2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2013,2014,2015,2016,2017 by the GROMACS development team.
7 * Copyright (c) 2018,2019,2020, by the GROMACS development team, led by
8 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
9 * and including many others, as listed in the AUTHORS file in the
10 * top-level source directory and at http://www.gromacs.org.
12 * GROMACS is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public License
14 * as published by the Free Software Foundation; either version 2.1
15 * of the License, or (at your option) any later version.
17 * GROMACS is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with GROMACS; if not, see
24 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
25 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
27 * If you want to redistribute modifications to GROMACS, please
28 * consider that scientific software is very special. Version
29 * control is crucial - bugs must be traceable. We will be happy to
30 * consider code for inclusion in the official distribution, but
31 * derived work must not be called official GROMACS. Details are found
32 * in the README & COPYING files - if they are missing, get the
33 * official version at http://www.gromacs.org.
35 * To help us fund GROMACS development, we humbly ask that you cite
36 * the research papers on the package. Check out http://www.gromacs.org.
39 * \brief Defines SETTLE code.
41 * \author Berk Hess <hess@kth.se>
42 * \author Mark Abraham <mark.j.abraham@gmail.com>
43 * \ingroup module_mdlib
55 #include "gromacs/math/arrayrefwithpadding.h"
56 #include "gromacs/math/functions.h"
57 #include "gromacs/math/invertmatrix.h"
58 #include "gromacs/math/vec.h"
59 #include "gromacs/mdlib/constr.h"
60 #include "gromacs/mdtypes/mdatom.h"
61 #include "gromacs/pbcutil/ishift.h"
62 #include "gromacs/pbcutil/pbc.h"
63 #include "gromacs/pbcutil/pbc_simd.h"
64 #include "gromacs/simd/simd.h"
65 #include "gromacs/simd/simd_math.h"
66 #include "gromacs/topology/idef.h"
67 #include "gromacs/topology/ifunc.h"
68 #include "gromacs/topology/mtop_util.h"
69 #include "gromacs/utility/fatalerror.h"
70 #include "gromacs/utility/gmxassert.h"
71 #include "gromacs/utility/smalloc.h"
97 settleparam_t massw; /* Parameters for SETTLE for coordinates */
98 settleparam_t mass1; /* Parameters with all masses 1, for forces */
100 int nsettle; /* The number of settles on our rank */
101 int* ow1; /* Index to OW1 atoms, size nsettle + SIMD padding */
102 int* hw2; /* Index to HW2 atoms, size nsettle + SIMD padding */
103 int* hw3; /* Index to HW3 atoms, size nsettle + SIMD padding */
104 real* virfac; /* Virial factor 0 or 1, size nsettle + SIMD pad. */
105 int nalloc; /* Allocation size of ow1, hw2, hw3, virfac */
107 bool bUseSimd; /* Use SIMD intrinsics code, if possible */
111 //! Initializes a projection matrix.
112 static void init_proj_matrix(real invmO, real invmH, real dOH, real dHH, matrix inverseCouplingMatrix)
114 /* We normalize the inverse masses with invmO for the matrix inversion.
115 * so we can keep using masses of almost zero for frozen particles,
116 * without running out of the float range in invertMatrix.
118 double invmORelative = 1.0;
119 double invmHRelative = invmH / static_cast<double>(invmO);
120 double distanceRatio = dHH / static_cast<double>(dOH);
122 /* Construct the constraint coupling matrix */
124 mat[0][0] = invmORelative + invmHRelative;
125 mat[0][1] = invmORelative * (1.0 - 0.5 * gmx::square(distanceRatio));
126 mat[0][2] = invmHRelative * 0.5 * distanceRatio;
127 mat[1][1] = mat[0][0];
128 mat[1][2] = mat[0][2];
129 mat[2][2] = invmHRelative + invmHRelative;
130 mat[1][0] = mat[0][1];
131 mat[2][0] = mat[0][2];
132 mat[2][1] = mat[1][2];
134 invertMatrix(mat, inverseCouplingMatrix);
136 msmul(inverseCouplingMatrix, 1 / invmO, inverseCouplingMatrix);
139 //! Initializes settle parameters.
140 static void settleparam_init(settleparam_t* p, real mO, real mH, real invmO, real invmH, real dOH, real dHH)
142 /* We calculate parameters in double precision to minimize errors.
143 * The velocity correction applied during SETTLE coordinate constraining
144 * introduces a systematic error of approximately 1 bit per atom,
145 * depending on what the compiler does with the code.
151 wohh = mO + 2.0 * mH;
155 double rc = dHH / 2.0;
156 double ra = 2.0 * mH * std::sqrt(dOH * dOH - rc * rc) / wohh;
157 p->rb = std::sqrt(dOH * dOH - rc * rc) - ra;
162 /* For projection: inverse masses and coupling matrix inversion */
166 p->invdOH = 1.0 / dOH;
167 p->invdHH = 1.0 / dHH;
169 init_proj_matrix(invmO, invmH, dOH, dHH, p->invmat);
173 fprintf(debug, "wh =%g, rc = %g, ra = %g\n", p->wh, p->rc, p->ra);
174 fprintf(debug, "rb = %g, irc2 = %g, dHH = %g, dOH = %g\n", p->rb, p->irc2, p->dHH, p->dOH);
178 settledata* settle_init(const gmx_mtop_t& mtop)
180 /* Check that we have only one settle type */
181 int settle_type = -1;
182 gmx_mtop_ilistloop_t iloop = gmx_mtop_ilistloop_init(mtop);
184 const int nral1 = 1 + NRAL(F_SETTLE);
185 while (const InteractionLists* ilists = gmx_mtop_ilistloop_next(iloop, &nmol))
187 const InteractionList& ilist = (*ilists)[F_SETTLE];
188 for (int i = 0; i < ilist.size(); i += nral1)
190 if (settle_type == -1)
192 settle_type = ilist.iatoms[i];
194 else if (ilist.iatoms[i] != settle_type)
197 "The [molecules] section of your topology specifies more than one block "
199 "a [moleculetype] with a [settles] block. Only one such is allowed.\n"
200 "If you are trying to partition your solvent into different *groups*\n"
201 "(e.g. for freezing, T-coupling, etc.), you are using the wrong "
203 "files specify groups. Otherwise, you may wish to change the least-used\n"
204 "block of molecules with SETTLE constraints into 3 normal constraints.");
208 GMX_RELEASE_ASSERT(settle_type >= 0, "settle_init called without settles");
214 /* We will not initialize the normal SETTLE parameters here yet,
215 * since the atom (inv)masses can depend on the integrator and
216 * free-energy perturbation. We set mO=-1 to trigger later initialization.
218 settled->massw.mO = -1;
220 real dOH = mtop.ffparams.iparams[settle_type].settle.doh;
221 real dHH = mtop.ffparams.iparams[settle_type].settle.dhh;
222 settleparam_init(&settled->mass1, 1.0, 1.0, 1.0, 1.0, dOH, dHH);
224 settled->ow1 = nullptr;
225 settled->hw2 = nullptr;
226 settled->hw3 = nullptr;
227 settled->virfac = nullptr;
230 /* Without SIMD configured, this bool is not used */
231 settled->bUseSimd = (getenv("GMX_DISABLE_SIMD_KERNELS") == nullptr);
236 void settle_free(settledata* settled)
238 sfree_aligned(settled->ow1);
239 sfree_aligned(settled->hw2);
240 sfree_aligned(settled->hw3);
241 sfree_aligned(settled->virfac);
245 void settle_set_constraints(settledata* settled, const t_ilist* il_settle, const t_mdatoms& mdatoms)
247 #if GMX_SIMD_HAVE_REAL
248 const int pack_size = GMX_SIMD_REAL_WIDTH;
250 const int pack_size = 1;
253 const int nral1 = 1 + NRAL(F_SETTLE);
254 int nsettle = il_settle->nr / nral1;
255 settled->nsettle = nsettle;
259 const t_iatom* iatoms = il_settle->iatoms;
261 /* Here we initialize the normal SETTLE parameters */
262 if (settled->massw.mO < 0)
264 int firstO = iatoms[1];
265 int firstH = iatoms[2];
266 settleparam_init(&settled->massw, mdatoms.massT[firstO], mdatoms.massT[firstH],
267 mdatoms.invmass[firstO], mdatoms.invmass[firstH], settled->mass1.dOH,
271 if (nsettle + pack_size > settled->nalloc)
273 settled->nalloc = over_alloc_dd(nsettle + pack_size);
274 sfree_aligned(settled->ow1);
275 sfree_aligned(settled->hw2);
276 sfree_aligned(settled->hw3);
277 sfree_aligned(settled->virfac);
278 snew_aligned(settled->ow1, settled->nalloc, 64);
279 snew_aligned(settled->hw2, settled->nalloc, 64);
280 snew_aligned(settled->hw3, settled->nalloc, 64);
281 snew_aligned(settled->virfac, settled->nalloc, 64);
284 for (int i = 0; i < nsettle; i++)
286 settled->ow1[i] = iatoms[i * nral1 + 1];
287 settled->hw2[i] = iatoms[i * nral1 + 2];
288 settled->hw3[i] = iatoms[i * nral1 + 3];
289 /* We should avoid double counting of virial contributions for
290 * SETTLEs that appear in multiple DD domains, so we only count
291 * the contribution on the home range of the oxygen atom.
293 settled->virfac[i] = (iatoms[i * nral1 + 1] < mdatoms.homenr ? 1 : 0);
296 /* Pack the index array to the full SIMD width with copies from
297 * the last normal entry, but with no virial contribution.
299 int end_packed = ((nsettle + pack_size - 1) / pack_size) * pack_size;
300 for (int i = nsettle; i < end_packed; i++)
302 settled->ow1[i] = settled->ow1[nsettle - 1];
303 settled->hw2[i] = settled->hw2[nsettle - 1];
304 settled->hw3[i] = settled->hw3[nsettle - 1];
305 settled->virfac[i] = 0;
310 void settle_proj(settledata* settled,
311 ConstraintVariable econq,
313 const t_iatom iatoms[],
315 ArrayRef<const RVec> x,
318 int calcvir_atom_end,
321 /* Settle for projection out constraint components
322 * of derivatives of the coordinates.
323 * Berk Hess 2008-1-10
327 real imO, imH, dOH, dHH, invdOH, invdHH;
329 int i, m, m2, ow1, hw2, hw3;
330 rvec roh2, roh3, rhh, dc, fc;
332 calcvir_atom_end *= DIM;
334 if (econq == ConstraintVariable::Force)
344 copy_mat(p->invmat, invmat);
350 const int nral1 = 1 + NRAL(F_SETTLE);
352 for (i = 0; i < nsettle; i++)
354 ow1 = iatoms[i * nral1 + 1];
355 hw2 = iatoms[i * nral1 + 2];
356 hw3 = iatoms[i * nral1 + 3];
360 rvec_sub(x[ow1], x[hw2], roh2);
361 rvec_sub(x[ow1], x[hw3], roh3);
362 rvec_sub(x[hw2], x[hw3], rhh);
366 pbc_dx_aiuc(pbc, x[ow1], x[hw2], roh2);
367 pbc_dx_aiuc(pbc, x[ow1], x[hw3], roh3);
368 pbc_dx_aiuc(pbc, x[hw2], x[hw3], rhh);
370 svmul(invdOH, roh2, roh2);
371 svmul(invdOH, roh3, roh3);
372 svmul(invdHH, rhh, rhh);
375 /* Determine the projections of der on the bonds */
377 for (m = 0; m < DIM; m++)
379 dc[0] += (der[ow1][m] - der[hw2][m]) * roh2[m];
380 dc[1] += (der[ow1][m] - der[hw3][m]) * roh3[m];
381 dc[2] += (der[hw2][m] - der[hw3][m]) * rhh[m];
385 /* Determine the correction for the three bonds */
386 mvmul(invmat, dc, fc);
389 /* Subtract the corrections from derp */
390 for (m = 0; m < DIM; m++)
392 derp[ow1][m] -= imO * (fc[0] * roh2[m] + fc[1] * roh3[m]);
393 derp[hw2][m] -= imH * (-fc[0] * roh2[m] + fc[2] * rhh[m]);
394 derp[hw3][m] -= imH * (-fc[1] * roh3[m] - fc[2] * rhh[m]);
399 if (ow1 < calcvir_atom_end)
401 /* Determining r \dot m der is easy,
402 * since fc contains the mass weighted corrections for der.
405 for (m = 0; m < DIM; m++)
407 for (m2 = 0; m2 < DIM; m2++)
409 vir_r_m_dder[m][m2] += dOH * roh2[m] * roh2[m2] * fc[0]
410 + dOH * roh3[m] * roh3[m2] * fc[1]
411 + dHH * rhh[m] * rhh[m2] * fc[2];
419 /*! \brief The actual settle code, templated for real/SimdReal and for optimization */
420 template<typename T, typename TypeBool, int packSize, typename TypePbc, bool bCorrectVelocity, bool bCalcVirial>
421 static void settleTemplate(const settledata* settled,
428 real* gmx_restrict v,
430 bool* bErrorHasOccurred)
432 /* ******************************************************************* */
434 /* Original code by Shuichi Miyamoto, last update Oct. 1, 1992 ** */
436 /* Algorithm changes by Berk Hess: ** */
437 /* 2004-07-15 Convert COM to double precision to avoid drift ** */
438 /* 2006-10-16 Changed velocity update to use differences ** */
439 /* 2012-09-24 Use oxygen as reference instead of COM ** */
440 /* 2016-02 Complete rewrite of the code for SIMD ** */
442 /* Reference for the SETTLE algorithm ** */
443 /* S. Miyamoto et al., J. Comp. Chem., 13, 952 (1992). ** */
445 /* ******************************************************************* */
447 assert(settleStart % packSize == 0);
448 assert(settleEnd % packSize == 0);
450 TypeBool bError = TypeBool(false);
452 const settleparam_t* p = &settled->massw;
461 T almost_zero = T(1e-12);
463 T sum_r_m_dr[DIM][DIM];
467 for (int d2 = 0; d2 < DIM; d2++)
469 for (int d = 0; d < DIM; d++)
471 sum_r_m_dr[d2][d] = T(0);
476 for (int i = settleStart; i < settleEnd; i += packSize)
478 /* Here we pad up to packSize with copies from the last valid entry.
479 * This gives correct results, since we store (not increment) all
480 * output, so we store the same output multiple times.
482 const int* ow1 = settled->ow1 + i;
483 const int* hw2 = settled->hw2 + i;
484 const int* hw3 = settled->hw3 + i;
486 T x_ow1[DIM], x_hw2[DIM], x_hw3[DIM];
488 gatherLoadUTranspose<3>(x, ow1, &x_ow1[XX], &x_ow1[YY], &x_ow1[ZZ]);
489 gatherLoadUTranspose<3>(x, hw2, &x_hw2[XX], &x_hw2[YY], &x_hw2[ZZ]);
490 gatherLoadUTranspose<3>(x, hw3, &x_hw3[XX], &x_hw3[YY], &x_hw3[ZZ]);
492 T xprime_ow1[DIM], xprime_hw2[DIM], xprime_hw3[DIM];
494 gatherLoadUTranspose<3>(xprime, ow1, &xprime_ow1[XX], &xprime_ow1[YY], &xprime_ow1[ZZ]);
495 gatherLoadUTranspose<3>(xprime, hw2, &xprime_hw2[XX], &xprime_hw2[YY], &xprime_hw2[ZZ]);
496 gatherLoadUTranspose<3>(xprime, hw3, &xprime_hw3[XX], &xprime_hw3[YY], &xprime_hw3[ZZ]);
498 T dist21[DIM], dist31[DIM];
499 T doh2[DIM], doh3[DIM];
500 T sh_hw2[DIM], sh_hw3[DIM];
502 pbc_dx_aiuc(pbc, x_hw2, x_ow1, dist21);
504 pbc_dx_aiuc(pbc, x_hw3, x_ow1, dist31);
506 /* Tedious way of doing pbc */
507 pbc_dx_aiuc(pbc, xprime_hw2, xprime_ow1, doh2);
508 for (int d = 0; d < DIM; d++)
510 sh_hw2[d] = xprime_hw2[d] - (xprime_ow1[d] + doh2[d]);
511 xprime_hw2[d] = xprime_hw2[d] - sh_hw2[d];
513 pbc_dx_aiuc(pbc, xprime_hw3, xprime_ow1, doh3);
514 for (int d = 0; d < DIM; d++)
516 sh_hw3[d] = xprime_hw3[d] - (xprime_ow1[d] + doh3[d]);
517 xprime_hw3[d] = xprime_hw3[d] - sh_hw3[d];
520 /* Not calculating the center of mass using the oxygen position
521 * and the O-H distances, as done below, will make SETTLE
522 * the largest source of energy drift for simulations of water,
523 * as then the oxygen coordinate is multiplied by 0.89 at every step,
524 * which can then transfer a systematic rounding to the oxygen velocity.
527 for (int d = 0; d < DIM; d++)
529 a1[d] = -(doh2[d] + doh3[d]) * wh;
530 com[d] = xprime_ow1[d] - a1[d];
533 for (int d = 0; d < DIM; d++)
535 b1[d] = xprime_hw2[d] - com[d];
538 for (int d = 0; d < DIM; d++)
540 c1[d] = xprime_hw3[d] - com[d];
544 T xakszd = dist21[YY] * dist31[ZZ] - dist21[ZZ] * dist31[YY];
545 T yakszd = dist21[ZZ] * dist31[XX] - dist21[XX] * dist31[ZZ];
546 T zakszd = dist21[XX] * dist31[YY] - dist21[YY] * dist31[XX];
547 T xaksxd = a1[YY] * zakszd - a1[ZZ] * yakszd;
548 T yaksxd = a1[ZZ] * xakszd - a1[XX] * zakszd;
549 T zaksxd = a1[XX] * yakszd - a1[YY] * xakszd;
550 T xaksyd = yakszd * zaksxd - zakszd * yaksxd;
551 T yaksyd = zakszd * xaksxd - xakszd * zaksxd;
552 T zaksyd = xakszd * yaksxd - yakszd * xaksxd;
555 T axlng = gmx::invsqrt(xaksxd * xaksxd + yaksxd * yaksxd + zaksxd * zaksxd);
556 T aylng = gmx::invsqrt(xaksyd * xaksyd + yaksyd * yaksyd + zaksyd * zaksyd);
557 T azlng = gmx::invsqrt(xakszd * xakszd + yakszd * yakszd + zakszd * zakszd);
559 T trns1[DIM], trns2[DIM], trns3[DIM];
561 trns1[XX] = xaksxd * axlng;
562 trns2[XX] = yaksxd * axlng;
563 trns3[XX] = zaksxd * axlng;
564 trns1[YY] = xaksyd * aylng;
565 trns2[YY] = yaksyd * aylng;
566 trns3[YY] = zaksyd * aylng;
567 trns1[ZZ] = xakszd * azlng;
568 trns2[ZZ] = yakszd * azlng;
569 trns3[ZZ] = zakszd * azlng;
574 for (int d = 0; d < 2; d++)
576 b0d[d] = trns1[d] * dist21[XX] + trns2[d] * dist21[YY] + trns3[d] * dist21[ZZ];
577 c0d[d] = trns1[d] * dist31[XX] + trns2[d] * dist31[YY] + trns3[d] * dist31[ZZ];
580 T a1d_z, b1d[DIM], c1d[DIM];
582 a1d_z = trns1[ZZ] * a1[XX] + trns2[ZZ] * a1[YY] + trns3[ZZ] * a1[ZZ];
583 for (int d = 0; d < DIM; d++)
585 b1d[d] = trns1[d] * b1[XX] + trns2[d] * b1[YY] + trns3[d] * b1[ZZ];
586 c1d[d] = trns1[d] * c1[XX] + trns2[d] * c1[YY] + trns3[d] * c1[ZZ];
592 T sinphi = a1d_z * gmx::invsqrt(ra * ra);
593 tmp2 = 1.0 - sinphi * sinphi;
595 /* If tmp2 gets close to or beyond zero we have severly distorted
596 * water molecules and we should terminate the simulation.
597 * Below we take the max with almost_zero to continue the loop.
599 bError = bError || (tmp2 <= almost_zero);
601 tmp2 = max(tmp2, almost_zero);
602 tmp = gmx::invsqrt(tmp2);
603 T cosphi = tmp2 * tmp;
604 T sinpsi = (b1d[ZZ] - c1d[ZZ]) * irc2 * tmp;
605 tmp2 = 1.0 - sinpsi * sinpsi;
607 T cospsi = tmp2 * gmx::invsqrt(tmp2);
610 T a2d_y = ra * cosphi;
611 T b2d_x = -rc * cospsi;
613 T t2 = rc * sinpsi * sinphi;
618 /* --- Step3 al,be,ga --- */
619 T alpha = b2d_x * (b0d[XX] - c0d[XX]) + b0d[YY] * b2d_y + c0d[YY] * c2d_y;
620 T beta = b2d_x * (c0d[YY] - b0d[YY]) + b0d[XX] * b2d_y + c0d[XX] * c2d_y;
621 T gamma = b0d[XX] * b1d[YY] - b1d[XX] * b0d[YY] + c0d[XX] * c1d[YY] - c1d[XX] * c0d[YY];
622 T al2be2 = alpha * alpha + beta * beta;
623 tmp2 = (al2be2 - gamma * gamma);
624 T sinthe = (alpha * gamma - beta * tmp2 * gmx::invsqrt(tmp2)) * gmx::invsqrt(al2be2 * al2be2);
627 /* --- Step4 A3' --- */
628 tmp2 = 1.0 - sinthe * sinthe;
629 T costhe = tmp2 * gmx::invsqrt(tmp2);
631 T a3d[DIM], b3d[DIM], c3d[DIM];
633 a3d[XX] = -a2d_y * sinthe;
634 a3d[YY] = a2d_y * costhe;
636 b3d[XX] = b2d_x * costhe - b2d_y * sinthe;
637 b3d[YY] = b2d_x * sinthe + b2d_y * costhe;
639 c3d[XX] = -b2d_x * costhe - c2d_y * sinthe;
640 c3d[YY] = -b2d_x * sinthe + c2d_y * costhe;
644 /* --- Step5 A3 --- */
645 T a3[DIM], b3[DIM], c3[DIM];
647 a3[XX] = trns1[XX] * a3d[XX] + trns1[YY] * a3d[YY] + trns1[ZZ] * a3d[ZZ];
648 a3[YY] = trns2[XX] * a3d[XX] + trns2[YY] * a3d[YY] + trns2[ZZ] * a3d[ZZ];
649 a3[ZZ] = trns3[XX] * a3d[XX] + trns3[YY] * a3d[YY] + trns3[ZZ] * a3d[ZZ];
650 b3[XX] = trns1[XX] * b3d[XX] + trns1[YY] * b3d[YY] + trns1[ZZ] * b3d[ZZ];
651 b3[YY] = trns2[XX] * b3d[XX] + trns2[YY] * b3d[YY] + trns2[ZZ] * b3d[ZZ];
652 b3[ZZ] = trns3[XX] * b3d[XX] + trns3[YY] * b3d[YY] + trns3[ZZ] * b3d[ZZ];
653 c3[XX] = trns1[XX] * c3d[XX] + trns1[YY] * c3d[YY] + trns1[ZZ] * c3d[ZZ];
654 c3[YY] = trns2[XX] * c3d[XX] + trns2[YY] * c3d[YY] + trns2[ZZ] * c3d[ZZ];
655 c3[ZZ] = trns3[XX] * c3d[XX] + trns3[YY] * c3d[YY] + trns3[ZZ] * c3d[ZZ];
658 /* Compute and store the corrected new coordinate */
659 for (int d = 0; d < DIM; d++)
661 xprime_ow1[d] = com[d] + a3[d];
663 for (int d = 0; d < DIM; d++)
665 xprime_hw2[d] = com[d] + b3[d] + sh_hw2[d];
667 for (int d = 0; d < DIM; d++)
669 xprime_hw3[d] = com[d] + c3[d] + sh_hw3[d];
671 /* 9 flops + 6 pbc flops */
673 transposeScatterStoreU<3>(xprime, ow1, xprime_ow1[XX], xprime_ow1[YY], xprime_ow1[ZZ]);
674 transposeScatterStoreU<3>(xprime, hw2, xprime_hw2[XX], xprime_hw2[YY], xprime_hw2[ZZ]);
675 transposeScatterStoreU<3>(xprime, hw3, xprime_hw3[XX], xprime_hw3[YY], xprime_hw3[ZZ]);
677 if (bCorrectVelocity || bCalcVirial)
679 T da[DIM], db[DIM], dc[DIM];
680 for (int d = 0; d < DIM; d++)
682 da[d] = a3[d] - a1[d];
684 for (int d = 0; d < DIM; d++)
686 db[d] = b3[d] - b1[d];
688 for (int d = 0; d < DIM; d++)
690 dc[d] = c3[d] - c1[d];
694 if (bCorrectVelocity)
696 T v_ow1[DIM], v_hw2[DIM], v_hw3[DIM];
698 gatherLoadUTranspose<3>(v, ow1, &v_ow1[XX], &v_ow1[YY], &v_ow1[ZZ]);
699 gatherLoadUTranspose<3>(v, hw2, &v_hw2[XX], &v_hw2[YY], &v_hw2[ZZ]);
700 gatherLoadUTranspose<3>(v, hw3, &v_hw3[XX], &v_hw3[YY], &v_hw3[ZZ]);
702 /* Add the position correction divided by dt to the velocity */
703 for (int d = 0; d < DIM; d++)
705 v_ow1[d] = gmx::fma(da[d], invdt, v_ow1[d]);
707 for (int d = 0; d < DIM; d++)
709 v_hw2[d] = gmx::fma(db[d], invdt, v_hw2[d]);
711 for (int d = 0; d < DIM; d++)
713 v_hw3[d] = gmx::fma(dc[d], invdt, v_hw3[d]);
717 transposeScatterStoreU<3>(v, ow1, v_ow1[XX], v_ow1[YY], v_ow1[ZZ]);
718 transposeScatterStoreU<3>(v, hw2, v_hw2[XX], v_hw2[YY], v_hw2[ZZ]);
719 transposeScatterStoreU<3>(v, hw3, v_hw3[XX], v_hw3[YY], v_hw3[ZZ]);
724 /* Filter out the non-local settles */
725 T filter = load<T>(settled->virfac + i);
729 T mdo[DIM], mdb[DIM], mdc[DIM];
731 for (int d = 0; d < DIM; d++)
733 mdb[d] = mHf * db[d];
734 mdc[d] = mHf * dc[d];
735 mdo[d] = mOf * da[d] + mdb[d] + mdc[d];
738 for (int d2 = 0; d2 < DIM; d2++)
740 for (int d = 0; d < DIM; d++)
744 - (x_ow1[d2] * mdo[d] + dist21[d2] * mdb[d] + dist31[d2] * mdc[d]);
754 for (int d2 = 0; d2 < DIM; d2++)
756 for (int d = 0; d < DIM; d++)
758 vir_r_m_dr[d2][d] += reduce(sum_r_m_dr[d2][d]);
763 *bErrorHasOccurred = anyTrue(bError);
766 /*! \brief Wrapper template function that divides the settles over threads
767 * and instantiates the core template with instantiated booleans.
769 template<typename T, typename TypeBool, int packSize, typename TypePbc>
770 static void settleTemplateWrapper(settledata* settled,
780 bool* bErrorHasOccurred)
782 /* We need to assign settles to threads in groups of pack_size */
783 int numSettlePacks = (settled->nsettle + packSize - 1) / packSize;
784 /* Round the end value up to give thread 0 more work */
785 int settleStart = ((numSettlePacks * thread + nthread - 1) / nthread) * packSize;
786 int settleEnd = ((numSettlePacks * (thread + 1) + nthread - 1) / nthread) * packSize;
792 settleTemplate<T, TypeBool, packSize, TypePbc, true, false>(
793 settled, settleStart, settleEnd, pbc, x, xprime, invdt, v, nullptr, bErrorHasOccurred);
797 settleTemplate<T, TypeBool, packSize, TypePbc, true, true>(
798 settled, settleStart, settleEnd, pbc, x, xprime, invdt, v, vir_r_m_dr, bErrorHasOccurred);
805 settleTemplate<T, TypeBool, packSize, TypePbc, false, false>(
806 settled, settleStart, settleEnd, pbc, x, xprime, invdt, v, nullptr, bErrorHasOccurred);
810 settleTemplate<T, TypeBool, packSize, TypePbc, false, true>(
811 settled, settleStart, settleEnd, pbc, x, xprime, invdt, v, vir_r_m_dr, bErrorHasOccurred);
816 void csettle(settledata* settled,
820 ArrayRefWithPadding<const RVec> x,
821 ArrayRefWithPadding<RVec> xprime,
823 ArrayRefWithPadding<RVec> v,
826 bool* bErrorHasOccurred)
828 const real* xPtr = as_rvec_array(x.paddedArrayRef().data())[0];
829 real* xprimePtr = as_rvec_array(xprime.paddedArrayRef().data())[0];
830 real* vPtr = as_rvec_array(v.paddedArrayRef().data())[0];
832 #if GMX_SIMD_HAVE_REAL
833 if (settled->bUseSimd)
835 /* Convert the pbc struct for SIMD */
836 alignas(GMX_SIMD_ALIGNMENT) real pbcSimd[9 * GMX_SIMD_REAL_WIDTH];
837 set_pbc_simd(pbc, pbcSimd);
839 settleTemplateWrapper<SimdReal, SimdBool, GMX_SIMD_REAL_WIDTH, const real*>(
840 settled, nthread, thread, pbcSimd, xPtr, xprimePtr, invdt, vPtr, bCalcVirial,
841 vir_r_m_dr, bErrorHasOccurred);
846 /* This construct is needed because pbc_dx_aiuc doesn't accept pbc=NULL */
848 const t_pbc* pbcNonNull;
856 set_pbc(&pbcNo, PbcType::No, nullptr);
860 settleTemplateWrapper<real, bool, 1, const t_pbc*>(settled, nthread, thread, pbcNonNull,
861 &xPtr[0], &xprimePtr[0], invdt, &vPtr[0],
862 bCalcVirial, vir_r_m_dr, bErrorHasOccurred);