/usr/include/fflas-ffpack/fflas/fflas_simd/simd128_double.inl is in fflas-ffpack-common 2.2.2-5.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 | /* -*- mode: C++; tab-width: 4; indent-tabs-mode: t; c-basic-offset: 4 -*- */
// vim:sts=4:sw=4:ts=4:noet:sr:cino=>s,f0,{0,g0,(0,\:0,t0,+0,=s
/*
* Copyright (C) 2014 the FFLAS-FFPACK group
*
* Written by Bastien Vialla<bastien.vialla@lirmm.fr>
* Brice Boyer (briceboyer) <boyer.brice@gmail.com>
*
*
* ========LICENCE========
* This file is part of the library FFLAS-FFPACK.
*
* FFLAS-FFPACK is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
* ========LICENCE========
*.
*/
#ifndef __FFLASFFPACK_fflas_ffpack_utils_simd128_double_INL
#define __FFLASFFPACK_fflas_ffpack_utils_simd128_double_INL
/*
* Simd128 specialized for double
*/
template <> struct Simd128_impl<true, false, true, 8> {
#if defined(__FFLASFFPACK_HAVE_SSE4_1_INSTRUCTIONS)
/*
* alias to 128 bit simd register
*/
using vect_t = __m128d;
/*
* define the scalar type corresponding to the specialization
*/
using scalar_t = double;
/*
* number of scalar_t in a simd register
*/
static const constexpr size_t vect_size = 2;
/*
* alignement required by scalar_t pointer to be loaded in a vect_t
*/
static const constexpr size_t alignment = 16;
/*
* Check if the pointer p is a multiple of alignemnt
*/
template <class T> static constexpr bool valid(T *p) { return (int64_t)p % alignment == 0; }
/*
* Check if the number n is a multiple of vect_size
*/
template <class T> static constexpr bool compliant(T n) { return n % vect_size == 0; }
/*
* Return vector of type vect_t with all elements set to zero.
* Return [0,0]
*/
static INLINE CONST vect_t zero() { return _mm_setzero_pd(); }
/*
* Broadcast double-precision (64-bit) floating-point value a to all elements of vect_t.
* Return [x,x]
*/
static INLINE CONST vect_t set1(const scalar_t x) { return _mm_set1_pd(x); }
/*
* Set packed double-precision (64-bit) floating-point elements in vect_t with the supplied values.
* Return [x1,x2]
*/
static INLINE CONST vect_t set(const scalar_t x1, const scalar_t x2) { return _mm_set_pd(x2, x1); }
/*
* Gather double-precision (64-bit) floating-point elements with indexes idx[0], ..., idx[3] from the address p in
* vect_t.
* Return [p[idx[0]], p[idx[1]]]
*/
template <class T> static INLINE PURE vect_t gather(const scalar_t *const p, const T *const idx) {
return _mm_set_pd(p[idx[1]], p[idx[0]]);
}
/*
* Load 128-bits (composed of 2 packed double-precision (64-bit) floating-point elements) from memory into vect_t.
* p must be aligned on a 16-byte boundary or a general-protection exception will be generated.
* Return [p[0], p[1]]
*/
static INLINE PURE vect_t load(const scalar_t *const p) { return _mm_load_pd(p); }
/*
* Load 128-bits (composed of 2 packed double-precision (64-bit) floating-point elements) from memory into vect_t.
* p does not need to be aligned on any particular boundary.
* Return [p[0], p[1]]
*/
static INLINE PURE vect_t loadu(const scalar_t *const p) { return _mm_loadu_pd(p); }
/*
* Store 128-bits (composed of 2 packed double-precision (64-bit) floating-point elements) from p into memory.
* p must be aligned on a 16-byte boundary or a general-protection exception will be generated.
*/
static INLINE void store(const scalar_t *p, const vect_t v) { _mm_store_pd(const_cast<scalar_t *>(p), v); }
/*
* Store 128-bits (composed of 2 packed double-precision (64-bit) floating-point elements) from p into memory.
* p must be aligned on a 16-byte boundary or a general-protection exception will be generated.
*/
static INLINE void storeu(const scalar_t *p, const vect_t v) { _mm_storeu_pd(const_cast<scalar_t *>(p), v); }
/*
* Store 128-bits (composed of 2 packed double-precision (64-bit) floating-point elements) from a into memory using
* a non-temporal memory hint.
* p must be aligned on a 16-byte boundary or a general-protection exception may be generated.
*/
static INLINE void stream(const scalar_t *p, const vect_t v) { _mm_stream_pd(const_cast<scalar_t *>(p), v); }
/*
* Shuffle double-precision (64-bit) floating-point elements using the control in s,
* and store the results in dst.
* Args : [a0, a1] double
* Return : [a[s[0]], a[s[1]]] double
*/
#if defined(__FFLASFFPACK_HAVE_AVX_INSTRUCTIONS)
template<uint8_t s>
static INLINE CONST vect_t shuffle(const vect_t a) {
return _mm_permute_pd(a, s);
}
#endif
/*
* Unpack and interleave double-precision (64-bit) floating-point elements from the low half of a and b, and store the results in dst.
* Args : [a0, a1] double
[b0, b1] double
* Return : [a0, b0] double
*/
static INLINE CONST vect_t unpacklo(const vect_t a, const vect_t b) { return _mm_unpacklo_pd(a, b); }
/*
* Unpack and interleave double-precision (64-bit) floating-point elements from the high half of a and b, and store the results in dst.
* Args : [a0, a1] double
[b0, b1] double
* Return : [a1, b1] double
*/
static INLINE CONST vect_t unpackhi(const vect_t a, const vect_t b) { return _mm_unpackhi_pd(a, b); }
/*
* Blend packed double-precision (64-bit) floating-point elements from a and b using control mask s,
* and store the results in dst.
* Args : [a0, a1] double
[b0, b1] double
* Return : [s[0]?a0:b0, s[1]?a1:b1] double
*/
template<uint8_t s>
static INLINE CONST vect_t blend(const vect_t a, const vect_t b) {
return _mm_blend_pd(a, b, s);
}
/*
* Blend packed double-precision (64-bit) floating-point elements from a and b using mask,
* and store the results in dst.
* Args : [a0, a1] double
[b0, b1] double
* Return : [mask[63]?a0:b0, mask[127]?a1:b1] double
*/
static INLINE CONST vect_t blendv(const vect_t a, const vect_t b, const vect_t mask) {
return _mm_blendv_pd(a, b, mask);
}
/*
* Add packed double-precision (64-bit) floating-point elements in a and b, and store the results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0+b0, a1+b1]
*/
static INLINE CONST vect_t add(const vect_t a, const vect_t b) { return _mm_add_pd(a, b); }
static INLINE vect_t addin(vect_t &a, const vect_t b) { return a = add(a, b); }
/*
* Subtract packed double-precision (64-bit) floating-point elements in b from packed double-precision (64-bit)
* floating-point elements in a, and store the results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0-b0, a1-b1]
*/
static INLINE CONST vect_t sub(const vect_t a, const vect_t b) { return _mm_sub_pd(a, b); }
static INLINE CONST vect_t subin(vect_t &a, const vect_t b) { return a = sub(a, b); }
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, and store the results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0*b0, a1*b1]
*/
static INLINE CONST vect_t mul(const vect_t a, const vect_t b) { return _mm_mul_pd(a, b); }
static INLINE CONST vect_t mulin(vect_t &a, const vect_t b) { return a = mul(a, b); }
/*
* Divide packed double-precision (64-bit) floating-point elements in a by packed elements in b,
* and store the results in dst.
* Args : [a0, a1], [b0, b1]
* Return : [a0/b0, a1/b1]
*/
static INLINE CONST vect_t div(const vect_t a, const vect_t b) { return _mm_div_pd(a, b); }
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, add the intermediate result to
* packed elements in c, and store the results in vect_t.
* Args : [a0, a1], [b0, b1], [c0, c1]
* Return : [a0*b0+c0, a1*b1+c1]
*/
static INLINE CONST vect_t fmadd(const vect_t c, const vect_t a, const vect_t b) {
#ifdef __FMA__
return _mm_fmadd_pd(a, b, c);
#else
return add(c, mul(a, b));
#endif
}
static INLINE CONST vect_t fmaddin(vect_t &c, const vect_t a, const vect_t b) { return c = fmadd(c, a, b); }
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, add the negated intermediate result
* to packed elements in c, and store the results in vect_t.
* Args : [a0, a1], [b0, b1], [c0, c1]
* Return : [-(a0*b0)+c0, -(a1*b1)+c1]
*/
static INLINE CONST vect_t fnmadd(const vect_t c, const vect_t a, const vect_t b) {
#ifdef __FMA__
return _mm_fnmadd_pd(a, b, c);
#else
return sub(c, mul(a, b));
#endif
}
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, add the negated intermediate result
* to packed elements in c, and store the results in vect_t.
* Args : [a0, a1], [b0, b1], [c0, c1]
* Return : [-(a0*b0)+c0, -(a1*b1)+c1]
*/
static INLINE CONST vect_t nmadd(const vect_t c, const vect_t a, const vect_t b) { return fnmadd(c, a, b); }
static INLINE CONST vect_t fnmaddin(vect_t &c, const vect_t a, const vect_t b) { return c = fnmadd(c, a, b); }
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, subtract packed elements in c from
* the intermediate result, and store the results in vect_t.
* Args : [a0, a1], [b0, b1], [c0, c1]
* Return : [a0*b0-c0, a1*b1-c1]
*/
static INLINE CONST vect_t fmsub(const vect_t c, const vect_t a, const vect_t b) {
#ifdef __FMA__
return _mm_fmsub_pd(a, b, c);
#else
return sub(mul(a, b), c);
#endif
}
/*
* Multiply packed double-precision (64-bit) floating-point elements in a and b, subtract packed elements in c from
* the intermediate result, and store the results in vect_t.
* Args : [a0, a1], [b0, b1], [c0, c1]
* Return : [a0*b0-c0, a1*b1-c1]
*/
static INLINE CONST vect_t msub(const vect_t c, const vect_t a, const vect_t b) { return fmsub(c, a, b); }
static INLINE CONST vect_t fmsubin(vect_t &c, const vect_t a, const vect_t b) { return c = fmsub(c, a, b); }
/*
* Compare packed double-precision (64-bit) floating-point elements in a and b for equality, and store the results
in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [(a0==b0) ? 0xFFFFFFFFFFFFFFFF : 0,
(a1==b1) ? 0xFFFFFFFFFFFFFFFF : 0]
*/
static INLINE CONST vect_t eq(const vect_t a, const vect_t b) { return _mm_cmpeq_pd(a, b); }
/*
* Compare packed double-precision (64-bit) floating-point elements in a and b for lesser-than, and store the
results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [(a0<b0) ? 0xFFFFFFFFFFFFFFFF : 0,
(a1<b1) ? 0xFFFFFFFFFFFFFFFF : 0]
*/
static INLINE CONST vect_t lesser(const vect_t a, const vect_t b) { return _mm_cmplt_pd(a, b); }
/*
* Compare packed double-precision (64-bit) floating-point elements in a and b for lesser or equal than, and store
the results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [(a0<=b0) ? 0xFFFFFFFFFFFFFFFF : 0,
(a1<=b1) ? 0xFFFFFFFFFFFFFFFF : 0]
*/
static INLINE CONST vect_t lesser_eq(const vect_t a, const vect_t b) { return _mm_cmple_pd(a, b); }
/*
* Compare packed double-precision (64-bit) floating-point elements in a and b for greater-than, and store the
results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [(a0>b0) ? 0xFFFFFFFFFFFFFFFF : 0,
(a1>b1) ? 0xFFFFFFFFFFFFFFFF : 0]
*/
static INLINE CONST vect_t greater(const vect_t a, const vect_t b) { return _mm_cmpgt_pd(a, b); }
/*
* Compare packed double-precision (64-bit) floating-point elements in a and b for greater or equal than, and store
the results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [(a0>=b0) ? 0xFFFFFFFFFFFFFFFF : 0,
(a1>=b1) ? 0xFFFFFFFFFFFFFFFF : 0]
*/
static INLINE CONST vect_t greater_eq(const vect_t a, const vect_t b) { return _mm_cmpge_pd(a, b); }
/*
* Compute the bitwise AND of packed double-precision (64-bit) floating-point elements in a and b, and store the
* results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0 AND b0, a1 AND b1]
*/
static INLINE CONST vect_t vand(const vect_t a, const vect_t b) { return _mm_and_pd(a, b); }
/*
* Compute the bitwise OR of packed double-precision (64-bit) floating-point elements in a and b, and store the
* results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0 OR b0, a1 OR b1]
*/
static INLINE CONST vect_t vor(const vect_t a, const vect_t b) { return _mm_or_pd(a, b); }
/*
* Compute the bitwise XOR of packed double-precision (64-bit) floating-point elements in a and b, and store the
* results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0 XOR b0, a1 XOR b1]
*/
static INLINE CONST vect_t vxor(const vect_t a, const vect_t b) { return _mm_xor_pd(a, b); }
/*
* Compute the bitwise AND NOT of packed double-precision (64-bit) floating-point elements in a and b, and store the
* results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0 AND NOT b0, a1 AND NOT b1]
*/
static INLINE CONST vect_t vandnot(const vect_t a, const vect_t b) { return _mm_andnot_pd(a, b); }
/*
* Round the packed double-precision (64-bit) floating-point elements in a down to an integer value, and store the
* results as packed double-precision floating-point elements in vect_t.
* Args : [a0, a1]
* Return : [floor(a0), floor(a1)]
*/
static INLINE CONST vect_t floor(const vect_t a) { return _mm_floor_pd(a); }
/*
* Round the packed double-precision (64-bit) floating-point elements in a up to an integer value, and store the
* results as packed double-precision floating-point elements in vect_t.
* Args : [a0, a1]
* Return : [ceil(a0), ceil(a1)]
*/
static INLINE CONST vect_t ceil(const vect_t a) { return _mm_ceil_pd(a); }
/*
* Round the packed double-precision (64-bit) floating-point elements in a, and store the results as packed
* double-precision floating-point elements in vect_t.
* Args : [a0, a1]
* Return : [round(a0), round(a1)]
*/
static INLINE CONST vect_t round(const vect_t a) {
return _mm_round_pd(a, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
/*
* Horizontally add adjacent pairs of double-precision (64-bit) floating-point elements in a and b, and pack the
* results in vect_t.
* Args : [a0, a1], [b0, b1]
* Return : [a0+a1, b0+b1]
*/
static INLINE CONST vect_t hadd(const vect_t a, const vect_t b) { return _mm_hadd_pd(a, b); }
/*
* Horizontally add double-precision (64-bit) floating-point elements in a.
* Args : [a0, a1]
* Return : a0+a1
*/
static INLINE CONST scalar_t hadd_to_scal(const vect_t a) {
return ((const scalar_t *)&a)[0] + ((const scalar_t *)&a)[1];
}
static INLINE vect_t mod(vect_t &C, const vect_t &P, const vect_t &INVP, const vect_t &NEGP, const vect_t &MIN,
const vect_t &MAX, vect_t &Q, vect_t &T) {
FLOAT_MOD(C, P, INVP, Q);
NORML_MOD(C, P, NEGP, MIN, MAX, Q, T);
return C;
}
#else // __FFLASFFPACK_HAVE_AVX_INSTRUCTIONS
#error "You need SSE instructions to perform 128bits operations on double"
#endif
};
#endif // __FFLASFFPACK_fflas_ffpack_utils_simd128_double_INL
|