src/gromacs/nbnxm/kerneldispatch.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2016,2017,2018,2019, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35
  36 #include "gmxpre.h"
  37
  38 #include "gromacs/gmxlib/nrnb.h"
  39 #include "gromacs/gmxlib/nonbonded/nb_free_energy.h"
  40 #include "gromacs/gmxlib/nonbonded/nb_kernel.h"
  41 #include "gromacs/gmxlib/nonbonded/nonbonded.h"
  42 #include "gromacs/math/vectypes.h"
  43 #include "gromacs/mdlib/force.h"
  44 #include "gromacs/mdlib/force_flags.h"
  45 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  46 #include "gromacs/mdtypes/enerdata.h"
  47 #include "gromacs/mdtypes/inputrec.h"
  48 #include "gromacs/mdtypes/interaction_const.h"
  49 #include "gromacs/mdtypes/md_enums.h"
  50 #include "gromacs/mdtypes/mdatom.h"
  51 #include "gromacs/nbnxm/gpu_data_mgmt.h"
  52 #include "gromacs/nbnxm/nbnxm.h"
  53 #include "gromacs/nbnxm/nbnxm_simd.h"
  54 #include "gromacs/nbnxm/pairlist.h"
  55 #include "gromacs/nbnxm/kernels_reference/kernel_gpu_ref.h"
  56 #include "gromacs/simd/simd.h"
  57 #include "gromacs/utility/gmxassert.h"
  58 #include "gromacs/utility/real.h"
  59
  60 #include "kernel_common.h"
  61 #define INCLUDE_KERNELFUNCTION_TABLES
  62 #include "gromacs/nbnxm/kernels_reference/kernel_ref.h"
  63 #ifdef GMX_NBNXN_SIMD_2XNN
  64 #include "gromacs/nbnxm/kernels_simd_2xmm/kernels.h"
  65 #endif
  66 #ifdef GMX_NBNXN_SIMD_4XN
  67 #include "gromacs/nbnxm/kernels_simd_4xm/kernels.h"
  68 #endif
  69 #undef INCLUDE_FUNCTION_TABLES
  70
  71 /*! \brief Clears the energy group output buffers
  72  *
  73  * \param[in,out] out  nbnxn kernel output struct
  74  */
  75 static void clearGroupEnergies(nbnxn_atomdata_output_t *out)
  76 {
  77     std::fill(out->Vvdw.begin(), out->Vvdw.end(), 0.0_real);
  78     std::fill(out->Vc.begin(), out->Vc.end(), 0.0_real);
  79     std::fill(out->VSvdw.begin(), out->VSvdw.end(), 0.0_real);
  80     std::fill(out->VSc.begin(), out->VSc.end(), 0.0_real);
  81 }
  82
  83 /*! \brief Reduce the group-pair energy buffers produced by a SIMD kernel
  84  * to single terms in the output buffers.
  85  *
  86  * The SIMD kernels produce a large number of energy buffer in SIMD registers
  87  * to avoid scattered reads and writes.
  88  *
  89  * \tparam        unrollj         The unroll size for j-particles in the SIMD kernel
  90  * \param[in]     numGroups       The number of energy groups
  91  * \param[in]     numGroups_2log  Log2 of numGroups, rounded up
  92  * \param[in,out] out             Struct with energy buffers
  93  */
  94 template <int unrollj> static void
  95 reduceGroupEnergySimdBuffers(int                       numGroups,
  96                              int                       numGroups_2log,
  97                              nbnxn_atomdata_output_t  *out)
  98 {
  99     const int                 unrollj_half     = unrollj/2;
 100     /* Energies are stored in SIMD registers with size 2^numGroups_2log */
 101     const int                 numGroupsStorage = (1 << numGroups_2log);
 102
 103     const real * gmx_restrict vVdwSimd     = out->VSvdw.data();
 104     const real * gmx_restrict vCoulombSimd = out->VSc.data();
 105     real * gmx_restrict       vVdw         = out->Vvdw.data();
 106     real * gmx_restrict       vCoulomb     = out->Vc.data();
 107
 108     /* The size of the SIMD energy group buffer array is:
 109      * numGroups*numGroups*numGroupsStorage*unrollj_half*simd_width
 110      */
 111     for (int i = 0; i < numGroups; i++)
 112     {
 113         for (int j1 = 0; j1 < numGroups; j1++)
 114         {
 115             for (int j0 = 0; j0 < numGroups; j0++)
 116             {
 117                 int c = ((i*numGroups + j1)*numGroupsStorage + j0)*unrollj_half*unrollj;
 118                 for (int s = 0; s < unrollj_half; s++)
 119                 {
 120                     vVdw    [i*numGroups + j0] += vVdwSimd    [c + 0];
 121                     vVdw    [i*numGroups + j1] += vVdwSimd    [c + 1];
 122                     vCoulomb[i*numGroups + j0] += vCoulombSimd[c + 0];
 123                     vCoulomb[i*numGroups + j1] += vCoulombSimd[c + 1];
 124                     c                          += unrollj + 2;
 125                 }
 126             }
 127         }
 128     }
 129 }
 130
 131 /*! \brief Dispatches the non-bonded N versus M atom cluster CPU kernels.
 132  *
 133  * OpenMP parallelization is performed within this function.
 134  * Energy reduction, but not force and shift force reduction, is performed
 135  * within this function.
 136  *
 137  * \param[in]     pairlistSet   Pairlists with local or non-local interactions to compute
 138  * \param[in]     kernelSetup   The non-bonded kernel setup
 139  * \param[in,out] nbat          The atomdata for the interactions
 140  * \param[in]     ic            Non-bonded interaction constants
 141  * \param[in]     shiftVectors  The PBC shift vectors
 142  * \param[in]     forceFlags    Flags that tell what to compute
 143  * \param[in]     clearF        Enum that tells if to clear the force output buffer
 144  * \param[out]    fshift        Shift force output buffer
 145  * \param[out]    vCoulomb      Output buffer for Coulomb energies
 146  * \param[out]    vVdw          Output buffer for Van der Waals energies
 147  */
 148 static void
 149 nbnxn_kernel_cpu(const PairlistSet              &pairlistSet,
 150                  const Nbnxm::KernelSetup       &kernelSetup,
 151                  nbnxn_atomdata_t               *nbat,
 152                  const interaction_const_t      &ic,
 153                  rvec                           *shiftVectors,
 154                  int                             forceFlags,
 155                  int                             clearF,
 156                  real                           *fshift,
 157                  real                           *vCoulomb,
 158                  real                           *vVdw)
 159 {
 160
 161     int                      coulkt;
 162     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 163     {
 164         coulkt = coulktRF;
 165     }
 166     else
 167     {
 168         if (kernelSetup.ewaldExclusionType == Nbnxm::EwaldExclusionType::Table)
 169         {
 170             if (ic.rcoulomb == ic.rvdw)
 171             {
 172                 coulkt = coulktTAB;
 173             }
 174             else
 175             {
 176                 coulkt = coulktTAB_TWIN;
 177             }
 178         }
 179         else
 180         {
 181             if (ic.rcoulomb == ic.rvdw)
 182             {
 183                 coulkt = coulktEWALD;
 184             }
 185             else
 186             {
 187                 coulkt = coulktEWALD_TWIN;
 188             }
 189         }
 190     }
 191
 192     const nbnxn_atomdata_t::Params &nbatParams = nbat->params();
 193
 194     int vdwkt = 0;
 195     if (ic.vdwtype == evdwCUT)
 196     {
 197         switch (ic.vdw_modifier)
 198         {
 199             case eintmodNONE:
 200             case eintmodPOTSHIFT:
 201                 switch (nbatParams.comb_rule)
 202                 {
 203                     case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
 204                     case ljcrLB:   vdwkt = vdwktLJCUT_COMBLB;   break;
 205                     case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
 206                     default:
 207                         GMX_RELEASE_ASSERT(false, "Unknown combination rule");
 208                 }
 209                 break;
 210             case eintmodFORCESWITCH:
 211                 vdwkt = vdwktLJFORCESWITCH;
 212                 break;
 213             case eintmodPOTSWITCH:
 214                 vdwkt = vdwktLJPOTSWITCH;
 215                 break;
 216             default:
 217                 GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction modifier");
 218         }
 219     }
 220     else if (ic.vdwtype == evdwPME)
 221     {
 222         if (ic.ljpme_comb_rule == eljpmeGEOM)
 223         {
 224             vdwkt = vdwktLJEWALDCOMBGEOM;
 225         }
 226         else
 227         {
 228             vdwkt = vdwktLJEWALDCOMBLB;
 229             /* At setup we (should have) selected the C reference kernel */
 230             GMX_RELEASE_ASSERT(kernelSetup.kernelType == Nbnxm::KernelType::Cpu4x4_PlainC, "Only the C reference nbnxn SIMD kernel supports LJ-PME with LB combination rules");
 231         }
 232     }
 233     else
 234     {
 235         GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction type");
 236     }
 237
 238     gmx::ArrayRef<const NbnxnPairlistCpu> pairlists = pairlistSet.cpuLists();
 239
 240     int gmx_unused                        nthreads = gmx_omp_nthreads_get(emntNonbonded);
 241 #pragma omp parallel for schedule(static) num_threads(nthreads)
 242     for (int nb = 0; nb < pairlists.ssize(); nb++)
 243     {
 244         // Presently, the kernels do not call C++ code that can throw,
 245         // so no need for a try/catch pair in this OpenMP region.
 246         nbnxn_atomdata_output_t *out = &nbat->out[nb];
 247
 248         if (clearF == enbvClearFYes)
 249         {
 250             clear_f(nbat, nb, out->f.data());
 251         }
 252
 253         real *fshift_p;
 254         if ((forceFlags & GMX_FORCE_VIRIAL) && pairlists.ssize() == 1)
 255         {
 256             fshift_p = fshift;
 257         }
 258         else
 259         {
 260             fshift_p = out->fshift.data();
 261
 262             if (clearF == enbvClearFYes)
 263             {
 264                 clear_fshift(fshift_p);
 265             }
 266         }
 267
 268         // TODO: Change to reference
 269         const NbnxnPairlistCpu *pairlist = &pairlists[nb];
 270
 271         if (!(forceFlags & GMX_FORCE_ENERGY))
 272         {
 273             /* Don't calculate energies */
 274             switch (kernelSetup.kernelType)
 275             {
 276                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 277                     nbnxn_kernel_noener_ref[coulkt][vdwkt](pairlist, nbat,
 278                                                            &ic,
 279                                                            shiftVectors,
 280                                                            out->f.data(),
 281                                                            fshift_p);
 282                     break;
 283 #ifdef GMX_NBNXN_SIMD_2XNN
 284                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 285                     nbnxm_kernel_noener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 286                                                                  &ic,
 287                                                                  shiftVectors,
 288                                                                  out->f.data(),
 289                                                                  fshift_p);
 290                     break;
 291 #endif
 292 #ifdef GMX_NBNXN_SIMD_4XN
 293                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 294                     nbnxm_kernel_noener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 295                                                                 &ic,
 296                                                                 shiftVectors,
 297                                                                 out->f.data(),
 298                                                                 fshift_p);
 299                     break;
 300 #endif
 301                 default:
 302                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 303             }
 304         }
 305         else if (out->Vvdw.size() == 1)
 306         {
 307             /* A single energy group (pair) */
 308             out->Vvdw[0] = 0;
 309             out->Vc[0]   = 0;
 310
 311             switch (kernelSetup.kernelType)
 312             {
 313                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 314                     nbnxn_kernel_ener_ref[coulkt][vdwkt](pairlist, nbat,
 315                                                          &ic,
 316                                                          shiftVectors,
 317                                                          out->f.data(),
 318                                                          fshift_p,
 319                                                          out->Vvdw.data(),
 320                                                          out->Vc.data());
 321                     break;
 322 #ifdef GMX_NBNXN_SIMD_2XNN
 323                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 324                     nbnxm_kernel_ener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 325                                                                &ic,
 326                                                                shiftVectors,
 327                                                                out->f.data(),
 328                                                                fshift_p,
 329                                                                out->Vvdw.data(),
 330                                                                out->Vc.data());
 331                     break;
 332 #endif
 333 #ifdef GMX_NBNXN_SIMD_4XN
 334                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 335                     nbnxm_kernel_ener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 336                                                               &ic,
 337                                                               shiftVectors,
 338                                                               out->f.data(),
 339                                                               fshift_p,
 340                                                               out->Vvdw.data(),
 341                                                               out->Vc.data());
 342                     break;
 343 #endif
 344                 default:
 345                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 346             }
 347         }
 348         else
 349         {
 350             /* Calculate energy group contributions */
 351             clearGroupEnergies(out);
 352
 353             int unrollj = 0;
 354
 355             switch (kernelSetup.kernelType)
 356             {
 357                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 358                     unrollj = c_nbnxnCpuIClusterSize;
 359                     nbnxn_kernel_energrp_ref[coulkt][vdwkt](pairlist, nbat,
 360                                                             &ic,
 361                                                             shiftVectors,
 362                                                             out->f.data(),
 363                                                             fshift_p,
 364                                                             out->Vvdw.data(),
 365                                                             out->Vc.data());
 366                     break;
 367 #ifdef GMX_NBNXN_SIMD_2XNN
 368                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 369                     unrollj = GMX_SIMD_REAL_WIDTH/2;
 370                     nbnxm_kernel_energrp_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 371                                                                   &ic,
 372                                                                   shiftVectors,
 373                                                                   out->f.data(),
 374                                                                   fshift_p,
 375                                                                   out->VSvdw.data(),
 376                                                                   out->VSc.data());
 377                     break;
 378 #endif
 379 #ifdef GMX_NBNXN_SIMD_4XN
 380                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 381                     unrollj = GMX_SIMD_REAL_WIDTH;
 382                     nbnxm_kernel_energrp_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 383                                                                  &ic,
 384                                                                  shiftVectors,
 385                                                                  out->f.data(),
 386                                                                  fshift_p,
 387                                                                  out->VSvdw.data(),
 388                                                                  out->VSc.data());
 389                     break;
 390 #endif
 391                 default:
 392                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 393             }
 394
 395             if (kernelSetup.kernelType != Nbnxm::KernelType::Cpu4x4_PlainC)
 396             {
 397                 switch (unrollj)
 398                 {
 399                     case 2:
 400                         reduceGroupEnergySimdBuffers<2>(nbatParams.nenergrp,
 401                                                         nbatParams.neg_2log,
 402                                                         out);
 403                         break;
 404                     case 4:
 405                         reduceGroupEnergySimdBuffers<4>(nbatParams.nenergrp,
 406                                                         nbatParams.neg_2log,
 407                                                         out);
 408                         break;
 409                     case 8:
 410                         reduceGroupEnergySimdBuffers<8>(nbatParams.nenergrp,
 411                                                         nbatParams.neg_2log,
 412                                                         out);
 413                         break;
 414                     default:
 415                         GMX_RELEASE_ASSERT(false, "Unsupported j-unroll size");
 416                 }
 417             }
 418         }
 419     }
 420
 421     if (forceFlags & GMX_FORCE_ENERGY)
 422     {
 423         reduce_energies_over_lists(nbat, pairlists.ssize(), vVdw, vCoulomb);
 424     }
 425 }
 426
 427 static void accountFlops(t_nrnb                           *nrnb,
 428                          const PairlistSet                &pairlistSet,
 429                          const nonbonded_verlet_t         &nbv,
 430                          const interaction_const_t        &ic,
 431                          const int                         forceFlags)
 432 {
 433     const bool usingGpuKernels = nbv.useGpu();
 434
 435     int        enr_nbnxn_kernel_ljc;
 436     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 437     {
 438         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_RF;
 439     }
 440     else if ((!usingGpuKernels && nbv.kernelSetup().ewaldExclusionType == Nbnxm::EwaldExclusionType::Analytical) ||
 441              (usingGpuKernels && Nbnxm::gpu_is_kernel_ewald_analytical(nbv.gpu_nbv)))
 442     {
 443         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_EWALD;
 444     }
 445     else
 446     {
 447         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_TAB;
 448     }
 449     int enr_nbnxn_kernel_lj = eNR_NBNXN_LJ;
 450     if (forceFlags & GMX_FORCE_ENERGY)
 451     {
 452         /* In eNR_??? the nbnxn F+E kernels are always the F kernel + 1 */
 453         enr_nbnxn_kernel_ljc += 1;
 454         enr_nbnxn_kernel_lj  += 1;
 455     }
 456
 457     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc,
 458              pairlistSet.natpair_ljq_);
 459     inc_nrnb(nrnb, enr_nbnxn_kernel_lj,
 460              pairlistSet.natpair_lj_);
 461     /* The Coulomb-only kernels are offset -eNR_NBNXN_LJ_RF+eNR_NBNXN_RF */
 462     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc-eNR_NBNXN_LJ_RF+eNR_NBNXN_RF,
 463              pairlistSet.natpair_q_);
 464
 465     const bool calcEnergy = ((forceFlags & GMX_FORCE_ENERGY) != 0);
 466     if (ic.vdw_modifier == eintmodFORCESWITCH)
 467     {
 468         /* We add up the switch cost separately */
 469         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_FSW + (calcEnergy ? 1 : 0),
 470                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 471     }
 472     if (ic.vdw_modifier == eintmodPOTSWITCH)
 473     {
 474         /* We add up the switch cost separately */
 475         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_PSW + (calcEnergy ? 1 : 0),
 476                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 477     }
 478     if (ic.vdwtype == evdwPME)
 479     {
 480         /* We add up the LJ Ewald cost separately */
 481         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_EWALD + (calcEnergy ? 1 : 0),
 482                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 483     }
 484 }
 485
 486 void
 487 nonbonded_verlet_t::dispatchNonbondedKernel(Nbnxm::InteractionLocality iLocality,
 488                                             const interaction_const_t &ic,
 489                                             int                        forceFlags,
 490                                             int                        clearF,
 491                                             t_forcerec                *fr,
 492                                             gmx_enerdata_t            *enerd,
 493                                             t_nrnb                    *nrnb)
 494 {
 495     const PairlistSet &pairlistSet = pairlistSets().pairlistSet(iLocality);
 496
 497     switch (kernelSetup().kernelType)
 498     {
 499         case Nbnxm::KernelType::Cpu4x4_PlainC:
 500         case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 501         case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 502             nbnxn_kernel_cpu(pairlistSet,
 503                              kernelSetup(),
 504                              nbat.get(),
 505                              ic,
 506                              fr->shift_vec,
 507                              forceFlags,
 508                              clearF,
 509                              fr->fshift[0],
 510                              enerd->grpp.ener[egCOULSR],
 511                              fr->bBHAM ?
 512                              enerd->grpp.ener[egBHAMSR] :
 513                              enerd->grpp.ener[egLJSR]);
 514             break;
 515
 516         case Nbnxm::KernelType::Gpu8x8x8:
 517             Nbnxm::gpu_launch_kernel(gpu_nbv, forceFlags, iLocality);
 518             break;
 519
 520         case Nbnxm::KernelType::Cpu8x8x8_PlainC:
 521             nbnxn_kernel_gpu_ref(pairlistSet.gpuList(),
 522                                  nbat.get(), &ic,
 523                                  fr->shift_vec,
 524                                  forceFlags,
 525                                  clearF,
 526                                  nbat->out[0].f,
 527                                  fr->fshift[0],
 528                                  enerd->grpp.ener[egCOULSR],
 529                                  fr->bBHAM ?
 530                                  enerd->grpp.ener[egBHAMSR] :
 531                                  enerd->grpp.ener[egLJSR]);
 532             break;
 533
 534         default:
 535             GMX_RELEASE_ASSERT(false, "Invalid nonbonded kernel type passed!");
 536
 537     }
 538
 539     accountFlops(nrnb, pairlistSet, *this, ic, forceFlags);
 540 }
 541
 542 void
 543 nonbonded_verlet_t::dispatchFreeEnergyKernel(Nbnxm::InteractionLocality  iLocality,
 544                                              t_forcerec                 *fr,
 545                                              rvec                        x[],
 546                                              rvec                        f[],
 547                                              const t_mdatoms            &mdatoms,
 548                                              t_lambda                   *fepvals,
 549                                              real                       *lambda,
 550                                              gmx_enerdata_t             *enerd,
 551                                              const int                   forceFlags,
 552                                              t_nrnb                     *nrnb)
 553 {
 554     const gmx::ArrayRef<t_nblist const * const > nbl_fep = pairlistSets().pairlistSet(iLocality).fepLists();
 555
 556     /* When the first list is empty, all are empty and there is nothing to do */
 557     if (!pairlistSets().params().haveFep || nbl_fep[0]->nrj == 0)
 558     {
 559         return;
 560     }
 561
 562     int donb_flags = 0;
 563     /* Add short-range interactions */
 564     donb_flags |= GMX_NONBONDED_DO_SR;
 565
 566     /* Currently all group scheme kernels always calculate (shift-)forces */
 567     if (forceFlags & GMX_FORCE_FORCES)
 568     {
 569         donb_flags |= GMX_NONBONDED_DO_FORCE;
 570     }
 571     if (forceFlags & GMX_FORCE_VIRIAL)
 572     {
 573         donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 574     }
 575     if (forceFlags & GMX_FORCE_ENERGY)
 576     {
 577         donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 578     }
 579
 580     nb_kernel_data_t kernel_data;
 581     real             dvdl_nb[efptNR] = { 0 };
 582     kernel_data.flags  = donb_flags;
 583     kernel_data.lambda = lambda;
 584     kernel_data.dvdl   = dvdl_nb;
 585
 586     kernel_data.energygrp_elec = enerd->grpp.ener[egCOULSR];
 587     kernel_data.energygrp_vdw  = enerd->grpp.ener[egLJSR];
 588
 589     GMX_ASSERT(gmx_omp_nthreads_get(emntNonbonded) == nbl_fep.ssize(), "Number of lists should be same as number of NB threads");
 590
 591 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 592     for (int th = 0; th < nbl_fep.ssize(); th++)
 593     {
 594         try
 595         {
 596             gmx_nb_free_energy_kernel(nbl_fep[th],
 597                                       x, f, fr, &mdatoms, &kernel_data, nrnb);
 598         }
 599         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 600     }
 601
 602     if (fepvals->sc_alpha != 0)
 603     {
 604         enerd->dvdl_nonlin[efptVDW]  += dvdl_nb[efptVDW];
 605         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 606     }
 607     else
 608     {
 609         enerd->dvdl_lin[efptVDW]  += dvdl_nb[efptVDW];
 610         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 611     }
 612
 613     /* If we do foreign lambda and we have soft-core interactions
 614      * we have to recalculate the (non-linear) energies contributions.
 615      */
 616     if (fepvals->n_lambda > 0 && (forceFlags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 617     {
 618         real lam_i[efptNR];
 619         kernel_data.flags          = (donb_flags & ~(GMX_NONBONDED_DO_FORCE | GMX_NONBONDED_DO_SHIFTFORCE)) | GMX_NONBONDED_DO_FOREIGNLAMBDA;
 620         kernel_data.lambda         = lam_i;
 621         kernel_data.energygrp_elec = enerd->foreign_grpp.ener[egCOULSR];
 622         kernel_data.energygrp_vdw  = enerd->foreign_grpp.ener[egLJSR];
 623         /* Note that we add to kernel_data.dvdl, but ignore the result */
 624
 625         for (int i = 0; i < enerd->n_lambda; i++)
 626         {
 627             for (int j = 0; j < efptNR; j++)
 628             {
 629                 lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 630             }
 631             reset_foreign_enerdata(enerd);
 632 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 633             for (int th = 0; th < nbl_fep.ssize(); th++)
 634             {
 635                 try
 636                 {
 637                     gmx_nb_free_energy_kernel(nbl_fep[th],
 638                                               x, f, fr, &mdatoms, &kernel_data, nrnb);
 639                 }
 640                 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 641             }
 642
 643             sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 644             enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 645         }
 646     }
 647 }