src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/kernelutil_x86_sse4_1_double.h

   1 /*
   2  *                This source code is part of
   3  *
   4  *                 G   R   O   M   A   C   S
   5  *
   6  * Copyright (c) 2011-2012, The GROMACS Development Team
   7  *
   8  * Gromacs is a library for molecular simulation and trajectory analysis,
   9  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  10  * a full list of developers and information, check out http://www.gromacs.org
  11  *
  12  * This program is free software; you can redistribute it and/or modify it under
  13  * the terms of the GNU Lesser General Public License as published by the Free
  14  * Software Foundation; either version 2 of the License, or (at your option) any
  15  * later version.
  16  * As a special exception, you may use this file as part of a free software
  17  * library without restriction.  Specifically, if other files instantiate
  18  * templates or use macros or inline functions from this file, or you compile
  19  * this file and link it with other files to produce an executable, this
  20  * file does not by itself cause the resulting executable to be covered by
  21  * the GNU Lesser General Public License.
  22  *
  23  * In plain-speak: do not worry about classes/macros/templates either - only
  24  * changes to the library have to be LGPL, not an application linking with it.
  25  *
  26  * To help fund GROMACS development, we humbly ask that you cite
  27  * the papers people have written on it - you can find them on the website!
  28  */
  29 #ifndef _kernelutil_x86_sse4_1_double_h_
  30 #define _kernelutil_x86_sse4_1_double_h_
  31
  32 #include <math.h>
  33
  34 #include "gmx_x86_sse4_1.h"
  35
  36 #include <stdio.h>
  37
  38
  39 /* Normal sum of four ymm registers */
  40 #define gmx_mm_sum4_pd(t0, t1, t2, t3)  _mm_add_pd(_mm_add_pd(t0, t1), _mm_add_pd(t2, t3))
  41
  42 static int
  43 gmx_mm_any_lt(__m128d a, __m128d b)
  44 {
  45     return _mm_movemask_pd(_mm_cmplt_pd(a, b));
  46 }
  47
  48 static gmx_inline __m128d
  49 gmx_mm_calc_rsq_pd(__m128d dx, __m128d dy, __m128d dz)
  50 {
  51     return _mm_add_pd( _mm_add_pd( _mm_mul_pd(dx, dx), _mm_mul_pd(dy, dy) ), _mm_mul_pd(dz, dz) );
  52 }
  53
  54
  55 /* Load a double value from 1-2 places, merge into xmm register */
  56 static gmx_inline __m128d
  57 gmx_mm_load_2real_swizzle_pd(const double * gmx_restrict ptrA,
  58                              const double * gmx_restrict ptrB)
  59 {
  60     return _mm_unpacklo_pd(_mm_load_sd(ptrA), _mm_load_sd(ptrB));
  61 }
  62
  63 static gmx_inline __m128d
  64 gmx_mm_load_1real_pd(const double * gmx_restrict ptrA)
  65 {
  66     return _mm_load_sd(ptrA);
  67 }
  68
  69
  70 static gmx_inline void
  71 gmx_mm_store_2real_swizzle_pd(double * gmx_restrict ptrA,
  72                               double * gmx_restrict ptrB,
  73                               __m128d               xmm1)
  74 {
  75     __m128d t2;
  76
  77     t2       = _mm_unpackhi_pd(xmm1, xmm1);
  78     _mm_store_sd(ptrA, xmm1);
  79     _mm_store_sd(ptrB, t2);
  80 }
  81
  82 static gmx_inline void
  83 gmx_mm_store_1real_pd(double * gmx_restrict ptrA, __m128d xmm1)
  84 {
  85     _mm_store_sd(ptrA, xmm1);
  86 }
  87
  88
  89 /* Similar to store, but increments value in memory */
  90 static gmx_inline void
  91 gmx_mm_increment_2real_swizzle_pd(double * gmx_restrict ptrA,
  92                                   double * gmx_restrict ptrB, __m128d xmm1)
  93 {
  94     __m128d t1;
  95
  96     t1   = _mm_unpackhi_pd(xmm1, xmm1);
  97     xmm1 = _mm_add_sd(xmm1, _mm_load_sd(ptrA));
  98     t1   = _mm_add_sd(t1, _mm_load_sd(ptrB));
  99     _mm_store_sd(ptrA, xmm1);
 100     _mm_store_sd(ptrB, t1);
 101 }
 102
 103 static gmx_inline void
 104 gmx_mm_increment_1real_pd(double * gmx_restrict ptrA, __m128d xmm1)
 105 {
 106     __m128d tmp;
 107
 108     tmp = gmx_mm_load_1real_pd(ptrA);
 109     tmp = _mm_add_sd(tmp, xmm1);
 110     gmx_mm_store_1real_pd(ptrA, tmp);
 111 }
 112
 113
 114 static gmx_inline void
 115 gmx_mm_load_2pair_swizzle_pd(const double * gmx_restrict p1,
 116                              const double * gmx_restrict p2,
 117                              __m128d * gmx_restrict      c6,
 118                              __m128d * gmx_restrict      c12)
 119 {
 120     __m128d t1, t2, t3;
 121
 122     t1   = _mm_loadu_pd(p1);
 123     t2   = _mm_loadu_pd(p2);
 124     *c6  = _mm_unpacklo_pd(t1, t2);
 125     *c12 = _mm_unpackhi_pd(t1, t2);
 126 }
 127
 128 static gmx_inline void
 129 gmx_mm_load_1pair_swizzle_pd(const double * gmx_restrict p1,
 130                              __m128d * gmx_restrict      c6,
 131                              __m128d * gmx_restrict      c12)
 132 {
 133     *c6     = _mm_load_sd(p1);
 134     *c12    = _mm_load_sd(p1+1);
 135 }
 136
 137
 138
 139 static gmx_inline void
 140 gmx_mm_load_shift_and_1rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 141                                          const double * gmx_restrict xyz,
 142                                          __m128d * gmx_restrict      x1,
 143                                          __m128d * gmx_restrict      y1,
 144                                          __m128d * gmx_restrict      z1)
 145 {
 146     __m128d mem_xy, mem_z, mem_sxy, mem_sz;
 147
 148     mem_xy  = _mm_loadu_pd(xyz);
 149     mem_z   = _mm_load_sd(xyz+2);
 150     mem_sxy = _mm_loadu_pd(xyz_shift);
 151     mem_sz  = _mm_load_sd(xyz_shift+2);
 152
 153     mem_xy  = _mm_add_pd(mem_xy, mem_sxy);
 154     mem_z   = _mm_add_pd(mem_z, mem_sz);
 155
 156     *x1  = _mm_shuffle_pd(mem_xy, mem_xy, _MM_SHUFFLE2(0, 0));
 157     *y1  = _mm_shuffle_pd(mem_xy, mem_xy, _MM_SHUFFLE2(1, 1));
 158     *z1  = _mm_shuffle_pd(mem_z, mem_z, _MM_SHUFFLE2(0, 0));
 159 }
 160
 161
 162 static gmx_inline void
 163 gmx_mm_load_shift_and_3rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 164                                          const double * gmx_restrict xyz,
 165                                          __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 166                                          __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 167                                          __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 168 {
 169     __m128d t1, t2, t3, t4, t5, sxy, sz, szx, syz;
 170
 171     t1  = _mm_loadu_pd(xyz);
 172     t2  = _mm_loadu_pd(xyz+2);
 173     t3  = _mm_loadu_pd(xyz+4);
 174     t4  = _mm_loadu_pd(xyz+6);
 175     t5  = _mm_load_sd(xyz+8);
 176
 177     sxy = _mm_loadu_pd(xyz_shift);
 178     sz  = _mm_load_sd(xyz_shift+2);
 179     szx = _mm_shuffle_pd(sz, sxy, _MM_SHUFFLE2(0, 0));
 180     syz = _mm_shuffle_pd(sxy, sz, _MM_SHUFFLE2(0, 1));
 181
 182     t1  = _mm_add_pd(t1, sxy);
 183     t2  = _mm_add_pd(t2, szx);
 184     t3  = _mm_add_pd(t3, syz);
 185     t4  = _mm_add_pd(t4, sxy);
 186     t5  = _mm_add_sd(t5, sz);
 187
 188     *x1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(0, 0));
 189     *y1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(1, 1));
 190     *z1  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(0, 0));
 191     *x2  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(1, 1));
 192     *y2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(0, 0));
 193     *z2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(1, 1));
 194     *x3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(0, 0));
 195     *y3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(1, 1));
 196     *z3  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(0, 0));
 197 }
 198
 199
 200 static gmx_inline void
 201 gmx_mm_load_shift_and_4rvec_broadcast_pd(const double * gmx_restrict xyz_shift,
 202                                          const double * gmx_restrict xyz,
 203                                          __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 204                                          __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 205                                          __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 206                                          __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 207 {
 208     __m128d t1, t2, t3, t4, t5, t6, sxy, sz, szx, syz;
 209
 210     t1  = _mm_loadu_pd(xyz);
 211     t2  = _mm_loadu_pd(xyz+2);
 212     t3  = _mm_loadu_pd(xyz+4);
 213     t4  = _mm_loadu_pd(xyz+6);
 214     t5  = _mm_loadu_pd(xyz+8);
 215     t6  = _mm_loadu_pd(xyz+10);
 216
 217     sxy = _mm_loadu_pd(xyz_shift);
 218     sz  = _mm_load_sd(xyz_shift+2);
 219     szx = _mm_shuffle_pd(sz, sxy, _MM_SHUFFLE2(0, 0));
 220     syz = _mm_shuffle_pd(sxy, sz, _MM_SHUFFLE2(0, 1));
 221
 222     t1  = _mm_add_pd(t1, sxy);
 223     t2  = _mm_add_pd(t2, szx);
 224     t3  = _mm_add_pd(t3, syz);
 225     t4  = _mm_add_pd(t4, sxy);
 226     t5  = _mm_add_pd(t5, szx);
 227     t6  = _mm_add_pd(t6, syz);
 228
 229     *x1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(0, 0));
 230     *y1  = _mm_shuffle_pd(t1, t1, _MM_SHUFFLE2(1, 1));
 231     *z1  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(0, 0));
 232     *x2  = _mm_shuffle_pd(t2, t2, _MM_SHUFFLE2(1, 1));
 233     *y2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(0, 0));
 234     *z2  = _mm_shuffle_pd(t3, t3, _MM_SHUFFLE2(1, 1));
 235     *x3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(0, 0));
 236     *y3  = _mm_shuffle_pd(t4, t4, _MM_SHUFFLE2(1, 1));
 237     *z3  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(0, 0));
 238     *x4  = _mm_shuffle_pd(t5, t5, _MM_SHUFFLE2(1, 1));
 239     *y4  = _mm_shuffle_pd(t6, t6, _MM_SHUFFLE2(0, 0));
 240     *z4  = _mm_shuffle_pd(t6, t6, _MM_SHUFFLE2(1, 1));
 241 }
 242
 243
 244
 245
 246 static gmx_inline void
 247 gmx_mm_load_1rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 248                                   __m128d * gmx_restrict x, __m128d * gmx_restrict y, __m128d * gmx_restrict z)
 249 {
 250     *x            = _mm_load_sd(p1);
 251     *y            = _mm_load_sd(p1+1);
 252     *z            = _mm_load_sd(p1+2);
 253 }
 254
 255 static gmx_inline void
 256 gmx_mm_load_3rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 257                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 258                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 259                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 260 {
 261     *x1            = _mm_load_sd(p1);
 262     *y1            = _mm_load_sd(p1+1);
 263     *z1            = _mm_load_sd(p1+2);
 264     *x2            = _mm_load_sd(p1+3);
 265     *y2            = _mm_load_sd(p1+4);
 266     *z2            = _mm_load_sd(p1+5);
 267     *x3            = _mm_load_sd(p1+6);
 268     *y3            = _mm_load_sd(p1+7);
 269     *z3            = _mm_load_sd(p1+8);
 270 }
 271
 272 static gmx_inline void
 273 gmx_mm_load_4rvec_1ptr_swizzle_pd(const double * gmx_restrict p1,
 274                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 275                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 276                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 277                                   __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 278 {
 279     *x1            = _mm_load_sd(p1);
 280     *y1            = _mm_load_sd(p1+1);
 281     *z1            = _mm_load_sd(p1+2);
 282     *x2            = _mm_load_sd(p1+3);
 283     *y2            = _mm_load_sd(p1+4);
 284     *z2            = _mm_load_sd(p1+5);
 285     *x3            = _mm_load_sd(p1+6);
 286     *y3            = _mm_load_sd(p1+7);
 287     *z3            = _mm_load_sd(p1+8);
 288     *x4            = _mm_load_sd(p1+9);
 289     *y4            = _mm_load_sd(p1+10);
 290     *z4            = _mm_load_sd(p1+11);
 291 }
 292
 293
 294 static gmx_inline void
 295 gmx_mm_load_1rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA,
 296                                   const double * gmx_restrict ptrB,
 297                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1)
 298 {
 299     __m128d t1, t2, t3, t4;
 300     t1           = _mm_loadu_pd(ptrA);
 301     t2           = _mm_loadu_pd(ptrB);
 302     t3           = _mm_load_sd(ptrA+2);
 303     t4           = _mm_load_sd(ptrB+2);
 304     GMX_MM_TRANSPOSE2_PD(t1, t2);
 305     *x1          = t1;
 306     *y1          = t2;
 307     *z1          = _mm_unpacklo_pd(t3, t4);
 308 }
 309
 310
 311 static gmx_inline void
 312 gmx_mm_load_3rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA, const double * gmx_restrict ptrB,
 313                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 314                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 315                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3)
 316 {
 317     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10;
 318     t1           = _mm_loadu_pd(ptrA);
 319     t2           = _mm_loadu_pd(ptrB);
 320     t3           = _mm_loadu_pd(ptrA+2);
 321     t4           = _mm_loadu_pd(ptrB+2);
 322     t5           = _mm_loadu_pd(ptrA+4);
 323     t6           = _mm_loadu_pd(ptrB+4);
 324     t7           = _mm_loadu_pd(ptrA+6);
 325     t8           = _mm_loadu_pd(ptrB+6);
 326     t9           = _mm_load_sd(ptrA+8);
 327     t10          = _mm_load_sd(ptrB+8);
 328     GMX_MM_TRANSPOSE2_PD(t1, t2);
 329     GMX_MM_TRANSPOSE2_PD(t3, t4);
 330     GMX_MM_TRANSPOSE2_PD(t5, t6);
 331     GMX_MM_TRANSPOSE2_PD(t7, t8);
 332     *x1          = t1;
 333     *y1          = t2;
 334     *z1          = t3;
 335     *x2          = t4;
 336     *y2          = t5;
 337     *z2          = t6;
 338     *x3          = t7;
 339     *y3          = t8;
 340     *z3          = _mm_unpacklo_pd(t9, t10);
 341 }
 342
 343
 344 static gmx_inline void
 345 gmx_mm_load_4rvec_2ptr_swizzle_pd(const double * gmx_restrict ptrA, const double * gmx_restrict ptrB,
 346                                   __m128d * gmx_restrict x1, __m128d * gmx_restrict y1, __m128d * gmx_restrict z1,
 347                                   __m128d * gmx_restrict x2, __m128d * gmx_restrict y2, __m128d * gmx_restrict z2,
 348                                   __m128d * gmx_restrict x3, __m128d * gmx_restrict y3, __m128d * gmx_restrict z3,
 349                                   __m128d * gmx_restrict x4, __m128d * gmx_restrict y4, __m128d * gmx_restrict z4)
 350 {
 351     __m128d t1, t2, t3, t4, t5, t6;
 352     t1           = _mm_loadu_pd(ptrA);
 353     t2           = _mm_loadu_pd(ptrB);
 354     t3           = _mm_loadu_pd(ptrA+2);
 355     t4           = _mm_loadu_pd(ptrB+2);
 356     t5           = _mm_loadu_pd(ptrA+4);
 357     t6           = _mm_loadu_pd(ptrB+4);
 358     GMX_MM_TRANSPOSE2_PD(t1, t2);
 359     GMX_MM_TRANSPOSE2_PD(t3, t4);
 360     GMX_MM_TRANSPOSE2_PD(t5, t6);
 361     *x1          = t1;
 362     *y1          = t2;
 363     *z1          = t3;
 364     *x2          = t4;
 365     *y2          = t5;
 366     *z2          = t6;
 367     t1           = _mm_loadu_pd(ptrA+6);
 368     t2           = _mm_loadu_pd(ptrB+6);
 369     t3           = _mm_loadu_pd(ptrA+8);
 370     t4           = _mm_loadu_pd(ptrB+8);
 371     t5           = _mm_loadu_pd(ptrA+10);
 372     t6           = _mm_loadu_pd(ptrB+10);
 373     GMX_MM_TRANSPOSE2_PD(t1, t2);
 374     GMX_MM_TRANSPOSE2_PD(t3, t4);
 375     GMX_MM_TRANSPOSE2_PD(t5, t6);
 376     *x3          = t1;
 377     *y3          = t2;
 378     *z3          = t3;
 379     *x4          = t4;
 380     *y4          = t5;
 381     *z4          = t6;
 382 }
 383
 384
 385 /* Routines to decrement rvec in memory, typically use for j particle force updates */
 386 static gmx_inline void
 387 gmx_mm_decrement_1rvec_1ptr_noswizzle_pd(double * gmx_restrict ptrA,
 388                                          __m128d xy, __m128d z)
 389 {
 390     __m128d t1, t2;
 391
 392     t1 = _mm_loadu_pd(ptrA);
 393     t2 = _mm_load_sd(ptrA+2);
 394
 395     t1 = _mm_sub_pd(t1, xy);
 396     t2 = _mm_sub_sd(t2, z);
 397
 398     _mm_storeu_pd(ptrA, t1);
 399     _mm_store_sd(ptrA+2, t2);
 400 }
 401
 402
 403 static gmx_inline void
 404 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 405                                        __m128d x1, __m128d y1, __m128d z1)
 406 {
 407     __m128d t1, t2, t3;
 408
 409     t1           = _mm_load_sd(ptrA);
 410     t2           = _mm_load_sd(ptrA+1);
 411     t3           = _mm_load_sd(ptrA+2);
 412
 413     t1           = _mm_sub_sd(t1, x1);
 414     t2           = _mm_sub_sd(t2, y1);
 415     t3           = _mm_sub_sd(t3, z1);
 416     _mm_store_sd(ptrA, t1);
 417     _mm_store_sd(ptrA+1, t2);
 418     _mm_store_sd(ptrA+2, t3);
 419 }
 420
 421
 422 #if defined (_MSC_VER) && defined(_M_IX86)
 423 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 424 #define gmx_mm_decrement_3rvec_1ptr_swizzle_pd(ptrA, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3) \
 425     { \
 426         __m128d _t1, _t2, _t3, _t4, _t5; \
 427         _t1          = _mm_loadu_pd(ptrA); \
 428         _t2          = _mm_loadu_pd(ptrA+2); \
 429         _t3          = _mm_loadu_pd(ptrA+4); \
 430         _t4          = _mm_loadu_pd(ptrA+6); \
 431         _t5          = _mm_load_sd(ptrA+8); \
 432         _x1          = _mm_unpacklo_pd(_x1, _y1); \
 433         _z1          = _mm_unpacklo_pd(_z1, _x2); \
 434         _y2          = _mm_unpacklo_pd(_y2, _z2); \
 435         _x3          = _mm_unpacklo_pd(_x3, _y3); \
 436         _t1          = _mm_sub_pd(_t1, _x1); \
 437         _t2          = _mm_sub_pd(_t2, _z1); \
 438         _t3          = _mm_sub_pd(_t3, _y2); \
 439         _t4          = _mm_sub_pd(_t4, _x3); \
 440         _t5          = _mm_sub_sd(_t5, _z3); \
 441         _mm_storeu_pd(ptrA, _t1); \
 442         _mm_storeu_pd(ptrA+2, _t2); \
 443         _mm_storeu_pd(ptrA+4, _t3); \
 444         _mm_storeu_pd(ptrA+6, _t4); \
 445         _mm_store_sd(ptrA+8, _t5); \
 446     }
 447 #else
 448 /* Real function for sane compilers */
 449 static gmx_inline void
 450 gmx_mm_decrement_3rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 451                                        __m128d x1, __m128d y1, __m128d z1,
 452                                        __m128d x2, __m128d y2, __m128d z2,
 453                                        __m128d x3, __m128d y3, __m128d z3)
 454 {
 455     __m128d t1, t2, t3, t4, t5;
 456
 457     t1          = _mm_loadu_pd(ptrA);
 458     t2          = _mm_loadu_pd(ptrA+2);
 459     t3          = _mm_loadu_pd(ptrA+4);
 460     t4          = _mm_loadu_pd(ptrA+6);
 461     t5          = _mm_load_sd(ptrA+8);
 462
 463     x1          = _mm_unpacklo_pd(x1, y1);
 464     z1          = _mm_unpacklo_pd(z1, x2);
 465     y2          = _mm_unpacklo_pd(y2, z2);
 466     x3          = _mm_unpacklo_pd(x3, y3);
 467     /* nothing to be done for z3 */
 468
 469     t1          = _mm_sub_pd(t1, x1);
 470     t2          = _mm_sub_pd(t2, z1);
 471     t3          = _mm_sub_pd(t3, y2);
 472     t4          = _mm_sub_pd(t4, x3);
 473     t5          = _mm_sub_sd(t5, z3);
 474     _mm_storeu_pd(ptrA, t1);
 475     _mm_storeu_pd(ptrA+2, t2);
 476     _mm_storeu_pd(ptrA+4, t3);
 477     _mm_storeu_pd(ptrA+6, t4);
 478     _mm_store_sd(ptrA+8, t5);
 479 }
 480 #endif
 481
 482
 483 #if defined (_MSC_VER) && defined(_M_IX86)
 484 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 485 #define gmx_mm_decrement_4rvec_1ptr_swizzle_pd(ptrA, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3, _x4, _y4, _z4) \
 486     { \
 487         __m128d _t1, _t2, _t3, _t4, _t5, _t6; \
 488         _t1          = _mm_loadu_pd(ptrA); \
 489         _t2          = _mm_loadu_pd(ptrA+2); \
 490         _t3          = _mm_loadu_pd(ptrA+4); \
 491         _t4          = _mm_loadu_pd(ptrA+6); \
 492         _t5          = _mm_loadu_pd(ptrA+8); \
 493         _t6          = _mm_loadu_pd(ptrA+10); \
 494         _x1          = _mm_unpacklo_pd(_x1, _y1); \
 495         _z1          = _mm_unpacklo_pd(_z1, _x2); \
 496         _y2          = _mm_unpacklo_pd(_y2, _z2); \
 497         _x3          = _mm_unpacklo_pd(_x3, _y3); \
 498         _z3          = _mm_unpacklo_pd(_z3, _x4); \
 499         _y4          = _mm_unpacklo_pd(_y4, _z4); \
 500         _mm_storeu_pd(ptrA,    _mm_sub_pd( _t1, _x1 )); \
 501         _mm_storeu_pd(ptrA+2,  _mm_sub_pd( _t2, _z1 )); \
 502         _mm_storeu_pd(ptrA+4,  _mm_sub_pd( _t3, _y2 )); \
 503         _mm_storeu_pd(ptrA+6,  _mm_sub_pd( _t4, _x3 )); \
 504         _mm_storeu_pd(ptrA+8,  _mm_sub_pd( _t5, _z3 )); \
 505         _mm_storeu_pd(ptrA+10, _mm_sub_pd( _t6, _y4 )); \
 506     }
 507 #else
 508 /* Real function for sane compilers */
 509 static gmx_inline void
 510 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(double * gmx_restrict ptrA,
 511                                        __m128d x1, __m128d y1, __m128d z1,
 512                                        __m128d x2, __m128d y2, __m128d z2,
 513                                        __m128d x3, __m128d y3, __m128d z3,
 514                                        __m128d x4, __m128d y4, __m128d z4)
 515 {
 516     __m128d t1, t2, t3, t4, t5, t6;
 517
 518     t1          = _mm_loadu_pd(ptrA);
 519     t2          = _mm_loadu_pd(ptrA+2);
 520     t3          = _mm_loadu_pd(ptrA+4);
 521     t4          = _mm_loadu_pd(ptrA+6);
 522     t5          = _mm_loadu_pd(ptrA+8);
 523     t6          = _mm_loadu_pd(ptrA+10);
 524
 525     x1          = _mm_unpacklo_pd(x1, y1);
 526     z1          = _mm_unpacklo_pd(z1, x2);
 527     y2          = _mm_unpacklo_pd(y2, z2);
 528     x3          = _mm_unpacklo_pd(x3, y3);
 529     z3          = _mm_unpacklo_pd(z3, x4);
 530     y4          = _mm_unpacklo_pd(y4, z4);
 531
 532     _mm_storeu_pd(ptrA,    _mm_sub_pd( t1, x1 ));
 533     _mm_storeu_pd(ptrA+2,  _mm_sub_pd( t2, z1 ));
 534     _mm_storeu_pd(ptrA+4,  _mm_sub_pd( t3, y2 ));
 535     _mm_storeu_pd(ptrA+6,  _mm_sub_pd( t4, x3 ));
 536     _mm_storeu_pd(ptrA+8,  _mm_sub_pd( t5, z3 ));
 537     _mm_storeu_pd(ptrA+10, _mm_sub_pd( t6, y4 ));
 538 }
 539 #endif
 540
 541
 542 static gmx_inline void
 543 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 544                                        __m128d x1, __m128d y1, __m128d z1)
 545 {
 546     __m128d t1, t2, t3, t4, t5, t6, t7;
 547
 548     t1          = _mm_loadu_pd(ptrA);
 549     t2          = _mm_load_sd(ptrA+2);
 550     t3          = _mm_loadu_pd(ptrB);
 551     t4          = _mm_load_sd(ptrB+2);
 552
 553     t5          = _mm_unpacklo_pd(x1, y1);
 554     t6          = _mm_unpackhi_pd(x1, y1);
 555     t7          = _mm_unpackhi_pd(z1, z1);
 556
 557     t1          = _mm_sub_pd(t1, t5);
 558     t2          = _mm_sub_sd(t2, z1);
 559
 560     t3          = _mm_sub_pd(t3, t6);
 561     t4          = _mm_sub_sd(t4, t7);
 562
 563     _mm_storeu_pd(ptrA, t1);
 564     _mm_store_sd(ptrA+2, t2);
 565     _mm_storeu_pd(ptrB, t3);
 566     _mm_store_sd(ptrB+2, t4);
 567 }
 568
 569 #if defined (_MSC_VER) && defined(_M_IX86)
 570 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 571 #define gmx_mm_decrement_3rvec_2ptr_swizzle_pd(ptrA, ptrB, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3) \
 572     { \
 573         __m128d _t1, _t2, _t3, _t4, _t5, _t6, _t7, _t8, _t9, _t10; \
 574         __m128d _tA, _tB, _tC, _tD, _tE, _tF, _tG, _tH, _tI; \
 575         _t1          = _mm_loadu_pd(ptrA); \
 576         _t2          = _mm_loadu_pd(ptrA+2); \
 577         _t3          = _mm_loadu_pd(ptrA+4); \
 578         _t4          = _mm_loadu_pd(ptrA+6); \
 579         _t5          = _mm_load_sd(ptrA+8); \
 580         _t6          = _mm_loadu_pd(ptrB); \
 581         _t7          = _mm_loadu_pd(ptrB+2); \
 582         _t8          = _mm_loadu_pd(ptrB+4); \
 583         _t9          = _mm_loadu_pd(ptrB+6); \
 584         _t10         = _mm_load_sd(ptrB+8); \
 585         _tA          = _mm_unpacklo_pd(_x1, _y1); \
 586         _tB          = _mm_unpackhi_pd(_x1, _y1); \
 587         _tC          = _mm_unpacklo_pd(_z1, _x2); \
 588         _tD          = _mm_unpackhi_pd(_z1, _x2); \
 589         _tE          = _mm_unpacklo_pd(_y2, _z2); \
 590         _tF          = _mm_unpackhi_pd(_y2, _z2); \
 591         _tG          = _mm_unpacklo_pd(_x3, _y3); \
 592         _tH          = _mm_unpackhi_pd(_x3, _y3); \
 593         _tI          = _mm_unpackhi_pd(_z3, _z3); \
 594         _t1          = _mm_sub_pd(_t1, _tA); \
 595         _t2          = _mm_sub_pd(_t2, _tC); \
 596         _t3          = _mm_sub_pd(_t3, _tE); \
 597         _t4          = _mm_sub_pd(_t4, _tG); \
 598         _t5          = _mm_sub_sd(_t5, _z3); \
 599         _t6          = _mm_sub_pd(_t6, _tB); \
 600         _t7          = _mm_sub_pd(_t7, _tD); \
 601         _t8          = _mm_sub_pd(_t8, _tF); \
 602         _t9          = _mm_sub_pd(_t9, _tH); \
 603         _t10         = _mm_sub_sd(_t10, _tI); \
 604         _mm_storeu_pd(ptrA, _t1); \
 605         _mm_storeu_pd(ptrA+2, _t2); \
 606         _mm_storeu_pd(ptrA+4, _t3); \
 607         _mm_storeu_pd(ptrA+6, _t4); \
 608         _mm_store_sd(ptrA+8, _t5); \
 609         _mm_storeu_pd(ptrB, _t6); \
 610         _mm_storeu_pd(ptrB+2, _t7); \
 611         _mm_storeu_pd(ptrB+4, _t8); \
 612         _mm_storeu_pd(ptrB+6, _t9); \
 613         _mm_store_sd(ptrB+8, _t10); \
 614     }
 615 #else
 616 /* Real function for sane compilers */
 617 static gmx_inline void
 618 gmx_mm_decrement_3rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 619                                        __m128d x1, __m128d y1, __m128d z1,
 620                                        __m128d x2, __m128d y2, __m128d z2,
 621                                        __m128d x3, __m128d y3, __m128d z3)
 622 {
 623     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10;
 624     __m128d tA, tB, tC, tD, tE, tF, tG, tH, tI;
 625
 626     t1          = _mm_loadu_pd(ptrA);
 627     t2          = _mm_loadu_pd(ptrA+2);
 628     t3          = _mm_loadu_pd(ptrA+4);
 629     t4          = _mm_loadu_pd(ptrA+6);
 630     t5          = _mm_load_sd(ptrA+8);
 631     t6          = _mm_loadu_pd(ptrB);
 632     t7          = _mm_loadu_pd(ptrB+2);
 633     t8          = _mm_loadu_pd(ptrB+4);
 634     t9          = _mm_loadu_pd(ptrB+6);
 635     t10         = _mm_load_sd(ptrB+8);
 636
 637     tA          = _mm_unpacklo_pd(x1, y1);
 638     tB          = _mm_unpackhi_pd(x1, y1);
 639     tC          = _mm_unpacklo_pd(z1, x2);
 640     tD          = _mm_unpackhi_pd(z1, x2);
 641     tE          = _mm_unpacklo_pd(y2, z2);
 642     tF          = _mm_unpackhi_pd(y2, z2);
 643     tG          = _mm_unpacklo_pd(x3, y3);
 644     tH          = _mm_unpackhi_pd(x3, y3);
 645     tI          = _mm_unpackhi_pd(z3, z3);
 646
 647     t1          = _mm_sub_pd(t1, tA);
 648     t2          = _mm_sub_pd(t2, tC);
 649     t3          = _mm_sub_pd(t3, tE);
 650     t4          = _mm_sub_pd(t4, tG);
 651     t5          = _mm_sub_sd(t5, z3);
 652
 653     t6          = _mm_sub_pd(t6, tB);
 654     t7          = _mm_sub_pd(t7, tD);
 655     t8          = _mm_sub_pd(t8, tF);
 656     t9          = _mm_sub_pd(t9, tH);
 657     t10         = _mm_sub_sd(t10, tI);
 658
 659     _mm_storeu_pd(ptrA, t1);
 660     _mm_storeu_pd(ptrA+2, t2);
 661     _mm_storeu_pd(ptrA+4, t3);
 662     _mm_storeu_pd(ptrA+6, t4);
 663     _mm_store_sd(ptrA+8, t5);
 664     _mm_storeu_pd(ptrB, t6);
 665     _mm_storeu_pd(ptrB+2, t7);
 666     _mm_storeu_pd(ptrB+4, t8);
 667     _mm_storeu_pd(ptrB+6, t9);
 668     _mm_store_sd(ptrB+8, t10);
 669 }
 670 #endif
 671
 672
 673 #if defined (_MSC_VER) && defined(_M_IX86)
 674 /* Macro work-around since 32-bit MSVC cannot handle >3 xmm/ymm parameters */
 675 #define gmx_mm_decrement_4rvec_2ptr_swizzle_pd(ptrA, ptrB, _x1, _y1, _z1, _x2, _y2, _z2, _x3, _y3, _z3, _x4, _y4, _z4) \
 676     { \
 677         __m128d _t1, _t2, _t3, _t4, _t5, _t6, _t7, _t8, _t9, _t10, _t11, _t12; \
 678         __m128d _tA, _tB, _tC, _tD, _tE, _tF, _tG, _tH, _tI, _tJ, _tK, _tL; \
 679         _t1          = _mm_loadu_pd(ptrA); \
 680         _t2          = _mm_loadu_pd(ptrA+2); \
 681         _t3          = _mm_loadu_pd(ptrA+4); \
 682         _t4          = _mm_loadu_pd(ptrA+6); \
 683         _t5          = _mm_loadu_pd(ptrA+8); \
 684         _t6          = _mm_loadu_pd(ptrA+10); \
 685         _t7          = _mm_loadu_pd(ptrB); \
 686         _t8          = _mm_loadu_pd(ptrB+2); \
 687         _t9          = _mm_loadu_pd(ptrB+4); \
 688         _t10         = _mm_loadu_pd(ptrB+6); \
 689         _t11         = _mm_loadu_pd(ptrB+8); \
 690         _t12         = _mm_loadu_pd(ptrB+10); \
 691         _tA          = _mm_unpacklo_pd(_x1, _y1); \
 692         _tB          = _mm_unpackhi_pd(_x1, _y1); \
 693         _tC          = _mm_unpacklo_pd(_z1, _x2); \
 694         _tD          = _mm_unpackhi_pd(_z1, _x2); \
 695         _tE          = _mm_unpacklo_pd(_y2, _z2); \
 696         _tF          = _mm_unpackhi_pd(_y2, _z2); \
 697         _tG          = _mm_unpacklo_pd(_x3, _y3); \
 698         _tH          = _mm_unpackhi_pd(_x3, _y3); \
 699         _tI          = _mm_unpacklo_pd(_z3, _x4); \
 700         _tJ          = _mm_unpackhi_pd(_z3, _x4); \
 701         _tK          = _mm_unpacklo_pd(_y4, _z4); \
 702         _tL          = _mm_unpackhi_pd(_y4, _z4); \
 703         _t1          = _mm_sub_pd(_t1, _tA); \
 704         _t2          = _mm_sub_pd(_t2, _tC); \
 705         _t3          = _mm_sub_pd(_t3, _tE); \
 706         _t4          = _mm_sub_pd(_t4, _tG); \
 707         _t5          = _mm_sub_pd(_t5, _tI); \
 708         _t6          = _mm_sub_pd(_t6, _tK); \
 709         _t7          = _mm_sub_pd(_t7, _tB); \
 710         _t8          = _mm_sub_pd(_t8, _tD); \
 711         _t9          = _mm_sub_pd(_t9, _tF); \
 712         _t10         = _mm_sub_pd(_t10, _tH); \
 713         _t11         = _mm_sub_pd(_t11, _tJ); \
 714         _t12         = _mm_sub_pd(_t12, _tL); \
 715         _mm_storeu_pd(ptrA,  _t1); \
 716         _mm_storeu_pd(ptrA+2, _t2); \
 717         _mm_storeu_pd(ptrA+4, _t3); \
 718         _mm_storeu_pd(ptrA+6, _t4); \
 719         _mm_storeu_pd(ptrA+8, _t5); \
 720         _mm_storeu_pd(ptrA+10, _t6); \
 721         _mm_storeu_pd(ptrB,  _t7); \
 722         _mm_storeu_pd(ptrB+2, _t8); \
 723         _mm_storeu_pd(ptrB+4, _t9); \
 724         _mm_storeu_pd(ptrB+6, _t10); \
 725         _mm_storeu_pd(ptrB+8, _t11); \
 726         _mm_storeu_pd(ptrB+10, _t12); \
 727     }
 728 #else
 729 /* Real function for sane compilers */
 730 static gmx_inline void
 731 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(double * gmx_restrict ptrA, double * gmx_restrict ptrB,
 732                                        __m128d x1, __m128d y1, __m128d z1,
 733                                        __m128d x2, __m128d y2, __m128d z2,
 734                                        __m128d x3, __m128d y3, __m128d z3,
 735                                        __m128d x4, __m128d y4, __m128d z4)
 736 {
 737     __m128d t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12;
 738     __m128d tA, tB, tC, tD, tE, tF, tG, tH, tI, tJ, tK, tL;
 739
 740     t1          = _mm_loadu_pd(ptrA);
 741     t2          = _mm_loadu_pd(ptrA+2);
 742     t3          = _mm_loadu_pd(ptrA+4);
 743     t4          = _mm_loadu_pd(ptrA+6);
 744     t5          = _mm_loadu_pd(ptrA+8);
 745     t6          = _mm_loadu_pd(ptrA+10);
 746     t7          = _mm_loadu_pd(ptrB);
 747     t8          = _mm_loadu_pd(ptrB+2);
 748     t9          = _mm_loadu_pd(ptrB+4);
 749     t10         = _mm_loadu_pd(ptrB+6);
 750     t11         = _mm_loadu_pd(ptrB+8);
 751     t12         = _mm_loadu_pd(ptrB+10);
 752
 753     tA          = _mm_unpacklo_pd(x1, y1);
 754     tB          = _mm_unpackhi_pd(x1, y1);
 755     tC          = _mm_unpacklo_pd(z1, x2);
 756     tD          = _mm_unpackhi_pd(z1, x2);
 757     tE          = _mm_unpacklo_pd(y2, z2);
 758     tF          = _mm_unpackhi_pd(y2, z2);
 759     tG          = _mm_unpacklo_pd(x3, y3);
 760     tH          = _mm_unpackhi_pd(x3, y3);
 761     tI          = _mm_unpacklo_pd(z3, x4);
 762     tJ          = _mm_unpackhi_pd(z3, x4);
 763     tK          = _mm_unpacklo_pd(y4, z4);
 764     tL          = _mm_unpackhi_pd(y4, z4);
 765
 766     t1          = _mm_sub_pd(t1, tA);
 767     t2          = _mm_sub_pd(t2, tC);
 768     t3          = _mm_sub_pd(t3, tE);
 769     t4          = _mm_sub_pd(t4, tG);
 770     t5          = _mm_sub_pd(t5, tI);
 771     t6          = _mm_sub_pd(t6, tK);
 772
 773     t7          = _mm_sub_pd(t7, tB);
 774     t8          = _mm_sub_pd(t8, tD);
 775     t9          = _mm_sub_pd(t9, tF);
 776     t10         = _mm_sub_pd(t10, tH);
 777     t11         = _mm_sub_pd(t11, tJ);
 778     t12         = _mm_sub_pd(t12, tL);
 779
 780     _mm_storeu_pd(ptrA,  t1);
 781     _mm_storeu_pd(ptrA+2, t2);
 782     _mm_storeu_pd(ptrA+4, t3);
 783     _mm_storeu_pd(ptrA+6, t4);
 784     _mm_storeu_pd(ptrA+8, t5);
 785     _mm_storeu_pd(ptrA+10, t6);
 786     _mm_storeu_pd(ptrB,  t7);
 787     _mm_storeu_pd(ptrB+2, t8);
 788     _mm_storeu_pd(ptrB+4, t9);
 789     _mm_storeu_pd(ptrB+6, t10);
 790     _mm_storeu_pd(ptrB+8, t11);
 791     _mm_storeu_pd(ptrB+10, t12);
 792 }
 793 #endif
 794
 795
 796
 797
 798 static gmx_inline void
 799 gmx_mm_update_iforce_1atom_swizzle_pd(__m128d fix1, __m128d fiy1, __m128d fiz1,
 800                                       double * gmx_restrict fptr,
 801                                       double * gmx_restrict fshiftptr)
 802 {
 803     fix1 = _mm_hadd_pd(fix1, fiy1);
 804     fiz1 = _mm_hadd_pd(fiz1, fiz1);
 805
 806     _mm_storeu_pd( fptr, _mm_add_pd( _mm_loadu_pd(fptr), fix1 ));
 807     _mm_store_sd( fptr+2, _mm_add_sd( _mm_load_sd(fptr+2), fiz1 ));
 808
 809     _mm_storeu_pd( fshiftptr, _mm_add_pd( _mm_loadu_pd(fshiftptr), fix1 ));
 810     _mm_store_sd( fshiftptr+2, _mm_add_sd( _mm_load_sd(fshiftptr+2), fiz1 ));
 811 }
 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 static gmx_inline void
 946 gmx_mm_update_1pot_pd(__m128d pot1, double * gmx_restrict ptrA)
 947 {
 948     pot1 = _mm_hadd_pd(pot1, pot1);
 949     _mm_store_sd(ptrA, _mm_add_sd(pot1, _mm_load_sd(ptrA)));
 950 }
 951
 952 static gmx_inline void
 953 gmx_mm_update_2pot_pd(__m128d pot1, double * gmx_restrict ptrA,
 954                       __m128d pot2, double * gmx_restrict ptrB)
 955 {
 956     pot1 = _mm_hadd_pd(pot1, pot2);
 957     pot2 = _mm_unpackhi_pd(pot1, pot1);
 958
 959     _mm_store_sd(ptrA, _mm_add_sd(pot1, _mm_load_sd(ptrA)));
 960     _mm_store_sd(ptrB, _mm_add_sd(pot2, _mm_load_sd(ptrB)));
 961 }
 962
 963
 964 #endif /* _kernelutil_x86_sse4_1_double_h_ */