src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/kernelutil_x86_avx_128_fma_double.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 #ifndef _kernelutil_x86_avx_128_fma_double_h_
  36 #define _kernelutil_x86_avx_128_fma_double_h_
  37
  38 #include <math.h>
  39 #include <immintrin.h>
  40 #ifdef _MSC_VER
  41 #    include <intrin.h>
  42 #else
  43 #    include <x86intrin.h>
  44 #endif
  45
  46 #define gmx_mm_castsi128_pd   _mm_castsi128_pd
  47 #define gmx_mm_extract_epi32  _mm_extract_epi32
  48
  49 #define GMX_MM_TRANSPOSE2_PD(row0, row1) {           \
  50         __m128d __gmx_t1 = row0;                         \
  51         row0           = _mm_unpacklo_pd(row0, row1);     \
  52         row1           = _mm_unpackhi_pd(__gmx_t1, row1); \
  53 }
  54
  55 static int
  56 gmx_mm_any_lt(__m128d a, __m128d b)
  57 {
  58     return _mm_movemask_pd(_mm_cmplt_pd(a, b));
  59 }
  60
  61
  62 static gmx_inline __m128d
  63 gmx_mm_calc_rsq_pd(__m128d dx, __m128d dy, __m128d dz)
  64 {
  65     return _mm_macc_pd(dx, dx, _mm_macc_pd(dy, dy, _mm_mul_pd(dz, dz)));
  66 }
  67
  68 /* Normal sum of four ymm registers */
  69 #define gmx_mm_sum4_pd(t0, t1, t2, t3)  _mm_add_pd(_mm_add_pd(t0, t1), _mm_add_pd(t2, t3))
  70
  71
  72
  73 /* Load a double value from 1-2 places, merge into xmm register */
  74
  75
  76 static __m128d
  77 gmx_mm_load_2real_swizzle_pd(const double * gmx_restrict ptrA,
  78                              const double * gmx_restrict ptrB)
  79 {
  80     return _mm_unpacklo_pd(_mm_load_sd(ptrA), _mm_load_sd(ptrB));
  81 }
  82
  83 static __m128d
  84 gmx_mm_load_1real_pd(const double * gmx_restrict ptrA)
  85 {
  86     return _mm_load_sd(ptrA);
  87 }
  88
  89
  90 static void
  91 gmx_mm_store_2real_swizzle_pd(double * gmx_restrict ptrA,
  92                               double * gmx_restrict ptrB,
  93                               __m128d               xmm1)
  94 {
  95     __m128d t2;
  96
  97     t2       = _mm_unpackhi_pd(xmm1, xmm1);
  98     _mm_store_sd(ptrA, xmm1);
  99     _mm_store_sd(ptrB, t2);
 100 }
 101
 102 static void
 103 gmx_mm_store_1real_pd(double * gmx_restrict ptrA, __m128d xmm1)
 104 {
 105     _mm_store_sd(ptrA, xmm1);
 106 }
 107
 108
 109 /* Similar to store, but increments value in memory */
 110 static void
 111 gmx_mm_increment_2real_swizzle_pd(double * gmx_restrict ptrA,
 112                                   double * gmx_restrict ptrB, __m128d xmm1)
 113 {
 114     __m128d t1;
 115
 116     t1   = _mm_unpackhi_pd(xmm1, xmm1);
 117     xmm1 = _mm_add_sd(xmm1, _mm_load_sd(ptrA));
 118     t1   = _mm_add_sd(t1, _mm_load_sd(ptrB));
 119     _mm_store_sd(ptrA, xmm1);
 120     _mm_store_sd(ptrB, t1);
 121 }
 122
 123 static void
 124 gmx_mm_increment_1real_pd(double * gmx_restrict ptrA, __m128d xmm1)
 125 {
 126     __m128d tmp;
 127
 128     tmp = gmx_mm_load_1real_pd(ptrA);
 129     tmp = _mm_add_sd(tmp, xmm1);
 130     gmx_mm_store_1real_pd(ptrA, tmp);
 131 }
 132
 133
 134
 135 static gmx_inline void
 136 gmx_mm_load_2pair_swizzle_pd(const double * gmx_restrict p1,
 137                              const double * gmx_restrict p2,
 138                              __m128d * gmx_restrict      c6,
 139                              __m128d * gmx_restrict      c12)
 140 {
 141     __m128d t1, t2, t3;
 142
 143     /* The c6/c12 array should be aligned */
 144     t1   = _mm_loadu_pd(p1);
 145     t2   = _mm_loadu_pd(p2);
 146     *c6  = _mm_unpacklo_pd(t1, t2);
 147     *c12 = _mm_unpackhi_pd(t1, t2);
 148 }
 149
 150 static gmx_inline void
 151 gmx_mm_load_1pair_swizzle_pd(const double * gmx_restrict p1,
 152                              __m128d * gmx_restrict      c6,
 153                              __m128d * gmx_restrict      c12)
 154 {
 155     *c6     = _mm_load_sd(p1);
 156     *c12    = _mm_load_sd(p1+1);
 157 }
 158
 159
 160 static gmx_inline void
 161 gmx_mm_load_shift_and_1rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 162                                          const double * gmx_restrict xyz,
 163                                          __m128d * gmx_restrict      x1,
 164                                          __m128d * gmx_restrict      y1,
 165                                          __m128d * gmx_restrict      z1)
 166 {
 167     __m128d mem_xy, mem_z, mem_sxy, mem_sz;
 168
 169     mem_xy  = _mm_loadu_pd(xyz);
 170     mem_z   = _mm_load_sd(xyz+2);
 171     mem_sxy = _mm_loadu_pd(xyz_shift);
 172     mem_sz  = _mm_load_sd(xyz_shift+2);
 173
 174     mem_xy  = _mm_add_pd(mem_xy, mem_sxy);
 175     mem_z   = _mm_add_pd(mem_z, mem_sz);
 176
 177     *x1  = _mm_shuffle_pd(mem_xy, mem_xy, _MM_SHUFFLE2(0, 0));
 178     *y1  = _mm_shuffle_pd(mem_xy, mem_xy, _MM_SHUFFLE2(1, 1));
 179     *z1  = _mm_shuffle_pd(mem_z, mem_z, _MM_SHUFFLE2(0, 0));
 180 }
 181
 182
 183 static gmx_inline void
 184 gmx_mm_load_shift_and_3rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 185                                          const double * gmx_restrict xyz,
 186                                          __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 187                                          __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 188                                          __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 189 {
 190     __m128d t1, t2, t3, t4, t5, sxy, sz, szx, syz;
 191
 192     t1  = _mm_loadu_pd(xyz);
 193     t2  = _mm_loadu_pd(xyz+2);
 194     t3  = _mm_loadu_pd(xyz+4);
 195     t4  = _mm_loadu_pd(xyz+6);
 196     t5  = _mm_load_sd(xyz+8);
 197
 198     sxy = _mm_loadu_pd(xyz_shift);
 199     sz  = _mm_load_sd(xyz_shift+2);
 200     szx = _mm_shuffle_pd(sz, sxy, _MM_SHUFFLE2(0, 0));
 201     syz = _mm_shuffle_pd(sxy, sz, _MM_SHUFFLE2(0, 1));
 202
 203     t1  = _mm_add_pd(t1, sxy);
 204     t2  = _mm_add_pd(t2, szx);
 205     t3  = _mm_add_pd(t3, syz);
 206     t4  = _mm_add_pd(t4, sxy);
 207     t5  = _mm_add_sd(t5, sz);
 208
 209     *x1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(0, 0));
 210     *y1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(1, 1));
 211     *z1  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(0, 0));
 212     *x2  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(1, 1));
 213     *y2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(0, 0));
 214     *z2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(1, 1));
 215     *x3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(0, 0));
 216     *y3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(1, 1));
 217     *z3  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(0, 0));
 218 }
 219
 220
 221 static gmx_inline void
 222 gmx_mm_load_shift_and_4rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 223                                          const double * gmx_restrict xyz,
 224                                          __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 225                                          __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 226                                          __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 227                                          __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 228 {
 229     __m128d t1, t2, t3, t4, t5, t6, sxy, sz, szx, syz;
 230
 231     t1  = _mm_loadu_pd(xyz);
 232     t2  = _mm_loadu_pd(xyz+2);
 233     t3  = _mm_loadu_pd(xyz+4);
 234     t4  = _mm_loadu_pd(xyz+6);
 235     t5  = _mm_loadu_pd(xyz+8);
 236     t6  = _mm_loadu_pd(xyz+10);
 237
 238     sxy = _mm_loadu_pd(xyz_shift);
 239     sz  = _mm_load_sd(xyz_shift+2);
 240     szx = _mm_shuffle_pd(sz, sxy, _MM_SHUFFLE2(0, 0));
 241     syz = _mm_shuffle_pd(sxy, sz, _MM_SHUFFLE2(0, 1));
 242
 243     t1  = _mm_add_pd(t1, sxy);
 244     t2  = _mm_add_pd(t2, szx);
 245     t3  = _mm_add_pd(t3, syz);
 246     t4  = _mm_add_pd(t4, sxy);
 247     t5  = _mm_add_pd(t5, szx);
 248     t6  = _mm_add_pd(t6, syz);
 249
 250     *x1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(0, 0));
 251     *y1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(1, 1));
 252     *z1  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(0, 0));
 253     *x2  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(1, 1));
 254     *y2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(0, 0));
 255     *z2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(1, 1));
 256     *x3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(0, 0));
 257     *y3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(1, 1));
 258     *z3  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(0, 0));
 259     *x4  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(1, 1));
 260     *y4  = _mm_shuffle_pd(t6, t6, _MM_SHUFFLE2(0, 0));
 261     *z4  = _mm_shuffle_pd(t6, t6, _MM_SHUFFLE2(1, 1));
 262 }
 263
 264
 265
 266 static gmx_inline void
 267 gmx_mm_load_1rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 268                                   __m128d * gmx_restrict x, __m128d * gmx_restrict y, __m128d * gmx_restrict z)
 269 {
 270     *x            = _mm_load_sd(p1);
 271     *y            = _mm_load_sd(p1+1);
 272     *z            = _mm_load_sd(p1+2);
 273 }
 274
 275 static gmx_inline void
 276 gmx_mm_load_3rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 277                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 278                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 279                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 280 {
 281     *x1            = _mm_load_sd(p1);
 282     *y1            = _mm_load_sd(p1+1);
 283     *z1            = _mm_load_sd(p1+2);
 284     *x2            = _mm_load_sd(p1+3);
 285     *y2            = _mm_load_sd(p1+4);
 286     *z2            = _mm_load_sd(p1+5);
 287     *x3            = _mm_load_sd(p1+6);
 288     *y3            = _mm_load_sd(p1+7);
 289     *z3            = _mm_load_sd(p1+8);
 290 }
 291
 292 static gmx_inline void
 293 gmx_mm_load_4rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 294                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 295                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 296                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 297                                   __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 298 {
 299     *x1            = _mm_load_sd(p1);
 300     *y1            = _mm_load_sd(p1+1);
 301     *z1            = _mm_load_sd(p1+2);
 302     *x2            = _mm_load_sd(p1+3);
 303     *y2            = _mm_load_sd(p1+4);
 304     *z2            = _mm_load_sd(p1+5);
 305     *x3            = _mm_load_sd(p1+6);
 306     *y3            = _mm_load_sd(p1+7);
 307     *z3            = _mm_load_sd(p1+8);
 308     *x4            = _mm_load_sd(p1+9);
 309     *y4            = _mm_load_sd(p1+10);
 310     *z4            = _mm_load_sd(p1+11);
 311 }
 312
 313
 314 static gmx_inline void
 315 gmx_mm_load_1rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA,
 316                                   const double * gmx_restrict ptrB,
 317                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1)
 318 {
 319     __m128d t1, t2, t3, t4;
 320     t1           = _mm_loadu_pd(ptrA);
 321     t2           = _mm_loadu_pd(ptrB);
 322     t3           = _mm_load_sd(ptrA+2);
 323     t4           = _mm_load_sd(ptrB+2);
 324     GMX_MM_TRANSPOSE2_PD(t1, t2);
 325     *x1          = t1;
 326     *y1          = t2;
 327     *z1          = _mm_unpacklo_pd(t3, t4);
 328 }
 329
 330 static gmx_inline void
 331 gmx_mm_load_3rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA, const double * gmx_restrict ptrB,
 332                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 333                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 334                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 335 {
 336     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10;
 337     t1           = _mm_loadu_pd(ptrA);
 338     t2           = _mm_loadu_pd(ptrB);
 339     t3           = _mm_loadu_pd(ptrA+2);
 340     t4           = _mm_loadu_pd(ptrB+2);
 341     t5           = _mm_loadu_pd(ptrA+4);
 342     t6           = _mm_loadu_pd(ptrB+4);
 343     t7           = _mm_loadu_pd(ptrA+6);
 344     t8           = _mm_loadu_pd(ptrB+6);
 345     t9           = _mm_load_sd(ptrA+8);
 346     t10          = _mm_load_sd(ptrB+8);
 347     GMX_MM_TRANSPOSE2_PD(t1, t2);
 348     GMX_MM_TRANSPOSE2_PD(t3, t4);
 349     GMX_MM_TRANSPOSE2_PD(t5, t6);
 350     GMX_MM_TRANSPOSE2_PD(t7, t8);
 351     *x1          = t1;
 352     *y1          = t2;
 353     *z1          = t3;
 354     *x2          = t4;
 355     *y2          = t5;
 356     *z2          = t6;
 357     *x3          = t7;
 358     *y3          = t8;
 359     *z3          = _mm_unpacklo_pd(t9, t10);
 360 }
 361
 362
 363 static gmx_inline void
 364 gmx_mm_load_4rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA, const double * gmx_restrict ptrB,
 365                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 366                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 367                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 368                                   __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 369 {
 370     __m128d t1, t2, t3, t4, t5, t6;
 371     t1           = _mm_loadu_pd(ptrA);
 372     t2           = _mm_loadu_pd(ptrB);
 373     t3           = _mm_loadu_pd(ptrA+2);
 374     t4           = _mm_loadu_pd(ptrB+2);
 375     t5           = _mm_loadu_pd(ptrA+4);
 376     t6           = _mm_loadu_pd(ptrB+4);
 377     GMX_MM_TRANSPOSE2_PD(t1, t2);
 378     GMX_MM_TRANSPOSE2_PD(t3, t4);
 379     GMX_MM_TRANSPOSE2_PD(t5, t6);
 380     *x1          = t1;
 381     *y1          = t2;
 382     *z1          = t3;
 383     *x2          = t4;
 384     *y2          = t5;
 385     *z2          = t6;
 386     t1           = _mm_loadu_pd(ptrA+6);
 387     t2           = _mm_loadu_pd(ptrB+6);
 388     t3           = _mm_loadu_pd(ptrA+8);
 389     t4           = _mm_loadu_pd(ptrB+8);
 390     t5           = _mm_loadu_pd(ptrA+10);
 391     t6           = _mm_loadu_pd(ptrB+10);
 392     GMX_MM_TRANSPOSE2_PD(t1, t2);
 393     GMX_MM_TRANSPOSE2_PD(t3, t4);
 394     GMX_MM_TRANSPOSE2_PD(t5, t6);
 395     *x3          = t1;
 396     *y3          = t2;
 397     *z3          = t3;
 398     *x4          = t4;
 399     *y4          = t5;
 400     *z4          = t6;
 401 }
 402
 403
 404 /* Routines to decrement rvec in memory, typically use for j particle force updates */
 405 static void
 406 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 407                                        __m128d x1, __m128d y1, __m128d z1)
 408 {
 409     __m128d t1, t2, t3;
 410
 411     t1           = _mm_load_sd(ptrA);
 412     t2           = _mm_load_sd(ptrA+1);
 413     t3           = _mm_load_sd(ptrA+2);
 414
 415     t1           = _mm_sub_sd(t1, x1);
 416     t2           = _mm_sub_sd(t2, y1);
 417     t3           = _mm_sub_sd(t3, z1);
 418     _mm_store_sd(ptrA, t1);
 419     _mm_store_sd(ptrA+1, t2);
 420     _mm_store_sd(ptrA+2, t3);
 421 }
 422
 423
 424 #if defined (_MSC_VER) && defined(_M_IX86)
 425 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 426 #define gmx_mm_decrement_3rvec_1ptr_swizzle_pd(ptrA, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3) \
 427     { \
 428         __m128d _t1, _t2, _t3, _t4, _t5; \
 429         _t1          = _mm_loadu_pd(ptrA); \
 430         _t2          = _mm_loadu_pd(ptrA+2); \
 431         _t3          = _mm_loadu_pd(ptrA+4); \
 432         _t4          = _mm_loadu_pd(ptrA+6); \
 433         _t5          = _mm_load_sd(ptrA+8); \
 434         _x1          = _mm_unpacklo_pd(_x1, _y1); \
 435         _z1          = _mm_unpacklo_pd(_z1, _x2); \
 436         _y2          = _mm_unpacklo_pd(_y2, _z2); \
 437         _x3          = _mm_unpacklo_pd(_x3, _y3); \
 438         _t1          = _mm_sub_pd(_t1, _x1); \
 439         _t2          = _mm_sub_pd(_t2, _z1); \
 440         _t3          = _mm_sub_pd(_t3, _y2); \
 441         _t4          = _mm_sub_pd(_t4, _x3); \
 442         _t5          = _mm_sub_sd(_t5, _z3); \
 443         _mm_storeu_pd(ptrA, _t1); \
 444         _mm_storeu_pd(ptrA+2, _t2); \
 445         _mm_storeu_pd(ptrA+4, _t3); \
 446         _mm_storeu_pd(ptrA+6, _t4); \
 447         _mm_store_sd(ptrA+8, _t5); \
 448     }
 449 #else
 450 /* Real function for sane compilers */
 451 static void
 452 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 453                                        __m128d x1, __m128d y1, __m128d z1,
 454                                        __m128d x2, __m128d y2, __m128d z2,
 455                                        __m128d x3, __m128d y3, __m128d z3)
 456 {
 457     __m128d t1, t2, t3, t4, t5;
 458
 459     t1          = _mm_loadu_pd(ptrA);
 460     t2          = _mm_loadu_pd(ptrA+2);
 461     t3          = _mm_loadu_pd(ptrA+4);
 462     t4          = _mm_loadu_pd(ptrA+6);
 463     t5          = _mm_load_sd(ptrA+8);
 464
 465     x1          = _mm_unpacklo_pd(x1, y1);
 466     z1          = _mm_unpacklo_pd(z1, x2);
 467     y2          = _mm_unpacklo_pd(y2, z2);
 468     x3          = _mm_unpacklo_pd(x3, y3);
 469     /* nothing to be done for z3 */
 470
 471     t1          = _mm_sub_pd(t1, x1);
 472     t2          = _mm_sub_pd(t2, z1);
 473     t3          = _mm_sub_pd(t3, y2);
 474     t4          = _mm_sub_pd(t4, x3);
 475     t5          = _mm_sub_sd(t5, z3);
 476     _mm_storeu_pd(ptrA, t1);
 477     _mm_storeu_pd(ptrA+2, t2);
 478     _mm_storeu_pd(ptrA+4, t3);
 479     _mm_storeu_pd(ptrA+6, t4);
 480     _mm_store_sd(ptrA+8, t5);
 481 }
 482 #endif
 483
 484
 485 #if defined (_MSC_VER) && defined(_M_IX86)
 486 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 487 #define gmx_mm_decrement_4rvec_1ptr_swizzle_pd(ptrA, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3, _x4, _y4, _z4) \
 488     { \
 489         __m128d _t1, _t2, _t3, _t4, _t5, _t6; \
 490         _t1          = _mm_loadu_pd(ptrA); \
 491         _t2          = _mm_loadu_pd(ptrA+2); \
 492         _t3          = _mm_loadu_pd(ptrA+4); \
 493         _t4          = _mm_loadu_pd(ptrA+6); \
 494         _t5          = _mm_loadu_pd(ptrA+8); \
 495         _t6          = _mm_loadu_pd(ptrA+10); \
 496         _x1          = _mm_unpacklo_pd(_x1, _y1); \
 497         _z1          = _mm_unpacklo_pd(_z1, _x2); \
 498         _y2          = _mm_unpacklo_pd(_y2, _z2); \
 499         _x3          = _mm_unpacklo_pd(_x3, _y3); \
 500         _z3          = _mm_unpacklo_pd(_z3, _x4); \
 501         _y4          = _mm_unpacklo_pd(_y4, _z4); \
 502         _mm_storeu_pd(ptrA,    _mm_sub_pd( _t1, _x1 )); \
 503         _mm_storeu_pd(ptrA+2,  _mm_sub_pd( _t2, _z1 )); \
 504         _mm_storeu_pd(ptrA+4,  _mm_sub_pd( _t3, _y2 )); \
 505         _mm_storeu_pd(ptrA+6,  _mm_sub_pd( _t4, _x3 )); \
 506         _mm_storeu_pd(ptrA+8,  _mm_sub_pd( _t5, _z3 )); \
 507         _mm_storeu_pd(ptrA+10, _mm_sub_pd( _t6, _y4 )); \
 508     }
 509 #else
 510 /* Real function for sane compilers */
 511 static void
 512 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 513                                        __m128d x1, __m128d y1, __m128d z1,
 514                                        __m128d x2, __m128d y2, __m128d z2,
 515                                        __m128d x3, __m128d y3, __m128d z3,
 516                                        __m128d x4, __m128d y4, __m128d z4)
 517 {
 518     __m128d t1, t2, t3, t4, t5, t6;
 519
 520     t1          = _mm_loadu_pd(ptrA);
 521     t2          = _mm_loadu_pd(ptrA+2);
 522     t3          = _mm_loadu_pd(ptrA+4);
 523     t4          = _mm_loadu_pd(ptrA+6);
 524     t5          = _mm_loadu_pd(ptrA+8);
 525     t6          = _mm_loadu_pd(ptrA+10);
 526
 527     x1          = _mm_unpacklo_pd(x1, y1);
 528     z1          = _mm_unpacklo_pd(z1, x2);
 529     y2          = _mm_unpacklo_pd(y2, z2);
 530     x3          = _mm_unpacklo_pd(x3, y3);
 531     z3          = _mm_unpacklo_pd(z3, x4);
 532     y4          = _mm_unpacklo_pd(y4, z4);
 533
 534     _mm_storeu_pd(ptrA,    _mm_sub_pd( t1, x1 ));
 535     _mm_storeu_pd(ptrA+2,  _mm_sub_pd( t2, z1 ));
 536     _mm_storeu_pd(ptrA+4,  _mm_sub_pd( t3, y2 ));
 537     _mm_storeu_pd(ptrA+6,  _mm_sub_pd( t4, x3 ));
 538     _mm_storeu_pd(ptrA+8,  _mm_sub_pd( t5, z3 ));
 539     _mm_storeu_pd(ptrA+10, _mm_sub_pd( t6, y4 ));
 540 }
 541 #endif
 542
 543
 544 static void
 545 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 546                                        __m128d x1, __m128d y1, __m128d z1)
 547 {
 548     __m128d t1, t2, t3, t4, t5, t6, t7;
 549
 550     t1          = _mm_loadu_pd(ptrA);
 551     t2          = _mm_load_sd(ptrA+2);
 552     t3          = _mm_loadu_pd(ptrB);
 553     t4          = _mm_load_sd(ptrB+2);
 554
 555     t5          = _mm_unpacklo_pd(x1, y1);
 556     t6          = _mm_unpackhi_pd(x1, y1);
 557     t7          = _mm_unpackhi_pd(z1, z1);
 558
 559     t1          = _mm_sub_pd(t1, t5);
 560     t2          = _mm_sub_sd(t2, z1);
 561
 562     t3          = _mm_sub_pd(t3, t6);
 563     t4          = _mm_sub_sd(t4, t7);
 564
 565     _mm_storeu_pd(ptrA, t1);
 566     _mm_store_sd(ptrA+2, t2);
 567     _mm_storeu_pd(ptrB, t3);
 568     _mm_store_sd(ptrB+2, t4);
 569 }
 570
 571
 572 #if defined (_MSC_VER) && defined(_M_IX86)
 573 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 574 #define gmx_mm_decrement_3rvec_2ptr_swizzle_pd(ptrA, ptrB, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3) \
 575     { \
 576         __m128d _t1, _t2, _t3, _t4, _t5, _t6, _t7, _t8, _t9, _t10; \
 577         __m128d _tA, _tB, _tC, _tD, _tE, _tF, _tG, _tH, _tI; \
 578         _t1          = _mm_loadu_pd(ptrA); \
 579         _t2          = _mm_loadu_pd(ptrA+2); \
 580         _t3          = _mm_loadu_pd(ptrA+4); \
 581         _t4          = _mm_loadu_pd(ptrA+6); \
 582         _t5          = _mm_load_sd(ptrA+8); \
 583         _t6          = _mm_loadu_pd(ptrB); \
 584         _t7          = _mm_loadu_pd(ptrB+2); \
 585         _t8          = _mm_loadu_pd(ptrB+4); \
 586         _t9          = _mm_loadu_pd(ptrB+6); \
 587         _t10         = _mm_load_sd(ptrB+8); \
 588         _tA          = _mm_unpacklo_pd(_x1, _y1); \
 589         _tB          = _mm_unpackhi_pd(_x1, _y1); \
 590         _tC          = _mm_unpacklo_pd(_z1, _x2); \
 591         _tD          = _mm_unpackhi_pd(_z1, _x2); \
 592         _tE          = _mm_unpacklo_pd(_y2, _z2); \
 593         _tF          = _mm_unpackhi_pd(_y2, _z2); \
 594         _tG          = _mm_unpacklo_pd(_x3, _y3); \
 595         _tH          = _mm_unpackhi_pd(_x3, _y3); \
 596         _tI          = _mm_unpackhi_pd(_z3, _z3); \
 597         _t1          = _mm_sub_pd(_t1, _tA); \
 598         _t2          = _mm_sub_pd(_t2, _tC); \
 599         _t3          = _mm_sub_pd(_t3, _tE); \
 600         _t4          = _mm_sub_pd(_t4, _tG); \
 601         _t5          = _mm_sub_sd(_t5, _z3); \
 602         _t6          = _mm_sub_pd(_t6, _tB); \
 603         _t7          = _mm_sub_pd(_t7, _tD); \
 604         _t8          = _mm_sub_pd(_t8, _tF); \
 605         _t9          = _mm_sub_pd(_t9, _tH); \
 606         _t10         = _mm_sub_sd(_t10, _tI); \
 607         _mm_storeu_pd(ptrA, _t1); \
 608         _mm_storeu_pd(ptrA+2, _t2); \
 609         _mm_storeu_pd(ptrA+4, _t3); \
 610         _mm_storeu_pd(ptrA+6, _t4); \
 611         _mm_store_sd(ptrA+8, _t5); \
 612         _mm_storeu_pd(ptrB, _t6); \
 613         _mm_storeu_pd(ptrB+2, _t7); \
 614         _mm_storeu_pd(ptrB+4, _t8); \
 615         _mm_storeu_pd(ptrB+6, _t9); \
 616         _mm_store_sd(ptrB+8, _t10); \
 617     }
 618 #else
 619 /* Real function for sane compilers */
 620 static void
 621 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 622                                        __m128d x1, __m128d y1, __m128d z1,
 623                                        __m128d x2, __m128d y2, __m128d z2,
 624                                        __m128d x3, __m128d y3, __m128d z3)
 625 {
 626     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10;
 627     __m128d tA, tB, tC, tD, tE, tF, tG, tH, tI;
 628
 629     t1          = _mm_loadu_pd(ptrA);
 630     t2          = _mm_loadu_pd(ptrA+2);
 631     t3          = _mm_loadu_pd(ptrA+4);
 632     t4          = _mm_loadu_pd(ptrA+6);
 633     t5          = _mm_load_sd(ptrA+8);
 634     t6          = _mm_loadu_pd(ptrB);
 635     t7          = _mm_loadu_pd(ptrB+2);
 636     t8          = _mm_loadu_pd(ptrB+4);
 637     t9          = _mm_loadu_pd(ptrB+6);
 638     t10         = _mm_load_sd(ptrB+8);
 639
 640     tA          = _mm_unpacklo_pd(x1, y1);
 641     tB          = _mm_unpackhi_pd(x1, y1);
 642     tC          = _mm_unpacklo_pd(z1, x2);
 643     tD          = _mm_unpackhi_pd(z1, x2);
 644     tE          = _mm_unpacklo_pd(y2, z2);
 645     tF          = _mm_unpackhi_pd(y2, z2);
 646     tG          = _mm_unpacklo_pd(x3, y3);
 647     tH          = _mm_unpackhi_pd(x3, y3);
 648     tI          = _mm_unpackhi_pd(z3, z3);
 649
 650     t1          = _mm_sub_pd(t1, tA);
 651     t2          = _mm_sub_pd(t2, tC);
 652     t3          = _mm_sub_pd(t3, tE);
 653     t4          = _mm_sub_pd(t4, tG);
 654     t5          = _mm_sub_sd(t5, z3);
 655
 656     t6          = _mm_sub_pd(t6, tB);
 657     t7          = _mm_sub_pd(t7, tD);
 658     t8          = _mm_sub_pd(t8, tF);
 659     t9          = _mm_sub_pd(t9, tH);
 660     t10         = _mm_sub_sd(t10, tI);
 661
 662     _mm_storeu_pd(ptrA, t1);
 663     _mm_storeu_pd(ptrA+2, t2);
 664     _mm_storeu_pd(ptrA+4, t3);
 665     _mm_storeu_pd(ptrA+6, t4);
 666     _mm_store_sd(ptrA+8, t5);
 667     _mm_storeu_pd(ptrB, t6);
 668     _mm_storeu_pd(ptrB+2, t7);
 669     _mm_storeu_pd(ptrB+4, t8);
 670     _mm_storeu_pd(ptrB+6, t9);
 671     _mm_store_sd(ptrB+8, t10);
 672 }
 673 #endif
 674
 675
 676 #if defined (_MSC_VER) && defined(_M_IX86)
 677 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 678 #define gmx_mm_decrement_4rvec_2ptr_swizzle_pd(ptrA, ptrB, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3, _x4, _y4, _z4) \
 679     { \
 680         __m128d _t1, _t2, _t3, _t4, _t5, _t6, _t7, _t8, _t9, _t10, _t11, _t12; \
 681         __m128d _tA, _tB, _tC, _tD, _tE, _tF, _tG, _tH, _tI, _tJ, _tK, _tL; \
 682         _t1          = _mm_loadu_pd(ptrA); \
 683         _t2          = _mm_loadu_pd(ptrA+2); \
 684         _t3          = _mm_loadu_pd(ptrA+4); \
 685         _t4          = _mm_loadu_pd(ptrA+6); \
 686         _t5          = _mm_loadu_pd(ptrA+8); \
 687         _t6          = _mm_loadu_pd(ptrA+10); \
 688         _t7          = _mm_loadu_pd(ptrB); \
 689         _t8          = _mm_loadu_pd(ptrB+2); \
 690         _t9          = _mm_loadu_pd(ptrB+4); \
 691         _t10         = _mm_loadu_pd(ptrB+6); \
 692         _t11         = _mm_loadu_pd(ptrB+8); \
 693         _t12         = _mm_loadu_pd(ptrB+10); \
 694         _tA          = _mm_unpacklo_pd(_x1, _y1); \
 695         _tB          = _mm_unpackhi_pd(_x1, _y1); \
 696         _tC          = _mm_unpacklo_pd(_z1, _x2); \
 697         _tD          = _mm_unpackhi_pd(_z1, _x2); \
 698         _tE          = _mm_unpacklo_pd(_y2, _z2); \
 699         _tF          = _mm_unpackhi_pd(_y2, _z2); \
 700         _tG          = _mm_unpacklo_pd(_x3, _y3); \
 701         _tH          = _mm_unpackhi_pd(_x3, _y3); \
 702         _tI          = _mm_unpacklo_pd(_z3, _x4); \
 703         _tJ          = _mm_unpackhi_pd(_z3, _x4); \
 704         _tK          = _mm_unpacklo_pd(_y4, _z4); \
 705         _tL          = _mm_unpackhi_pd(_y4, _z4); \
 706         _t1          = _mm_sub_pd(_t1, _tA); \
 707         _t2          = _mm_sub_pd(_t2, _tC); \
 708         _t3          = _mm_sub_pd(_t3, _tE); \
 709         _t4          = _mm_sub_pd(_t4, _tG); \
 710         _t5          = _mm_sub_pd(_t5, _tI); \
 711         _t6          = _mm_sub_pd(_t6, _tK); \
 712         _t7          = _mm_sub_pd(_t7, _tB); \
 713         _t8          = _mm_sub_pd(_t8, _tD); \
 714         _t9          = _mm_sub_pd(_t9, _tF); \
 715         _t10         = _mm_sub_pd(_t10, _tH); \
 716         _t11         = _mm_sub_pd(_t11, _tJ); \
 717         _t12         = _mm_sub_pd(_t12, _tL); \
 718         _mm_storeu_pd(ptrA,  _t1); \
 719         _mm_storeu_pd(ptrA+2, _t2); \
 720         _mm_storeu_pd(ptrA+4, _t3); \
 721         _mm_storeu_pd(ptrA+6, _t4); \
 722         _mm_storeu_pd(ptrA+8, _t5); \
 723         _mm_storeu_pd(ptrA+10, _t6); \
 724         _mm_storeu_pd(ptrB,  _t7); \
 725         _mm_storeu_pd(ptrB+2, _t8); \
 726         _mm_storeu_pd(ptrB+4, _t9); \
 727         _mm_storeu_pd(ptrB+6, _t10); \
 728         _mm_storeu_pd(ptrB+8, _t11); \
 729         _mm_storeu_pd(ptrB+10, _t12); \
 730     }
 731 #else
 732 /* Real function for sane compilers */
 733 static void
 734 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 735                                        __m128d x1, __m128d y1, __m128d z1,
 736                                        __m128d x2, __m128d y2, __m128d z2,
 737                                        __m128d x3, __m128d y3, __m128d z3,
 738                                        __m128d x4, __m128d y4, __m128d z4)
 739 {
 740     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12;
 741     __m128d tA, tB, tC, tD, tE, tF, tG, tH, tI, tJ, tK, tL;
 742
 743     t1          = _mm_loadu_pd(ptrA);
 744     t2          = _mm_loadu_pd(ptrA+2);
 745     t3          = _mm_loadu_pd(ptrA+4);
 746     t4          = _mm_loadu_pd(ptrA+6);
 747     t5          = _mm_loadu_pd(ptrA+8);
 748     t6          = _mm_loadu_pd(ptrA+10);
 749     t7          = _mm_loadu_pd(ptrB);
 750     t8          = _mm_loadu_pd(ptrB+2);
 751     t9          = _mm_loadu_pd(ptrB+4);
 752     t10         = _mm_loadu_pd(ptrB+6);
 753     t11         = _mm_loadu_pd(ptrB+8);
 754     t12         = _mm_loadu_pd(ptrB+10);
 755
 756     tA          = _mm_unpacklo_pd(x1, y1);
 757     tB          = _mm_unpackhi_pd(x1, y1);
 758     tC          = _mm_unpacklo_pd(z1, x2);
 759     tD          = _mm_unpackhi_pd(z1, x2);
 760     tE          = _mm_unpacklo_pd(y2, z2);
 761     tF          = _mm_unpackhi_pd(y2, z2);
 762     tG          = _mm_unpacklo_pd(x3, y3);
 763     tH          = _mm_unpackhi_pd(x3, y3);
 764     tI          = _mm_unpacklo_pd(z3, x4);
 765     tJ          = _mm_unpackhi_pd(z3, x4);
 766     tK          = _mm_unpacklo_pd(y4, z4);
 767     tL          = _mm_unpackhi_pd(y4, z4);
 768
 769     t1          = _mm_sub_pd(t1, tA);
 770     t2          = _mm_sub_pd(t2, tC);
 771     t3          = _mm_sub_pd(t3, tE);
 772     t4          = _mm_sub_pd(t4, tG);
 773     t5          = _mm_sub_pd(t5, tI);
 774     t6          = _mm_sub_pd(t6, tK);
 775
 776     t7          = _mm_sub_pd(t7, tB);
 777     t8          = _mm_sub_pd(t8, tD);
 778     t9          = _mm_sub_pd(t9, tF);
 779     t10         = _mm_sub_pd(t10, tH);
 780     t11         = _mm_sub_pd(t11, tJ);
 781     t12         = _mm_sub_pd(t12, tL);
 782
 783     _mm_storeu_pd(ptrA,  t1);
 784     _mm_storeu_pd(ptrA+2, t2);
 785     _mm_storeu_pd(ptrA+4, t3);
 786     _mm_storeu_pd(ptrA+6, t4);
 787     _mm_storeu_pd(ptrA+8, t5);
 788     _mm_storeu_pd(ptrA+10, t6);
 789     _mm_storeu_pd(ptrB,  t7);
 790     _mm_storeu_pd(ptrB+2, t8);
 791     _mm_storeu_pd(ptrB+4, t9);
 792     _mm_storeu_pd(ptrB+6, t10);
 793     _mm_storeu_pd(ptrB+8, t11);
 794     _mm_storeu_pd(ptrB+10, t12);
 795 }
 796 #endif
 797
 798
 799 static gmx_inline void
 800 gmx_mm_update_iforce_1atom_swizzle_pd(__m128d fix1, __m128d fiy1, __m128d fiz1,
 801                                       double * gmx_restrict fptr,
 802                                       double * gmx_restrict fshiftptr)
 803 {
 804     fix1 = _mm_hadd_pd(fix1, fiy1);
 805     fiz1 = _mm_hadd_pd(fiz1, fiz1);
 806
 807     _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr), fix1 ));
 808     _mm_store_sd( fptr+2, _mm_add_sd( _mm_load_sd(fptr+2), fiz1 ));
 809
 810     _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 ));
 811     _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 ));
 812 }
 813
 814 #if defined (_MSC_VER) && defined(_M_IX86)
 815 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 816 #define gmx_mm_update_iforce_3atom_swizzle_pd(fix1, fiy1, fiz1, fix2, fiy2, fiz2, fix3, fiy3, fiz3, \
 817                                               fptr, fshiftptr) \
 818     { \
 819         __m128d _t1, _t2; \
 820         fix1 = _mm_hadd_pd(fix1, fiy1); \
 821         fiz1 = _mm_hadd_pd(fiz1, fix2); \
 822         fiy2 = _mm_hadd_pd(fiy2, fiz2); \
 823         fix3 = _mm_hadd_pd(fix3, fiy3); \
 824         fiz3 = _mm_hadd_pd(fiz3, fiz3); \
 825         _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr), fix1 )); \
 826         _mm_storeu_pd( fptr+2, _mm_add_pd( _mm_loadu_pd(fptr+2), fiz1 )); \
 827         _mm_storeu_pd( fptr+4, _mm_add_pd( _mm_loadu_pd(fptr+4), fiy2 )); \
 828         _mm_storeu_pd( fptr+6, _mm_add_pd( _mm_loadu_pd(fptr+6), fix3 )); \
 829         _mm_store_sd( fptr+8, _mm_add_sd( _mm_load_sd(fptr+8), fiz3 )); \
 830         fix1  = _mm_add_pd(fix1, fix3); \
 831         _t1   = _mm_shuffle_pd(fiz1, fiy2, _MM_SHUFFLE2(0, 1)); \
 832         fix1  = _mm_add_pd(fix1, _t1); \
 833         _t2   = _mm_shuffle_pd(fiy2, fiy2, _MM_SHUFFLE2(1, 1)); \
 834         fiz1  = _mm_add_sd(fiz1, fiz3); \
 835         fiz1  = _mm_add_sd(fiz1, _t2); \
 836         _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 )); \
 837         _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 )); \
 838     }
 839 #else
 840 /* Real function for sane compilers */
 841 static gmx_inline void
 842 gmx_mm_update_iforce_3atom_swizzle_pd(__m128d fix1, __m128d fiy1, __m128d fiz1,
 843                                       __m128d fix2, __m128d fiy2, __m128d fiz2,
 844                                       __m128d fix3, __m128d fiy3, __m128d fiz3,
 845                                       double * gmx_restrict fptr,
 846                                       double * gmx_restrict fshiftptr)
 847 {
 848     __m128d t1, t2;
 849
 850     fix1 = _mm_hadd_pd(fix1, fiy1);
 851     fiz1 = _mm_hadd_pd(fiz1, fix2);
 852     fiy2 = _mm_hadd_pd(fiy2, fiz2);
 853     fix3 = _mm_hadd_pd(fix3, fiy3);
 854     fiz3 = _mm_hadd_pd(fiz3, fiz3);
 855
 856     _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr), fix1 ));
 857     _mm_storeu_pd( fptr+2, _mm_add_pd( _mm_loadu_pd(fptr+2), fiz1 ));
 858     _mm_storeu_pd( fptr+4, _mm_add_pd( _mm_loadu_pd(fptr+4), fiy2 ));
 859     _mm_storeu_pd( fptr+6, _mm_add_pd( _mm_loadu_pd(fptr+6), fix3 ));
 860     _mm_store_sd( fptr+8, _mm_add_sd( _mm_load_sd(fptr+8), fiz3 ));
 861
 862     fix1 = _mm_add_pd(fix1, fix3);
 863     t1   = _mm_shuffle_pd(fiz1, fiy2, _MM_SHUFFLE2(0, 1));
 864     fix1 = _mm_add_pd(fix1, t1); /* x and y sums */
 865
 866     t2   = _mm_shuffle_pd(fiy2, fiy2, _MM_SHUFFLE2(1, 1));
 867     fiz1 = _mm_add_sd(fiz1, fiz3);
 868     fiz1 = _mm_add_sd(fiz1, t2); /* z sum */
 869
 870     _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 ));
 871     _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 ));
 872 }
 873 #endif
 874
 875 #if defined (_MSC_VER) && defined(_M_IX86)
 876 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 877 #define gmx_mm_update_iforce_4atom_swizzle_pd(fix1, fiy1, fiz1, fix2, fiy2, fiz2, fix3, fiy3, fiz3, fix4, fiy4, fiz4, \
 878                                               fptr, fshiftptr) \
 879     { \
 880         __m128d _t1, _t2; \
 881         fix1 = _mm_hadd_pd(fix1, fiy1); \
 882         fiz1 = _mm_hadd_pd(fiz1, fix2); \
 883         fiy2 = _mm_hadd_pd(fiy2, fiz2); \
 884         fix3 = _mm_hadd_pd(fix3, fiy3); \
 885         fiz3 = _mm_hadd_pd(fiz3, fix4); \
 886         fiy4 = _mm_hadd_pd(fiy4, fiz4); \
 887         _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr),       fix1 )); \
 888         _mm_storeu_pd( fptr+2, _mm_add_pd( _mm_loadu_pd(fptr+2),   fiz1 )); \
 889         _mm_storeu_pd( fptr+4, _mm_add_pd( _mm_loadu_pd(fptr+4),   fiy2 )); \
 890         _mm_storeu_pd( fptr+6, _mm_add_pd( _mm_loadu_pd(fptr+6),   fix3 )); \
 891         _mm_storeu_pd( fptr+8, _mm_add_pd( _mm_loadu_pd(fptr+8),   fiz3 )); \
 892         _mm_storeu_pd( fptr+10, _mm_add_pd( _mm_loadu_pd(fptr+10), fiy4 )); \
 893         _t1  = _mm_shuffle_pd(fiz1, fiy2, _MM_SHUFFLE2(0, 1)); \
 894         fix1 = _mm_add_pd(fix1, _t1); \
 895         _t2  = _mm_shuffle_pd(fiz3, fiy4, _MM_SHUFFLE2(0, 1)); \
 896         fix3 = _mm_add_pd(fix3, _t2); \
 897         fix1 = _mm_add_pd(fix1, fix3); \
 898         fiz1 = _mm_add_sd(fiz1, _mm_unpackhi_pd(fiy2, fiy2)); \
 899         fiz3 = _mm_add_sd(fiz3, _mm_unpackhi_pd(fiy4, fiy4)); \
 900         fiz1 = _mm_add_sd(fiz1, fiz3); \
 901         _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 )); \
 902         _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 )); \
 903     }
 904 #else
 905 /* Real function for sane compilers */
 906 static gmx_inline void
 907 gmx_mm_update_iforce_4atom_swizzle_pd(__m128d fix1, __m128d fiy1, __m128d fiz1,
 908                                       __m128d fix2, __m128d fiy2, __m128d fiz2,
 909                                       __m128d fix3, __m128d fiy3, __m128d fiz3,
 910                                       __m128d fix4, __m128d fiy4, __m128d fiz4,
 911                                       double * gmx_restrict fptr,
 912                                       double * gmx_restrict fshiftptr)
 913 {
 914     __m128d t1, t2;
 915
 916     fix1 = _mm_hadd_pd(fix1, fiy1);
 917     fiz1 = _mm_hadd_pd(fiz1, fix2);
 918     fiy2 = _mm_hadd_pd(fiy2, fiz2);
 919     fix3 = _mm_hadd_pd(fix3, fiy3);
 920     fiz3 = _mm_hadd_pd(fiz3, fix4);
 921     fiy4 = _mm_hadd_pd(fiy4, fiz4);
 922
 923     _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr),       fix1 ));
 924     _mm_storeu_pd( fptr+2, _mm_add_pd( _mm_loadu_pd(fptr+2),   fiz1 ));
 925     _mm_storeu_pd( fptr+4, _mm_add_pd( _mm_loadu_pd(fptr+4),   fiy2 ));
 926     _mm_storeu_pd( fptr+6, _mm_add_pd( _mm_loadu_pd(fptr+6),   fix3 ));
 927     _mm_storeu_pd( fptr+8, _mm_add_pd( _mm_loadu_pd(fptr+8),   fiz3 ));
 928     _mm_storeu_pd( fptr+10, _mm_add_pd( _mm_loadu_pd(fptr+10), fiy4 ));
 929
 930     t1   = _mm_shuffle_pd(fiz1, fiy2, _MM_SHUFFLE2(0, 1));
 931     fix1 = _mm_add_pd(fix1, t1);
 932     t2   = _mm_shuffle_pd(fiz3, fiy4, _MM_SHUFFLE2(0, 1));
 933     fix3 = _mm_add_pd(fix3, t2);
 934     fix1 = _mm_add_pd(fix1, fix3); /* x and y sums */
 935
 936     fiz1 = _mm_add_sd(fiz1, _mm_unpackhi_pd(fiy2, fiy2));
 937     fiz3 = _mm_add_sd(fiz3, _mm_unpackhi_pd(fiy4, fiy4));
 938     fiz1 = _mm_add_sd(fiz1, fiz3); /* z sum */
 939
 940     _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 ));
 941     _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 ));
 942 }
 943 #endif
 944
 945
 946 static gmx_inline void
 947 gmx_mm_update_1pot_pd(__m128d pot1, double * gmx_restrict ptrA)
 948 {
 949     pot1 = _mm_hadd_pd(pot1, pot1);
 950     _mm_store_sd(ptrA, _mm_add_sd(pot1, _mm_load_sd(ptrA)));
 951 }
 952
 953 static gmx_inline void
 954 gmx_mm_update_2pot_pd(__m128d pot1, double * gmx_restrict ptrA,
 955                       __m128d pot2, double * gmx_restrict ptrB)
 956 {
 957     pot1 = _mm_hadd_pd(pot1, pot2);
 958     pot2 = _mm_unpackhi_pd(pot1, pot1);
 959
 960     _mm_store_sd(ptrA, _mm_add_sd(pot1, _mm_load_sd(ptrA)));
 961     _mm_store_sd(ptrB, _mm_add_sd(pot2, _mm_load_sd(ptrB)));
 962 }
 963
 964
 965 #endif /* _kernelutil_x86_avx_128_fma_double_h_ */