1 /*===---- pmmintrin.h - SSE3 intrinsics ------------------------------------===
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:
10 * The above copyright notice and this permission notice shall be included in
11 * all copies or substantial portions of the Software.
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
21 *===-----------------------------------------------------------------------===
27 #include <emmintrin.h>
29 /* Define the default attributes for the functions in this file. */
30 #define __DEFAULT_FN_ATTRS \
31 __attribute__((__always_inline__, __nodebug__, __target__("sse3")))
33 /// \brief Loads data from an unaligned memory location to elements in a 128-bit
34 /// vector. If the address of the data is not 16-byte aligned, the
35 /// instruction may read two adjacent aligned blocks of memory to retrieve
36 /// the requested data.
38 /// \headerfile <x86intrin.h>
40 /// This intrinsic corresponds to the <c> VLDDQU </c> instruction.
43 /// A pointer to a 128-bit integer vector containing integer values.
44 /// \returns A 128-bit vector containing the moved values.
45 static __inline__ __m128i __DEFAULT_FN_ATTRS
46 _mm_lddqu_si128(__m128i const *__p)
48 return (__m128i)__builtin_ia32_lddqu((char const *)__p);
51 /// \brief Adds the even-indexed values and subtracts the odd-indexed values of
52 /// two 128-bit vectors of [4 x float].
54 /// \headerfile <x86intrin.h>
56 /// This intrinsic corresponds to the <c> VADDSUBPS </c> instruction.
59 /// A 128-bit vector of [4 x float] containing the left source operand.
61 /// A 128-bit vector of [4 x float] containing the right source operand.
62 /// \returns A 128-bit vector of [4 x float] containing the alternating sums and
63 /// differences of both operands.
64 static __inline__ __m128 __DEFAULT_FN_ATTRS
65 _mm_addsub_ps(__m128 __a, __m128 __b)
67 return __builtin_ia32_addsubps((__v4sf)__a, (__v4sf)__b);
70 /// \brief Horizontally adds the adjacent pairs of values contained in two
71 /// 128-bit vectors of [4 x float].
73 /// \headerfile <x86intrin.h>
75 /// This intrinsic corresponds to the <c> VHADDPS </c> instruction.
78 /// A 128-bit vector of [4 x float] containing one of the source operands.
79 /// The horizontal sums of the values are stored in the lower bits of the
82 /// A 128-bit vector of [4 x float] containing one of the source operands.
83 /// The horizontal sums of the values are stored in the upper bits of the
85 /// \returns A 128-bit vector of [4 x float] containing the horizontal sums of
87 static __inline__ __m128 __DEFAULT_FN_ATTRS
88 _mm_hadd_ps(__m128 __a, __m128 __b)
90 return __builtin_ia32_haddps((__v4sf)__a, (__v4sf)__b);
93 /// \brief Horizontally subtracts the adjacent pairs of values contained in two
94 /// 128-bit vectors of [4 x float].
96 /// \headerfile <x86intrin.h>
98 /// This intrinsic corresponds to the <c> VHSUBPS </c> instruction.
101 /// A 128-bit vector of [4 x float] containing one of the source operands.
102 /// The horizontal differences between the values are stored in the lower
103 /// bits of the destination.
105 /// A 128-bit vector of [4 x float] containing one of the source operands.
106 /// The horizontal differences between the values are stored in the upper
107 /// bits of the destination.
108 /// \returns A 128-bit vector of [4 x float] containing the horizontal
109 /// differences of both operands.
110 static __inline__ __m128 __DEFAULT_FN_ATTRS
111 _mm_hsub_ps(__m128 __a, __m128 __b)
113 return __builtin_ia32_hsubps((__v4sf)__a, (__v4sf)__b);
116 /// \brief Moves and duplicates high-order (odd-indexed) values from a 128-bit
117 /// vector of [4 x float] to float values stored in a 128-bit vector of
120 /// \headerfile <x86intrin.h>
122 /// This intrinsic corresponds to the <c> VMOVSHDUP </c> instruction.
125 /// A 128-bit vector of [4 x float]. \n
126 /// Bits [127:96] of the source are written to bits [127:96] and [95:64] of
127 /// the destination. \n
128 /// Bits [63:32] of the source are written to bits [63:32] and [31:0] of the
130 /// \returns A 128-bit vector of [4 x float] containing the moved and duplicated
132 static __inline__ __m128 __DEFAULT_FN_ATTRS
133 _mm_movehdup_ps(__m128 __a)
135 return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 1, 1, 3, 3);
138 /// \brief Duplicates low-order (even-indexed) values from a 128-bit vector of
139 /// [4 x float] to float values stored in a 128-bit vector of [4 x float].
141 /// \headerfile <x86intrin.h>
143 /// This intrinsic corresponds to the <c> VMOVSLDUP </c> instruction.
146 /// A 128-bit vector of [4 x float] \n
147 /// Bits [95:64] of the source are written to bits [127:96] and [95:64] of
148 /// the destination. \n
149 /// Bits [31:0] of the source are written to bits [63:32] and [31:0] of the
151 /// \returns A 128-bit vector of [4 x float] containing the moved and duplicated
153 static __inline__ __m128 __DEFAULT_FN_ATTRS
154 _mm_moveldup_ps(__m128 __a)
156 return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 0, 0, 2, 2);
159 /// \brief Adds the even-indexed values and subtracts the odd-indexed values of
160 /// two 128-bit vectors of [2 x double].
162 /// \headerfile <x86intrin.h>
164 /// This intrinsic corresponds to the <c> VADDSUBPD </c> instruction.
167 /// A 128-bit vector of [2 x double] containing the left source operand.
169 /// A 128-bit vector of [2 x double] containing the right source operand.
170 /// \returns A 128-bit vector of [2 x double] containing the alternating sums
171 /// and differences of both operands.
172 static __inline__ __m128d __DEFAULT_FN_ATTRS
173 _mm_addsub_pd(__m128d __a, __m128d __b)
175 return __builtin_ia32_addsubpd((__v2df)__a, (__v2df)__b);
178 /// \brief Horizontally adds the pairs of values contained in two 128-bit
179 /// vectors of [2 x double].
181 /// \headerfile <x86intrin.h>
183 /// This intrinsic corresponds to the <c> VHADDPD </c> instruction.
186 /// A 128-bit vector of [2 x double] containing one of the source operands.
187 /// The horizontal sum of the values is stored in the lower bits of the
190 /// A 128-bit vector of [2 x double] containing one of the source operands.
191 /// The horizontal sum of the values is stored in the upper bits of the
193 /// \returns A 128-bit vector of [2 x double] containing the horizontal sums of
195 static __inline__ __m128d __DEFAULT_FN_ATTRS
196 _mm_hadd_pd(__m128d __a, __m128d __b)
198 return __builtin_ia32_haddpd((__v2df)__a, (__v2df)__b);
201 /// \brief Horizontally subtracts the pairs of values contained in two 128-bit
202 /// vectors of [2 x double].
204 /// \headerfile <x86intrin.h>
206 /// This intrinsic corresponds to the <c> VHSUBPD </c> instruction.
209 /// A 128-bit vector of [2 x double] containing one of the source operands.
210 /// The horizontal difference of the values is stored in the lower bits of
213 /// A 128-bit vector of [2 x double] containing one of the source operands.
214 /// The horizontal difference of the values is stored in the upper bits of
216 /// \returns A 128-bit vector of [2 x double] containing the horizontal
217 /// differences of both operands.
218 static __inline__ __m128d __DEFAULT_FN_ATTRS
219 _mm_hsub_pd(__m128d __a, __m128d __b)
221 return __builtin_ia32_hsubpd((__v2df)__a, (__v2df)__b);
224 /// \brief Moves and duplicates one double-precision value to double-precision
225 /// values stored in a 128-bit vector of [2 x double].
227 /// \headerfile <x86intrin.h>
230 /// __m128d _mm_loaddup_pd(double const * dp);
233 /// This intrinsic corresponds to the <c> VMOVDDUP </c> instruction.
236 /// A pointer to a double-precision value to be moved and duplicated.
237 /// \returns A 128-bit vector of [2 x double] containing the moved and
238 /// duplicated values.
239 #define _mm_loaddup_pd(dp) _mm_load1_pd(dp)
241 /// \brief Moves and duplicates the double-precision value in the lower bits of
242 /// a 128-bit vector of [2 x double] to double-precision values stored in a
243 /// 128-bit vector of [2 x double].
245 /// \headerfile <x86intrin.h>
247 /// This intrinsic corresponds to the <c> VMOVDDUP </c> instruction.
250 /// A 128-bit vector of [2 x double]. Bits [63:0] are written to bits
251 /// [127:64] and [63:0] of the destination.
252 /// \returns A 128-bit vector of [2 x double] containing the moved and
253 /// duplicated values.
254 static __inline__ __m128d __DEFAULT_FN_ATTRS
255 _mm_movedup_pd(__m128d __a)
257 return __builtin_shufflevector((__v2df)__a, (__v2df)__a, 0, 0);
260 #define _MM_DENORMALS_ZERO_ON (0x0040)
261 #define _MM_DENORMALS_ZERO_OFF (0x0000)
263 #define _MM_DENORMALS_ZERO_MASK (0x0040)
265 #define _MM_GET_DENORMALS_ZERO_MODE() (_mm_getcsr() & _MM_DENORMALS_ZERO_MASK)
266 #define _MM_SET_DENORMALS_ZERO_MODE(x) (_mm_setcsr((_mm_getcsr() & ~_MM_DENORMALS_ZERO_MASK) | (x)))
268 /// \brief Establishes a linear address memory range to be monitored and puts
269 /// the processor in the monitor event pending state. Data stored in the
270 /// monitored address range causes the processor to exit the pending state.
272 /// \headerfile <x86intrin.h>
274 /// This intrinsic corresponds to the <c> MONITOR </c> instruction.
277 /// The memory range to be monitored. The size of the range is determined by
278 /// CPUID function 0000_0005h.
279 /// \param __extensions
280 /// Optional extensions for the monitoring state.
282 /// Optional hints for the monitoring state.
283 static __inline__ void __DEFAULT_FN_ATTRS
284 _mm_monitor(void const *__p, unsigned __extensions, unsigned __hints)
286 __builtin_ia32_monitor((void *)__p, __extensions, __hints);
289 /// \brief Used with the MONITOR instruction to wait while the processor is in
290 /// the monitor event pending state. Data stored in the monitored address
291 /// range causes the processor to exit the pending state.
293 /// \headerfile <x86intrin.h>
295 /// This intrinsic corresponds to the <c> MWAIT </c> instruction.
297 /// \param __extensions
298 /// Optional extensions for the monitoring state, which may vary by
301 /// Optional hints for the monitoring state, which may vary by processor.
302 static __inline__ void __DEFAULT_FN_ATTRS
303 _mm_mwait(unsigned __extensions, unsigned __hints)
305 __builtin_ia32_mwait(__extensions, __hints);
308 #undef __DEFAULT_FN_ATTRS
310 #endif /* __PMMINTRIN_H */