src/gromacs/mdlib/update.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   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) 2018,2019,2020,2021, 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.
  11  *
  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.
  16  *
  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.
  21  *
  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.
  26  *
  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.
  34  *
  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.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "update.h"
  41
  42 #include <cmath>
  43 #include <cstdio>
  44
  45 #include <algorithm>
  46 #include <memory>
  47
  48 #include "gromacs/domdec/domdec_struct.h"
  49 #include "gromacs/fileio/confio.h"
  50 #include "gromacs/gmxlib/network.h"
  51 #include "gromacs/gmxlib/nrnb.h"
  52 #include "gromacs/listed_forces/disre.h"
  53 #include "gromacs/listed_forces/orires.h"
  54 #include "gromacs/math/functions.h"
  55 #include "gromacs/math/invertmatrix.h"
  56 #include "gromacs/math/paddedvector.h"
  57 #include "gromacs/math/units.h"
  58 #include "gromacs/math/vec.h"
  59 #include "gromacs/math/vecdump.h"
  60 #include "gromacs/mdlib/boxdeformation.h"
  61 #include "gromacs/mdlib/constr.h"
  62 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  63 #include "gromacs/mdlib/mdatoms.h"
  64 #include "gromacs/mdlib/stat.h"
  65 #include "gromacs/mdlib/tgroup.h"
  66 #include "gromacs/mdtypes/commrec.h"
  67 #include "gromacs/mdtypes/fcdata.h"
  68 #include "gromacs/mdtypes/group.h"
  69 #include "gromacs/mdtypes/inputrec.h"
  70 #include "gromacs/mdtypes/md_enums.h"
  71 #include "gromacs/mdtypes/mdatom.h"
  72 #include "gromacs/mdtypes/state.h"
  73 #include "gromacs/pbcutil/boxutilities.h"
  74 #include "gromacs/pbcutil/pbc.h"
  75 #include "gromacs/pulling/pull.h"
  76 #include "gromacs/random/tabulatednormaldistribution.h"
  77 #include "gromacs/random/threefry.h"
  78 #include "gromacs/simd/simd.h"
  79 #include "gromacs/timing/wallcycle.h"
  80 #include "gromacs/topology/atoms.h"
  81 #include "gromacs/utility/exceptions.h"
  82 #include "gromacs/utility/fatalerror.h"
  83 #include "gromacs/utility/futil.h"
  84 #include "gromacs/utility/gmxassert.h"
  85 #include "gromacs/utility/gmxomp.h"
  86 #include "gromacs/utility/smalloc.h"
  87
  88 using namespace gmx; // TODO: Remove when this file is moved into gmx namespace
  89
  90 struct gmx_sd_const_t
  91 {
  92     double em = 0;
  93 };
  94
  95 struct gmx_sd_sigma_t
  96 {
  97     real V = 0;
  98 };
  99
 100 struct gmx_stochd_t
 101 {
 102     /* BD stuff */
 103     std::vector<real> bd_rf;
 104     /* SD stuff */
 105     std::vector<gmx_sd_const_t> sdc;
 106     std::vector<gmx_sd_sigma_t> sdsig;
 107     /* andersen temperature control stuff */
 108     std::vector<bool> randomize_group;
 109     std::vector<real> boltzfac;
 110
 111     explicit gmx_stochd_t(const t_inputrec& inputRecord);
 112 };
 113
 114 //! pImpled implementation for Update
 115 class Update::Impl
 116 {
 117 public:
 118     //! Constructor
 119     Impl(const t_inputrec& inputRecord, BoxDeformation* boxDeformation);
 120     //! Destructor
 121     ~Impl() = default;
 122
 123     void update_coords(const t_inputrec&                                inputRecord,
 124                        int64_t                                          step,
 125                        const t_mdatoms*                                 md,
 126                        t_state*                                         state,
 127                        const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
 128                        const t_fcdata&                                  fcdata,
 129                        const gmx_ekindata_t*                            ekind,
 130                        const matrix                                     M,
 131                        int                                              UpdatePart,
 132                        const t_commrec*                                 cr,
 133                        bool                                             haveConstraints);
 134
 135     void finish_update(const t_inputrec& inputRecord,
 136                        const t_mdatoms*  md,
 137                        t_state*          state,
 138                        gmx_wallcycle_t   wcycle,
 139                        bool              haveConstraints);
 140
 141     void update_sd_second_half(const t_inputrec& inputRecord,
 142                                int64_t           step,
 143                                real*             dvdlambda,
 144                                const t_mdatoms*  md,
 145                                t_state*          state,
 146                                const t_commrec*  cr,
 147                                t_nrnb*           nrnb,
 148                                gmx_wallcycle_t   wcycle,
 149                                gmx::Constraints* constr,
 150                                bool              do_log,
 151                                bool              do_ene);
 152
 153     void update_for_constraint_virial(const t_inputrec&                                inputRecord,
 154                                       const t_mdatoms&                                 md,
 155                                       const t_state&                                   state,
 156                                       const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
 157                                       const gmx_ekindata_t&                            ekind);
 158
 159     void update_temperature_constants(const t_inputrec& inputRecord);
 160
 161     const std::vector<bool>& getAndersenRandomizeGroup() const { return sd_.randomize_group; }
 162
 163     const std::vector<real>& getBoltzmanFactor() const { return sd_.boltzfac; }
 164
 165     PaddedVector<RVec>* xp() { return &xp_; }
 166
 167     BoxDeformation* deform() const { return deform_; }
 168
 169 private:
 170     //! stochastic dynamics struct
 171     gmx_stochd_t sd_;
 172     //! xprime for constraint algorithms
 173     PaddedVector<RVec> xp_;
 174     //! Box deformation handler (or nullptr if inactive).
 175     BoxDeformation* deform_ = nullptr;
 176 };
 177
 178 Update::Update(const t_inputrec& inputRecord, BoxDeformation* boxDeformation) :
 179     impl_(new Impl(inputRecord, boxDeformation)){};
 180
 181 Update::~Update(){};
 182
 183 const std::vector<bool>& Update::getAndersenRandomizeGroup() const
 184 {
 185     return impl_->getAndersenRandomizeGroup();
 186 }
 187
 188 const std::vector<real>& Update::getBoltzmanFactor() const
 189 {
 190     return impl_->getBoltzmanFactor();
 191 }
 192
 193 PaddedVector<RVec>* Update::xp()
 194 {
 195     return impl_->xp();
 196 }
 197
 198 BoxDeformation* Update::deform() const
 199 {
 200     return impl_->deform();
 201 }
 202
 203 void Update::update_coords(const t_inputrec&                                inputRecord,
 204                            int64_t                                          step,
 205                            const t_mdatoms*                                 md,
 206                            t_state*                                         state,
 207                            const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
 208                            const t_fcdata&                                  fcdata,
 209                            const gmx_ekindata_t*                            ekind,
 210                            const matrix                                     M,
 211                            int                                              updatePart,
 212                            const t_commrec*                                 cr,
 213                            const bool                                       haveConstraints)
 214 {
 215     return impl_->update_coords(
 216             inputRecord, step, md, state, f, fcdata, ekind, M, updatePart, cr, haveConstraints);
 217 }
 218
 219 void Update::finish_update(const t_inputrec& inputRecord,
 220                            const t_mdatoms*  md,
 221                            t_state*          state,
 222                            gmx_wallcycle_t   wcycle,
 223                            const bool        haveConstraints)
 224 {
 225     return impl_->finish_update(inputRecord, md, state, wcycle, haveConstraints);
 226 }
 227
 228 void Update::update_sd_second_half(const t_inputrec& inputRecord,
 229                                    int64_t           step,
 230                                    real*             dvdlambda,
 231                                    const t_mdatoms*  md,
 232                                    t_state*          state,
 233                                    const t_commrec*  cr,
 234                                    t_nrnb*           nrnb,
 235                                    gmx_wallcycle_t   wcycle,
 236                                    gmx::Constraints* constr,
 237                                    bool              do_log,
 238                                    bool              do_ene)
 239 {
 240     return impl_->update_sd_second_half(
 241             inputRecord, step, dvdlambda, md, state, cr, nrnb, wcycle, constr, do_log, do_ene);
 242 }
 243
 244 void Update::update_for_constraint_virial(const t_inputrec& inputRecord,
 245                                           const t_mdatoms&  md,
 246                                           const t_state&    state,
 247                                           const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
 248                                           const gmx_ekindata_t&                            ekind)
 249 {
 250     return impl_->update_for_constraint_virial(inputRecord, md, state, f, ekind);
 251 }
 252
 253 void Update::update_temperature_constants(const t_inputrec& inputRecord)
 254 {
 255     return impl_->update_temperature_constants(inputRecord);
 256 }
 257
 258 /*! \brief Sets the velocities of virtual sites to zero */
 259 static void clearVsiteVelocities(int start, int nrend, const unsigned short* particleType, rvec* gmx_restrict v)
 260 {
 261     for (int a = start; a < nrend; a++)
 262     {
 263         if (particleType[a] == eptVSite)
 264         {
 265             clear_rvec(v[a]);
 266         }
 267     }
 268 }
 269
 270 /*! \brief Sets whether we store the updated velocities */
 271 enum class StoreUpdatedVelocities
 272 {
 273     yes, //!< Store the updated velocities
 274     no   //!< Do not store the updated velocities
 275 };
 276
 277 /*! \brief Sets the number of different temperature coupling values */
 278 enum class NumTempScaleValues
 279 {
 280     single,  //!< Single T-scaling value (either one group or all values =1)
 281     multiple //!< Multiple T-scaling values, need to use T-group indices
 282 };
 283
 284 /*! \brief Sets if to apply no or diagonal Parrinello-Rahman pressure scaling
 285  *
 286  * Note that this enum is only used in updateMDLeapfrogSimple(), which does
 287  * not handle fully anistropic Parrinello-Rahman scaling, so we only have
 288  * options \p no and \p diagonal here and no anistropic option.
 289  */
 290 enum class ApplyParrinelloRahmanVScaling
 291 {
 292     no,      //!< Do not apply velocity scaling (not a PR-coupling run or step)
 293     diagonal //!< Apply velocity scaling using a diagonal matrix
 294 };
 295
 296 /*! \brief Integrate using leap-frog with T-scaling and optionally diagonal Parrinello-Rahman p-coupling
 297  *
 298  * \tparam       storeUpdatedVelocities Tells whether we should store the updated velocities
 299  * \tparam       numTempScaleValues     The number of different T-couple values
 300  * \tparam       applyPRVScaling        Apply Parrinello-Rahman velocity scaling
 301  * \param[in]    start                  Index of first atom to update
 302  * \param[in]    nrend                  Last atom to update: \p nrend - 1
 303  * \param[in]    dt                     The time step
 304  * \param[in]    dtPressureCouple       Time step for pressure coupling
 305  * \param[in]    invMassPerDim          1/mass per atom and dimension
 306  * \param[in]    tcstat                 Temperature coupling information
 307  * \param[in]    cTC                    T-coupling group index per atom
 308  * \param[in]    pRVScaleMatrixDiagonal Parrinello-Rahman v-scale matrix diagonal
 309  * \param[in]    x                      Input coordinates
 310  * \param[out]   xprime                 Updated coordinates
 311  * \param[inout] v                      Velocities
 312  * \param[in]    f                      Forces
 313  *
 314  * We expect this template to get good SIMD acceleration by most compilers,
 315  * unlike the more complex general template.
 316  * Note that we might get even better SIMD acceleration when we introduce
 317  * aligned (and padded) memory, possibly with some hints for the compilers.
 318  */
 319 template<StoreUpdatedVelocities storeUpdatedVelocities, NumTempScaleValues numTempScaleValues, ApplyParrinelloRahmanVScaling applyPRVScaling>
 320 static void updateMDLeapfrogSimple(int         start,
 321                                    int         nrend,
 322                                    real        dt,
 323                                    real        dtPressureCouple,
 324                                    const rvec* gmx_restrict          invMassPerDim,
 325                                    gmx::ArrayRef<const t_grp_tcstat> tcstat,
 326                                    const unsigned short*             cTC,
 327                                    const rvec                        pRVScaleMatrixDiagonal,
 328                                    const rvec* gmx_restrict x,
 329                                    rvec* gmx_restrict xprime,
 330                                    rvec* gmx_restrict v,
 331                                    const rvec* gmx_restrict f)
 332 {
 333     real lambdaGroup;
 334
 335     if (numTempScaleValues == NumTempScaleValues::single)
 336     {
 337         lambdaGroup = tcstat[0].lambda;
 338     }
 339
 340     for (int a = start; a < nrend; a++)
 341     {
 342         if (numTempScaleValues == NumTempScaleValues::multiple)
 343         {
 344             lambdaGroup = tcstat[cTC[a]].lambda;
 345         }
 346
 347         for (int d = 0; d < DIM; d++)
 348         {
 349             /* Note that using rvec invMassPerDim results in more efficient
 350              * SIMD code, but this increases the cache pressure.
 351              * For large systems with PME on the CPU this slows down the
 352              * (then already slow) update by 20%. If all data remains in cache,
 353              * using rvec is much faster.
 354              */
 355             real vNew = lambdaGroup * v[a][d] + f[a][d] * invMassPerDim[a][d] * dt;
 356
 357             if (applyPRVScaling == ApplyParrinelloRahmanVScaling::diagonal)
 358             {
 359                 vNew -= dtPressureCouple * pRVScaleMatrixDiagonal[d] * v[a][d];
 360             }
 361             if (storeUpdatedVelocities == StoreUpdatedVelocities::yes)
 362             {
 363                 v[a][d] = vNew;
 364             }
 365             xprime[a][d] = x[a][d] + vNew * dt;
 366         }
 367     }
 368 }
 369
 370 #if GMX_SIMD && GMX_SIMD_HAVE_REAL
 371 #    define GMX_HAVE_SIMD_UPDATE 1
 372 #else
 373 #    define GMX_HAVE_SIMD_UPDATE 0
 374 #endif
 375
 376 #if GMX_HAVE_SIMD_UPDATE
 377
 378 /*! \brief Load (aligned) the contents of GMX_SIMD_REAL_WIDTH rvec elements sequentially into 3 SIMD registers
 379  *
 380  * The loaded output is:
 381  * \p r0: { r[index][XX], r[index][YY], ... }
 382  * \p r1: { ... }
 383  * \p r2: { ..., r[index+GMX_SIMD_REAL_WIDTH-1][YY], r[index+GMX_SIMD_REAL_WIDTH-1][ZZ] }
 384  *
 385  * \param[in]  r      Real to an rvec array, has to be aligned to SIMD register width
 386  * \param[in]  index  Index of the first rvec triplet of reals to load
 387  * \param[out] r0     Pointer to first SIMD register
 388  * \param[out] r1     Pointer to second SIMD register
 389  * \param[out] r2     Pointer to third SIMD register
 390  */
 391 static inline void simdLoadRvecs(const rvec* r, int index, SimdReal* r0, SimdReal* r1, SimdReal* r2)
 392 {
 393     const real* realPtr = r[index];
 394
 395     GMX_ASSERT(isSimdAligned(realPtr), "Pointer should be SIMD aligned");
 396
 397     *r0 = simdLoad(realPtr + 0 * GMX_SIMD_REAL_WIDTH);
 398     *r1 = simdLoad(realPtr + 1 * GMX_SIMD_REAL_WIDTH);
 399     *r2 = simdLoad(realPtr + 2 * GMX_SIMD_REAL_WIDTH);
 400 }
 401
 402 /*! \brief Store (aligned) 3 SIMD registers sequentially to GMX_SIMD_REAL_WIDTH rvec elements
 403  *
 404  * The stored output is:
 405  * \p r[index] = { { r0[0], r0[1], ... }
 406  * ...
 407  * \p r[index+GMX_SIMD_REAL_WIDTH-1] =  { ... , r2[GMX_SIMD_REAL_WIDTH-2], r2[GMX_SIMD_REAL_WIDTH-1] }
 408  *
 409  * \param[out] r      Pointer to an rvec array, has to be aligned to SIMD register width
 410  * \param[in]  index  Index of the first rvec triplet of reals to store to
 411  * \param[in]  r0     First SIMD register
 412  * \param[in]  r1     Second SIMD register
 413  * \param[in]  r2     Third SIMD register
 414  */
 415 static inline void simdStoreRvecs(rvec* r, int index, SimdReal r0, SimdReal r1, SimdReal r2)
 416 {
 417     real* realPtr = r[index];
 418
 419     GMX_ASSERT(isSimdAligned(realPtr), "Pointer should be SIMD aligned");
 420
 421     store(realPtr + 0 * GMX_SIMD_REAL_WIDTH, r0);
 422     store(realPtr + 1 * GMX_SIMD_REAL_WIDTH, r1);
 423     store(realPtr + 2 * GMX_SIMD_REAL_WIDTH, r2);
 424 }
 425
 426 /*! \brief Integrate using leap-frog with single group T-scaling and SIMD
 427  *
 428  * \tparam       storeUpdatedVelocities Tells whether we should store the updated velocities
 429  * \param[in]    start                  Index of first atom to update
 430  * \param[in]    nrend                  Last atom to update: \p nrend - 1
 431  * \param[in]    dt                     The time step
 432  * \param[in]    invMass                1/mass per atom
 433  * \param[in]    tcstat                 Temperature coupling information
 434  * \param[in]    x                      Input coordinates
 435  * \param[out]   xprime                 Updated coordinates
 436  * \param[inout] v                      Velocities
 437  * \param[in]    f                      Forces
 438  */
 439 template<StoreUpdatedVelocities storeUpdatedVelocities>
 440 static void updateMDLeapfrogSimpleSimd(int         start,
 441                                        int         nrend,
 442                                        real        dt,
 443                                        const real* gmx_restrict          invMass,
 444                                        gmx::ArrayRef<const t_grp_tcstat> tcstat,
 445                                        const rvec* gmx_restrict x,
 446                                        rvec* gmx_restrict xprime,
 447                                        rvec* gmx_restrict v,
 448                                        const rvec* gmx_restrict f)
 449 {
 450     SimdReal timestep(dt);
 451     SimdReal lambdaSystem(tcstat[0].lambda);
 452
 453     /* We declare variables here, since code is often slower when declaring them inside the loop */
 454
 455     /* Note: We should implement a proper PaddedVector, so we don't need this check */
 456     GMX_ASSERT(isSimdAligned(invMass), "invMass should be aligned");
 457
 458     for (int a = start; a < nrend; a += GMX_SIMD_REAL_WIDTH)
 459     {
 460         SimdReal invMass0, invMass1, invMass2;
 461         expandScalarsToTriplets(simdLoad(invMass + a), &invMass0, &invMass1, &invMass2);
 462
 463         SimdReal v0, v1, v2;
 464         SimdReal f0, f1, f2;
 465         simdLoadRvecs(v, a, &v0, &v1, &v2);
 466         simdLoadRvecs(f, a, &f0, &f1, &f2);
 467
 468         v0 = fma(f0 * invMass0, timestep, lambdaSystem * v0);
 469         v1 = fma(f1 * invMass1, timestep, lambdaSystem * v1);
 470         v2 = fma(f2 * invMass2, timestep, lambdaSystem * v2);
 471
 472         if (storeUpdatedVelocities == StoreUpdatedVelocities::yes)
 473         {
 474             simdStoreRvecs(v, a, v0, v1, v2);
 475         }
 476
 477         SimdReal x0, x1, x2;
 478         simdLoadRvecs(x, a, &x0, &x1, &x2);
 479
 480         SimdReal xprime0 = fma(v0, timestep, x0);
 481         SimdReal xprime1 = fma(v1, timestep, x1);
 482         SimdReal xprime2 = fma(v2, timestep, x2);
 483
 484         simdStoreRvecs(xprime, a, xprime0, xprime1, xprime2);
 485     }
 486 }
 487
 488 #endif // GMX_HAVE_SIMD_UPDATE
 489
 490 /*! \brief Sets the NEMD acceleration type */
 491 enum class AccelerationType
 492 {
 493     none,
 494     cosine
 495 };
 496
 497 /*! \brief Integrate using leap-frog with support for everything.
 498  *
 499  * \tparam        accelerationType  Type of NEMD acceleration.
 500  * \param[in]     start             Index of first atom to update.
 501  * \param[in]     nrend             Last atom to update: \p nrend - 1.
 502  * \param[in]     doNoseHoover      If to apply Nose-Hoover T-coupling.
 503  * \param[in]     dt                The time step.
 504  * \param[in]     dtPressureCouple  Time step for pressure coupling, is 0 when no pressure
 505  *                                  coupling should be applied at this step.
 506  * \param[in]     md                Atom properties.
 507  * \param[in]     ekind             Kinetic energy data.
 508  * \param[in]     box               The box dimensions.
 509  * \param[in]     x                 Input coordinates.
 510  * \param[out]    xprime            Updated coordinates.
 511  * \param[inout]  v                 Velocities.
 512  * \param[in]     f                 Forces.
 513  * \param[in]     nh_vxi            Nose-Hoover velocity scaling factors.
 514  * \param[in]     nsttcouple        Frequency of the temperature coupling steps.
 515  * \param[in]     M                 Parrinello-Rahman scaling matrix.
 516  */
 517 template<AccelerationType accelerationType>
 518 static void updateMDLeapfrogGeneral(int                   start,
 519                                     int                   nrend,
 520                                     bool                  doNoseHoover,
 521                                     real                  dt,
 522                                     real                  dtPressureCouple,
 523                                     const t_mdatoms*      md,
 524                                     const gmx_ekindata_t* ekind,
 525                                     const matrix          box,
 526                                     const rvec* gmx_restrict x,
 527                                     rvec* gmx_restrict xprime,
 528                                     rvec* gmx_restrict v,
 529                                     const rvec* gmx_restrict f,
 530                                     const double* gmx_restrict nh_vxi,
 531                                     const int                  nsttcouple,
 532                                     const matrix               M)
 533 {
 534     /* This is a version of the leap-frog integrator that supports
 535      * all combinations of T-coupling, P-coupling and NEMD.
 536      * Nose-Hoover uses the reversible leap-frog integrator from
 537      * Holian et al. Phys Rev E 52(3) : 2338, 1995
 538      */
 539
 540     gmx::ArrayRef<const t_grp_tcstat> tcstat = ekind->tcstat;
 541     const unsigned short*             cTC    = md->cTC;
 542
 543     const rvec* gmx_restrict invMassPerDim = md->invMassPerDim;
 544
 545     /* Initialize group values, changed later when multiple groups are used */
 546     int gt = 0;
 547
 548     real omega_Z = 2 * static_cast<real>(M_PI) / box[ZZ][ZZ];
 549
 550     for (int n = start; n < nrend; n++)
 551     {
 552         if (cTC)
 553         {
 554             gt = cTC[n];
 555         }
 556         real lg = tcstat[gt].lambda;
 557
 558         rvec vRel;
 559         real cosineZ, vCosine;
 560 #ifdef __INTEL_COMPILER
 561 #    pragma warning(disable : 280)
 562 #endif
 563         switch (accelerationType)
 564         {
 565             case AccelerationType::none: copy_rvec(v[n], vRel); break;
 566             case AccelerationType::cosine:
 567                 cosineZ = std::cos(x[n][ZZ] * omega_Z);
 568                 vCosine = cosineZ * ekind->cosacc.vcos;
 569                 /* Avoid scaling the cosine profile velocity */
 570                 copy_rvec(v[n], vRel);
 571                 vRel[XX] -= vCosine;
 572                 break;
 573         }
 574
 575         real factorNH = 0.0;
 576         if (doNoseHoover)
 577         {
 578             /* Here we account for multiple time stepping, by increasing
 579              * the Nose-Hoover correction by nsttcouple
 580              * TODO: This can be pre-computed.
 581              */
 582             factorNH = 0.5 * nsttcouple * dt * nh_vxi[gt];
 583         }
 584
 585         for (int d = 0; d < DIM; d++)
 586         {
 587             real vNew = (lg * vRel[d]
 588                          + (f[n][d] * invMassPerDim[n][d] * dt - factorNH * vRel[d]
 589                             - dtPressureCouple * iprod(M[d], vRel)))
 590                         / (1 + factorNH);
 591             switch (accelerationType)
 592             {
 593                 case AccelerationType::none: break;
 594                 case AccelerationType::cosine:
 595                     if (d == XX)
 596                     {
 597                         /* Add back the mean velocity and apply acceleration */
 598                         vNew += vCosine + cosineZ * ekind->cosacc.cos_accel * dt;
 599                     }
 600                     break;
 601             }
 602             v[n][d]      = vNew;
 603             xprime[n][d] = x[n][d] + vNew * dt;
 604         }
 605     }
 606 }
 607
 608 /*! \brief Handles the Leap-frog MD x and v integration */
 609 static void do_update_md(int         start,
 610                          int         nrend,
 611                          real        dt,
 612                          int64_t     step,
 613                          const rvec* gmx_restrict x,
 614                          rvec* gmx_restrict xprime,
 615                          rvec* gmx_restrict v,
 616                          const rvec* gmx_restrict f,
 617                          const int                etc,
 618                          const int                epc,
 619                          const int                nsttcouple,
 620                          const int                nstpcouple,
 621                          const t_mdatoms*         md,
 622                          const gmx_ekindata_t*    ekind,
 623                          const matrix             box,
 624                          const double* gmx_restrict nh_vxi,
 625                          const matrix               M)
 626 {
 627     GMX_ASSERT(nrend == start || xprime != x,
 628                "For SIMD optimization certain compilers need to have xprime != x");
 629
 630     /* Note: Berendsen pressure scaling is handled after do_update_md() */
 631     bool doTempCouple = (etc != etcNO && do_per_step(step + nsttcouple - 1, nsttcouple));
 632     bool doNoseHoover = (etc == etcNOSEHOOVER && doTempCouple);
 633     bool doParrinelloRahman =
 634             (epc == epcPARRINELLORAHMAN && do_per_step(step + nstpcouple - 1, nstpcouple));
 635     bool doPROffDiagonal = (doParrinelloRahman && (M[YY][XX] != 0 || M[ZZ][XX] != 0 || M[ZZ][YY] != 0));
 636
 637     real dtPressureCouple = (doParrinelloRahman ? nstpcouple * dt : 0);
 638
 639     /* NEMD cosine acceleration is applied in updateMDLeapFrogGeneral */
 640     bool doAcceleration = (ekind->cosacc.cos_accel != 0);
 641
 642     if (doNoseHoover || doPROffDiagonal || doAcceleration)
 643     {
 644         matrix stepM;
 645         if (!doParrinelloRahman)
 646         {
 647             /* We should not apply PR scaling at this step */
 648             clear_mat(stepM);
 649         }
 650         else
 651         {
 652             copy_mat(M, stepM);
 653         }
 654
 655         if (!doAcceleration)
 656         {
 657             updateMDLeapfrogGeneral<AccelerationType::none>(
 658                     start, nrend, doNoseHoover, dt, dtPressureCouple, md, ekind, box, x, xprime, v, f, nh_vxi, nsttcouple, stepM);
 659         }
 660         else
 661         {
 662             updateMDLeapfrogGeneral<AccelerationType::cosine>(
 663                     start, nrend, doNoseHoover, dt, dtPressureCouple, md, ekind, box, x, xprime, v, f, nh_vxi, nsttcouple, stepM);
 664         }
 665     }
 666     else
 667     {
 668         /* Use a simple and thus more efficient integration loop. */
 669         /* The simple loop does not check for particle type (so it can
 670          * be vectorized), which means we need to clear the velocities
 671          * of virtual sites in advance, when present. Note that vsite
 672          * velocities are computed after update and constraints from
 673          * their displacement.
 674          */
 675         if (md->haveVsites)
 676         {
 677             /* Note: The overhead of this loop is completely neligible */
 678             clearVsiteVelocities(start, nrend, md->ptype, v);
 679         }
 680
 681         /* We determine if we have a single T-coupling lambda value for all
 682          * atoms. That allows for better SIMD acceleration in the template.
 683          * If we do not do temperature coupling (in the run or this step),
 684          * all scaling values are 1, so we effectively have a single value.
 685          */
 686         bool haveSingleTempScaleValue = (!doTempCouple || ekind->ngtc == 1);
 687
 688         /* Extract some pointers needed by all cases */
 689         const unsigned short*             cTC           = md->cTC;
 690         gmx::ArrayRef<const t_grp_tcstat> tcstat        = ekind->tcstat;
 691         const rvec*                       invMassPerDim = md->invMassPerDim;
 692
 693         if (doParrinelloRahman)
 694         {
 695             GMX_ASSERT(!doPROffDiagonal,
 696                        "updateMDLeapfrogSimple only support diagonal Parrinello-Rahman scaling "
 697                        "matrices");
 698
 699             rvec diagM;
 700             for (int d = 0; d < DIM; d++)
 701             {
 702                 diagM[d] = M[d][d];
 703             }
 704
 705             if (haveSingleTempScaleValue)
 706             {
 707                 updateMDLeapfrogSimple<StoreUpdatedVelocities::yes, NumTempScaleValues::single, ApplyParrinelloRahmanVScaling::diagonal>(
 708                         start, nrend, dt, dtPressureCouple, invMassPerDim, tcstat, cTC, diagM, x, xprime, v, f);
 709             }
 710             else
 711             {
 712                 updateMDLeapfrogSimple<StoreUpdatedVelocities::yes, NumTempScaleValues::multiple, ApplyParrinelloRahmanVScaling::diagonal>(
 713                         start, nrend, dt, dtPressureCouple, invMassPerDim, tcstat, cTC, diagM, x, xprime, v, f);
 714             }
 715         }
 716         else
 717         {
 718             if (haveSingleTempScaleValue)
 719             {
 720                 /* Note that modern compilers are pretty good at vectorizing
 721                  * updateMDLeapfrogSimple(). But the SIMD version will still
 722                  * be faster because invMass lowers the cache pressure
 723                  * compared to invMassPerDim.
 724                  */
 725 #if GMX_HAVE_SIMD_UPDATE
 726                 /* Check if we can use invmass instead of invMassPerDim */
 727                 if (!md->havePartiallyFrozenAtoms)
 728                 {
 729                     updateMDLeapfrogSimpleSimd<StoreUpdatedVelocities::yes>(
 730                             start, nrend, dt, md->invmass, tcstat, x, xprime, v, f);
 731                 }
 732                 else
 733 #endif
 734                 {
 735                     updateMDLeapfrogSimple<StoreUpdatedVelocities::yes, NumTempScaleValues::single, ApplyParrinelloRahmanVScaling::no>(
 736                             start, nrend, dt, dtPressureCouple, invMassPerDim, tcstat, cTC, nullptr, x, xprime, v, f);
 737                 }
 738             }
 739             else
 740             {
 741                 updateMDLeapfrogSimple<StoreUpdatedVelocities::yes, NumTempScaleValues::multiple, ApplyParrinelloRahmanVScaling::no>(
 742                         start, nrend, dt, dtPressureCouple, invMassPerDim, tcstat, cTC, nullptr, x, xprime, v, f);
 743             }
 744         }
 745     }
 746 }
 747 /*! \brief Handles the Leap-frog MD x and v integration */
 748 static void doUpdateMDDoNotUpdateVelocities(int         start,
 749                                             int         nrend,
 750                                             real        dt,
 751                                             const rvec* gmx_restrict x,
 752                                             rvec* gmx_restrict xprime,
 753                                             rvec* gmx_restrict v,
 754                                             const rvec* gmx_restrict f,
 755                                             const t_mdatoms&         md,
 756                                             const gmx_ekindata_t&    ekind)
 757 {
 758     GMX_ASSERT(nrend == start || xprime != x,
 759                "For SIMD optimization certain compilers need to have xprime != x");
 760
 761     gmx::ArrayRef<const t_grp_tcstat> tcstat = ekind.tcstat;
 762
 763     /* Check if we can use invmass instead of invMassPerDim */
 764 #if GMX_HAVE_SIMD_UPDATE
 765     if (!md.havePartiallyFrozenAtoms)
 766     {
 767         updateMDLeapfrogSimpleSimd<StoreUpdatedVelocities::no>(
 768                 start, nrend, dt, md.invmass, tcstat, x, xprime, v, f);
 769     }
 770     else
 771 #endif
 772     {
 773         updateMDLeapfrogSimple<StoreUpdatedVelocities::no, NumTempScaleValues::single, ApplyParrinelloRahmanVScaling::no>(
 774                 start, nrend, dt, dt, md.invMassPerDim, tcstat, nullptr, nullptr, x, xprime, v, f);
 775     }
 776 }
 777
 778 static void do_update_vv_vel(int                  start,
 779                              int                  nrend,
 780                              real                 dt,
 781                              const ivec           nFreeze[],
 782                              const real           invmass[],
 783                              const unsigned short ptype[],
 784                              const unsigned short cFREEZE[],
 785                              rvec                 v[],
 786                              const rvec           f[],
 787                              gmx_bool             bExtended,
 788                              real                 veta,
 789                              real                 alpha)
 790 {
 791     int  gf = 0;
 792     int  n, d;
 793     real g, mv1, mv2;
 794
 795     if (bExtended)
 796     {
 797         g   = 0.25 * dt * veta * alpha;
 798         mv1 = std::exp(-g);
 799         mv2 = gmx::series_sinhx(g);
 800     }
 801     else
 802     {
 803         mv1 = 1.0;
 804         mv2 = 1.0;
 805     }
 806     for (n = start; n < nrend; n++)
 807     {
 808         real w_dt = invmass[n] * dt;
 809         if (cFREEZE)
 810         {
 811             gf = cFREEZE[n];
 812         }
 813
 814         for (d = 0; d < DIM; d++)
 815         {
 816             if ((ptype[n] != eptVSite) && (ptype[n] != eptShell) && !nFreeze[gf][d])
 817             {
 818                 v[n][d] = mv1 * (mv1 * v[n][d] + 0.5 * (w_dt * mv2 * f[n][d]));
 819             }
 820             else
 821             {
 822                 v[n][d] = 0.0;
 823             }
 824         }
 825     }
 826 } /* do_update_vv_vel */
 827
 828 static void do_update_vv_pos(int                  start,
 829                              int                  nrend,
 830                              real                 dt,
 831                              const ivec           nFreeze[],
 832                              const unsigned short ptype[],
 833                              const unsigned short cFREEZE[],
 834                              const rvec           x[],
 835                              rvec                 xprime[],
 836                              const rvec           v[],
 837                              gmx_bool             bExtended,
 838                              real                 veta)
 839 {
 840     int  gf = 0;
 841     int  n, d;
 842     real g, mr1, mr2;
 843
 844     /* Would it make more sense if Parrinello-Rahman was put here? */
 845     if (bExtended)
 846     {
 847         g   = 0.5 * dt * veta;
 848         mr1 = std::exp(g);
 849         mr2 = gmx::series_sinhx(g);
 850     }
 851     else
 852     {
 853         mr1 = 1.0;
 854         mr2 = 1.0;
 855     }
 856
 857     for (n = start; n < nrend; n++)
 858     {
 859
 860         if (cFREEZE)
 861         {
 862             gf = cFREEZE[n];
 863         }
 864
 865         for (d = 0; d < DIM; d++)
 866         {
 867             if ((ptype[n] != eptVSite) && (ptype[n] != eptShell) && !nFreeze[gf][d])
 868             {
 869                 xprime[n][d] = mr1 * (mr1 * x[n][d] + mr2 * dt * v[n][d]);
 870             }
 871             else
 872             {
 873                 xprime[n][d] = x[n][d];
 874             }
 875         }
 876     }
 877 } /* do_update_vv_pos */
 878
 879 gmx_stochd_t::gmx_stochd_t(const t_inputrec& inputRecord)
 880 {
 881     const t_grpopts* opts = &inputRecord.opts;
 882     const int        ngtc = opts->ngtc;
 883
 884     if (inputRecord.eI == eiBD)
 885     {
 886         bd_rf.resize(ngtc);
 887     }
 888     else if (EI_SD(inputRecord.eI))
 889     {
 890         sdc.resize(ngtc);
 891         sdsig.resize(ngtc);
 892
 893         for (int gt = 0; gt < ngtc; gt++)
 894         {
 895             if (opts->tau_t[gt] > 0)
 896             {
 897                 sdc[gt].em = std::exp(-inputRecord.delta_t / opts->tau_t[gt]);
 898             }
 899             else
 900             {
 901                 /* No friction and noise on this group */
 902                 sdc[gt].em = 1;
 903             }
 904         }
 905     }
 906     else if (ETC_ANDERSEN(inputRecord.etc))
 907     {
 908         randomize_group.resize(ngtc);
 909         boltzfac.resize(ngtc);
 910
 911         /* for now, assume that all groups, if randomized, are randomized at the same rate, i.e. tau_t is the same. */
 912         /* since constraint groups don't necessarily match up with temperature groups! This is checked in readir.c */
 913
 914         for (int gt = 0; gt < ngtc; gt++)
 915         {
 916             real reft = std::max<real>(0, opts->ref_t[gt]);
 917             if ((opts->tau_t[gt] > 0)
 918                 && (reft > 0)) /* tau_t or ref_t = 0 means that no randomization is done */
 919             {
 920                 randomize_group[gt] = true;
 921                 boltzfac[gt]        = BOLTZ * opts->ref_t[gt];
 922             }
 923             else
 924             {
 925                 randomize_group[gt] = false;
 926             }
 927         }
 928     }
 929 }
 930
 931 void Update::Impl::update_temperature_constants(const t_inputrec& inputRecord)
 932 {
 933     if (inputRecord.eI == eiBD)
 934     {
 935         if (inputRecord.bd_fric != 0)
 936         {
 937             for (int gt = 0; gt < inputRecord.opts.ngtc; gt++)
 938             {
 939                 sd_.bd_rf[gt] = std::sqrt(2.0 * BOLTZ * inputRecord.opts.ref_t[gt]
 940                                           / (inputRecord.bd_fric * inputRecord.delta_t));
 941             }
 942         }
 943         else
 944         {
 945             for (int gt = 0; gt < inputRecord.opts.ngtc; gt++)
 946             {
 947                 sd_.bd_rf[gt] = std::sqrt(2.0 * BOLTZ * inputRecord.opts.ref_t[gt]);
 948             }
 949         }
 950     }
 951     if (inputRecord.eI == eiSD1)
 952     {
 953         for (int gt = 0; gt < inputRecord.opts.ngtc; gt++)
 954         {
 955             real kT = BOLTZ * inputRecord.opts.ref_t[gt];
 956             /* The mass is accounted for later, since this differs per atom */
 957             sd_.sdsig[gt].V = std::sqrt(kT * (1 - sd_.sdc[gt].em * sd_.sdc[gt].em));
 958         }
 959     }
 960 }
 961
 962 Update::Impl::Impl(const t_inputrec& inputRecord, BoxDeformation* boxDeformation) :
 963     sd_(inputRecord),
 964     deform_(boxDeformation)
 965 {
 966     update_temperature_constants(inputRecord);
 967     xp_.resizeWithPadding(0);
 968 }
 969
 970 void Update::setNumAtoms(int numAtoms)
 971 {
 972
 973     impl_->xp()->resizeWithPadding(numAtoms);
 974 }
 975
 976 /*! \brief Sets the SD update type */
 977 enum class SDUpdate : int
 978 {
 979     ForcesOnly,
 980     FrictionAndNoiseOnly,
 981     Combined
 982 };
 983
 984 /*! \brief SD integrator update
 985  *
 986  * Two phases are required in the general case of a constrained
 987  * update, the first phase from the contribution of forces, before
 988  * applying constraints, and then a second phase applying the friction
 989  * and noise, and then further constraining. For details, see
 990  * Goga2012.
 991  *
 992  * Without constraints, the two phases can be combined, for
 993  * efficiency.
 994  *
 995  * Thus three instantiations of this templated function will be made,
 996  * two with only one contribution, and one with both contributions. */
 997 template<SDUpdate updateType>
 998 static void doSDUpdateGeneral(const gmx_stochd_t&  sd,
 999                               int                  start,
1000                               int                  nrend,
1001                               real                 dt,
1002                               const ivec           nFreeze[],
1003                               const real           invmass[],
1004                               const unsigned short ptype[],
1005                               const unsigned short cFREEZE[],
1006                               const unsigned short cTC[],
1007                               const rvec           x[],
1008                               rvec                 xprime[],
1009                               rvec                 v[],
1010                               const rvec           f[],
1011                               int64_t              step,
1012                               int                  seed,
1013                               const int*           gatindex)
1014 {
1015     // cTC and cFREEZE can be nullptr any time, but various
1016     // instantiations do not make sense with particular pointer
1017     // values.
1018     if (updateType == SDUpdate::ForcesOnly)
1019     {
1020         GMX_ASSERT(f != nullptr, "SD update with only forces requires forces");
1021         GMX_ASSERT(cTC == nullptr, "SD update with only forces cannot handle temperature groups");
1022     }
1023     if (updateType == SDUpdate::FrictionAndNoiseOnly)
1024     {
1025         GMX_ASSERT(f == nullptr, "SD update with only noise cannot handle forces");
1026     }
1027     if (updateType == SDUpdate::Combined)
1028     {
1029         GMX_ASSERT(f != nullptr, "SD update with forces and noise requires forces");
1030     }
1031
1032     // Even 0 bits internal counter gives 2x64 ints (more than enough for three table lookups)
1033     gmx::ThreeFry2x64<0>                       rng(seed, gmx::RandomDomain::UpdateCoordinates);
1034     gmx::TabulatedNormalDistribution<real, 14> dist;
1035
1036     for (int n = start; n < nrend; n++)
1037     {
1038         int globalAtomIndex = gatindex ? gatindex[n] : n;
1039         rng.restart(step, globalAtomIndex);
1040         dist.reset();
1041
1042         real inverseMass = invmass[n];
1043         real invsqrtMass = std::sqrt(inverseMass);
1044
1045         int freezeGroup      = cFREEZE ? cFREEZE[n] : 0;
1046         int temperatureGroup = cTC ? cTC[n] : 0;
1047
1048         for (int d = 0; d < DIM; d++)
1049         {
1050             if ((ptype[n] != eptVSite) && (ptype[n] != eptShell) && !nFreeze[freezeGroup][d])
1051             {
1052                 if (updateType == SDUpdate::ForcesOnly)
1053                 {
1054                     real vn = v[n][d] + inverseMass * f[n][d] * dt;
1055                     v[n][d] = vn;
1056                     // Simple position update.
1057                     xprime[n][d] = x[n][d] + v[n][d] * dt;
1058                 }
1059                 else if (updateType == SDUpdate::FrictionAndNoiseOnly)
1060                 {
1061                     real vn = v[n][d];
1062                     v[n][d] = (vn * sd.sdc[temperatureGroup].em
1063                                + invsqrtMass * sd.sdsig[temperatureGroup].V * dist(rng));
1064                     // The previous phase already updated the
1065                     // positions with a full v*dt term that must
1066                     // now be half removed.
1067                     xprime[n][d] = xprime[n][d] + 0.5 * (v[n][d] - vn) * dt;
1068                 }
1069                 else
1070                 {
1071                     real vn = v[n][d] + inverseMass * f[n][d] * dt;
1072                     v[n][d] = (vn * sd.sdc[temperatureGroup].em
1073                                + invsqrtMass * sd.sdsig[temperatureGroup].V * dist(rng));
1074                     // Here we include half of the friction+noise
1075                     // update of v into the position update.
1076                     xprime[n][d] = x[n][d] + 0.5 * (vn + v[n][d]) * dt;
1077                 }
1078             }
1079             else
1080             {
1081                 // When using constraints, the update is split into
1082                 // two phases, but we only need to zero the update of
1083                 // virtual, shell or frozen particles in at most one
1084                 // of the phases.
1085                 if (updateType != SDUpdate::FrictionAndNoiseOnly)
1086                 {
1087                     v[n][d]      = 0.0;
1088                     xprime[n][d] = x[n][d];
1089                 }
1090             }
1091         }
1092     }
1093 }
1094
1095 static void do_update_sd(int         start,
1096                          int         nrend,
1097                          real        dt,
1098                          int64_t     step,
1099                          const rvec* gmx_restrict x,
1100                          rvec* gmx_restrict xprime,
1101                          rvec* gmx_restrict v,
1102                          const rvec* gmx_restrict f,
1103                          const ivec               nFreeze[],
1104                          const real               invmass[],
1105                          const unsigned short     ptype[],
1106                          const unsigned short     cFREEZE[],
1107                          const unsigned short     cTC[],
1108                          int                      seed,
1109                          const t_commrec*         cr,
1110                          const gmx_stochd_t&      sd,
1111                          bool                     haveConstraints)
1112 {
1113     if (haveConstraints)
1114     {
1115         // With constraints, the SD update is done in 2 parts
1116         doSDUpdateGeneral<SDUpdate::ForcesOnly>(
1117                 sd, start, nrend, dt, nFreeze, invmass, ptype, cFREEZE, nullptr, x, xprime, v, f, step, seed, nullptr);
1118     }
1119     else
1120     {
1121         doSDUpdateGeneral<SDUpdate::Combined>(sd,
1122                                               start,
1123                                               nrend,
1124                                               dt,
1125                                               nFreeze,
1126                                               invmass,
1127                                               ptype,
1128                                               cFREEZE,
1129                                               cTC,
1130                                               x,
1131                                               xprime,
1132                                               v,
1133                                               f,
1134                                               step,
1135                                               seed,
1136                                               DOMAINDECOMP(cr) ? cr->dd->globalAtomIndices.data() : nullptr);
1137     }
1138 }
1139
1140 static void do_update_bd(int         start,
1141                          int         nrend,
1142                          real        dt,
1143                          int64_t     step,
1144                          const rvec* gmx_restrict x,
1145                          rvec* gmx_restrict xprime,
1146                          rvec* gmx_restrict v,
1147                          const rvec* gmx_restrict f,
1148                          const ivec               nFreeze[],
1149                          const real               invmass[],
1150                          const unsigned short     ptype[],
1151                          const unsigned short     cFREEZE[],
1152                          const unsigned short     cTC[],
1153                          real                     friction_coefficient,
1154                          const real*              rf,
1155                          int                      seed,
1156                          const int*               gatindex)
1157 {
1158     /* note -- these appear to be full step velocities . . .  */
1159     int  gf = 0, gt = 0;
1160     real vn;
1161     real invfr = 0;
1162     int  n, d;
1163     // Use 1 bit of internal counters to give us 2*2 64-bits values per stream
1164     // Each 64-bit value is enough for 4 normal distribution table numbers.
1165     gmx::ThreeFry2x64<0>                       rng(seed, gmx::RandomDomain::UpdateCoordinates);
1166     gmx::TabulatedNormalDistribution<real, 14> dist;
1167
1168     if (friction_coefficient != 0)
1169     {
1170         invfr = 1.0 / friction_coefficient;
1171     }
1172
1173     for (n = start; (n < nrend); n++)
1174     {
1175         int ng = gatindex ? gatindex[n] : n;
1176
1177         rng.restart(step, ng);
1178         dist.reset();
1179
1180         if (cFREEZE)
1181         {
1182             gf = cFREEZE[n];
1183         }
1184         if (cTC)
1185         {
1186             gt = cTC[n];
1187         }
1188         for (d = 0; (d < DIM); d++)
1189         {
1190             if ((ptype[n] != eptVSite) && (ptype[n] != eptShell) && !nFreeze[gf][d])
1191             {
1192                 if (friction_coefficient != 0)
1193                 {
1194                     vn = invfr * f[n][d] + rf[gt] * dist(rng);
1195                 }
1196                 else
1197                 {
1198                     /* NOTE: invmass = 2/(mass*friction_constant*dt) */
1199                     vn = 0.5 * invmass[n] * f[n][d] * dt
1200                          + std::sqrt(0.5 * invmass[n]) * rf[gt] * dist(rng);
1201                 }
1202
1203                 v[n][d]      = vn;
1204                 xprime[n][d] = x[n][d] + vn * dt;
1205             }
1206             else
1207             {
1208                 v[n][d]      = 0.0;
1209                 xprime[n][d] = x[n][d];
1210             }
1211         }
1212     }
1213 }
1214
1215 extern void init_ekinstate(ekinstate_t* ekinstate, const t_inputrec* ir)
1216 {
1217     ekinstate->ekin_n = ir->opts.ngtc;
1218     snew(ekinstate->ekinh, ekinstate->ekin_n);
1219     snew(ekinstate->ekinf, ekinstate->ekin_n);
1220     snew(ekinstate->ekinh_old, ekinstate->ekin_n);
1221     ekinstate->ekinscalef_nhc.resize(ekinstate->ekin_n);
1222     ekinstate->ekinscaleh_nhc.resize(ekinstate->ekin_n);
1223     ekinstate->vscale_nhc.resize(ekinstate->ekin_n);
1224     ekinstate->dekindl          = 0;
1225     ekinstate->mvcos            = 0;
1226     ekinstate->hasReadEkinState = false;
1227 }
1228
1229 void update_ekinstate(ekinstate_t* ekinstate, const gmx_ekindata_t* ekind)
1230 {
1231     int i;
1232
1233     for (i = 0; i < ekinstate->ekin_n; i++)
1234     {
1235         copy_mat(ekind->tcstat[i].ekinh, ekinstate->ekinh[i]);
1236         copy_mat(ekind->tcstat[i].ekinf, ekinstate->ekinf[i]);
1237         copy_mat(ekind->tcstat[i].ekinh_old, ekinstate->ekinh_old[i]);
1238         ekinstate->ekinscalef_nhc[i] = ekind->tcstat[i].ekinscalef_nhc;
1239         ekinstate->ekinscaleh_nhc[i] = ekind->tcstat[i].ekinscaleh_nhc;
1240         ekinstate->vscale_nhc[i]     = ekind->tcstat[i].vscale_nhc;
1241     }
1242
1243     copy_mat(ekind->ekin, ekinstate->ekin_total);
1244     ekinstate->dekindl = ekind->dekindl;
1245     ekinstate->mvcos   = ekind->cosacc.mvcos;
1246 }
1247
1248 void restore_ekinstate_from_state(const t_commrec* cr, gmx_ekindata_t* ekind, const ekinstate_t* ekinstate)
1249 {
1250     int i, n;
1251
1252     if (MASTER(cr))
1253     {
1254         for (i = 0; i < ekinstate->ekin_n; i++)
1255         {
1256             copy_mat(ekinstate->ekinh[i], ekind->tcstat[i].ekinh);
1257             copy_mat(ekinstate->ekinf[i], ekind->tcstat[i].ekinf);
1258             copy_mat(ekinstate->ekinh_old[i], ekind->tcstat[i].ekinh_old);
1259             ekind->tcstat[i].ekinscalef_nhc = ekinstate->ekinscalef_nhc[i];
1260             ekind->tcstat[i].ekinscaleh_nhc = ekinstate->ekinscaleh_nhc[i];
1261             ekind->tcstat[i].vscale_nhc     = ekinstate->vscale_nhc[i];
1262         }
1263
1264         copy_mat(ekinstate->ekin_total, ekind->ekin);
1265
1266         ekind->dekindl      = ekinstate->dekindl;
1267         ekind->cosacc.mvcos = ekinstate->mvcos;
1268         n                   = ekinstate->ekin_n;
1269     }
1270
1271     if (PAR(cr))
1272     {
1273         gmx_bcast(sizeof(n), &n, cr->mpi_comm_mygroup);
1274         for (i = 0; i < n; i++)
1275         {
1276             gmx_bcast(DIM * DIM * sizeof(ekind->tcstat[i].ekinh[0][0]),
1277                       ekind->tcstat[i].ekinh[0],
1278                       cr->mpi_comm_mygroup);
1279             gmx_bcast(DIM * DIM * sizeof(ekind->tcstat[i].ekinf[0][0]),
1280                       ekind->tcstat[i].ekinf[0],
1281                       cr->mpi_comm_mygroup);
1282             gmx_bcast(DIM * DIM * sizeof(ekind->tcstat[i].ekinh_old[0][0]),
1283                       ekind->tcstat[i].ekinh_old[0],
1284                       cr->mpi_comm_mygroup);
1285
1286             gmx_bcast(sizeof(ekind->tcstat[i].ekinscalef_nhc),
1287                       &(ekind->tcstat[i].ekinscalef_nhc),
1288                       cr->mpi_comm_mygroup);
1289             gmx_bcast(sizeof(ekind->tcstat[i].ekinscaleh_nhc),
1290                       &(ekind->tcstat[i].ekinscaleh_nhc),
1291                       cr->mpi_comm_mygroup);
1292             gmx_bcast(sizeof(ekind->tcstat[i].vscale_nhc), &(ekind->tcstat[i].vscale_nhc), cr->mpi_comm_mygroup);
1293         }
1294         gmx_bcast(DIM * DIM * sizeof(ekind->ekin[0][0]), ekind->ekin[0], cr->mpi_comm_mygroup);
1295
1296         gmx_bcast(sizeof(ekind->dekindl), &ekind->dekindl, cr->mpi_comm_mygroup);
1297         gmx_bcast(sizeof(ekind->cosacc.mvcos), &ekind->cosacc.mvcos, cr->mpi_comm_mygroup);
1298     }
1299 }
1300
1301 void getThreadAtomRange(int numThreads, int threadIndex, int numAtoms, int* startAtom, int* endAtom)
1302 {
1303 #if GMX_HAVE_SIMD_UPDATE
1304     constexpr int blockSize = GMX_SIMD_REAL_WIDTH;
1305 #else
1306     constexpr int blockSize = 1;
1307 #endif
1308     int numBlocks = (numAtoms + blockSize - 1) / blockSize;
1309
1310     *startAtom = ((numBlocks * threadIndex) / numThreads) * blockSize;
1311     *endAtom   = ((numBlocks * (threadIndex + 1)) / numThreads) * blockSize;
1312     if (threadIndex == numThreads - 1)
1313     {
1314         *endAtom = numAtoms;
1315     }
1316 }
1317
1318 void Update::Impl::update_sd_second_half(const t_inputrec& inputRecord,
1319                                          int64_t           step,
1320                                          real*             dvdlambda,
1321                                          const t_mdatoms*  md,
1322                                          t_state*          state,
1323                                          const t_commrec*  cr,
1324                                          t_nrnb*           nrnb,
1325                                          gmx_wallcycle_t   wcycle,
1326                                          gmx::Constraints* constr,
1327                                          bool              do_log,
1328                                          bool              do_ene)
1329 {
1330     if (!constr)
1331     {
1332         return;
1333     }
1334     if (inputRecord.eI == eiSD1)
1335     {
1336         int homenr = md->homenr;
1337
1338         /* Cast delta_t from double to real to make the integrators faster.
1339          * The only reason for having delta_t double is to get accurate values
1340          * for t=delta_t*step when step is larger than float precision.
1341          * For integration dt the accuracy of real suffices, since with
1342          * integral += dt*integrand the increment is nearly always (much) smaller
1343          * than the integral (and the integrand has real precision).
1344          */
1345         real dt = inputRecord.delta_t;
1346
1347         wallcycle_start(wcycle, ewcUPDATE);
1348
1349         int nth = gmx_omp_nthreads_get(emntUpdate);
1350
1351 #pragma omp parallel for num_threads(nth) schedule(static)
1352         for (int th = 0; th < nth; th++)
1353         {
1354             try
1355             {
1356                 int start_th, end_th;
1357                 getThreadAtomRange(nth, th, homenr, &start_th, &end_th);
1358
1359                 doSDUpdateGeneral<SDUpdate::FrictionAndNoiseOnly>(
1360                         sd_,
1361                         start_th,
1362                         end_th,
1363                         dt,
1364                         inputRecord.opts.nFreeze,
1365                         md->invmass,
1366                         md->ptype,
1367                         md->cFREEZE,
1368                         md->cTC,
1369                         state->x.rvec_array(),
1370                         xp_.rvec_array(),
1371                         state->v.rvec_array(),
1372                         nullptr,
1373                         step,
1374                         inputRecord.ld_seed,
1375                         DOMAINDECOMP(cr) ? cr->dd->globalAtomIndices.data() : nullptr);
1376             }
1377             GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1378         }
1379         inc_nrnb(nrnb, eNR_UPDATE, homenr);
1380         wallcycle_stop(wcycle, ewcUPDATE);
1381
1382         /* Constrain the coordinates upd->xp for half a time step */
1383         bool computeVirial = false;
1384         constr->apply(do_log,
1385                       do_ene,
1386                       step,
1387                       1,
1388                       0.5,
1389                       state->x.arrayRefWithPadding(),
1390                       xp_.arrayRefWithPadding(),
1391                       ArrayRef<RVec>(),
1392                       state->box,
1393                       state->lambda[efptBONDED],
1394                       dvdlambda,
1395                       state->v.arrayRefWithPadding(),
1396                       computeVirial,
1397                       nullptr,
1398                       ConstraintVariable::Positions);
1399     }
1400 }
1401
1402 void Update::Impl::finish_update(const t_inputrec& inputRecord,
1403                                  const t_mdatoms*  md,
1404                                  t_state*          state,
1405                                  gmx_wallcycle_t   wcycle,
1406                                  const bool        haveConstraints)
1407 {
1408     /* NOTE: Currently we always integrate to a temporary buffer and
1409      * then copy the results back here.
1410      */
1411
1412     wallcycle_start_nocount(wcycle, ewcUPDATE);
1413
1414     const int homenr = md->homenr;
1415     auto      xp     = makeConstArrayRef(xp_).subArray(0, homenr);
1416     auto      x      = makeArrayRef(state->x).subArray(0, homenr);
1417
1418     if (md->havePartiallyFrozenAtoms && haveConstraints)
1419     {
1420         /* We have atoms that are frozen along some, but not all dimensions,
1421          * then constraints will have moved them also along the frozen dimensions.
1422          * To freeze such degrees of freedom we do not copy them back here.
1423          */
1424         const ivec* nFreeze = inputRecord.opts.nFreeze;
1425
1426         for (int i = 0; i < homenr; i++)
1427         {
1428             const int g = md->cFREEZE[i];
1429
1430             for (int d = 0; d < DIM; d++)
1431             {
1432                 if (nFreeze[g][d] == 0)
1433                 {
1434                     x[i][d] = xp[i][d];
1435                 }
1436             }
1437         }
1438     }
1439     else
1440     {
1441         /* We have no frozen atoms or fully frozen atoms which have not
1442          * been moved by the update, so we can simply copy all coordinates.
1443          */
1444         int gmx_unused nth = gmx_omp_nthreads_get(emntUpdate);
1445 #pragma omp parallel for num_threads(nth) schedule(static)
1446         for (int i = 0; i < homenr; i++)
1447         {
1448             // Trivial statement, does not throw
1449             x[i] = xp[i];
1450         }
1451     }
1452
1453     wallcycle_stop(wcycle, ewcUPDATE);
1454 }
1455
1456 void Update::Impl::update_coords(const t_inputrec&                                inputRecord,
1457                                  int64_t                                          step,
1458                                  const t_mdatoms*                                 md,
1459                                  t_state*                                         state,
1460                                  const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
1461                                  const t_fcdata&                                  fcdata,
1462                                  const gmx_ekindata_t*                            ekind,
1463                                  const matrix                                     M,
1464                                  int                                              updatePart,
1465                                  const t_commrec*                                 cr,
1466                                  const bool                                       haveConstraints)
1467 {
1468     /* Running the velocity half does nothing except for velocity verlet */
1469     if ((updatePart == etrtVELOCITY1 || updatePart == etrtVELOCITY2) && !EI_VV(inputRecord.eI))
1470     {
1471         gmx_incons("update_coords called for velocity without VV integrator");
1472     }
1473
1474     int homenr = md->homenr;
1475
1476     /* Cast to real for faster code, no loss in precision (see comment above) */
1477     real dt = inputRecord.delta_t;
1478
1479     /* We need to update the NMR restraint history when time averaging is used */
1480     if (state->flags & (1 << estDISRE_RM3TAV))
1481     {
1482         update_disres_history(*fcdata.disres, &state->hist);
1483     }
1484     if (state->flags & (1 << estORIRE_DTAV))
1485     {
1486         update_orires_history(*fcdata.orires, &state->hist);
1487     }
1488
1489     /* ############# START The update of velocities and positions ######### */
1490     int nth = gmx_omp_nthreads_get(emntUpdate);
1491
1492 #pragma omp parallel for num_threads(nth) schedule(static)
1493     for (int th = 0; th < nth; th++)
1494     {
1495         try
1496         {
1497             int start_th, end_th;
1498             getThreadAtomRange(nth, th, homenr, &start_th, &end_th);
1499
1500             const rvec* x_rvec  = state->x.rvec_array();
1501             rvec*       xp_rvec = xp_.rvec_array();
1502             rvec*       v_rvec  = state->v.rvec_array();
1503             const rvec* f_rvec  = as_rvec_array(f.unpaddedConstArrayRef().data());
1504
1505             switch (inputRecord.eI)
1506             {
1507                 case (eiMD):
1508                     do_update_md(start_th,
1509                                  end_th,
1510                                  dt,
1511                                  step,
1512                                  x_rvec,
1513                                  xp_rvec,
1514                                  v_rvec,
1515                                  f_rvec,
1516                                  inputRecord.etc,
1517                                  inputRecord.epc,
1518                                  inputRecord.nsttcouple,
1519                                  inputRecord.nstpcouple,
1520                                  md,
1521                                  ekind,
1522                                  state->box,
1523                                  state->nosehoover_vxi.data(),
1524                                  M);
1525                     break;
1526                 case (eiSD1):
1527                     do_update_sd(start_th,
1528                                  end_th,
1529                                  dt,
1530                                  step,
1531                                  x_rvec,
1532                                  xp_rvec,
1533                                  v_rvec,
1534                                  f_rvec,
1535                                  inputRecord.opts.nFreeze,
1536                                  md->invmass,
1537                                  md->ptype,
1538                                  md->cFREEZE,
1539                                  md->cTC,
1540                                  inputRecord.ld_seed,
1541                                  cr,
1542                                  sd_,
1543                                  haveConstraints);
1544                     break;
1545                 case (eiBD):
1546                     do_update_bd(start_th,
1547                                  end_th,
1548                                  dt,
1549                                  step,
1550                                  x_rvec,
1551                                  xp_rvec,
1552                                  v_rvec,
1553                                  f_rvec,
1554                                  inputRecord.opts.nFreeze,
1555                                  md->invmass,
1556                                  md->ptype,
1557                                  md->cFREEZE,
1558                                  md->cTC,
1559                                  inputRecord.bd_fric,
1560                                  sd_.bd_rf.data(),
1561                                  inputRecord.ld_seed,
1562                                  DOMAINDECOMP(cr) ? cr->dd->globalAtomIndices.data() : nullptr);
1563                     break;
1564                 case (eiVV):
1565                 case (eiVVAK):
1566                 {
1567                     gmx_bool bExtended = (inputRecord.etc == etcNOSEHOOVER || inputRecord.epc == epcPARRINELLORAHMAN
1568                                           || inputRecord.epc == epcMTTK);
1569
1570                     /* assuming barostat coupled to group 0 */
1571                     real alpha = 1.0 + DIM / static_cast<real>(inputRecord.opts.nrdf[0]);
1572                     switch (updatePart)
1573                     {
1574                         case etrtVELOCITY1:
1575                         case etrtVELOCITY2:
1576                             do_update_vv_vel(start_th,
1577                                              end_th,
1578                                              dt,
1579                                              inputRecord.opts.nFreeze,
1580                                              md->invmass,
1581                                              md->ptype,
1582                                              md->cFREEZE,
1583                                              v_rvec,
1584                                              f_rvec,
1585                                              bExtended,
1586                                              state->veta,
1587                                              alpha);
1588                             break;
1589                         case etrtPOSITION:
1590                             do_update_vv_pos(start_th,
1591                                              end_th,
1592                                              dt,
1593                                              inputRecord.opts.nFreeze,
1594                                              md->ptype,
1595                                              md->cFREEZE,
1596                                              x_rvec,
1597                                              xp_rvec,
1598                                              v_rvec,
1599                                              bExtended,
1600                                              state->veta);
1601                             break;
1602                     }
1603                     break;
1604                 }
1605                 default: gmx_fatal(FARGS, "Don't know how to update coordinates");
1606             }
1607         }
1608         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1609     }
1610 }
1611
1612 void Update::Impl::update_for_constraint_virial(const t_inputrec& inputRecord,
1613                                                 const t_mdatoms&  md,
1614                                                 const t_state&    state,
1615                                                 const gmx::ArrayRefWithPadding<const gmx::RVec>& f,
1616                                                 const gmx_ekindata_t& ekind)
1617 {
1618     GMX_ASSERT(inputRecord.eI == eiMD || inputRecord.eI == eiSD1,
1619                "Only leap-frog is supported here");
1620
1621     // Cast to real for faster code, no loss in precision
1622     const real dt = inputRecord.delta_t;
1623
1624     const int nth = gmx_omp_nthreads_get(emntUpdate);
1625
1626 #pragma omp parallel for num_threads(nth) schedule(static)
1627     for (int th = 0; th < nth; th++)
1628     {
1629         try
1630         {
1631             int start_th, end_th;
1632             getThreadAtomRange(nth, th, md.homenr, &start_th, &end_th);
1633
1634             const rvec* x_rvec  = state.x.rvec_array();
1635             rvec*       xp_rvec = xp_.rvec_array();
1636             rvec*       v_rvec  = const_cast<rvec*>(state.v.rvec_array());
1637             const rvec* f_rvec  = as_rvec_array(f.unpaddedConstArrayRef().data());
1638
1639             doUpdateMDDoNotUpdateVelocities(
1640                     start_th, end_th, dt, x_rvec, xp_rvec, v_rvec, f_rvec, md, ekind);
1641         }
1642         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1643     }
1644 }