __m128i __b)
2330 {
2331  return (__m128i)__builtin_ia32_pmaddwd128((__v8hi)__a, (__v8hi)__b);
2332 }
2333 
2334 /// Compares corresponding elements of two 128-bit signed [8 x i16]
2335 /// vectors, saving the greater value from each comparison in the
2336 /// corresponding element of a 128-bit result vector of [8 x i16].
2337 ///
2338 /// \headerfile <x86intrin.h>
2339 ///
2340 /// This intrinsic corresponds to the <c> VPMAXSW / PMAXSW </c> instruction.
2341 ///
2342 /// \param __a
2343 /// A 128-bit signed [8 x i16] vector.
2344 /// \param __b
2345 /// A 128-bit signed [8 x i16] vector.
2346 /// \returns A 128-bit signed [8 x i16] vector containing the greater value of
2347 /// each comparison.
2348 static __inline__ __m128i __DEFAULT_FN_ATTRS
2349 _mm_max_epi16(__m128i __a, __m128i __b)
2350 {
2351  return (__m128i)__builtin_ia32_pmaxsw128((__v8hi)__a, (__v8hi)__b);
2352 }
2353 
2354 /// Compares corresponding elements of two 128-bit unsigned [16 x i8]
2355 /// vectors, saving the greater value from each comparison in the
2356 /// corresponding element of a 128-bit result vector of [16 x i8].
2357 ///
2358 /// \headerfile <x86intrin.h>
2359 ///
2360 /// This intrinsic corresponds to the <c> VPMAXUB / PMAXUB </c> instruction.
2361 ///
2362 /// \param __a
2363 /// A 128-bit unsigned [16 x i8] vector.
2364 /// \param __b
2365 /// A 128-bit unsigned [16 x i8] vector.
2366 /// \returns A 128-bit unsigned [16 x i8] vector containing the greater value of
2367 /// each comparison.
2368 static __inline__ __m128i __DEFAULT_FN_ATTRS
2369 _mm_max_epu8(__m128i __a, __m128i __b)
2370 {
2371  return (__m128i)__builtin_ia32_pmaxub128((__v16qi)__a, (__v16qi)__b);
2372 }
2373 
2374 /// Compares corresponding elements of two 128-bit signed [8 x i16]
2375 /// vectors, saving the smaller value from each comparison in the
2376 /// corresponding element of a 128-bit result vector of [8 x i16].
2377 ///
2378 /// \headerfile <x86intrin.h>
2379 ///
2380 /// This intrinsic corresponds to the <c> VPMINSW / PMINSW </c> instruction.
2381 ///
2382 /// \param __a
2383 /// A 128-bit signed [8 x i16] vector.
2384 /// \param __b
2385 /// A 128-bit signed [8 x i16] vector.
2386 /// \returns A 128-bit signed [8 x i16] vector containing the smaller value of
2387 /// each comparison.
2388 static __inline__ __m128i __DEFAULT_FN_ATTRS
2389 _mm_min_epi16(__m128i __a, __m128i __b)
2390 {
2391  return (__m128i)__builtin_ia32_pminsw128((__v8hi)__a, (__v8hi)__b);
2392 }
2393 
2394 /// Compares corresponding elements of two 128-bit unsigned [16 x i8]
2395 /// vectors, saving the smaller value from each comparison in the
2396 /// corresponding element of a 128-bit result vector of [16 x i8].
2397 ///
2398 /// \headerfile <x86intrin.h>
2399 ///
2400 /// This intrinsic corresponds to the <c> VPMINUB / PMINUB </c> instruction.
2401 ///
2402 /// \param __a
2403 /// A 128-bit unsigned [16 x i8] vector.
2404 /// \param __b
2405 /// A 128-bit unsigned [16 x i8] vector.
2406 /// \returns A 128-bit unsigned [16 x i8] vector containing the smaller value of
2407 /// each comparison.
2408 static __inline__ __m128i __DEFAULT_FN_ATTRS
2409 _mm_min_epu8(__m128i __a, __m128i __b)
2410 {
2411  return (__m128i)__builtin_ia32_pminub128((__v16qi)__a, (__v16qi)__b);
2412 }
2413 
2414 /// Multiplies the corresponding elements of two signed [8 x i16]
2415 /// vectors, saving the upper 16 bits of each 32-bit product in the
2416 /// corresponding element of a 128-bit signed [8 x i16] result vector.
2417 ///
2418 /// \headerfile <x86intrin.h>
2419 ///
2420 /// This intrinsic corresponds to the <c> VPMULHW / PMULHW </c> instruction.
2421 ///
2422 /// \param __a
2423 /// A 128-bit signed [8 x i16] vector.
2424 /// \param __b
2425 /// A 128-bit signed [8 x i16] vector.
2426 /// \returns A 128-bit signed [8 x i16] vector containing the upper 16 bits of
2427 /// each of the eight 32-bit products.
2428 static __inline__ __m128i __DEFAULT_FN_ATTRS
2429 _mm_mulhi_epi16(__m128i __a, __m128i __b)
2430 {
2431  return (__m128i)__builtin_ia32_pmulhw128((__v8hi)__a, (__v8hi)__b);
2432 }
2433 
2434 /// Multiplies the corresponding elements of two unsigned [8 x i16]
2435 /// vectors, saving the upper 16 bits of each 32-bit product in the
2436 /// corresponding element of a 128-bit unsigned [8 x i16] result vector.
2437 ///
2438 /// \headerfile <x86intrin.h>
2439 ///
2440 /// This intrinsic corresponds to the <c> VPMULHUW / PMULHUW </c> instruction.
2441 ///
2442 /// \param __a
2443 /// A 128-bit unsigned [8 x i16] vector.
2444 /// \param __b
2445 /// A 128-bit unsigned [8 x i16] vector.
2446 /// \returns A 128-bit unsigned [8 x i16] vector containing the upper 16 bits
2447 /// of each of the eight 32-bit products.
2448 static __inline__ __m128i __DEFAULT_FN_ATTRS
2449 _mm_mulhi_epu16(__m128i __a, __m128i __b)
2450 {
2451  return (__m128i)__builtin_ia32_pmulhuw128((__v8hi)__a, (__v8hi)__b);
2452 }
2453 
2454 /// Multiplies the corresponding elements of two signed [8 x i16]
2455 /// vectors, saving the lower 16 bits of each 32-bit product in the
2456 /// corresponding element of a 128-bit signed [8 x i16] result vector.
2457 ///
2458 /// \headerfile <x86intrin.h>
2459 ///
2460 /// This intrinsic corresponds to the <c> VPMULLW / PMULLW </c> instruction.
2461 ///
2462 /// \param __a
2463 /// A 128-bit signed [8 x i16] vector.
2464 /// \param __b
2465 /// A 128-bit signed [8 x i16] vector.
2466 /// \returns A 128-bit signed [8 x i16] vector containing the lower 16 bits of
2467 /// each of the eight 32-bit products.
2468 static __inline__ __m128i __DEFAULT_FN_ATTRS
2469 _mm_mullo_epi16(__m128i __a, __m128i __b)
2470 {
2471  return (__m128i)((__v8hu)__a * (__v8hu)__b);
2472 }
2473 
2474 /// Multiplies 32-bit unsigned integer values contained in the lower bits
2475 /// of the two 64-bit integer vectors and returns the 64-bit unsigned
2476 /// product.
2477 ///
2478 /// \headerfile <x86intrin.h>
2479 ///
2480 /// This intrinsic corresponds to the <c> PMULUDQ </c> instruction.
2481 ///
2482 /// \param __a
2483 /// A 64-bit integer containing one of the source operands.
2484 /// \param __b
2485 /// A 64-bit integer containing one of the source operands.
2486 /// \returns A 64-bit integer vector containing the product of both operands.
2487 static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
2488 _mm_mul_su32(__m64 __a, __m64 __b)
2489 {
2490  return __builtin_ia32_pmuludq((__v2si)__a, (__v2si)__b);
2491 }
2492 
2493 /// Multiplies 32-bit unsigned integer values contained in the lower
2494 /// bits of the corresponding elements of two [2 x i64] vectors, and returns
2495 /// the 64-bit products in the corresponding elements of a [2 x i64] vector.
2496 ///
2497 /// \headerfile <x86intrin.h>
2498 ///
2499 /// This intrinsic corresponds to the <c> VPMULUDQ / PMULUDQ </c> instruction.
2500 ///
2501 /// \param __a
2502 /// A [2 x i64] vector containing one of the source operands.
2503 /// \param __b
2504 /// A [2 x i64] vector containing one of the source operands.
2505 /// \returns A [2 x i64] vector containing the product of both operands.
2506 static __inline__ __m128i __DEFAULT_FN_ATTRS
2507 _mm_mul_epu32(__m128i __a, __m128i __b)
2508 {
2509  return __builtin_ia32_pmuludq128((__v4si)__a, (__v4si)__b);
2510 }
2511 
2512 /// Computes the absolute differences of corresponding 8-bit integer
2513 /// values in two 128-bit vectors. Sums the first 8 absolute differences, and
2514 /// separately sums the second 8 absolute differences. Packs these two
2515 /// unsigned 16-bit integer sums into the upper and lower elements of a
2516 /// [2 x i64] vector.
2517 ///
2518 /// \headerfile <x86intrin.h>
2519 ///
2520 /// This intrinsic corresponds to the <c> VPSADBW / PSADBW </c> instruction.
2521 ///
2522 /// \param __a
2523 /// A 128-bit integer vector containing one of the source operands.
2524 /// \param __b
2525 /// A 128-bit integer vector containing one of the source operands.
2526 /// \returns A [2 x i64] vector containing the sums of the sets of absolute
2527 /// differences between both operands.
2528 static __inline__ __m128i __DEFAULT_FN_ATTRS
2529 _mm_sad_epu8(__m128i __a, __m128i __b)
2530 {
2531  return __builtin_ia32_psadbw128((__v16qi)__a, (__v16qi)__b);
2532 }
2533 
2534 /// Subtracts the corresponding 8-bit integer values in the operands.
2535 ///
2536 /// \headerfile <x86intrin.h>
2537 ///
2538 /// This intrinsic corresponds to the <c> VPSUBB / PSUBB </c> instruction.
2539 ///
2540 /// \param __a
2541 /// A 128-bit integer vector containing the minuends.
2542 /// \param __b
2543 /// A 128-bit integer vector containing the subtrahends.
2544 /// \returns A 128-bit integer vector containing the differences of the values
2545 /// in the operands.
2546 static __inline__ __m128i __DEFAULT_FN_ATTRS
2547 _mm_sub_epi8(__m128i __a, __m128i __b)
2548 {
2549  return (__m128i)((__v16qu)__a - (__v16qu)__b);
2550 }
2551 
2552 /// Subtracts the corresponding 16-bit integer values in the operands.
2553 ///
2554 /// \headerfile <x86intrin.h>
2555 ///
2556 /// This intrinsic corresponds to the <c> VPSUBW / PSUBW </c> instruction.
2557 ///
2558 /// \param __a
2559 /// A 128-bit integer vector containing the minuends.
2560 /// \param __b
2561 /// A 128-bit integer vector containing the subtrahends.
2562 /// \returns A 128-bit integer vector containing the differences of the values
2563 /// in the operands.
2564 static __inline__ __m128i __DEFAULT_FN_ATTRS
2565 _mm_sub_epi16(__m128i __a, __m128i __b)
2566 {
2567  return (__m128i)((__v8hu)__a - (__v8hu)__b);
2568 }
2569 
2570 /// Subtracts the corresponding 32-bit integer values in the operands.
2571 ///
2572 /// \headerfile <x86intrin.h>
2573 ///
2574 /// This intrinsic corresponds to the <c> VPSUBD / PSUBD </c> instruction.
2575 ///
2576 /// \param __a
2577 /// A 128-bit integer vector containing the minuends.
2578 /// \param __b
2579 /// A 128-bit integer vector containing the subtrahends.
2580 /// \returns A 128-bit integer vector containing the differences of the values
2581 /// in the operands.
2582 static __inline__ __m128i __DEFAULT_FN_ATTRS
2583 _mm_sub_epi32(__m128i __a, __m128i __b)
2584 {
2585  return (__m128i)((__v4su)__a - (__v4su)__b);
2586 }
2587 
2588 /// Subtracts signed or unsigned 64-bit integer values and writes the
2589 /// difference to the corresponding bits in the destination.
2590 ///
2591 /// \headerfile <x86intrin.h>
2592 ///
2593 /// This intrinsic corresponds to the <c> PSUBQ </c> instruction.
2594 ///
2595 /// \param __a
2596 /// A 64-bit integer vector containing the minuend.
2597 /// \param __b
2598 /// A 64-bit integer vector containing the subtrahend.
2599 /// \returns A 64-bit integer vector containing the difference of the values in
2600 /// the operands.
2601 static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
2602 _mm_sub_si64(__m64 __a, __m64 __b)
2603 {
2604  return (__m64)__builtin_ia32_psubq((__v1di)__a, (__v1di)__b);
2605 }
2606 
2607 /// Subtracts the corresponding elements of two [2 x i64] vectors.
2608 ///
2609 /// \headerfile <x86intrin.h>
2610 ///
2611 /// This intrinsic corresponds to the <c> VPSUBQ / PSUBQ </c> instruction.
2612 ///
2613 /// \param __a
2614 /// A 128-bit integer vector containing the minuends.
2615 /// \param __b
2616 /// A 128-bit integer vector containing the subtrahends.
2617 /// \returns A 128-bit integer vector containing the differences of the values
2618 /// in the operands.
2619 static __inline__ __m128i __DEFAULT_FN_ATTRS
2620 _mm_sub_epi64(__m128i __a, __m128i __b)
2621 {
2622  return (__m128i)((__v2du)__a - (__v2du)__b);
2623 }
2624 
2625 /// Subtracts corresponding 8-bit signed integer values in the input and
2626 /// returns the differences in the corresponding bytes in the destination.
2627 /// Differences greater than 0x7F are saturated to 0x7F, and differences less
2628 /// than 0x80 are saturated to 0x80.
2629 ///
2630 /// \headerfile <x86intrin.h>
2631 ///
2632 /// This intrinsic corresponds to the <c> VPSUBSB / PSUBSB </c> instruction.
2633 ///
2634 /// \param __a
2635 /// A 128-bit integer vector containing the minuends.
2636 /// \param __b
2637 /// A 128-bit integer vector containing the subtrahends.
2638 /// \returns A 128-bit integer vector containing the differences of the values
2639 /// in the operands.
2640 static __inline__ __m128i __DEFAULT_FN_ATTRS
2641 _mm_subs_epi8(__m128i __a, __m128i __b)
2642 {
2643  return (__m128i)__builtin_ia32_psubsb128((__v16qi)__a, (__v16qi)__b);
2644 }
2645 
2646 /// Subtracts corresponding 16-bit signed integer values in the input and
2647 /// returns the differences in the corresponding bytes in the destination.
2648 /// Differences greater than 0x7FFF are saturated to 0x7FFF, and values less
2649 /// than 0x8000 are saturated to 0x8000.
2650 ///
2651 /// \headerfile <x86intrin.h>
2652 ///
2653 /// This intrinsic corresponds to the <c> VPSUBSW / PSUBSW </c> instruction.
2654 ///
2655 /// \param __a
2656 /// A 128-bit integer vector containing the minuends.
2657 /// \param __b
2658 /// A 128-bit integer vector containing the subtrahends.
2659 /// \returns A 128-bit integer vector containing the differences of the values
2660 /// in the operands.
2661 static __inline__ __m128i __DEFAULT_FN_ATTRS
2662 _mm_subs_epi16(__m128i __a, __m128i __b)
2663 {
2664  return (__m128i)__builtin_ia32_psubsw128((__v8hi)__a, (__v8hi)__b);
2665 }
2666 
2667 /// Subtracts corresponding 8-bit unsigned integer values in the input
2668 /// and returns the differences in the corresponding bytes in the
2669 /// destination. Differences less than 0x00 are saturated to 0x00.
2670 ///
2671 /// \headerfile <x86intrin.h>
2672 ///
2673 /// This intrinsic corresponds to the <c> VPSUBUSB / PSUBUSB </c> instruction.
2674 ///
2675 /// \param __a
2676 /// A 128-bit integer vector containing the minuends.
2677 /// \param __b
2678 /// A 128-bit integer vector containing the subtrahends.
2679 /// \returns A 128-bit integer vector containing the unsigned integer
2680 /// differences of the values in the operands.
2681 static __inline__ __m128i __DEFAULT_FN_ATTRS
2682 _mm_subs_epu8(__m128i __a, __m128i __b)
2683 {
2684  return (__m128i)__builtin_ia32_psubusb128((__v16qi)__a, (__v16qi)__b);
2685 }
2686 
2687 /// Subtracts corresponding 16-bit unsigned integer values in the input
2688 /// and returns the differences in the corresponding bytes in the
2689 /// destination. Differences less than 0x0000 are saturated to 0x0000.
2690 ///
2691 /// \headerfile <x86intrin.h>
2692 ///
2693 /// This intrinsic corresponds to the <c> VPSUBUSW / PSUBUSW </c> instruction.
2694 ///
2695 /// \param __a
2696 /// A 128-bit integer vector containing the minuends.
2697 /// \param __b
2698 /// A 128-bit integer vector containing the subtrahends.
2699 /// \returns A 128-bit integer vector containing the unsigned integer
2700 /// differences of the values in the operands.
2701 static __inline__ __m128i __DEFAULT_FN_ATTRS
2702 _mm_subs_epu16(__m128i __a, __m128i __b)
2703 {
2704  return (__m128i)__builtin_ia32_psubusw128((__v8hi)__a, (__v8hi)__b);
2705 }
2706 
2707 /// Performs a bitwise AND of two 128-bit integer vectors.
2708 ///
2709 /// \headerfile <x86intrin.h>
2710 ///
2711 /// This intrinsic corresponds to the <c> VPAND / PAND </c> instruction.
2712 ///
2713 /// \param __a
2714 /// A 128-bit integer vector containing one of the source operands.
2715 /// \param __b
2716 /// A 128-bit integer vector containing one of the source operands.
2717 /// \returns A 128-bit integer vector containing the bitwise AND of the values
2718 /// in both operands.
2719 static __inline__ __m128i __DEFAULT_FN_ATTRS
2720 _mm_and_si128(__m128i __a, __m128i __b)
2721 {
2722  return (__m128i)((__v2du)__a & (__v2du)__b);
2723 }
2724 
2725 /// Performs a bitwise AND of two 128-bit integer vectors, using the
2726 /// one's complement of the values contained in the first source operand.
2727 ///
2728 /// \headerfile <x86intrin.h>
2729 ///
2730 /// This intrinsic corresponds to the <c> VPANDN / PANDN </c> instruction.
2731 ///
2732 /// \param __a
2733 /// A 128-bit vector containing the left source operand. The one's complement
2734 /// of this value is used in the bitwise AND.
2735 /// \param __b
2736 /// A 128-bit vector containing the right source operand.
2737 /// \returns A 128-bit integer vector containing the bitwise AND of the one's
2738 /// complement of the first operand and the values in the second operand.
2739 static __inline__ __m128i __DEFAULT_FN_ATTRS
2740 _mm_andnot_si128(__m128i __a, __m128i __b)
2741 {
2742  return (__m128i)(~(__v2du)__a & (__v2du)__b);
2743 }
2744 /// Performs a bitwise OR of two 128-bit integer vectors.
2745 ///
2746 /// \headerfile <x86intrin.h>
2747 ///
2748 /// This intrinsic corresponds to the <c> VPOR / POR </c> instruction.
2749 ///
2750 /// \param __a
2751 /// A 128-bit integer vector containing one of the source operands.
2752 /// \param __b
2753 /// A 128-bit integer vector containing one of the source operands.
2754 /// \returns A 128-bit integer vector containing the bitwise OR of the values
2755 /// in both operands.
2756 static __inline__ __m128i __DEFAULT_FN_ATTRS
2757 _mm_or_si128(__m128i __a, __m128i __b)
2758 {
2759  return (__m128i)((__v2du)__a | (__v2du)__b);
2760 }
2761 
2762 /// Performs a bitwise exclusive OR of two 128-bit integer vectors.
2763 ///
2764 /// \headerfile <x86intrin.h>
2765 ///
2766 /// This intrinsic corresponds to the <c> VPXOR / PXOR </c> instruction.
2767 ///
2768 /// \param __a
2769 /// A 128-bit integer vector containing one of the source operands.
2770 /// \param __b
2771 /// A 128-bit integer vector containing one of the source operands.
2772 /// \returns A 128-bit integer vector containing the bitwise exclusive OR of the
2773 /// values in both operands.
2774 static __inline__ __m128i __DEFAULT_FN_ATTRS
2775 _mm_xor_si128(__m128i __a, __m128i __b)
2776 {
2777  return (__m128i)((__v2du)__a ^ (__v2du)__b);
2778 }
2779 
2780 /// Left-shifts the 128-bit integer vector operand by the specified
2781 /// number of bytes. Low-order bits are cleared.
2782 ///
2783 /// \headerfile <x86intrin.h>
2784 ///
2785 /// \code
2786 /// __m128i _mm_slli_si128(__m128i a, const int imm);
2787 /// \endcode
2788 ///
2789 /// This intrinsic corresponds to the <c> VPSLLDQ / PSLLDQ </c> instruction.
2790 ///
2791 /// \param a
2792 /// A 128-bit integer vector containing the source operand.
2793 /// \param imm
2794 /// An immediate value specifying the number of bytes to left-shift operand
2795 /// \a a.
2796 /// \returns A 128-bit integer vector containing the left-shifted value.
2797 #define _mm_slli_si128(a, imm) \
2798  (__m128i)__builtin_ia32_pslldqi128_byteshift((__v2di)(__m128i)(a), (int)(imm))
2799 
2800 #define _mm_bslli_si128(a, imm) \
2801  (__m128i)__builtin_ia32_pslldqi128_byteshift((__v2di)(__m128i)(a), (int)(imm))
2802 
2803 /// Left-shifts each 16-bit value in the 128-bit integer vector operand
2804 /// by the specified number of bits. Low-order bits are cleared.
2805 ///
2806 /// \headerfile <x86intrin.h>
2807 ///
2808 /// This intrinsic corresponds to the <c> VPSLLW / PSLLW </c> instruction.
2809 ///
2810 /// \param __a
2811 /// A 128-bit integer vector containing the source operand.
2812 /// \param __count
2813 /// An integer value specifying the number of bits to left-shift each value
2814 /// in operand \a __a.
2815 /// \returns A 128-bit integer vector containing the left-shifted values.
2816 static __inline__ __m128i __DEFAULT_FN_ATTRS
2817 _mm_slli_epi16(__m128i __a, int __count)
2818 {
2819  return (__m128i)__builtin_ia32_psllwi128((__v8hi)__a, __count);
2820 }
2821 
2822 /// Left-shifts each 16-bit value in the 128-bit integer vector operand
2823 /// by the specified number of bits. Low-order bits are cleared.
2824 ///
2825 /// \headerfile <x86intrin.h>
2826 ///
2827 /// This intrinsic corresponds to the <c> VPSLLW / PSLLW </c> instruction.
2828 ///
2829 /// \param __a
2830 /// A 128-bit integer vector containing the source operand.
2831 /// \param __count
2832 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
2833 /// to left-shift each value in operand \a __a.
2834 /// \returns A 128-bit integer vector containing the left-shifted values.
2835 static __inline__ __m128i __DEFAULT_FN_ATTRS
2836 _mm_sll_epi16(__m128i __a, __m128i __count)
2837 {
2838  return (__m128i)__builtin_ia32_psllw128((__v8hi)__a, (__v8hi)__count);
2839 }
2840 
2841 /// Left-shifts each 32-bit value in the 128-bit integer vector operand
2842 /// by the specified number of bits. Low-order bits are cleared.
2843 ///
2844 /// \headerfile <x86intrin.h>
2845 ///
2846 /// This intrinsic corresponds to the <c> VPSLLD / PSLLD </c> instruction.
2847 ///
2848 /// \param __a
2849 /// A 128-bit integer vector containing the source operand.
2850 /// \param __count
2851 /// An integer value specifying the number of bits to left-shift each value
2852 /// in operand \a __a.
2853 /// \returns A 128-bit integer vector containing the left-shifted values.
2854 static __inline__ __m128i __DEFAULT_FN_ATTRS
2855 _mm_slli_epi32(__m128i __a, int __count)
2856 {
2857  return (__m128i)__builtin_ia32_pslldi128((__v4si)__a, __count);
2858 }
2859 
2860 /// Left-shifts each 32-bit value in the 128-bit integer vector operand
2861 /// by the specified number of bits. Low-order bits are cleared.
2862 ///
2863 /// \headerfile <x86intrin.h>
2864 ///
2865 /// This intrinsic corresponds to the <c> VPSLLD / PSLLD </c> instruction.
2866 ///
2867 /// \param __a
2868 /// A 128-bit integer vector containing the source operand.
2869 /// \param __count
2870 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
2871 /// to left-shift each value in operand \a __a.
2872 /// \returns A 128-bit integer vector containing the left-shifted values.
2873 static __inline__ __m128i __DEFAULT_FN_ATTRS
2874 _mm_sll_epi32(__m128i __a, __m128i __count)
2875 {
2876  return (__m128i)__builtin_ia32_pslld128((__v4si)__a, (__v4si)__count);
2877 }
2878 
2879 /// Left-shifts each 64-bit value in the 128-bit integer vector operand
2880 /// by the specified number of bits. Low-order bits are cleared.
2881 ///
2882 /// \headerfile <x86intrin.h>
2883 ///
2884 /// This intrinsic corresponds to the <c> VPSLLQ / PSLLQ </c> instruction.
2885 ///
2886 /// \param __a
2887 /// A 128-bit integer vector containing the source operand.
2888 /// \param __count
2889 /// An integer value specifying the number of bits to left-shift each value
2890 /// in operand \a __a.
2891 /// \returns A 128-bit integer vector containing the left-shifted values.
2892 static __inline__ __m128i __DEFAULT_FN_ATTRS
2893 _mm_slli_epi64(__m128i __a, int __count)
2894 {
2895  return __builtin_ia32_psllqi128((__v2di)__a, __count);
2896 }
2897 
2898 /// Left-shifts each 64-bit value in the 128-bit integer vector operand
2899 /// by the specified number of bits. Low-order bits are cleared.
2900 ///
2901 /// \headerfile <x86intrin.h>
2902 ///
2903 /// This intrinsic corresponds to the <c> VPSLLQ / PSLLQ </c> instruction.
2904 ///
2905 /// \param __a
2906 /// A 128-bit integer vector containing the source operand.
2907 /// \param __count
2908 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
2909 /// to left-shift each value in operand \a __a.
2910 /// \returns A 128-bit integer vector containing the left-shifted values.
2911 static __inline__ __m128i __DEFAULT_FN_ATTRS
2912 _mm_sll_epi64(__m128i __a, __m128i __count)
2913 {
2914  return __builtin_ia32_psllq128((__v2di)__a, (__v2di)__count);
2915 }
2916 
2917 /// Right-shifts each 16-bit value in the 128-bit integer vector operand
2918 /// by the specified number of bits. High-order bits are filled with the sign
2919 /// bit of the initial value.
2920 ///
2921 /// \headerfile <x86intrin.h>
2922 ///
2923 /// This intrinsic corresponds to the <c> VPSRAW / PSRAW </c> instruction.
2924 ///
2925 /// \param __a
2926 /// A 128-bit integer vector containing the source operand.
2927 /// \param __count
2928 /// An integer value specifying the number of bits to right-shift each value
2929 /// in operand \a __a.
2930 /// \returns A 128-bit integer vector containing the right-shifted values.
2931 static __inline__ __m128i __DEFAULT_FN_ATTRS
2932 _mm_srai_epi16(__m128i __a, int __count)
2933 {
2934  return (__m128i)__builtin_ia32_psrawi128((__v8hi)__a, __count);
2935 }
2936 
2937 /// Right-shifts each 16-bit value in the 128-bit integer vector operand
2938 /// by the specified number of bits. High-order bits are filled with the sign
2939 /// bit of the initial value.
2940 ///
2941 /// \headerfile <x86intrin.h>
2942 ///
2943 /// This intrinsic corresponds to the <c> VPSRAW / PSRAW </c> instruction.
2944 ///
2945 /// \param __a
2946 /// A 128-bit integer vector containing the source operand.
2947 /// \param __count
2948 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
2949 /// to right-shift each value in operand \a __a.
2950 /// \returns A 128-bit integer vector containing the right-shifted values.
2951 static __inline__ __m128i __DEFAULT_FN_ATTRS
2952 _mm_sra_epi16(__m128i __a, __m128i __count)
2953 {
2954  return (__m128i)__builtin_ia32_psraw128((__v8hi)__a, (__v8hi)__count);
2955 }
2956 
2957 /// Right-shifts each 32-bit value in the 128-bit integer vector operand
2958 /// by the specified number of bits. High-order bits are filled with the sign
2959 /// bit of the initial value.
2960 ///
2961 /// \headerfile <x86intrin.h>
2962 ///
2963 /// This intrinsic corresponds to the <c> VPSRAD / PSRAD </c> instruction.
2964 ///
2965 /// \param __a
2966 /// A 128-bit integer vector containing the source operand.
2967 /// \param __count
2968 /// An integer value specifying the number of bits to right-shift each value
2969 /// in operand \a __a.
2970 /// \returns A 128-bit integer vector containing the right-shifted values.
2971 static __inline__ __m128i __DEFAULT_FN_ATTRS
2972 _mm_srai_epi32(__m128i __a, int __count)
2973 {
2974  return (__m128i)__builtin_ia32_psradi128((__v4si)__a, __count);
2975 }
2976 
2977 /// Right-shifts each 32-bit value in the 128-bit integer vector operand
2978 /// by the specified number of bits. High-order bits are filled with the sign
2979 /// bit of the initial value.
2980 ///
2981 /// \headerfile <x86intrin.h>
2982 ///
2983 /// This intrinsic corresponds to the <c> VPSRAD / PSRAD </c> instruction.
2984 ///
2985 /// \param __a
2986 /// A 128-bit integer vector containing the source operand.
2987 /// \param __count
2988 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
2989 /// to right-shift each value in operand \a __a.
2990 /// \returns A 128-bit integer vector containing the right-shifted values.
2991 static __inline__ __m128i __DEFAULT_FN_ATTRS
2992 _mm_sra_epi32(__m128i __a, __m128i __count)
2993 {
2994  return (__m128i)__builtin_ia32_psrad128((__v4si)__a, (__v4si)__count);
2995 }
2996 
2997 /// Right-shifts the 128-bit integer vector operand by the specified
2998 /// number of bytes. High-order bits are cleared.
2999 ///
3000 /// \headerfile <x86intrin.h>
3001 ///
3002 /// \code
3003 /// __m128i _mm_srli_si128(__m128i a, const int imm);
3004 /// \endcode
3005 ///
3006 /// This intrinsic corresponds to the <c> VPSRLDQ / PSRLDQ </c> instruction.
3007 ///
3008 /// \param a
3009 /// A 128-bit integer vector containing the source operand.
3010 /// \param imm
3011 /// An immediate value specifying the number of bytes to right-shift operand
3012 /// \a a.
3013 /// \returns A 128-bit integer vector containing the right-shifted value.
3014 #define _mm_srli_si128(a, imm) \
3015  (__m128i)__builtin_ia32_psrldqi128_byteshift((__v2di)(__m128i)(a), (int)(imm))
3016 
3017 #define _mm_bsrli_si128(a, imm) \
3018  (__m128i)__builtin_ia32_psrldqi128_byteshift((__v2di)(__m128i)(a), (int)(imm))
3019 
3020 /// Right-shifts each of 16-bit values in the 128-bit integer vector
3021 /// operand by the specified number of bits. High-order bits are cleared.
3022 ///
3023 /// \headerfile <x86intrin.h>
3024 ///
3025 /// This intrinsic corresponds to the <c> VPSRLW / PSRLW </c> instruction.
3026 ///
3027 /// \param __a
3028 /// A 128-bit integer vector containing the source operand.
3029 /// \param __count
3030 /// An integer value specifying the number of bits to right-shift each value
3031 /// in operand \a __a.
3032 /// \returns A 128-bit integer vector containing the right-shifted values.
3033 static __inline__ __m128i __DEFAULT_FN_ATTRS
3034 _mm_srli_epi16(__m128i __a, int __count)
3035 {
3036  return (__m128i)__builtin_ia32_psrlwi128((__v8hi)__a, __count);
3037 }
3038 
3039 /// Right-shifts each of 16-bit values in the 128-bit integer vector
3040 /// operand by the specified number of bits. High-order bits are cleared.
3041 ///
3042 /// \headerfile <x86intrin.h>
3043 ///
3044 /// This intrinsic corresponds to the <c> VPSRLW / PSRLW </c> instruction.
3045 ///
3046 /// \param __a
3047 /// A 128-bit integer vector containing the source operand.
3048 /// \param __count
3049 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
3050 /// to right-shift each value in operand \a __a.
3051 /// \returns A 128-bit integer vector containing the right-shifted values.
3052 static __inline__ __m128i __DEFAULT_FN_ATTRS
3053 _mm_srl_epi16(__m128i __a, __m128i __count)
3054 {
3055  return (__m128i)__builtin_ia32_psrlw128((__v8hi)__a, (__v8hi)__count);
3056 }
3057 
3058 /// Right-shifts each of 32-bit values in the 128-bit integer vector
3059 /// operand by the specified number of bits. High-order bits are cleared.
3060 ///
3061 /// \headerfile <x86intrin.h>
3062 ///
3063 /// This intrinsic corresponds to the <c> VPSRLD / PSRLD </c> instruction.
3064 ///
3065 /// \param __a
3066 /// A 128-bit integer vector containing the source operand.
3067 /// \param __count
3068 /// An integer value specifying the number of bits to right-shift each value
3069 /// in operand \a __a.
3070 /// \returns A 128-bit integer vector containing the right-shifted values.
3071 static __inline__ __m128i __DEFAULT_FN_ATTRS
3072 _mm_srli_epi32(__m128i __a, int __count)
3073 {
3074  return (__m128i)__builtin_ia32_psrldi128((__v4si)__a, __count);
3075 }
3076 
3077 /// Right-shifts each of 32-bit values in the 128-bit integer vector
3078 /// operand by the specified number of bits. High-order bits are cleared.
3079 ///
3080 /// \headerfile <x86intrin.h>
3081 ///
3082 /// This intrinsic corresponds to the <c> VPSRLD / PSRLD </c> instruction.
3083 ///
3084 /// \param __a
3085 /// A 128-bit integer vector containing the source operand.
3086 /// \param __count
3087 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
3088 /// to right-shift each value in operand \a __a.
3089 /// \returns A 128-bit integer vector containing the right-shifted values.
3090 static __inline__ __m128i __DEFAULT_FN_ATTRS
3091 _mm_srl_epi32(__m128i __a, __m128i __count)
3092 {
3093  return (__m128i)__builtin_ia32_psrld128((__v4si)__a, (__v4si)__count);
3094 }
3095 
3096 /// Right-shifts each of 64-bit values in the 128-bit integer vector
3097 /// operand by the specified number of bits. High-order bits are cleared.
3098 ///
3099 /// \headerfile <x86intrin.h>
3100 ///
3101 /// This intrinsic corresponds to the <c> VPSRLQ / PSRLQ </c> instruction.
3102 ///
3103 /// \param __a
3104 /// A 128-bit integer vector containing the source operand.
3105 /// \param __count
3106 /// An integer value specifying the number of bits to right-shift each value
3107 /// in operand \a __a.
3108 /// \returns A 128-bit integer vector containing the right-shifted values.
3109 static __inline__ __m128i __DEFAULT_FN_ATTRS
3110 _mm_srli_epi64(__m128i __a, int __count)
3111 {
3112  return __builtin_ia32_psrlqi128((__v2di)__a, __count);
3113 }
3114 
3115 /// Right-shifts each of 64-bit values in the 128-bit integer vector
3116 /// operand by the specified number of bits. High-order bits are cleared.
3117 ///
3118 /// \headerfile <x86intrin.h>
3119 ///
3120 /// This intrinsic corresponds to the <c> VPSRLQ / PSRLQ </c> instruction.
3121 ///
3122 /// \param __a
3123 /// A 128-bit integer vector containing the source operand.
3124 /// \param __count
3125 /// A 128-bit integer vector in which bits [63:0] specify the number of bits
3126 /// to right-shift each value in operand \a __a.
3127 /// \returns A 128-bit integer vector containing the right-shifted values.
3128 static __inline__ __m128i __DEFAULT_FN_ATTRS
3129 _mm_srl_epi64(__m128i __a, __m128i __count)
3130 {
3131  return __builtin_ia32_psrlq128((__v2di)__a, (__v2di)__count);
3132 }
3133 
3134 /// Compares each of the corresponding 8-bit values of the 128-bit
3135 /// integer vectors for equality. Each comparison yields 0x0 for false, 0xFF
3136 /// for true.
3137 ///
3138 /// \headerfile <x86intrin.h>
3139 ///
3140 /// This intrinsic corresponds to the <c> VPCMPEQB / PCMPEQB </c> instruction.
3141 ///
3142 /// \param __a
3143 /// A 128-bit integer vector.
3144 /// \param __b
3145 /// A 128-bit integer vector.
3146 /// \returns A 128-bit integer vector containing the comparison results.
3147 static __inline__ __m128i __DEFAULT_FN_ATTRS
3148 _mm_cmpeq_epi8(__m128i __a, __m128i __b)
3149 {
3150  return (__m128i)((__v16qi)__a == (__v16qi)__b);
3151 }
3152 
3153 /// Compares each of the corresponding 16-bit values of the 128-bit
3154 /// integer vectors for equality. Each comparison yields 0x0 for false,
3155 /// 0xFFFF for true.
3156 ///
3157 /// \headerfile <x86intrin.h>
3158 ///
3159 /// This intrinsic corresponds to the <c> VPCMPEQW / PCMPEQW </c> instruction.
3160 ///
3161 /// \param __a
3162 /// A 128-bit integer vector.
3163 /// \param __b
3164 /// A 128-bit integer vector.
3165 /// \returns A 128-bit integer vector containing the comparison results.
3166 static __inline__ __m128i __DEFAULT_FN_ATTRS
3167 _mm_cmpeq_epi16(__m128i __a, __m128i __b)
3168 {
3169  return (__m128i)((__v8hi)__a == (__v8hi)__b);
3170 }
3171 
3172 /// Compares each of the corresponding 32-bit values of the 128-bit
3173 /// integer vectors for equality. Each comparison yields 0x0 for false,
3174 /// 0xFFFFFFFF for true.
3175 ///
3176 /// \headerfile <x86intrin.h>
3177 ///
3178 /// This intrinsic corresponds to the <c> VPCMPEQD / PCMPEQD </c> instruction.
3179 ///
3180 /// \param __a
3181 /// A 128-bit integer vector.
3182 /// \param __b
3183 /// A 128-bit integer vector.
3184 /// \returns A 128-bit integer vector containing the comparison results.
3185 static __inline__ __m128i __DEFAULT_FN_ATTRS
3186 _mm_cmpeq_epi32(__m128i __a, __m128i __b)
3187 {
3188  return (__m128i)((__v4si)__a == (__v4si)__b);
3189 }
3190 
3191 /// Compares each of the corresponding signed 8-bit values of the 128-bit
3192 /// integer vectors to determine if the values in the first operand are
3193 /// greater than those in the second operand. Each comparison yields 0x0 for
3194 /// false, 0xFF for true.
3195 ///
3196 /// \headerfile <x86intrin.h>
3197 ///
3198 /// This intrinsic corresponds to the <c> VPCMPGTB / PCMPGTB </c> instruction.
3199 ///
3200 /// \param __a
3201 /// A 128-bit integer vector.
3202 /// \param __b
3203 /// A 128-bit integer vector.
3204 /// \returns A 128-bit integer vector containing the comparison results.
3205 static __inline__ __m128i __DEFAULT_FN_ATTRS
3206 _mm_cmpgt_epi8(__m128i __a, __m128i __b)
3207 {
3208  /* This function always performs a signed comparison, but __v16qi is a char
3209  which may be signed or unsigned, so use __v16qs. */
3210  return (__m128i)((__v16qs)__a > (__v16qs)__b);
3211 }
3212 
3213 /// Compares each of the corresponding signed 16-bit values of the
3214 /// 128-bit integer vectors to determine if the values in the first operand
3215 /// are greater than those in the second operand.
3216 ///
3217 /// Each comparison yields 0x0 for false, 0xFFFF for true.
3218 ///
3219 /// \headerfile <x86intrin.h>
3220 ///
3221 /// This intrinsic corresponds to the <c> VPCMPGTW / PCMPGTW </c> instruction.
3222 ///
3223 /// \param __a
3224 /// A 128-bit integer vector.
3225 /// \param __b
3226 /// A 128-bit integer vector.
3227 /// \returns A 128-bit integer vector containing the comparison results.
3228 static __inline__ __m128i __DEFAULT_FN_ATTRS
3229 _mm_cmpgt_epi16(__m128i __a, __m128i __b)
3230 {
3231  return (__m128i)((__v8hi)__a > (__v8hi)__b);
3232 }
3233 
3234 /// Compares each of the corresponding signed 32-bit values of the
3235 /// 128-bit integer vectors to determine if the values in the first operand
3236 /// are greater than those in the second operand.
3237 ///
3238 /// Each comparison yields 0x0 for false, 0xFFFFFFFF for true.
3239 ///
3240 /// \headerfile <x86intrin.h>
3241 ///
3242 /// This intrinsic corresponds to the <c> VPCMPGTD / PCMPGTD </c> instruction.
3243 ///
3244 /// \param __a
3245 /// A 128-bit integer vector.
3246 /// \param __b
3247 /// A 128-bit integer vector.
3248 /// \returns A 128-bit integer vector containing the comparison results.
3249 static __inline__ __m128i __DEFAULT_FN_ATTRS
3250 _mm_cmpgt_epi32(__m128i __a, __m128i __b)
3251 {
3252  return (__m128i)((__v4si)__a > (__v4si)__b);
3253 }
3254 
3255 /// Compares each of the corresponding signed 8-bit values of the 128-bit
3256 /// integer vectors to determine if the values in the first operand are less
3257 /// than those in the second operand.
3258 ///
3259 /// Each comparison yields 0x0 for false, 0xFF for true.
3260 ///
3261 /// \headerfile <x86intrin.h>
3262 ///
3263 /// This intrinsic corresponds to the <c> VPCMPGTB / PCMPGTB </c> instruction.
3264 ///
3265 /// \param __a
3266 /// A 128-bit integer vector.
3267 /// \param __b
3268 /// A 128-bit integer vector.
3269 /// \returns A 128-bit integer vector containing the comparison results.
3270 static __inline__ __m128i __DEFAULT_FN_ATTRS
3271 _mm_cmplt_epi8(__m128i __a, __m128i __b)
3272 {
3273  return _mm_cmpgt_epi8(__b, __a);
3274 }
3275 
3276 /// Compares each of the corresponding signed 16-bit values of the
3277 /// 128-bit integer vectors to determine if the values in the first operand
3278 /// are less than those in the second operand.
3279 ///
3280 /// Each comparison yields 0x0 for false, 0xFFFF for true.
3281 ///
3282 /// \headerfile <x86intrin.h>
3283 ///
3284 /// This intrinsic corresponds to the <c> VPCMPGTW / PCMPGTW </c> instruction.
3285 ///
3286 /// \param __a
3287 /// A 128-bit integer vector.
3288 /// \param __b
3289 /// A 128-bit integer vector.
3290 /// \returns A 128-bit integer vector containing the comparison results.
3291 static __inline__ __m128i __DEFAULT_FN_ATTRS
3292 _mm_cmplt_epi16(__m128i __a, __m128i __b)
3293 {
3294  return _mm_cmpgt_epi16(__b, __a);
3295 }
3296 
3297 /// Compares each of the corresponding signed 32-bit values of the
3298 /// 128-bit integer vectors to determine if the values in the first operand
3299 /// are less than those in the second operand.
3300 ///
3301 /// Each comparison yields 0x0 for false, 0xFFFFFFFF for true.
3302 ///
3303 /// \headerfile <x86intrin.h>
3304 ///
3305 /// This intrinsic corresponds to the <c> VPCMPGTD / PCMPGTD </c> instruction.
3306 ///
3307 /// \param __a
3308 /// A 128-bit integer vector.
3309 /// \param __b
3310 /// A 128-bit integer vector.
3311 /// \returns A 128-bit integer vector containing the comparison results.
3312 static __inline__ __m128i __DEFAULT_FN_ATTRS
3313 _mm_cmplt_epi32(__m128i __a, __m128i __b)
3314 {
3315  return _mm_cmpgt_epi32(__b, __a);
3316 }
3317 
3318 #ifdef __x86_64__
3319 /// Converts a 64-bit signed integer value from the second operand into a
3320 /// double-precision value and returns it in the lower element of a [2 x
3321 /// double] vector; the upper element of the returned vector is copied from
3322 /// the upper element of the first operand.
3323 ///
3324 /// \headerfile <x86intrin.h>
3325 ///
3326 /// This intrinsic corresponds to the <c> VCVTSI2SD / CVTSI2SD </c> instruction.
3327 ///
3328 /// \param __a
3329 /// A 128-bit vector of [2 x double]. The upper 64 bits of this operand are
3330 /// copied to the upper 64 bits of the destination.
3331 /// \param __b
3332 /// A 64-bit signed integer operand containing the value to be converted.
3333 /// \returns A 128-bit vector of [2 x double] whose lower 64 bits contain the
3334 /// converted value of the second operand. The upper 64 bits are copied from
3335 /// the upper 64 bits of the first operand.
3336 static __inline__ __m128d __DEFAULT_FN_ATTRS
3337 _mm_cvtsi64_sd(__m128d __a, long long __b)
3338 {
3339  __a[0] = __b;
3340  return __a;
3341 }
3342 
3343 /// Converts the first (lower) element of a vector of [2 x double] into a
3344 /// 64-bit signed integer value, according to the current rounding mode.
3345 ///
3346 /// \headerfile <x86intrin.h>
3347 ///
3348 /// This intrinsic corresponds to the <c> VCVTSD2SI / CVTSD2SI </c> instruction.
3349 ///
3350 /// \param __a
3351 /// A 128-bit vector of [2 x double]. The lower 64 bits are used in the
3352 /// conversion.
3353 /// \returns A 64-bit signed integer containing the converted value.
3354 static __inline__ long long __DEFAULT_FN_ATTRS
3355 _mm_cvtsd_si64(__m128d __a)
3356 {
3357  return __builtin_ia32_cvtsd2si64((__v2df)__a);
3358 }
3359 
3360 /// Converts the first (lower) element of a vector of [2 x double] into a
3361 /// 64-bit signed integer value, truncating the result when it is inexact.
3362 ///
3363 /// \headerfile <x86intrin.h>
3364 ///
3365 /// This intrinsic corresponds to the <c> VCVTTSD2SI / CVTTSD2SI </c>
3366 /// instruction.
3367 ///
3368 /// \param __a
3369 /// A 128-bit vector of [2 x double]. The lower 64 bits are used in the
3370 /// conversion.
3371 /// \returns A 64-bit signed integer containing the converted value.
3372 static __inline__ long long __DEFAULT_FN_ATTRS
3373 _mm_cvttsd_si64(__m128d __a)
3374 {
3375  return __builtin_ia32_cvttsd2si64((__v2df)__a);
3376 }
3377 #endif
3378 
3379 /// Converts a vector of [4 x i32] into a vector of [4 x float].
3380 ///
3381 /// \headerfile <x86intrin.h>
3382 ///
3383 /// This intrinsic corresponds to the <c> VCVTDQ2PS / CVTDQ2PS </c> instruction.
3384 ///
3385 /// \param __a
3386 /// A 128-bit integer vector.
3387 /// \returns A 128-bit vector of [4 x float] containing the converted values.
3388 static __inline__ __m128 __DEFAULT_FN_ATTRS
3389 _mm_cvtepi32_ps(__m128i __a)
3390 {
3391  return (__m128)__builtin_convertvector((__v4si)__a, __v4sf);
3392 }
3393 
3394 /// Converts a vector of [4 x float] into a vector of [4 x i32].
3395 ///
3396 /// \headerfile <x86intrin.h>
3397 ///
3398 /// This intrinsic corresponds to the <c> VCVTPS2DQ / CVTPS2DQ </c> instruction.
3399 ///
3400 /// \param __a
3401 /// A 128-bit vector of [4 x float].
3402 /// \returns A 128-bit integer vector of [4 x i32] containing the converted
3403 /// values.
3404 static __inline__ __m128i __DEFAULT_FN_ATTRS
3405 _mm_cvtps_epi32(__m128 __a)
3406 {
3407  return (__m128i)__builtin_ia32_cvtps2dq((__v4sf)__a);
3408 }
3409 
3410 /// Converts a vector of [4 x float] into a vector of [4 x i32],
3411 /// truncating the result when it is inexact.
3412 ///
3413 /// \headerfile <x86intrin.h>
3414 ///
3415 /// This intrinsic corresponds to the <c> VCVTTPS2DQ / CVTTPS2DQ </c>
3416 /// instruction.
3417 ///
3418 /// \param __a
3419 /// A 128-bit vector of [4 x float].
3420 /// \returns A 128-bit vector of [4 x i32] containing the converted values.
3421 static __inline__ __m128i __DEFAULT_FN_ATTRS
3422 _mm_cvttps_epi32(__m128 __a)
3423 {
3424  return (__m128i)__builtin_ia32_cvttps2dq((__v4sf)__a);
3425 }
3426 
3427 /// Returns a vector of [4 x i32] where the lowest element is the input
3428 /// operand and the remaining elements are zero.
3429 ///
3430 /// \headerfile <x86intrin.h>
3431 ///
3432 /// This intrinsic corresponds to the <c> VMOVD / MOVD </c> instruction.
3433 ///
3434 /// \param __a
3435 /// A 32-bit signed integer operand.
3436 /// \returns A 128-bit vector of [4 x i32].
3437 static __inline__ __m128i __DEFAULT_FN_ATTRS
3439 {
3440  return __extension__ (__m128i)(__v4si){ __a, 0, 0, 0 };
3441 }
3442 
3443 #ifdef __x86_64__
3444 /// Returns a vector of [2 x i64] where the lower element is the input
3445 /// operand and the upper element is zero.
3446 ///
3447 /// \headerfile <x86intrin.h>
3448 ///
3449 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
3450 ///
3451 /// \param __a
3452 /// A 64-bit signed integer operand containing the value to be converted.
3453 /// \returns A 128-bit vector of [2 x i64] containing the converted value.
3454 static __inline__ __m128i __DEFAULT_FN_ATTRS
3455 _mm_cvtsi64_si128(long long __a)
3456 {
3457  return __extension__ (__m128i)(__v2di){ __a, 0 };
3458 }
3459 #endif
3460 
3461 /// Moves the least significant 32 bits of a vector of [4 x i32] to a
3462 /// 32-bit signed integer value.
3463 ///
3464 /// \headerfile <x86intrin.h>
3465 ///
3466 /// This intrinsic corresponds to the <c> VMOVD / MOVD </c> instruction.
3467 ///
3468 /// \param __a
3469 /// A vector of [4 x i32]. The least significant 32 bits are moved to the
3470 /// destination.
3471 /// \returns A 32-bit signed integer containing the moved value.
3472 static __inline__ int __DEFAULT_FN_ATTRS
3473 _mm_cvtsi128_si32(__m128i __a)
3474 {
3475  __v4si __b = (__v4si)__a;
3476  return __b[0];
3477 }
3478 
3479 #ifdef __x86_64__
3480 /// Moves the least significant 64 bits of a vector of [2 x i64] to a
3481 /// 64-bit signed integer value.
3482 ///
3483 /// \headerfile <x86intrin.h>
3484 ///
3485 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
3486 ///
3487 /// \param __a
3488 /// A vector of [2 x i64]. The least significant 64 bits are moved to the
3489 /// destination.
3490 /// \returns A 64-bit signed integer containing the moved value.
3491 static __inline__ long long __DEFAULT_FN_ATTRS
3492 _mm_cvtsi128_si64(__m128i __a)
3493 {
3494  return __a[0];
3495 }
3496 #endif
3497 
3498 /// Moves packed integer values from an aligned 128-bit memory location
3499 /// to elements in a 128-bit integer vector.
3500 ///
3501 /// \headerfile <x86intrin.h>
3502 ///
3503 /// This intrinsic corresponds to the <c> VMOVDQA / MOVDQA </c> instruction.
3504 ///
3505 /// \param __p
3506 /// An aligned pointer to a memory location containing integer values.
3507 /// \returns A 128-bit integer vector containing the moved values.
3508 static __inline__ __m128i __DEFAULT_FN_ATTRS
3509 _mm_load_si128(__m128i const *__p)
3510 {
3511  return *__p;
3512 }
3513 
3514 /// Moves packed integer values from an unaligned 128-bit memory location
3515 /// to elements in a 128-bit integer vector.
3516 ///
3517 /// \headerfile <x86intrin.h>
3518 ///
3519 /// This intrinsic corresponds to the <c> VMOVDQU / MOVDQU </c> instruction.
3520 ///
3521 /// \param __p
3522 /// A pointer to a memory location containing integer values.
3523 /// \returns A 128-bit integer vector containing the moved values.
3524 static __inline__ __m128i __DEFAULT_FN_ATTRS
3525 _mm_loadu_si128(__m128i const *__p)
3526 {
3527  struct __loadu_si128 {
3528  __m128i __v;
3529  } __attribute__((__packed__, __may_alias__));
3530  return ((struct __loadu_si128*)__p)->__v;
3531 }
3532 
3533 /// Returns a vector of [2 x i64] where the lower element is taken from
3534 /// the lower element of the operand, and the upper element is zero.
3535 ///
3536 /// \headerfile <x86intrin.h>
3537 ///
3538 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
3539 ///
3540 /// \param __p
3541 /// A 128-bit vector of [2 x i64]. Bits [63:0] are written to bits [63:0] of
3542 /// the destination.
3543 /// \returns A 128-bit vector of [2 x i64]. The lower order bits contain the
3544 /// moved value. The higher order bits are cleared.
3545 static __inline__ __m128i __DEFAULT_FN_ATTRS
3546 _mm_loadl_epi64(__m128i const *__p)
3547 {
3548  struct __mm_loadl_epi64_struct {
3549  long long __u;
3550  } __attribute__((__packed__, __may_alias__));
3551  return __extension__ (__m128i) { ((struct __mm_loadl_epi64_struct*)__p)->__u, 0};
3552 }
3553 
3554 /// Generates a 128-bit vector of [4 x i32] with unspecified content.
3555 /// This could be used as an argument to another intrinsic function where the
3556 /// argument is required but the value is not actually used.
3557 ///
3558 /// \headerfile <x86intrin.h>
3559 ///
3560 /// This intrinsic has no corresponding instruction.
3561 ///
3562 /// \returns A 128-bit vector of [4 x i32] with unspecified content.
3563 static __inline__ __m128i __DEFAULT_FN_ATTRS
3565 {
3566  return (__m128i)__builtin_ia32_undef128();
3567 }
3568 
3569 /// Initializes both 64-bit values in a 128-bit vector of [2 x i64] with
3570 /// the specified 64-bit integer values.
3571 ///
3572 /// \headerfile <x86intrin.h>
3573 ///
3574 /// This intrinsic is a utility function and does not correspond to a specific
3575 /// instruction.
3576 ///
3577 /// \param __q1
3578 /// A 64-bit integer value used to initialize the upper 64 bits of the
3579 /// destination vector of [2 x i64].
3580 /// \param __q0
3581 /// A 64-bit integer value used to initialize the lower 64 bits of the
3582 /// destination vector of [2 x i64].
3583 /// \returns An initialized 128-bit vector of [2 x i64] containing the values
3584 /// provided in the operands.
3585 static __inline__ __m128i __DEFAULT_FN_ATTRS
3586 _mm_set_epi64x(long long __q1, long long __q0)
3587 {
3588  return __extension__ (__m128i)(__v2di){ __q0, __q1 };
3589 }
3590 
3591 /// Initializes both 64-bit values in a 128-bit vector of [2 x i64] with
3592 /// the specified 64-bit integer values.
3593 ///
3594 /// \headerfile <x86intrin.h>
3595 ///
3596 /// This intrinsic is a utility function and does not correspond to a specific
3597 /// instruction.
3598 ///
3599 /// \param __q1
3600 /// A 64-bit integer value used to initialize the upper 64 bits of the
3601 /// destination vector of [2 x i64].
3602 /// \param __q0
3603 /// A 64-bit integer value used to initialize the lower 64 bits of the
3604 /// destination vector of [2 x i64].
3605 /// \returns An initialized 128-bit vector of [2 x i64] containing the values
3606 /// provided in the operands.
3607 static __inline__ __m128i __DEFAULT_FN_ATTRS
3608 _mm_set_epi64(__m64 __q1, __m64 __q0)
3609 {
3610  return _mm_set_epi64x((long long)__q1, (long long)__q0);
3611 }
3612 
3613 /// Initializes the 32-bit values in a 128-bit vector of [4 x i32] with
3614 /// the specified 32-bit integer values.
3615 ///
3616 /// \headerfile <x86intrin.h>
3617 ///
3618 /// This intrinsic is a utility function and does not correspond to a specific
3619 /// instruction.
3620 ///
3621 /// \param __i3
3622 /// A 32-bit integer value used to initialize bits [127:96] of the
3623 /// destination vector.
3624 /// \param __i2
3625 /// A 32-bit integer value used to initialize bits [95:64] of the destination
3626 /// vector.
3627 /// \param __i1
3628 /// A 32-bit integer value used to initialize bits [63:32] of the destination
3629 /// vector.
3630 /// \param __i0
3631 /// A 32-bit integer value used to initialize bits [31:0] of the destination
3632 /// vector.
3633 /// \returns An initialized 128-bit vector of [4 x i32] containing the values
3634 /// provided in the operands.
3635 static __inline__ __m128i __DEFAULT_FN_ATTRS
3636 _mm_set_epi32(int __i3, int __i2, int __i1, int __i0)
3637 {
3638  return __extension__ (__m128i)(__v4si){ __i0, __i1, __i2, __i3};
3639 }
3640 
3641 /// Initializes the 16-bit values in a 128-bit vector of [8 x i16] with
3642 /// the specified 16-bit integer values.
3643 ///
3644 /// \headerfile <x86intrin.h>
3645 ///
3646 /// This intrinsic is a utility function and does not correspond to a specific
3647 /// instruction.
3648 ///
3649 /// \param __w7
3650 /// A 16-bit integer value used to initialize bits [127:112] of the
3651 /// destination vector.
3652 /// \param __w6
3653 /// A 16-bit integer value used to initialize bits [111:96] of the
3654 /// destination vector.
3655 /// \param __w5
3656 /// A 16-bit integer value used to initialize bits [95:80] of the destination
3657 /// vector.
3658 /// \param __w4
3659 /// A 16-bit integer value used to initialize bits [79:64] of the destination
3660 /// vector.
3661 /// \param __w3
3662 /// A 16-bit integer value used to initialize bits [63:48] of the destination
3663 /// vector.
3664 /// \param __w2
3665 /// A 16-bit integer value used to initialize bits [47:32] of the destination
3666 /// vector.
3667 /// \param __w1
3668 /// A 16-bit integer value used to initialize bits [31:16] of the destination
3669 /// vector.
3670 /// \param __w0
3671 /// A 16-bit integer value used to initialize bits [15:0] of the destination
3672 /// vector.
3673 /// \returns An initialized 128-bit vector of [8 x i16] containing the values
3674 /// provided in the operands.
3675 static __inline__ __m128i __DEFAULT_FN_ATTRS
3676 _mm_set_epi16(short __w7, short __w6, short __w5, short __w4, short __w3, short __w2, short __w1, short __w0)
3677 {
3678  return __extension__ (__m128i)(__v8hi){ __w0, __w1, __w2, __w3, __w4, __w5, __w6, __w7 };
3679 }
3680 
3681 /// Initializes the 8-bit values in a 128-bit vector of [16 x i8] with
3682 /// the specified 8-bit integer values.
3683 ///
3684 /// \headerfile <x86intrin.h>
3685 ///
3686 /// This intrinsic is a utility function and does not correspond to a specific
3687 /// instruction.
3688 ///
3689 /// \param __b15
3690 /// Initializes bits [127:120] of the destination vector.
3691 /// \param __b14
3692 /// Initializes bits [119:112] of the destination vector.
3693 /// \param __b13
3694 /// Initializes bits [111:104] of the destination vector.
3695 /// \param __b12
3696 /// Initializes bits [103:96] of the destination vector.
3697 /// \param __b11
3698 /// Initializes bits [95:88] of the destination vector.
3699 /// \param __b10
3700 /// Initializes bits [87:80] of the destination vector.
3701 /// \param __b9
3702 /// Initializes bits [79:72] of the destination vector.
3703 /// \param __b8
3704 /// Initializes bits [71:64] of the destination vector.
3705 /// \param __b7
3706 /// Initializes bits [63:56] of the destination vector.
3707 /// \param __b6
3708 /// Initializes bits [55:48] of the destination vector.
3709 /// \param __b5
3710 /// Initializes bits [47:40] of the destination vector.
3711 /// \param __b4
3712 /// Initializes bits [39:32] of the destination vector.
3713 /// \param __b3
3714 /// Initializes bits [31:24] of the destination vector.
3715 /// \param __b2
3716 /// Initializes bits [23:16] of the destination vector.
3717 /// \param __b1
3718 /// Initializes bits [15:8] of the destination vector.
3719 /// \param __b0
3720 /// Initializes bits [7:0] of the destination vector.
3721 /// \returns An initialized 128-bit vector of [16 x i8] containing the values
3722 /// provided in the operands.
3723 static __inline__ __m128i __DEFAULT_FN_ATTRS
3724 _mm_set_epi8(char __b15, char __b14, char __b13, char __b12, char __b11, char __b10, char __b9, char __b8, char __b7, char __b6, char __b5, char __b4, char __b3, char __b2, char __b1, char __b0)
3725 {
3726  return __extension__ (__m128i)(__v16qi){ __b0, __b1, __b2, __b3, __b4, __b5, __b6, __b7, __b8, __b9, __b10, __b11, __b12, __b13, __b14, __b15 };
3727 }
3728 
3729 /// Initializes both values in a 128-bit integer vector with the
3730 /// specified 64-bit integer value.
3731 ///
3732 /// \headerfile <x86intrin.h>
3733 ///
3734 /// This intrinsic is a utility function and does not correspond to a specific
3735 /// instruction.
3736 ///
3737 /// \param __q
3738 /// Integer value used to initialize the elements of the destination integer
3739 /// vector.
3740 /// \returns An initialized 128-bit integer vector of [2 x i64] with both
3741 /// elements containing the value provided in the operand.
3742 static __inline__ __m128i __DEFAULT_FN_ATTRS
3743 _mm_set1_epi64x(long long __q)
3744 {
3745  return _mm_set_epi64x(__q, __q);
3746 }
3747 
3748 /// Initializes both values in a 128-bit vector of [2 x i64] with the
3749 /// specified 64-bit value.
3750 ///
3751 /// \headerfile <x86intrin.h>
3752 ///
3753 /// This intrinsic is a utility function and does not correspond to a specific
3754 /// instruction.
3755 ///
3756 /// \param __q
3757 /// A 64-bit value used to initialize the elements of the destination integer
3758 /// vector.
3759 /// \returns An initialized 128-bit vector of [2 x i64] with all elements
3760 /// containing the value provided in the operand.
3761 static __inline__ __m128i __DEFAULT_FN_ATTRS
3762 _mm_set1_epi64(__m64 __q)
3763 {
3764  return _mm_set_epi64(__q, __q);
3765 }
3766 
3767 /// Initializes all values in a 128-bit vector of [4 x i32] with the
3768 /// specified 32-bit value.
3769 ///
3770 /// \headerfile <x86intrin.h>
3771 ///
3772 /// This intrinsic is a utility function and does not correspond to a specific
3773 /// instruction.
3774 ///
3775 /// \param __i
3776 /// A 32-bit value used to initialize the elements of the destination integer
3777 /// vector.
3778 /// \returns An initialized 128-bit vector of [4 x i32] with all elements
3779 /// containing the value provided in the operand.
3780 static __inline__ __m128i __DEFAULT_FN_ATTRS
3782 {
3783  return _mm_set_epi32(__i, __i, __i, __i);
3784 }
3785 
3786 /// Initializes all values in a 128-bit vector of [8 x i16] with the
3787 /// specified 16-bit value.
3788 ///
3789 /// \headerfile <x86intrin.h>
3790 ///
3791 /// This intrinsic is a utility function and does not correspond to a specific
3792 /// instruction.
3793 ///
3794 /// \param __w
3795 /// A 16-bit value used to initialize the elements of the destination integer
3796 /// vector.
3797 /// \returns An initialized 128-bit vector of [8 x i16] with all elements
3798 /// containing the value provided in the operand.
3799 static __inline__ __m128i __DEFAULT_FN_ATTRS
3800 _mm_set1_epi16(short __w)
3801 {
3802  return _mm_set_epi16(__w, __w, __w, __w, __w, __w, __w, __w);
3803 }
3804 
3805 /// Initializes all values in a 128-bit vector of [16 x i8] with the
3806 /// specified 8-bit value.
3807 ///
3808 /// \headerfile <x86intrin.h>
3809 ///
3810 /// This intrinsic is a utility function and does not correspond to a specific
3811 /// instruction.
3812 ///
3813 /// \param __b
3814 /// An 8-bit value used to initialize the elements of the destination integer
3815 /// vector.
3816 /// \returns An initialized 128-bit vector of [16 x i8] with all elements
3817 /// containing the value provided in the operand.
3818 static __inline__ __m128i __DEFAULT_FN_ATTRS
3819 _mm_set1_epi8(char __b)
3820 {
3821  return _mm_set_epi8(__b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b, __b);
3822 }
3823 
3824 /// Constructs a 128-bit integer vector, initialized in reverse order
3825 /// with the specified 64-bit integral values.
3826 ///
3827 /// \headerfile <x86intrin.h>
3828 ///
3829 /// This intrinsic does not correspond to a specific instruction.
3830 ///
3831 /// \param __q0
3832 /// A 64-bit integral value used to initialize the lower 64 bits of the
3833 /// result.
3834 /// \param __q1
3835 /// A 64-bit integral value used to initialize the upper 64 bits of the
3836 /// result.
3837 /// \returns An initialized 128-bit integer vector.
3838 static __inline__ __m128i __DEFAULT_FN_ATTRS
3839 _mm_setr_epi64(__m64 __q0, __m64 __q1)
3840 {
3841  return _mm_set_epi64(__q1, __q0);
3842 }
3843 
3844 /// Constructs a 128-bit integer vector, initialized in reverse order
3845 /// with the specified 32-bit integral values.
3846 ///
3847 /// \headerfile <x86intrin.h>
3848 ///
3849 /// This intrinsic is a utility function and does not correspond to a specific
3850 /// instruction.
3851 ///
3852 /// \param __i0
3853 /// A 32-bit integral value used to initialize bits [31:0] of the result.
3854 /// \param __i1
3855 /// A 32-bit integral value used to initialize bits [63:32] of the result.
3856 /// \param __i2
3857 /// A 32-bit integral value used to initialize bits [95:64] of the result.
3858 /// \param __i3
3859 /// A 32-bit integral value used to initialize bits [127:96] of the result.
3860 /// \returns An initialized 128-bit integer vector.
3861 static __inline__ __m128i __DEFAULT_FN_ATTRS
3862 _mm_setr_epi32(int __i0, int __i1, int __i2, int __i3)
3863 {
3864  return _mm_set_epi32(__i3, __i2, __i1, __i0);
3865 }
3866 
3867 /// Constructs a 128-bit integer vector, initialized in reverse order
3868 /// with the specified 16-bit integral values.
3869 ///
3870 /// \headerfile <x86intrin.h>
3871 ///
3872 /// This intrinsic is a utility function and does not correspond to a specific
3873 /// instruction.
3874 ///
3875 /// \param __w0
3876 /// A 16-bit integral value used to initialize bits [15:0] of the result.
3877 /// \param __w1
3878 /// A 16-bit integral value used to initialize bits [31:16] of the result.
3879 /// \param __w2
3880 /// A 16-bit integral value used to initialize bits [47:32] of the result.
3881 /// \param __w3
3882 /// A 16-bit integral value used to initialize bits [63:48] of the result.
3883 /// \param __w4
3884 /// A 16-bit integral value used to initialize bits [79:64] of the result.
3885 /// \param __w5
3886 /// A 16-bit integral value used to initialize bits [95:80] of the result.
3887 /// \param __w6
3888 /// A 16-bit integral value used to initialize bits [111:96] of the result.
3889 /// \param __w7
3890 /// A 16-bit integral value used to initialize bits [127:112] of the result.
3891 /// \returns An initialized 128-bit integer vector.
3892 static __inline__ __m128i __DEFAULT_FN_ATTRS
3893 _mm_setr_epi16(short __w0, short __w1, short __w2, short __w3, short __w4, short __w5, short __w6, short __w7)
3894 {
3895  return _mm_set_epi16(__w7, __w6, __w5, __w4, __w3, __w2, __w1, __w0);
3896 }
3897 
3898 /// Constructs a 128-bit integer vector, initialized in reverse order
3899 /// with the specified 8-bit integral values.
3900 ///
3901 /// \headerfile <x86intrin.h>
3902 ///
3903 /// This intrinsic is a utility function and does not correspond to a specific
3904 /// instruction.
3905 ///
3906 /// \param __b0
3907 /// An 8-bit integral value used to initialize bits [7:0] of the result.
3908 /// \param __b1
3909 /// An 8-bit integral value used to initialize bits [15:8] of the result.
3910 /// \param __b2
3911 /// An 8-bit integral value used to initialize bits [23:16] of the result.
3912 /// \param __b3
3913 /// An 8-bit integral value used to initialize bits [31:24] of the result.
3914 /// \param __b4
3915 /// An 8-bit integral value used to initialize bits [39:32] of the result.
3916 /// \param __b5
3917 /// An 8-bit integral value used to initialize bits [47:40] of the result.
3918 /// \param __b6
3919 /// An 8-bit integral value used to initialize bits [55:48] of the result.
3920 /// \param __b7
3921 /// An 8-bit integral value used to initialize bits [63:56] of the result.
3922 /// \param __b8
3923 /// An 8-bit integral value used to initialize bits [71:64] of the result.
3924 /// \param __b9
3925 /// An 8-bit integral value used to initialize bits [79:72] of the result.
3926 /// \param __b10
3927 /// An 8-bit integral value used to initialize bits [87:80] of the result.
3928 /// \param __b11
3929 /// An 8-bit integral value used to initialize bits [95:88] of the result.
3930 /// \param __b12
3931 /// An 8-bit integral value used to initialize bits [103:96] of the result.
3932 /// \param __b13
3933 /// An 8-bit integral value used to initialize bits [111:104] of the result.
3934 /// \param __b14
3935 /// An 8-bit integral value used to initialize bits [119:112] of the result.
3936 /// \param __b15
3937 /// An 8-bit integral value used to initialize bits [127:120] of the result.
3938 /// \returns An initialized 128-bit integer vector.
3939 static __inline__ __m128i __DEFAULT_FN_ATTRS
3940 _mm_setr_epi8(char __b0, char __b1, char __b2, char __b3, char __b4, char __b5, char __b6, char __b7, char __b8, char __b9, char __b10, char __b11, char __b12, char __b13, char __b14, char __b15)
3941 {
3942  return _mm_set_epi8(__b15, __b14, __b13, __b12, __b11, __b10, __b9, __b8, __b7, __b6, __b5, __b4, __b3, __b2, __b1, __b0);
3943 }
3944 
3945 /// Creates a 128-bit integer vector initialized to zero.
3946 ///
3947 /// \headerfile <x86intrin.h>
3948 ///
3949 /// This intrinsic corresponds to the <c> VXORPS / XORPS </c> instruction.
3950 ///
3951 /// \returns An initialized 128-bit integer vector with all elements set to
3952 /// zero.
3953 static __inline__ __m128i __DEFAULT_FN_ATTRS
3955 {
3956  return __extension__ (__m128i)(__v2di){ 0LL, 0LL };
3957 }
3958 
3959 /// Stores a 128-bit integer vector to a memory location aligned on a
3960 /// 128-bit boundary.
3961 ///
3962 /// \headerfile <x86intrin.h>
3963 ///
3964 /// This intrinsic corresponds to the <c> VMOVAPS / MOVAPS </c> instruction.
3965 ///
3966 /// \param __p
3967 /// A pointer to an aligned memory location that will receive the integer
3968 /// values.
3969 /// \param __b
3970 /// A 128-bit integer vector containing the values to be moved.
3971 static __inline__ void __DEFAULT_FN_ATTRS
3972 _mm_store_si128(__m128i *__p, __m128i __b)
3973 {
3974  *__p = __b;
3975 }
3976 
3977 /// Stores a 128-bit integer vector to an unaligned memory location.
3978 ///
3979 /// \headerfile <x86intrin.h>
3980 ///
3981 /// This intrinsic corresponds to the <c> VMOVUPS / MOVUPS </c> instruction.
3982 ///
3983 /// \param __p
3984 /// A pointer to a memory location that will receive the integer values.
3985 /// \param __b
3986 /// A 128-bit integer vector containing the values to be moved.
3987 static __inline__ void __DEFAULT_FN_ATTRS
3988 _mm_storeu_si128(__m128i *__p, __m128i __b)
3989 {
3990  struct __storeu_si128 {
3991  __m128i __v;
3992  } __attribute__((__packed__, __may_alias__));
3993  ((struct __storeu_si128*)__p)->__v = __b;
3994 }
3995 
3996 /// Moves bytes selected by the mask from the first operand to the
3997 /// specified unaligned memory location. When a mask bit is 1, the
3998 /// corresponding byte is written, otherwise it is not written.
3999 ///
4000 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
4001 /// used again soon). Exception and trap behavior for elements not selected
4002 /// for storage to memory are implementation dependent.
4003 ///
4004 /// \headerfile <x86intrin.h>
4005 ///
4006 /// This intrinsic corresponds to the <c> VMASKMOVDQU / MASKMOVDQU </c>
4007 /// instruction.
4008 ///
4009 /// \param __d
4010 /// A 128-bit integer vector containing the values to be moved.
4011 /// \param __n
4012 /// A 128-bit integer vector containing the mask. The most significant bit of
4013 /// each byte represents the mask bits.
4014 /// \param __p
4015 /// A pointer to an unaligned 128-bit memory location where the specified
4016 /// values are moved.
4017 static __inline__ void __DEFAULT_FN_ATTRS
4018 _mm_maskmoveu_si128(__m128i __d, __m128i __n, char *__p)
4019 {
4020  __builtin_ia32_maskmovdqu((__v16qi)__d, (__v16qi)__n, __p);
4021 }
4022 
4023 /// Stores the lower 64 bits of a 128-bit integer vector of [2 x i64] to
4024 /// a memory location.
4025 ///
4026 /// \headerfile <x86intrin.h>
4027 ///
4028 /// This intrinsic corresponds to the <c> VMOVLPS / MOVLPS </c> instruction.
4029 ///
4030 /// \param __p
4031 /// A pointer to a 64-bit memory location that will receive the lower 64 bits
4032 /// of the integer vector parameter.
4033 /// \param __a
4034 /// A 128-bit integer vector of [2 x i64]. The lower 64 bits contain the
4035 /// value to be stored.
4036 static __inline__ void __DEFAULT_FN_ATTRS
4037 _mm_storel_epi64(__m128i *__p, __m128i __a)
4038 {
4039  struct __mm_storel_epi64_struct {
4040  long long __u;
4041  } __attribute__((__packed__, __may_alias__));
4042  ((struct __mm_storel_epi64_struct*)__p)->__u = __a[0];
4043 }
4044 
4045 /// Stores a 128-bit floating point vector of [2 x double] to a 128-bit
4046 /// aligned memory location.
4047 ///
4048 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
4049 /// used again soon).
4050 ///
4051 /// \headerfile <x86intrin.h>
4052 ///
4053 /// This intrinsic corresponds to the <c> VMOVNTPS / MOVNTPS </c> instruction.
4054 ///
4055 /// \param __p
4056 /// A pointer to the 128-bit aligned memory location used to store the value.
4057 /// \param __a
4058 /// A vector of [2 x double] containing the 64-bit values to be stored.
4059 static __inline__ void __DEFAULT_FN_ATTRS
4060 _mm_stream_pd(double *__p, __m128d __a)
4061 {
4062  __builtin_nontemporal_store((__v2df)__a, (__v2df*)__p);
4063 }
4064 
4065 /// Stores a 128-bit integer vector to a 128-bit aligned memory location.
4066 ///
4067 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
4068 /// used again soon).
4069 ///
4070 /// \headerfile <x86intrin.h>
4071 ///
4072 /// This intrinsic corresponds to the <c> VMOVNTPS / MOVNTPS </c> instruction.
4073 ///
4074 /// \param __p
4075 /// A pointer to the 128-bit aligned memory location used to store the value.
4076 /// \param __a
4077 /// A 128-bit integer vector containing the values to be stored.
4078 static __inline__ void __DEFAULT_FN_ATTRS
4079 _mm_stream_si128(__m128i *__p, __m128i __a)
4080 {
4081  __builtin_nontemporal_store((__v2di)__a, (__v2di*)__p);
4082 }
4083 
4084 /// Stores a 32-bit integer value in the specified memory location.
4085 ///
4086 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
4087 /// used again soon).
4088 ///
4089 /// \headerfile <x86intrin.h>
4090 ///
4091 /// This intrinsic corresponds to the <c> MOVNTI </c> instruction.
4092 ///
4093 /// \param __p
4094 /// A pointer to the 32-bit memory location used to store the value.
4095 /// \param __a
4096 /// A 32-bit integer containing the value to be stored.
4097 static __inline__ void __attribute__((__always_inline__, __nodebug__, __target__("sse2")))
4098 _mm_stream_si32(int *__p, int __a)
4100  __builtin_ia32_movnti(__p, __a);
4101 }
4102 
4103 #ifdef __x86_64__
4104 /// Stores a 64-bit integer value in the specified memory location.
4105 ///
4106 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
4107 /// used again soon).
4108 ///
4109 /// \headerfile <x86intrin.h>
4110 ///
4111 /// This intrinsic corresponds to the <c> MOVNTIQ </c> instruction.
4112 ///
4113 /// \param __p
4114 /// A pointer to the 64-bit memory location used to store the value.
4115 /// \param __a
4116 /// A 64-bit integer containing the value to be stored.
4117 static __inline__ void __attribute__((__always_inline__, __nodebug__, __target__("sse2")))
4118 _mm_stream_si64(long long *__p, long long __a)
4119 {
4120  __builtin_ia32_movnti64(__p, __a);
4121 }
4122 #endif
4123 
4124 #if defined(__cplusplus)
4125 extern "C" {
4126 #endif
4127 
4128 /// The cache line containing \a __p is flushed and invalidated from all
4129 /// caches in the coherency domain.
4130 ///
4131 /// \headerfile <x86intrin.h>
4132 ///
4133 /// This intrinsic corresponds to the <c> CLFLUSH </c> instruction.
4134 ///
4135 /// \param __p
4136 /// A pointer to the memory location used to identify the cache line to be
4137 /// flushed.
4138 void _mm_clflush(void const * __p);
4139 
4140 /// Forces strong memory ordering (serialization) between load
4141 /// instructions preceding this instruction and load instructions following
4142 /// this instruction, ensuring the system completes all previous loads before
4143 /// executing subsequent loads.
4144 ///
4145 /// \headerfile <x86intrin.h>
4146 ///
4147 /// This intrinsic corresponds to the <c> LFENCE </c> instruction.
4148 ///
4149 void _mm_lfence(void);
4150 
4151 /// Forces strong memory ordering (serialization) between load and store
4152 /// instructions preceding this instruction and load and store instructions
4153 /// following this instruction, ensuring that the system completes all
4154 /// previous memory accesses before executing subsequent memory accesses.
4155 ///
4156 /// \headerfile <x86intrin.h>
4157 ///
4158 /// This intrinsic corresponds to the <c> MFENCE </c> instruction.
4159 ///
4160 void _mm_mfence(void);
4161 
4162 #if defined(__cplusplus)
4163 } // extern "C"
4164 #endif
4165 
4166 /// Converts 16-bit signed integers from both 128-bit integer vector
4167 /// operands into 8-bit signed integers, and packs the results into the
4168 /// destination. Positive values greater than 0x7F are saturated to 0x7F.
4169 /// Negative values less than 0x80 are saturated to 0x80.
4170 ///
4171 /// \headerfile <x86intrin.h>
4172 ///
4173 /// This intrinsic corresponds to the <c> VPACKSSWB / PACKSSWB </c> instruction.
4174 ///
4175 /// \param __a
4176 /// A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
4177 /// a signed integer and is converted to a 8-bit signed integer with
4178 /// saturation. Values greater than 0x7F are saturated to 0x7F. Values less
4179 /// than 0x80 are saturated to 0x80. The converted [8 x i8] values are
4180 /// written to the lower 64 bits of the result.
4181 /// \param __b
4182 /// A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
4183 /// a signed integer and is converted to a 8-bit signed integer with
4184 /// saturation. Values greater than 0x7F are saturated to 0x7F. Values less
4185 /// than 0x80 are saturated to 0x80. The converted [8 x i8] values are
4186 /// written to the higher 64 bits of the result.
4187 /// \returns A 128-bit vector of [16 x i8] containing the converted values.
4188 static __inline__ __m128i __DEFAULT_FN_ATTRS
4189 _mm_packs_epi16(__m128i __a, __m128i __b)
4190 {
4191  return (__m128i)__builtin_ia32_packsswb128((__v8hi)__a, (__v8hi)__b);
4192 }
4193 
4194 /// Converts 32-bit signed integers from both 128-bit integer vector
4195 /// operands into 16-bit signed integers, and packs the results into the
4196 /// destination. Positive values greater than 0x7FFF are saturated to 0x7FFF.
4197 /// Negative values less than 0x8000 are saturated to 0x8000.
4198 ///
4199 /// \headerfile <x86intrin.h>
4200 ///
4201 /// This intrinsic corresponds to the <c> VPACKSSDW / PACKSSDW </c> instruction.
4202 ///
4203 /// \param __a
4204 /// A 128-bit integer vector of [4 x i32]. Each 32-bit element is treated as
4205 /// a signed integer and is converted to a 16-bit signed integer with
4206 /// saturation. Values greater than 0x7FFF are saturated to 0x7FFF. Values
4207 /// less than 0x8000 are saturated to 0x8000. The converted [4 x i16] values
4208 /// are written to the lower 64 bits of the result.
4209 /// \param __b
4210 /// A 128-bit integer vector of [4 x i32]. Each 32-bit element is treated as
4211 /// a signed integer and is converted to a 16-bit signed integer with
4212 /// saturation. Values greater than 0x7FFF are saturated to 0x7FFF. Values
4213 /// less than 0x8000 are saturated to 0x8000. The converted [4 x i16] values
4214 /// are written to the higher 64 bits of the result.
4215 /// \returns A 128-bit vector of [8 x i16] containing the converted values.
4216 static __inline__ __m128i __DEFAULT_FN_ATTRS
4217 _mm_packs_epi32(__m128i __a, __m128i __b)
4218 {
4219  return (__m128i)__builtin_ia32_packssdw128((__v4si)__a, (__v4si)__b);
4220 }
4221 
4222 /// Converts 16-bit signed integers from both 128-bit integer vector
4223 /// operands into 8-bit unsigned integers, and packs the results into the
4224 /// destination. Values greater than 0xFF are saturated to 0xFF. Values less
4225 /// than 0x00 are saturated to 0x00.
4226 ///
4227 /// \headerfile <x86intrin.h>
4228 ///
4229 /// This intrinsic corresponds to the <c> VPACKUSWB / PACKUSWB </c> instruction.
4230 ///
4231 /// \param __a
4232 /// A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
4233 /// a signed integer and is converted to an 8-bit unsigned integer with
4234 /// saturation. Values greater than 0xFF are saturated to 0xFF. Values less
4235 /// than 0x00 are saturated to 0x00. The converted [8 x i8] values are
4236 /// written to the lower 64 bits of the result.
4237 /// \param __b
4238 /// A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
4239 /// a signed integer and is converted to an 8-bit unsigned integer with
4240 /// saturation. Values greater than 0xFF are saturated to 0xFF. Values less
4241 /// than 0x00 are saturated to 0x00. The converted [8 x i8] values are
4242 /// written to the higher 64 bits of the result.
4243 /// \returns A 128-bit vector of [16 x i8] containing the converted values.
4244 static __inline__ __m128i __DEFAULT_FN_ATTRS
4245 _mm_packus_epi16(__m128i __a, __m128i __b)
4246 {
4247  return (__m128i)__builtin_ia32_packuswb128((__v8hi)__a, (__v8hi)__b);
4248 }
4249 
4250 /// Extracts 16 bits from a 128-bit integer vector of [8 x i16], using
4251 /// the immediate-value parameter as a selector.
4252 ///
4253 /// \headerfile <x86intrin.h>
4254 ///
4255 /// This intrinsic corresponds to the <c> VPEXTRW / PEXTRW </c> instruction.
4256 ///
4257 /// \param __a
4258 /// A 128-bit integer vector.
4259 /// \param __imm
4260 /// An immediate value. Bits [2:0] selects values from \a __a to be assigned
4261 /// to bits[15:0] of the result. \n
4262 /// 000: assign values from bits [15:0] of \a __a. \n
4263 /// 001: assign values from bits [31:16] of \a __a. \n
4264 /// 010: assign values from bits [47:32] of \a __a. \n
4265 /// 011: assign values from bits [63:48] of \a __a. \n
4266 /// 100: assign values from bits [79:64] of \a __a. \n
4267 /// 101: assign values from bits [95:80] of \a __a. \n
4268 /// 110: assign values from bits [111:96] of \a __a. \n
4269 /// 111: assign values from bits [127:112] of \a __a.
4270 /// \returns An integer, whose lower 16 bits are selected from the 128-bit
4271 /// integer vector parameter and the remaining bits are assigned zeros.
4272 #define _mm_extract_epi16(a, imm) \
4273  (int)(unsigned short)__builtin_ia32_vec_ext_v8hi((__v8hi)(__m128i)(a), \
4274  (int)(imm))
4275 
4276 /// Constructs a 128-bit integer vector by first making a copy of the
4277 /// 128-bit integer vector parameter, and then inserting the lower 16 bits
4278 /// of an integer parameter into an offset specified by the immediate-value
4279 /// parameter.
4280 ///
4281 /// \headerfile <x86intrin.h>
4282 ///
4283 /// This intrinsic corresponds to the <c> VPINSRW / PINSRW </c> instruction.
4284 ///
4285 /// \param __a
4286 /// A 128-bit integer vector of [8 x i16]. This vector is copied to the
4287 /// result and then one of the eight elements in the result is replaced by
4288 /// the lower 16 bits of \a __b.
4289 /// \param __b
4290 /// An integer. The lower 16 bits of this parameter are written to the
4291 /// result beginning at an offset specified by \a __imm.
4292 /// \param __imm
4293 /// An immediate value specifying the bit offset in the result at which the
4294 /// lower 16 bits of \a __b are written.
4295 /// \returns A 128-bit integer vector containing the constructed values.
4296 #define _mm_insert_epi16(a, b, imm) \
4297  (__m128i)__builtin_ia32_vec_set_v8hi((__v8hi)(__m128i)(a), (int)(b), \
4298  (int)(imm))
4299 
4300 /// Copies the values of the most significant bits from each 8-bit
4301 /// element in a 128-bit integer vector of [16 x i8] to create a 16-bit mask
4302 /// value, zero-extends the value, and writes it to the destination.
4303 ///
4304 /// \headerfile <x86intrin.h>
4305 ///
4306 /// This intrinsic corresponds to the <c> VPMOVMSKB / PMOVMSKB </c> instruction.
4307 ///
4308 /// \param __a
4309 /// A 128-bit integer vector containing the values with bits to be extracted.
4310 /// \returns The most significant bits from each 8-bit element in \a __a,
4311 /// written to bits [15:0]. The other bits are assigned zeros.
4312 static __inline__ int __DEFAULT_FN_ATTRS
4313 _mm_movemask_epi8(__m128i __a)
4314 {
4315  return __builtin_ia32_pmovmskb128((__v16qi)__a);
4316 }
4317 
4318 /// Constructs a 128-bit integer vector by shuffling four 32-bit
4319 /// elements of a 128-bit integer vector parameter, using the immediate-value
4320 /// parameter as a specifier.
4321 ///
4322 /// \headerfile <x86intrin.h>
4323 ///
4324 /// \code
4325 /// __m128i _mm_shuffle_epi32(__m128i a, const int imm);
4326 /// \endcode
4327 ///
4328 /// This intrinsic corresponds to the <c> VPSHUFD / PSHUFD </c> instruction.
4329 ///
4330 /// \param a
4331 /// A 128-bit integer vector containing the values to be copied.
4332 /// \param imm
4333 /// An immediate value containing an 8-bit value specifying which elements to
4334 /// copy from a. The destinations within the 128-bit destination are assigned
4335 /// values as follows: \n
4336 /// Bits [1:0] are used to assign values to bits [31:0] of the result. \n
4337 /// Bits [3:2] are used to assign values to bits [63:32] of the result. \n
4338 /// Bits [5:4] are used to assign values to bits [95:64] of the result. \n
4339 /// Bits [7:6] are used to assign values to bits [127:96] of the result. \n
4340 /// Bit value assignments: \n
4341 /// 00: assign values from bits [31:0] of \a a. \n
4342 /// 01: assign values from bits [63:32] of \a a. \n
4343 /// 10: assign values from bits [95:64] of \a a. \n
4344 /// 11: assign values from bits [127:96] of \a a.
4345 /// \returns A 128-bit integer vector containing the shuffled values.
4346 #define _mm_shuffle_epi32(a, imm) \
4347  (__m128i)__builtin_ia32_pshufd((__v4si)(__m128i)(a), (int)(imm))
4348 
4349 /// Constructs a 128-bit integer vector by shuffling four lower 16-bit
4350 /// elements of a 128-bit integer vector of [8 x i16], using the immediate
4351 /// value parameter as a specifier.
4352 ///
4353 /// \headerfile <x86intrin.h>
4354 ///
4355 /// \code
4356 /// __m128i _mm_shufflelo_epi16(__m128i a, const int imm);
4357 /// \endcode
4358 ///
4359 /// This intrinsic corresponds to the <c> VPSHUFLW / PSHUFLW </c> instruction.
4360 ///
4361 /// \param a
4362 /// A 128-bit integer vector of [8 x i16]. Bits [127:64] are copied to bits
4363 /// [127:64] of the result.
4364 /// \param imm
4365 /// An 8-bit immediate value specifying which elements to copy from \a a. \n
4366 /// Bits[1:0] are used to assign values to bits [15:0] of the result. \n
4367 /// Bits[3:2] are used to assign values to bits [31:16] of the result. \n
4368 /// Bits[5:4] are used to assign values to bits [47:32] of the result. \n
4369 /// Bits[7:6] are used to assign values to bits [63:48] of the result. \n
4370 /// Bit value assignments: \n
4371 /// 00: assign values from bits [15:0] of \a a. \n
4372 /// 01: assign values from bits [31:16] of \a a. \n
4373 /// 10: assign values from bits [47:32] of \a a. \n
4374 /// 11: assign values from bits [63:48] of \a a. \n
4375 /// \returns A 128-bit integer vector containing the shuffled values.
4376 #define _mm_shufflelo_epi16(a, imm) \
4377  (__m128i)__builtin_ia32_pshuflw((__v8hi)(__m128i)(a), (int)(imm))
4378 
4379 /// Constructs a 128-bit integer vector by shuffling four upper 16-bit
4380 /// elements of a 128-bit integer vector of [8 x i16], using the immediate
4381 /// value parameter as a specifier.
4382 ///
4383 /// \headerfile <x86intrin.h>
4384 ///
4385 /// \code
4386 /// __m128i _mm_shufflehi_epi16(__m128i a, const int imm);
4387 /// \endcode
4388 ///
4389 /// This intrinsic corresponds to the <c> VPSHUFHW / PSHUFHW </c> instruction.
4390 ///
4391 /// \param a
4392 /// A 128-bit integer vector of [8 x i16]. Bits [63:0] are copied to bits
4393 /// [63:0] of the result.
4394 /// \param imm
4395 /// An 8-bit immediate value specifying which elements to copy from \a a. \n
4396 /// Bits[1:0] are used to assign values to bits [79:64] of the result. \n
4397 /// Bits[3:2] are used to assign values to bits [95:80] of the result. \n
4398 /// Bits[5:4] are used to assign values to bits [111:96] of the result. \n
4399 /// Bits[7:6] are used to assign values to bits [127:112] of the result. \n
4400 /// Bit value assignments: \n
4401 /// 00: assign values from bits [79:64] of \a a. \n
4402 /// 01: assign values from bits [95:80] of \a a. \n
4403 /// 10: assign values from bits [111:96] of \a a. \n
4404 /// 11: assign values from bits [127:112] of \a a. \n
4405 /// \returns A 128-bit integer vector containing the shuffled values.
4406 #define _mm_shufflehi_epi16(a, imm) \
4407  (__m128i)__builtin_ia32_pshufhw((__v8hi)(__m128i)(a), (int)(imm))
4408 
4409 /// Unpacks the high-order (index 8-15) values from two 128-bit vectors
4410 /// of [16 x i8] and interleaves them into a 128-bit vector of [16 x i8].
4411 ///
4412 /// \headerfile <x86intrin.h>
4413 ///
4414 /// This intrinsic corresponds to the <c> VPUNPCKHBW / PUNPCKHBW </c>
4415 /// instruction.
4416 ///
4417 /// \param __a
4418 /// A 128-bit vector of [16 x i8].
4419 /// Bits [71:64] are written to bits [7:0] of the result. \n
4420 /// Bits [79:72] are written to bits [23:16] of the result. \n
4421 /// Bits [87:80] are written to bits [39:32] of the result. \n
4422 /// Bits [95:88] are written to bits [55:48] of the result. \n
4423 /// Bits [103:96] are written to bits [71:64] of the result. \n
4424 /// Bits [111:104] are written to bits [87:80] of the result. \n
4425 /// Bits [119:112] are written to bits [103:96] of the result. \n
4426 /// Bits [127:120] are written to bits [119:112] of the result.
4427 /// \param __b
4428 /// A 128-bit vector of [16 x i8]. \n
4429 /// Bits [71:64] are written to bits [15:8] of the result. \n
4430 /// Bits [79:72] are written to bits [31:24] of the result. \n
4431 /// Bits [87:80] are written to bits [47:40] of the result. \n
4432 /// Bits [95:88] are written to bits [63:56] of the result. \n
4433 /// Bits [103:96] are written to bits [79:72] of the result. \n
4434 /// Bits [111:104] are written to bits [95:88] of the result. \n
4435 /// Bits [119:112] are written to bits [111:104] of the result. \n
4436 /// Bits [127:120] are written to bits [127:120] of the result.
4437 /// \returns A 128-bit vector of [16 x i8] containing the interleaved values.
4438 static __inline__ __m128i __DEFAULT_FN_ATTRS
4439 _mm_unpackhi_epi8(__m128i __a, __m128i __b)
4440 {
4441  return (__m128i)__builtin_shufflevector((__v16qi)__a, (__v16qi)__b, 8, 16+8, 9, 16+9, 10, 16+10, 11, 16+11, 12, 16+12, 13, 16+13, 14, 16+14, 15, 16+15);
4442 }
4443 
4444 /// Unpacks the high-order (index 4-7) values from two 128-bit vectors of
4445 /// [8 x i16] and interleaves them into a 128-bit vector of [8 x i16].
4446 ///
4447 /// \headerfile <x86intrin.h>
4448 ///
4449 /// This intrinsic corresponds to the <c> VPUNPCKHWD / PUNPCKHWD </c>
4450 /// instruction.
4451 ///
4452 /// \param __a
4453 /// A 128-bit vector of [8 x i16].
4454 /// Bits [79:64] are written to bits [15:0] of the result. \n
4455 /// Bits [95:80] are written to bits [47:32] of the result. \n
4456 /// Bits [111:96] are written to bits [79:64] of the result. \n
4457 /// Bits [127:112] are written to bits [111:96] of the result.
4458 /// \param __b
4459 /// A 128-bit vector of [8 x i16].
4460 /// Bits [79:64] are written to bits [31:16] of the result. \n
4461 /// Bits [95:80] are written to bits [63:48] of the result. \n
4462 /// Bits [111:96] are written to bits [95:80] of the result. \n
4463 /// Bits [127:112] are written to bits [127:112] of the result.
4464 /// \returns A 128-bit vector of [8 x i16] containing the interleaved values.
4465 static __inline__ __m128i __DEFAULT_FN_ATTRS
4466 _mm_unpackhi_epi16(__m128i __a, __m128i __b)
4467 {
4468  return (__m128i)__builtin_shufflevector((__v8hi)__a, (__v8hi)__b, 4, 8+4, 5, 8+5, 6, 8+6, 7, 8+7);
4469 }
4470 
4471 /// Unpacks the high-order (index 2,3) values from two 128-bit vectors of
4472 /// [4 x i32] and interleaves them into a 128-bit vector of [4 x i32].
4473 ///
4474 /// \headerfile <x86intrin.h>
4475 ///
4476 /// This intrinsic corresponds to the <c> VPUNPCKHDQ / PUNPCKHDQ </c>
4477 /// instruction.
4478 ///
4479 /// \param __a
4480 /// A 128-bit vector of [4 x i32]. \n
4481 /// Bits [95:64] are written to bits [31:0] of the destination. \n
4482 /// Bits [127:96] are written to bits [95:64] of the destination.
4483 /// \param __b
4484 /// A 128-bit vector of [4 x i32]. \n
4485 /// Bits [95:64] are written to bits [64:32] of the destination. \n
4486 /// Bits [127:96] are written to bits [127:96] of the destination.
4487 /// \returns A 128-bit vector of [4 x i32] containing the interleaved values.
4488 static __inline__ __m128i __DEFAULT_FN_ATTRS
4489 _mm_unpackhi_epi32(__m128i __a, __m128i __b)
4490 {
4491  return (__m128i)__builtin_shufflevector((__v4si)__a, (__v4si)__b, 2, 4+2, 3, 4+3);
4492 }
4493 
4494 /// Unpacks the high-order 64-bit elements from two 128-bit vectors of
4495 /// [2 x i64] and interleaves them into a 128-bit vector of [2 x i64].
4496 ///
4497 /// \headerfile <x86intrin.h>
4498 ///
4499 /// This intrinsic corresponds to the <c> VPUNPCKHQDQ / PUNPCKHQDQ </c>
4500 /// instruction.
4501 ///
4502 /// \param __a
4503 /// A 128-bit vector of [2 x i64]. \n
4504 /// Bits [127:64] are written to bits [63:0] of the destination.
4505 /// \param __b
4506 /// A 128-bit vector of [2 x i64]. \n
4507 /// Bits [127:64] are written to bits [127:64] of the destination.
4508 /// \returns A 128-bit vector of [2 x i64] containing the interleaved values.
4509 static __inline__ __m128i __DEFAULT_FN_ATTRS
4510 _mm_unpackhi_epi64(__m128i __a, __m128i __b)
4511 {
4512  return (__m128i)__builtin_shufflevector((__v2di)__a, (__v2di)__b, 1, 2+1);
4513 }
4514 
4515 /// Unpacks the low-order (index 0-7) values from two 128-bit vectors of
4516 /// [16 x i8] and interleaves them into a 128-bit vector of [16 x i8].
4517 ///
4518 /// \headerfile <x86intrin.h>
4519 ///
4520 /// This intrinsic corresponds to the <c> VPUNPCKLBW / PUNPCKLBW </c>
4521 /// instruction.
4522 ///
4523 /// \param __a
4524 /// A 128-bit vector of [16 x i8]. \n
4525 /// Bits [7:0] are written to bits [7:0] of the result. \n
4526 /// Bits [15:8] are written to bits [23:16] of the result. \n
4527 /// Bits [23:16] are written to bits [39:32] of the result. \n
4528 /// Bits [31:24] are written to bits [55:48] of the result. \n
4529 /// Bits [39:32] are written to bits [71:64] of the result. \n
4530 /// Bits [47:40] are written to bits [87:80] of the result. \n
4531 /// Bits [55:48] are written to bits [103:96] of the result. \n
4532 /// Bits [63:56] are written to bits [119:112] of the result.
4533 /// \param __b
4534 /// A 128-bit vector of [16 x i8].
4535 /// Bits [7:0] are written to bits [15:8] of the result. \n
4536 /// Bits [15:8] are written to bits [31:24] of the result. \n
4537 /// Bits [23:16] are written to bits [47:40] of the result. \n
4538 /// Bits [31:24] are written to bits [63:56] of the result. \n
4539 /// Bits [39:32] are written to bits [79:72] of the result. \n
4540 /// Bits [47:40] are written to bits [95:88] of the result. \n
4541 /// Bits [55:48] are written to bits [111:104] of the result. \n
4542 /// Bits [63:56] are written to bits [127:120] of the result.
4543 /// \returns A 128-bit vector of [16 x i8] containing the interleaved values.
4544 static __inline__ __m128i __DEFAULT_FN_ATTRS
4545 _mm_unpacklo_epi8(__m128i __a, __m128i __b)
4546 {
4547  return (__m128i)__builtin_shufflevector((__v16qi)__a, (__v16qi)__b, 0, 16+0, 1, 16+1, 2, 16+2, 3, 16+3, 4, 16+4, 5, 16+5, 6, 16+6, 7, 16+7);
4548 }
4549 
4550 /// Unpacks the low-order (index 0-3) values from each of the two 128-bit
4551 /// vectors of [8 x i16] and interleaves them into a 128-bit vector of
4552 /// [8 x i16].
4553 ///
4554 /// \headerfile <x86intrin.h>
4555 ///
4556 /// This intrinsic corresponds to the <c> VPUNPCKLWD / PUNPCKLWD </c>
4557 /// instruction.
4558 ///
4559 /// \param __a
4560 /// A 128-bit vector of [8 x i16].
4561 /// Bits [15:0] are written to bits [15:0] of the result. \n
4562 /// Bits [31:16] are written to bits [47:32] of the result. \n
4563 /// Bits [47:32] are written to bits [79:64] of the result. \n
4564 /// Bits [63:48] are written to bits [111:96] of the result.
4565 /// \param __b
4566 /// A 128-bit vector of [8 x i16].
4567 /// Bits [15:0] are written to bits [31:16] of the result. \n
4568 /// Bits [31:16] are written to bits [63:48] of the result. \n
4569 /// Bits [47:32] are written to bits [95:80] of the result. \n
4570 /// Bits [63:48] are written to bits [127:112] of the result.
4571 /// \returns A 128-bit vector of [8 x i16] containing the interleaved values.
4572 static __inline__ __m128i __DEFAULT_FN_ATTRS
4573 _mm_unpacklo_epi16(__m128i __a, __m128i __b)
4574 {
4575  return (__m128i)__builtin_shufflevector((__v8hi)__a, (__v8hi)__b, 0, 8+0, 1, 8+1, 2, 8+2, 3, 8+3);
4576 }
4577 
4578 /// Unpacks the low-order (index 0,1) values from two 128-bit vectors of
4579 /// [4 x i32] and interleaves them into a 128-bit vector of [4 x i32].
4580 ///
4581 /// \headerfile <x86intrin.h>
4582 ///
4583 /// This intrinsic corresponds to the <c> VPUNPCKLDQ / PUNPCKLDQ </c>
4584 /// instruction.
4585 ///
4586 /// \param __a
4587 /// A 128-bit vector of [4 x i32]. \n
4588 /// Bits [31:0] are written to bits [31:0] of the destination. \n
4589 /// Bits [63:32] are written to bits [95:64] of the destination.
4590 /// \param __b
4591 /// A 128-bit vector of [4 x i32]. \n
4592 /// Bits [31:0] are written to bits [64:32] of the destination. \n
4593 /// Bits [63:32] are written to bits [127:96] of the destination.
4594 /// \returns A 128-bit vector of [4 x i32] containing the interleaved values.
4595 static __inline__ __m128i __DEFAULT_FN_ATTRS
4596 _mm_unpacklo_epi32(__m128i __a, __m128i __b)
4597 {
4598  return (__m128i)__builtin_shufflevector((__v4si)__a, (__v4si)__b, 0, 4+0, 1, 4+1);
4599 }
4600 
4601 /// Unpacks the low-order 64-bit elements from two 128-bit vectors of
4602 /// [2 x i64] and interleaves them into a 128-bit vector of [2 x i64].
4603 ///
4604 /// \headerfile <x86intrin.h>
4605 ///
4606 /// This intrinsic corresponds to the <c> VPUNPCKLQDQ / PUNPCKLQDQ </c>
4607 /// instruction.
4608 ///
4609 /// \param __a
4610 /// A 128-bit vector of [2 x i64]. \n
4611 /// Bits [63:0] are written to bits [63:0] of the destination. \n
4612 /// \param __b
4613 /// A 128-bit vector of [2 x i64]. \n
4614 /// Bits [63:0] are written to bits [127:64] of the destination. \n
4615 /// \returns A 128-bit vector of [2 x i64] containing the interleaved values.
4616 static __inline__ __m128i __DEFAULT_FN_ATTRS
4617 _mm_unpacklo_epi64(__m128i __a, __m128i __b)
4618 {
4619  return (__m128i)__builtin_shufflevector((__v2di)__a, (__v2di)__b, 0, 2+0);
4620 }
4621 
4622 /// Returns the lower 64 bits of a 128-bit integer vector as a 64-bit
4623 /// integer.
4624 ///
4625 /// \headerfile <x86intrin.h>
4626 ///
4627 /// This intrinsic corresponds to the <c> MOVDQ2Q </c> instruction.
4628 ///
4629 /// \param __a
4630 /// A 128-bit integer vector operand. The lower 64 bits are moved to the
4631 /// destination.
4632 /// \returns A 64-bit integer containing the lower 64 bits of the parameter.
4633 static __inline__ __m64 __DEFAULT_FN_ATTRS
4634 _mm_movepi64_pi64(__m128i __a)
4635 {
4636  return (__m64)__a[0];
4637 }
4638 
4639 /// Moves the 64-bit operand to a 128-bit integer vector, zeroing the
4640 /// upper bits.
4641 ///
4642 /// \headerfile <x86intrin.h>
4643 ///
4644 /// This intrinsic corresponds to the <c> MOVD+VMOVQ </c> instruction.
4645 ///
4646 /// \param __a
4647 /// A 64-bit value.
4648 /// \returns A 128-bit integer vector. The lower 64 bits contain the value from
4649 /// the operand. The upper 64 bits are assigned zeros.
4650 static __inline__ __m128i __DEFAULT_FN_ATTRS
4652 {
4653  return __extension__ (__m128i)(__v2di){ (long long)__a, 0 };
4654 }
4655 
4656 /// Moves the lower 64 bits of a 128-bit integer vector to a 128-bit
4657 /// integer vector, zeroing the upper bits.
4658 ///
4659 /// \headerfile <x86intrin.h>
4660 ///
4661 /// This intrinsic corresponds to the <c> VMOVQ / MOVQ </c> instruction.
4662 ///
4663 /// \param __a
4664 /// A 128-bit integer vector operand. The lower 64 bits are moved to the
4665 /// destination.
4666 /// \returns A 128-bit integer vector. The lower 64 bits contain the value from
4667 /// the operand. The upper 64 bits are assigned zeros.
4668 static __inline__ __m128i __DEFAULT_FN_ATTRS
4669 _mm_move_epi64(__m128i __a)
4670 {
4671  return __builtin_shufflevector((__v2di)__a, _mm_setzero_si128(), 0, 2);
4672 }
4673 
4674 /// Unpacks the high-order 64-bit elements from two 128-bit vectors of
4675 /// [2 x double] and interleaves them into a 128-bit vector of [2 x
4676 /// double].
4677 ///
4678 /// \headerfile <x86intrin.h>
4679 ///
4680 /// This intrinsic corresponds to the <c> VUNPCKHPD / UNPCKHPD </c> instruction.
4681 ///
4682 /// \param __a
4683 /// A 128-bit vector of [2 x double]. \n
4684 /// Bits [127:64] are written to bits [63:0] of the destination.
4685 /// \param __b
4686 /// A 128-bit vector of [2 x double]. \n
4687 /// Bits [127:64] are written to bits [127:64] of the destination.
4688 /// \returns A 128-bit vector of [2 x double] containing the interleaved values.
4689 static __inline__ __m128d __DEFAULT_FN_ATTRS
4690 _mm_unpackhi_pd(__m128d __a, __m128d __b)
4691 {
4692  return __builtin_shufflevector((__v2df)__a, (__v2df)__b, 1, 2+1);
4693 }
4694 
4695 /// Unpacks the low-order 64-bit elements from two 128-bit vectors
4696 /// of [2 x double] and interleaves them into a 128-bit vector of [2 x
4697 /// double].
4698 ///
4699 /// \headerfile <x86intrin.h>
4700 ///
4701 /// This intrinsic corresponds to the <c> VUNPCKLPD / UNPCKLPD </c> instruction.
4702 ///
4703 /// \param __a
4704 /// A 128-bit vector of [2 x double]. \n
4705 /// Bits [63:0] are written to bits [63:0] of the destination.
4706 /// \param __b
4707 /// A 128-bit vector of [2 x double]. \n
4708 /// Bits [63:0] are written to bits [127:64] of the destination.
4709 /// \returns A 128-bit vector of [2 x double] containing the interleaved values.
4710 static __inline__ __m128d __DEFAULT_FN_ATTRS
4711 _mm_unpacklo_pd(__m128d __a, __m128d __b)
4712 {
4713  return __builtin_shufflevector((__v2df)__a, (__v2df)__b, 0, 2+0);
4714 }
4715 
4716 /// Extracts the sign bits of the double-precision values in the 128-bit
4717 /// vector of [2 x double], zero-extends the value, and writes it to the
4718 /// low-order bits of the destination.
4719 ///
4720 /// \headerfile <x86intrin.h>
4721 ///
4722 /// This intrinsic corresponds to the <c> VMOVMSKPD / MOVMSKPD </c> instruction.
4723 ///
4724 /// \param __a
4725 /// A 128-bit vector of [2 x double] containing the values with sign bits to
4726 /// be extracted.
4727 /// \returns The sign bits from each of the double-precision elements in \a __a,
4728 /// written to bits [1:0]. The remaining bits are assigned values of zero.
4729 static __inline__ int __DEFAULT_FN_ATTRS
4730 _mm_movemask_pd(__m128d __a)
4731 {
4732  return __builtin_ia32_movmskpd((__v2df)__a);
4733 }
4734 
4735 
4736 /// Constructs a 128-bit floating-point vector of [2 x double] from two
4737 /// 128-bit vector parameters of [2 x double], using the immediate-value
4738 /// parameter as a specifier.
4739 ///
4740 /// \headerfile <x86intrin.h>
4741 ///
4742 /// \code
4743 /// __m128d _mm_shuffle_pd(__m128d a, __m128d b, const int i);
4744 /// \endcode
4745 ///
4746 /// This intrinsic corresponds to the <c> VSHUFPD / SHUFPD </c> instruction.
4747 ///
4748 /// \param a
4749 /// A 128-bit vector of [2 x double].
4750 /// \param b
4751 /// A 128-bit vector of [2 x double].
4752 /// \param i
4753 /// An 8-bit immediate value. The least significant two bits specify which
4754 /// elements to copy from \a a and \a b: \n
4755 /// Bit[0] = 0: lower element of \a a copied to lower element of result. \n
4756 /// Bit[0] = 1: upper element of \a a copied to lower element of result. \n
4757 /// Bit[1] = 0: lower element of \a b copied to upper element of result. \n
4758 /// Bit[1] = 1: upper element of \a b copied to upper element of result. \n
4759 /// \returns A 128-bit vector of [2 x double] containing the shuffled values.
4760 #define _mm_shuffle_pd(a, b, i) \
4761  (__m128d)__builtin_ia32_shufpd((__v2df)(__m128d)(a), (__v2df)(__m128d)(b), \
4762  (int)(i))
4763 
4764 /// Casts a 128-bit floating-point vector of [2 x double] into a 128-bit
4765 /// floating-point vector of [4 x float].
4766 ///
4767 /// \headerfile <x86intrin.h>
4768 ///
4769 /// This intrinsic has no corresponding instruction.
4770 ///
4771 /// \param __a
4772 /// A 128-bit floating-point vector of [2 x double].
4773 /// \returns A 128-bit floating-point vector of [4 x float] containing the same
4774 /// bitwise pattern as the parameter.
4775 static __inline__ __m128 __DEFAULT_FN_ATTRS
4776 _mm_castpd_ps(__m128d __a)
4777 {
4778  return (__m128)__a;
4779 }
4780 
4781 /// Casts a 128-bit floating-point vector of [2 x double] into a 128-bit
4782 /// integer vector.
4783 ///
4784 /// \headerfile <x86intrin.h>
4785 ///
4786 /// This intrinsic has no corresponding instruction.
4787 ///
4788 /// \param __a
4789 /// A 128-bit floating-point vector of [2 x double].
4790 /// \returns A 128-bit integer vector containing the same bitwise pattern as the
4791 /// parameter.
4792 static __inline__ __m128i __DEFAULT_FN_ATTRS
4793 _mm_castpd_si128(__m128d __a)
4794 {
4795  return (__m128i)__a;
4796 }
4797 
4798 /// Casts a 128-bit floating-point vector of [4 x float] into a 128-bit
4799 /// floating-point vector of [2 x double].
4800 ///
4801 /// \headerfile <x86intrin.h>
4802 ///
4803 /// This intrinsic has no corresponding instruction.
4804 ///
4805 /// \param __a
4806 /// A 128-bit floating-point vector of [4 x float].
4807 /// \returns A 128-bit floating-point vector of [2 x double] containing the same
4808 /// bitwise pattern as the parameter.
4809 static __inline__ __m128d __DEFAULT_FN_ATTRS
4810 _mm_castps_pd(__m128 __a)
4811 {
4812  return (__m128d)__a;
4813 }
4814 
4815 /// Casts a 128-bit floating-point vector of [4 x float] into a 128-bit
4816 /// integer vector.
4817 ///
4818 /// \headerfile <x86intrin.h>
4819 ///
4820 /// This intrinsic has no corresponding instruction.
4821 ///
4822 /// \param __a
4823 /// A 128-bit floating-point vector of [4 x float].
4824 /// \returns A 128-bit integer vector containing the same bitwise pattern as the
4825 /// parameter.
4826 static __inline__ __m128i __DEFAULT_FN_ATTRS
4827 _mm_castps_si128(__m128 __a)
4828 {
4829  return (__m128i)__a;
4830 }
4831 
4832 /// Casts a 128-bit integer vector into a 128-bit floating-point vector
4833 /// of [4 x float].
4834 ///
4835 /// \headerfile <x86intrin.h>
4836 ///
4837 /// This intrinsic has no corresponding instruction.
4838 ///
4839 /// \param __a
4840 /// A 128-bit integer vector.
4841 /// \returns A 128-bit floating-point vector of [4 x float] containing the same
4842 /// bitwise pattern as the parameter.
4843 static __inline__ __m128 __DEFAULT_FN_ATTRS
4844 _mm_castsi128_ps(__m128i __a)
4845 {
4846  return (__m128)__a;
4847 }
4848 
4849 /// Casts a 128-bit integer vector into a 128-bit floating-point vector
4850 /// of [2 x double].
4851 ///
4852 /// \headerfile <x86intrin.h>
4853 ///
4854 /// This intrinsic has no corresponding instruction.
4855 ///
4856 /// \param __a
4857 /// A 128-bit integer vector.
4858 /// \returns A 128-bit floating-point vector of [2 x double] containing the same
4859 /// bitwise pattern as the parameter.
4860 static __inline__ __m128d __DEFAULT_FN_ATTRS
4861 _mm_castsi128_pd(__m128i __a)
4862 {
4863  return (__m128d)__a;
4864 }
4865 
4866 #if defined(__cplusplus)
4867 extern "C" {
4868 #endif
4869 
4870 /// Indicates that a spin loop is being executed for the purposes of
4871 /// optimizing power consumption during the loop.
4872 ///
4873 /// \headerfile <x86intrin.h>
4874 ///
4875 /// This intrinsic corresponds to the <c> PAUSE </c> instruction.
4876 ///
4877 void _mm_pause(void);
4878 
4879 #if defined(__cplusplus)
4880 } // extern "C"
4881 #endif
4882 #undef __DEFAULT_FN_ATTRS
4883 #undef __DEFAULT_FN_ATTRS_MMX
4884 
4885 #define _MM_SHUFFLE2(x, y) (((x) << 1) | (y))
4886 
4887 #define _MM_DENORMALS_ZERO_ON (0x0040)
4888 #define _MM_DENORMALS_ZERO_OFF (0x0000)
4889 
4890 #define _MM_DENORMALS_ZERO_MASK (0x0040)
4891 
4892 #define _MM_GET_DENORMALS_ZERO_MODE() (_mm_getcsr() & _MM_DENORMALS_ZERO_MASK)
4893 #define _MM_SET_DENORMALS_ZERO_MODE(x) (_mm_setcsr((_mm_getcsr() & ~_MM_DENORMALS_ZERO_MASK) | (x)))
4894 
4895 #endif /* __EMMINTRIN_H */
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpgt_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:499
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_slli_epi32(__m128i __a, int __count)
Left-shifts each 32-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2855
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srli_epi16(__m128i __a, int __count)
Right-shifts each of 16-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3034
static __inline__ int __DEFAULT_FN_ATTRS _mm_cvtsd_si32(__m128d __a)
Converts the low-order element of a 128-bit vector of [2 x double] into a 32-bit signed integer value...
Definition: emmintrin.h:1386
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomige_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1261
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set_epi16(short __w7, short __w6, short __w5, short __w4, short __w3, short __w2, short __w1, short __w0)
Initializes the 16-bit values in a 128-bit vector of [8 x i16] with the specified 16-bit integer valu...
Definition: emmintrin.h:3676
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_sqrt_pd(__m128d __a)
Calculates the square root of the each of two values stored in a 128-bit vector of [2 x double]...
Definition: emmintrin.h:255
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_setr_epi8(char __b0, char __b1, char __b2, char __b3, char __b4, char __b5, char __b6, char __b7, char __b8, char __b9, char __b10, char __b11, char __b12, char __b13, char __b14, char __b15)
Constructs a 128-bit integer vector, initialized in reverse order with the specified 8-bit integral v...
Definition: emmintrin.h:3940
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sll_epi32(__m128i __a, __m128i __count)
Left-shifts each 32-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2874
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srai_epi16(__m128i __a, int __count)
Right-shifts each 16-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2932
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomile_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1209
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvttpd_epi32(__m128d __a)
Converts the two double-precision floating-point elements of a 128-bit vector of [2 x double] into tw...
Definition: emmintrin.h:1484
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cvtepi32_pd(__m128i __a)
Converts the lower two integer elements of a 128-bit vector of [4 x i32] into two double-precision fl...
Definition: emmintrin.h:1349
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi64(__m64 __q)
Initializes both values in a 128-bit vector of [2 x i64] with the specified 64-bit value...
Definition: emmintrin.h:3762
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srli_epi32(__m128i __a, int __count)
Right-shifts each of 32-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3072
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_setr_epi32(int __i0, int __i1, int __i2, int __i3)
Constructs a 128-bit integer vector, initialized in reverse order with the specified 32-bit integral ...
Definition: emmintrin.h:3862
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packs_epi32(__m128i __a, __m128i __b)
Converts 32-bit signed integers from both 128-bit integer vector operands into 16-bit signed integers...
Definition: emmintrin.h:4217
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi64x(long long __q)
Initializes both values in a 128-bit integer vector with the specified 64-bit integer value...
Definition: emmintrin.h:3743
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_set_sd(double __w)
Constructs a 128-bit floating-point vector of [2 x double].
Definition: emmintrin.h:1788
static __inline__ int __DEFAULT_FN_ATTRS _mm_comile_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1055
static __inline__ void __DEFAULT_FN_ATTRS _mm_storel_pd(double *__dp, __m128d __a)
Stores the lower 64 bits of a 128-bit vector of [2 x double] to a memory location.
Definition: emmintrin.h:2061
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_castps_si128(__m128 __a)
Casts a 128-bit floating-point vector of [4 x float] into a 128-bit integer vector.
Definition: emmintrin.h:4827
void _mm_pause(void)
Indicates that a spin loop is being executed for the purposes of optimizing power consumption during ...
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_set1_pd(double __w)
Constructs a 128-bit floating-point vector of [2 x double], with each of the two double-precision flo...
Definition: emmintrin.h:1806
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi32(__m128i __a, __m128i __b)
Adds the corresponding elements of two 128-bit vectors of [4 x i32], saving the lower 32 bits of each...
Definition: emmintrin.h:2130
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_slli_epi16(__m128i __a, int __count)
Left-shifts each 16-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2817
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpgt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:770
static __inline__ int __DEFAULT_FN_ATTRS _mm_comige_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1107
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_loadu_si64(void const *__a)
Loads a 64-bit integer value to the low element of a 128-bit integer vector and clears the upper elem...
Definition: emmintrin.h:1672
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi64(__m128i __a, __m128i __b)
Subtracts the corresponding elements of two [2 x i64] vectors.
Definition: emmintrin.h:2620
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_max_sd(__m128d __a, __m128d __b)
Compares lower 64-bit double-precision values of both operands, and returns the greater of the pair o...
Definition: emmintrin.h:323
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srli_epi64(__m128i __a, int __count)
Right-shifts each of 64-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3110
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomigt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1235
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_unpacklo_pd(__m128d __a, __m128d __b)
Unpacks the low-order 64-bit elements from two 128-bit vectors of [2 x double] and interleaves them i...
Definition: emmintrin.h:4711
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_cvtpd_ps(__m128d __a)
Converts the two double-precision floating-point elements of a 128-bit vector of [2 x double] into tw...
Definition: emmintrin.h:1306
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi8(__m128i __a, __m128i __b)
Subtracts the corresponding 8-bit integer values in the operands.
Definition: emmintrin.h:2547
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtpd_epi32(__m128d __a)
Converts the two double-precision floating-point elements of a 128-bit vector of [2 x double] into tw...
Definition: emmintrin.h:1369
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_mul_su32(__m64 __a, __m64 __b)
Multiplies 32-bit unsigned integer values contained in the lower bits of the two 64-bit integer vecto...
Definition: emmintrin.h:2488
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_loadl_pd(__m128d __a, double const *__dp)
Loads a double-precision value into the low-order bits of a 128-bit vector of [2 x double]...
Definition: emmintrin.h:1747
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epu16(__m128i __a, __m128i __b)
Subtracts corresponding 16-bit unsigned integer values in the input and returns the differences in th...
Definition: emmintrin.h:2702
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpackhi_epi32(__m128i __a, __m128i __b)
Unpacks the high-order (index 2,3) values from two 128-bit vectors of [4 x i32] and interleaves them ...
Definition: emmintrin.h:4489
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi16(__m128i __a, __m128i __b)
Compares each of the corresponding 16-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:3167
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_setr_epi16(short __w0, short __w1, short __w2, short __w3, short __w4, short __w5, short __w6, short __w7)
Constructs a 128-bit integer vector, initialized in reverse order with the specified 16-bit integral ...
Definition: emmintrin.h:3893
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpngt_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:651
static __inline__ void __DEFAULT_FN_ATTRS _mm_maskmoveu_si128(__m128i __d, __m128i __n, char *__p)
Moves bytes selected by the mask from the first operand to the specified unaligned memory location...
Definition: emmintrin.h:4018
double __m128d __attribute__((__vector_size__(16)))
Definition: emmintrin.h:29
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi16(__m128i __a, __m128i __b)
Compares each of the corresponding signed 16-bit values of the 128-bit integer vectors to determine i...
Definition: emmintrin.h:3292
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set_epi64(__m64 __q1, __m64 __q0)
Initializes both 64-bit values in a 128-bit vector of [2 x i64] with the specified 64-bit integer val...
Definition: emmintrin.h:3608
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_castsi128_ps(__m128i __a)
Casts a 128-bit integer vector into a 128-bit floating-point vector of [4 x float].
Definition: emmintrin.h:4844
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_loadl_epi64(__m128i const *__p)
Returns a vector of [2 x i64] where the lower element is taken from the lower element of the operand...
Definition: emmintrin.h:3546
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomineq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1287
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_min_sd(__m128d __a, __m128d __b)
Compares lower 64-bit double-precision values of both operands, and returns the lesser of the pair of...
Definition: emmintrin.h:279
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_add_sd(__m128d __a, __m128d __b)
Adds lower double-precision values in both operands and returns the sum in the lower 64 bits of the r...
Definition: emmintrin.h:67
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi8(char __b)
Initializes all values in a 128-bit vector of [16 x i8] with the specified 8-bit value.
Definition: emmintrin.h:3819
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpge_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:796
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_movpi64_epi64(__m64 __a)
Moves the 64-bit operand to a 128-bit integer vector, zeroing the upper bits.
Definition: emmintrin.h:4651
static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_movepi64_pi64(__m128i __a)
Returns the lower 64 bits of a 128-bit integer vector as a 64-bit integer.
Definition: emmintrin.h:4634
static __inline__ __m128d __DEFAULT_FN_ATTRS_MMX _mm_cvtpi32_pd(__m64 __a)
Converts the two signed 32-bit integer elements of a 64-bit vector of [2 x i32] into two double-preci...
Definition: emmintrin.h:1556
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_loadu_pd(double const *__dp)
Loads a 128-bit floating-point vector of [2 x double] from an unaligned memory location.
Definition: emmintrin.h:1652
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sll_epi64(__m128i __a, __m128i __count)
Left-shifts each 64-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2912
void _mm_mfence(void)
Forces strong memory ordering (serialization) between load and store instructions preceding this inst...
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_load1_pd(double const *__dp)
Loads a double-precision floating-point value from a specified memory location and duplicates it to b...
Definition: emmintrin.h:1608
static __inline__ void int __a
Definition: emmintrin.h:4099
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_castpd_ps(__m128d __a)
Casts a 128-bit floating-point vector of [2 x double] into a 128-bit floating-point vector of [4 x fl...
Definition: emmintrin.h:4776
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi32(__m128i __a, __m128i __b)
Compares each of the corresponding signed 32-bit values of the 128-bit integer vectors to determine i...
Definition: emmintrin.h:3313
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpeq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:695
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_loadu_si128(__m128i const *__p)
Moves packed integer values from an unaligned 128-bit memory location to elements in a 128-bit intege...
Definition: emmintrin.h:3525
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set_epi64x(long long __q1, long long __q0)
Initializes both 64-bit values in a 128-bit vector of [2 x i64] with the specified 64-bit integer val...
Definition: emmintrin.h:3586
static __inline unsigned char unsigned int unsigned int unsigned int * __p
Definition: adxintrin.h:38
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi16(__m128i __a, __m128i __b)
Adds, with saturation, the corresponding elements of two 128-bit signed [8 x i16] vectors...
Definition: emmintrin.h:2213
static __inline__ void __DEFAULT_FN_ATTRS _mm_store_pd(double *__dp, __m128d __a)
Moves packed double-precision values from a 128-bit vector of [2 x double] to a memory location...
Definition: emmintrin.h:1941
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomilt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1183
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packs_epi16(__m128i __a, __m128i __b)
Converts 16-bit signed integers from both 128-bit integer vector operands into 8-bit signed integers...
Definition: emmintrin.h:4189
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srai_epi32(__m128i __a, int __count)
Right-shifts each 32-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2972
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi32(__m128i __a, __m128i __b)
Subtracts the corresponding 32-bit integer values in the operands.
Definition: emmintrin.h:2583
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvttpd_pi32(__m128d __a)
Converts the two double-precision floating-point elements of a 128-bit vector of [2 x double] into tw...
Definition: emmintrin.h:1539
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_load_sd(double const *__dp)
Loads a 64-bit double-precision value to the low element of a 128-bit integer vector and clears the u...
Definition: emmintrin.h:1693
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi32(__m128i __a, __m128i __b)
Compares each of the corresponding signed 32-bit values of the 128-bit integer vectors to determine i...
Definition: emmintrin.h:3250
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_setr_pd(double __w, double __x)
Constructs a 128-bit floating-point vector of [2 x double], initialized in reverse order with the spe...
Definition: emmintrin.h:1865
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtsi32_si128(int __a)
Returns a vector of [4 x i32] where the lowest element is the input operand and the remaining element...
Definition: emmintrin.h:3438
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnge_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:672
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_load_si128(__m128i const *__p)
Moves packed integer values from an aligned 128-bit memory location to elements in a 128-bit integer ...
Definition: emmintrin.h:3509
static __inline__ void __DEFAULT_FN_ATTRS _mm_storer_pd(double *__dp, __m128d __a)
Stores two double-precision values, in reverse order, from a 128-bit vector of [2 x double] to a 16-b...
Definition: emmintrin.h:2023
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_setzero_pd(void)
Constructs a 128-bit floating-point vector of [2 x double] initialized to zero.
Definition: emmintrin.h:1880
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_sub_pd(__m128d __a, __m128d __b)
Subtracts two 128-bit vectors of [2 x double].
Definition: emmintrin.h:128
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_castps_pd(__m128 __a)
Casts a 128-bit floating-point vector of [4 x float] into a 128-bit floating-point vector of [2 x dou...
Definition: emmintrin.h:4810
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_div_sd(__m128d __a, __m128d __b)
Divides the lower double-precision value of the first operand by the lower double-precision value of ...
Definition: emmintrin.h:192
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_cvtsd_ss(__m128 __a, __m128d __b)
Converts the lower double-precision floating-point element of a 128-bit vector of [2 x double]...
Definition: emmintrin.h:1411
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cvtps_pd(__m128 __a)
Converts the lower two single-precision floating-point elements of a 128-bit vector of [4 x float] in...
Definition: emmintrin.h:1326
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi64(__m128i __a, __m128i __b)
Adds the corresponding elements of two 128-bit vectors of [2 x i64], saving the lower 64 bits of each...
Definition: emmintrin.h:2170
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_set_pd1(double __w)
Constructs a 128-bit floating-point vector of [2 x double], with each of the two double-precision flo...
Definition: emmintrin.h:1824
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmplt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:720
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmple_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:478
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_sub_si64(__m64 __a, __m64 __b)
Subtracts signed or unsigned 64-bit integer values and writes the difference to the corresponding bit...
Definition: emmintrin.h:2602
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvttps_epi32(__m128 __a)
Converts a vector of [4 x float] into a vector of [4 x i32], truncating the result when it is inexact...
Definition: emmintrin.h:3422
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_set_pd(double __w, double __x)
Constructs a 128-bit floating-point vector of [2 x double] initialized with the specified double-prec...
Definition: emmintrin.h:1844
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_slli_epi64(__m128i __a, int __count)
Left-shifts each 64-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2893
static __inline__ int __DEFAULT_FN_ATTRS _mm_ucomieq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1157
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epi16(__m128i __a, __m128i __b)
Compares corresponding elements of two 128-bit signed [8 x i16] vectors, saving the smaller value fro...
Definition: emmintrin.h:2389
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_load_pd(double const *__dp)
Loads a 128-bit floating-point vector of [2 x double] from an aligned memory location.
Definition: emmintrin.h:1590
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sra_epi16(__m128i __a, __m128i __count)
Right-shifts each 16-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2952
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpngt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:952
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_add_pd(__m128d __a, __m128d __b)
Adds two 128-bit vectors of [2 x double].
Definition: emmintrin.h:86
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpord_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:824
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnge_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:978
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi8(__m128i __a, __m128i __b)
Adds the corresponding elements of two 128-bit vectors of [16 x i8], saving the lower 8 bits of each ...
Definition: emmintrin.h:2086
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi8(__m128i __a, __m128i __b)
Adds, with saturation, the corresponding elements of two 128-bit signed [16 x i8] vectors...
Definition: emmintrin.h:2191
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi32(__m128i __a, __m128i __b)
Compares each of the corresponding 32-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:3186
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_move_epi64(__m128i __a)
Moves the lower 64 bits of a 128-bit integer vector to a 128-bit integer vector, zeroing the upper bi...
Definition: emmintrin.h:4669
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_or_pd(__m128d __a, __m128d __b)
Performs a bitwise OR of two 128-bit vectors of [2 x double].
Definition: emmintrin.h:400
static __inline__ void __DEFAULT_FN_ATTRS _mm_store_sd(double *__dp, __m128d __a)
Stores the lower 64 bits of a 128-bit vector of [2 x double] to a memory location.
Definition: emmintrin.h:1919
static __inline__ vector float vector float __b
Definition: altivec.h:534
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnle_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:927
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding signed 8-bit values of the 128-bit integer vectors to determine if...
Definition: emmintrin.h:3206
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnlt_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:609
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_div_pd(__m128d __a, __m128d __b)
Performs an element-by-element division of two 128-bit vectors of [2 x double].
Definition: emmintrin.h:212
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpacklo_epi64(__m128i __a, __m128i __b)
Unpacks the low-order 64-bit elements from two 128-bit vectors of [2 x i64] and interleaves them into...
Definition: emmintrin.h:4617
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_sqrt_sd(__m128d __a, __m128d __b)
Calculates the square root of the lower double-precision value of the second operand and returns it i...
Definition: emmintrin.h:237
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set_epi32(int __i3, int __i2, int __i1, int __i0)
Initializes the 32-bit values in a 128-bit vector of [4 x i32] with the specified 32-bit integer valu...
Definition: emmintrin.h:3636
void _mm_clflush(void const *__p)
The cache line containing __p is flushed and invalidated from all caches in the coherency domain...
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi16(__m128i __a, __m128i __b)
Compares each of the corresponding signed 16-bit values of the 128-bit integer vectors to determine i...
Definition: emmintrin.h:3229
static __inline unsigned char unsigned int __x
Definition: adxintrin.h:36
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi16(__m128i __a, __m128i __b)
Adds the corresponding elements of two 128-bit vectors of [8 x i16], saving the lower 16 bits of each...
Definition: emmintrin.h:2108
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_unpackhi_pd(__m128d __a, __m128d __b)
Unpacks the high-order 64-bit elements from two 128-bit vectors of [2 x double] and interleaves them ...
Definition: emmintrin.h:4690
static __inline__ void __DEFAULT_FN_ATTRS _mm_stream_si128(__m128i *__p, __m128i __a)
Stores a 128-bit integer vector to a 128-bit aligned memory location.
Definition: emmintrin.h:4079
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_min_epu8(__m128i __a, __m128i __b)
Compares corresponding elements of two 128-bit unsigned [16 x i8] vectors, saving the smaller value f...
Definition: emmintrin.h:2409
static __inline__ void __DEFAULT_FN_ATTRS _mm_storeu_pd(double *__dp, __m128d __a)
Stores a 128-bit vector of [2 x double] into an unaligned memory location.
Definition: emmintrin.h:2000
static __inline__ double __DEFAULT_FN_ATTRS _mm_cvtsd_f64(__m128d __a)
Returns the low-order element of a 128-bit vector of [2 x double] as a double-precision floating-poin...
Definition: emmintrin.h:1573
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_mul_sd(__m128d __a, __m128d __b)
Multiplies lower double-precision values in both operands and returns the product in the lower 64 bit...
Definition: emmintrin.h:150
static __inline__ void __DEFAULT_FN_ATTRS _mm_storel_epi64(__m128i *__p, __m128i __a)
Stores the lower 64 bits of a 128-bit integer vector of [2 x i64] to a memory location.
Definition: emmintrin.h:4037
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sra_epi32(__m128i __a, __m128i __count)
Right-shifts each 32-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2992
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_sub_sd(__m128d __a, __m128d __b)
Subtracts the lower double-precision value of the second operand from the lower double-precision valu...
Definition: emmintrin.h:109
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_and_pd(__m128d __a, __m128d __b)
Performs a bitwise AND of two 128-bit vectors of [2 x double].
Definition: emmintrin.h:361
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_castpd_si128(__m128d __a)
Casts a 128-bit floating-point vector of [2 x double] into a 128-bit integer vector.
Definition: emmintrin.h:4793
static __inline__ int __DEFAULT_FN_ATTRS _mm_comineq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1133
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mullo_epi16(__m128i __a, __m128i __b)
Multiplies the corresponding elements of two signed [8 x i16] vectors, saving the lower 16 bits of ea...
Definition: emmintrin.h:2469
static __inline__ int __DEFAULT_FN_ATTRS _mm_cvttsd_si32(__m128d __a)
Converts the low-order element of a [2 x double] vector into a 32-bit signed integer value...
Definition: emmintrin.h:1502
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_and_si128(__m128i __a, __m128i __b)
Performs a bitwise AND of two 128-bit integer vectors.
Definition: emmintrin.h:2720
static __inline__ int __DEFAULT_FN_ATTRS _mm_cvtsi128_si32(__m128i __a)
Moves the least significant 32 bits of a vector of [4 x i32] to a 32-bit signed integer value...
Definition: emmintrin.h:3473
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set_epi8(char __b15, char __b14, char __b13, char __b12, char __b11, char __b10, char __b9, char __b8, char __b7, char __b6, char __b5, char __b4, char __b3, char __b2, char __b1, char __b0)
Initializes the 8-bit values in a 128-bit vector of [16 x i8] with the specified 8-bit integer values...
Definition: emmintrin.h:3724
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi32(__m128i __a, __m128i __count)
Right-shifts each of 32-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3091
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpacklo_epi32(__m128i __a, __m128i __b)
Unpacks the low-order (index 0,1) values from two 128-bit vectors of [4 x i32] and interleaves them i...
Definition: emmintrin.h:4596
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnlt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:902
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_castsi128_pd(__m128i __a)
Casts a 128-bit integer vector into a 128-bit floating-point vector of [2 x double].
Definition: emmintrin.h:4861
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_min_pd(__m128d __a, __m128d __b)
Performs element-by-element comparison of the two 128-bit vectors of [2 x double] and returns the vec...
Definition: emmintrin.h:299
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epi16(__m128i __a, __m128i __b)
Compares corresponding elements of two 128-bit signed [8 x i16] vectors, saving the greater value fro...
Definition: emmintrin.h:2349
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_max_epu8(__m128i __a, __m128i __b)
Compares corresponding elements of two 128-bit unsigned [16 x i8] vectors, saving the greater value f...
Definition: emmintrin.h:2369
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpord_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:543
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_xor_pd(__m128d __a, __m128d __b)
Performs a bitwise XOR of two 128-bit vectors of [2 x double].
Definition: emmintrin.h:418
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_setr_epi64(__m64 __q0, __m64 __q1)
Constructs a 128-bit integer vector, initialized in reverse order with the specified 64-bit integral ...
Definition: emmintrin.h:3839
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epu16(__m128i __a, __m128i __b)
Adds, with saturation, the corresponding elements of two 128-bit unsigned [8 x i16] vectors...
Definition: emmintrin.h:2255
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpacklo_epi8(__m128i __a, __m128i __b)
Unpacks the low-order (index 0-7) values from two 128-bit vectors of [16 x i8] and interleaves them i...
Definition: emmintrin.h:4545
static __inline__ void __DEFAULT_FN_ATTRS _mm_storeh_pd(double *__dp, __m128d __a)
Stores the upper 64 bits of a 128-bit vector of [2 x double] to a memory location.
Definition: emmintrin.h:2041
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cvtsi32_sd(__m128d __a, int __b)
Converts a 32-bit signed integer value, in the second parameter, into a double-precision floating-poi...
Definition: emmintrin.h:1434
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_setzero_si128(void)
Creates a 128-bit integer vector initialized to zero.
Definition: emmintrin.h:3954
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mul_epu32(__m128i __a, __m128i __b)
Multiplies 32-bit unsigned integer values contained in the lower bits of the corresponding elements o...
Definition: emmintrin.h:2507
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_andnot_pd(__m128d __a, __m128d __b)
Performs a bitwise AND of two 128-bit vectors of [2 x double], using the one&#39;s complement of the valu...
Definition: emmintrin.h:382
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpackhi_epi64(__m128i __a, __m128i __b)
Unpacks the high-order 64-bit elements from two 128-bit vectors of [2 x i64] and interleaves them int...
Definition: emmintrin.h:4510
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmple_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:745
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_madd_epi16(__m128i __a, __m128i __b)
Multiplies the corresponding elements of two 128-bit signed [8 x i16] vectors, producing eight interm...
Definition: emmintrin.h:2329
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi32(int __i)
Initializes all values in a 128-bit vector of [4 x i32] with the specified 32-bit value...
Definition: emmintrin.h:3781
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvtpd_pi32(__m128d __a)
Converts the two double-precision floating-point elements of a 128-bit vector of [2 x double] into tw...
Definition: emmintrin.h:1519
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpacklo_epi16(__m128i __a, __m128i __b)
Unpacks the low-order (index 0-3) values from each of the two 128-bit vectors of [8 x i16] and interl...
Definition: emmintrin.h:4573
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi16(__m128i __a, __m128i __b)
Subtracts corresponding 16-bit signed integer values in the input and returns the differences in the ...
Definition: emmintrin.h:2662
static __inline__ int __DEFAULT_FN_ATTRS _mm_comieq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1003
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhi_epi16(__m128i __a, __m128i __b)
Multiplies the corresponding elements of two signed [8 x i16] vectors, saving the upper 16 bits of ea...
Definition: emmintrin.h:2429
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packus_epi16(__m128i __a, __m128i __b)
Converts 16-bit signed integers from both 128-bit integer vector operands into 8-bit unsigned integer...
Definition: emmintrin.h:4245
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epu8(__m128i __a, __m128i __b)
Subtracts corresponding 8-bit unsigned integer values in the input and returns the differences in the...
Definition: emmintrin.h:2682
static __inline__ void __DEFAULT_FN_ATTRS _mm_store_si128(__m128i *__p, __m128i __b)
Stores a 128-bit integer vector to a memory location aligned on a 128-bit boundary.
Definition: emmintrin.h:3972
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding signed 8-bit values of the 128-bit integer vectors to determine if...
Definition: emmintrin.h:3271
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_undefined_si128(void)
Generates a 128-bit vector of [4 x i32] with unspecified content.
Definition: emmintrin.h:3564
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sad_epu8(__m128i __a, __m128i __b)
Computes the absolute differences of corresponding 8-bit integer values in two 128-bit vectors...
Definition: emmintrin.h:2529
static __inline__ int __DEFAULT_FN_ATTRS _mm_comigt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1081
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhi_epu16(__m128i __a, __m128i __b)
Multiplies the corresponding elements of two unsigned [8 x i16] vectors, saving the upper 16 bits of ...
Definition: emmintrin.h:2449
static __inline__ int __DEFAULT_FN_ATTRS _mm_comilt_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:1029
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpackhi_epi8(__m128i __a, __m128i __b)
Unpacks the high-order (index 8-15) values from two 128-bit vectors of [16 x i8] and interleaves them...
Definition: emmintrin.h:4439
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi64(__m128i __a, __m128i __count)
Right-shifts each of 64-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3129
static __inline__ void __DEFAULT_FN_ATTRS _mm_stream_pd(double *__p, __m128d __a)
Stores a 128-bit floating point vector of [2 x double] to a 128-bit aligned memory location...
Definition: emmintrin.h:4060
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi16(short __w)
Initializes all values in a 128-bit vector of [8 x i16] with the specified 16-bit value...
Definition: emmintrin.h:3800
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_move_sd(__m128d __a, __m128d __b)
Constructs a 128-bit floating-point vector of [2 x double].
Definition: emmintrin.h:1901
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpunord_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:852
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpge_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:520
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpneq_sd(__m128d __a, __m128d __b)
Compares the lower double-precision floating-point values in each of the two 128-bit floating-point v...
Definition: emmintrin.h:877
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_avg_epu16(__m128i __a, __m128i __b)
Computes the rounded avarages of corresponding elements of two 128-bit unsigned [8 x i16] vectors...
Definition: emmintrin.h:2299
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmplt_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:457
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi16(__m128i __a, __m128i __count)
Right-shifts each of 16-bit values in the 128-bit integer vector operand by the specified number of b...
Definition: emmintrin.h:3053
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sll_epi16(__m128i __a, __m128i __count)
Left-shifts each 16-bit value in the 128-bit integer vector operand by the specified number of bits...
Definition: emmintrin.h:2836
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_loadh_pd(__m128d __a, double const *__dp)
Loads a double-precision value into the high-order bits of a 128-bit vector of [2 x double]...
Definition: emmintrin.h:1720
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpneq_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:588
#define __DEFAULT_FN_ATTRS
Definition: emmintrin.h:48
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtps_epi32(__m128 __a)
Converts a vector of [4 x float] into a vector of [4 x i32].
Definition: emmintrin.h:3405
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi16(__m128i __a, __m128i __b)
Subtracts the corresponding 16-bit integer values in the operands.
Definition: emmintrin.h:2565
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpeq_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] for...
Definition: emmintrin.h:437
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_max_pd(__m128d __a, __m128d __b)
Performs element-by-element comparison of the two 128-bit vectors of [2 x double] and returns the vec...
Definition: emmintrin.h:343
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cvtss_sd(__m128d __a, __m128 __b)
Converts the lower single-precision floating-point element of a 128-bit vector of [4 x float]...
Definition: emmintrin.h:1460
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_xor_si128(__m128i __a, __m128i __b)
Performs a bitwise exclusive OR of two 128-bit integer vectors.
Definition: emmintrin.h:2775
static __inline__ int __DEFAULT_FN_ATTRS _mm_movemask_pd(__m128d __a)
Extracts the sign bits of the double-precision values in the 128-bit vector of [2 x double]...
Definition: emmintrin.h:4730
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi8(__m128i __a, __m128i __b)
Subtracts corresponding 8-bit signed integer values in the input and returns the differences in the c...
Definition: emmintrin.h:2641
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_add_si64(__m64 __a, __m64 __b)
Adds two signed or unsigned 64-bit integer values, returning the lower 64 bits of the sum...
Definition: emmintrin.h:2148
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_or_si128(__m128i __a, __m128i __b)
Performs a bitwise OR of two 128-bit integer vectors.
Definition: emmintrin.h:2757
static __inline__ void __DEFAULT_FN_ATTRS _mm_storeu_si128(__m128i *__p, __m128i __b)
Stores a 128-bit integer vector to an unaligned memory location.
Definition: emmintrin.h:3988
static __inline__ int __DEFAULT_FN_ATTRS _mm_movemask_epi8(__m128i __a)
Copies the values of the most significant bits from each 8-bit element in a 128-bit integer vector of...
Definition: emmintrin.h:4313
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_mul_pd(__m128d __a, __m128d __b)
Multiplies two 128-bit vectors of [2 x double].
Definition: emmintrin.h:169
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_loadr_pd(double const *__dp)
Loads two double-precision values, in reverse order, from an aligned memory location into a 128-bit v...
Definition: emmintrin.h:1634
static __inline__ void __DEFAULT_FN_ATTRS _mm_store1_pd(double *__dp, __m128d __a)
Moves the lower 64 bits of a 128-bit vector of [2 x double] twice to the upper and lower 64 bits of a...
Definition: emmintrin.h:1961
static __inline__ __m128 __DEFAULT_FN_ATTRS _mm_cvtepi32_ps(__m128i __a)
Converts a vector of [4 x i32] into a vector of [4 x float].
Definition: emmintrin.h:3389
#define __DEFAULT_FN_ATTRS_MMX
Definition: emmintrin.h:49
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_avg_epu8(__m128i __a, __m128i __b)
Computes the rounded avarages of corresponding elements of two 128-bit unsigned [16 x i8] vectors...
Definition: emmintrin.h:2275
static __inline__ vector float vector float vector float __c
Definition: altivec.h:4199
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epu8(__m128i __a, __m128i __b)
Adds, with saturation, the corresponding elements of two 128-bit unsigned [16 x i8] vectors...
Definition: emmintrin.h:2234
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_undefined_pd(void)
Constructs a 128-bit floating-point vector of [2 x double] with unspecified content.
Definition: emmintrin.h:1768
void _mm_lfence(void)
Forces strong memory ordering (serialization) between load instructions preceding this instruction an...
static __inline__ void __DEFAULT_FN_ATTRS _mm_store_pd1(double *__dp, __m128d __a)
Moves the lower 64 bits of a 128-bit vector of [2 x double] twice to the upper and lower 64 bits of a...
Definition: emmintrin.h:1982
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpnle_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:630
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_andnot_si128(__m128i __a, __m128i __b)
Performs a bitwise AND of two 128-bit integer vectors, using the one&#39;s complement of the values conta...
Definition: emmintrin.h:2740
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding 8-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:3148
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_unpackhi_epi16(__m128i __a, __m128i __b)
Unpacks the high-order (index 4-7) values from two 128-bit vectors of [8 x i16] and interleaves them ...
Definition: emmintrin.h:4466
static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cmpunord_pd(__m128d __a, __m128d __b)
Compares each of the corresponding double-precision values of the 128-bit vectors of [2 x double] to ...
Definition: emmintrin.h:567