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 #include "pairlistsets.h"
  62 #define INCLUDE_KERNELFUNCTION_TABLES
  63 #include "gromacs/nbnxm/kernels_reference/kernel_ref.h"
  64 #ifdef GMX_NBNXN_SIMD_2XNN
  65 #include "gromacs/nbnxm/kernels_simd_2xmm/kernels.h"
  66 #endif
  67 #ifdef GMX_NBNXN_SIMD_4XN
  68 #include "gromacs/nbnxm/kernels_simd_4xm/kernels.h"
  69 #endif
  70 #undef INCLUDE_FUNCTION_TABLES
  71
  72 /*! \brief Clears the energy group output buffers
  73  *
  74  * \param[in,out] out  nbnxn kernel output struct
  75  */
  76 static void clearGroupEnergies(nbnxn_atomdata_output_t *out)
  77 {
  78     std::fill(out->Vvdw.begin(), out->Vvdw.end(), 0.0_real);
  79     std::fill(out->Vc.begin(), out->Vc.end(), 0.0_real);
  80     std::fill(out->VSvdw.begin(), out->VSvdw.end(), 0.0_real);
  81     std::fill(out->VSc.begin(), out->VSc.end(), 0.0_real);
  82 }
  83
  84 /*! \brief Reduce the group-pair energy buffers produced by a SIMD kernel
  85  * to single terms in the output buffers.
  86  *
  87  * The SIMD kernels produce a large number of energy buffer in SIMD registers
  88  * to avoid scattered reads and writes.
  89  *
  90  * \tparam        unrollj         The unroll size for j-particles in the SIMD kernel
  91  * \param[in]     numGroups       The number of energy groups
  92  * \param[in]     numGroups_2log  Log2 of numGroups, rounded up
  93  * \param[in,out] out             Struct with energy buffers
  94  */
  95 template <int unrollj> static void
  96 reduceGroupEnergySimdBuffers(int                       numGroups,
  97                              int                       numGroups_2log,
  98                              nbnxn_atomdata_output_t  *out)
  99 {
 100     const int                 unrollj_half     = unrollj/2;
 101     /* Energies are stored in SIMD registers with size 2^numGroups_2log */
 102     const int                 numGroupsStorage = (1 << numGroups_2log);
 103
 104     const real * gmx_restrict vVdwSimd     = out->VSvdw.data();
 105     const real * gmx_restrict vCoulombSimd = out->VSc.data();
 106     real * gmx_restrict       vVdw         = out->Vvdw.data();
 107     real * gmx_restrict       vCoulomb     = out->Vc.data();
 108
 109     /* The size of the SIMD energy group buffer array is:
 110      * numGroups*numGroups*numGroupsStorage*unrollj_half*simd_width
 111      */
 112     for (int i = 0; i < numGroups; i++)
 113     {
 114         for (int j1 = 0; j1 < numGroups; j1++)
 115         {
 116             for (int j0 = 0; j0 < numGroups; j0++)
 117             {
 118                 int c = ((i*numGroups + j1)*numGroupsStorage + j0)*unrollj_half*unrollj;
 119                 for (int s = 0; s < unrollj_half; s++)
 120                 {
 121                     vVdw    [i*numGroups + j0] += vVdwSimd    [c + 0];
 122                     vVdw    [i*numGroups + j1] += vVdwSimd    [c + 1];
 123                     vCoulomb[i*numGroups + j0] += vCoulombSimd[c + 0];
 124                     vCoulomb[i*numGroups + j1] += vCoulombSimd[c + 1];
 125                     c                          += unrollj + 2;
 126                 }
 127             }
 128         }
 129     }
 130 }
 131
 132 /*! \brief Dispatches the non-bonded N versus M atom cluster CPU kernels.
 133  *
 134  * OpenMP parallelization is performed within this function.
 135  * Energy reduction, but not force and shift force reduction, is performed
 136  * within this function.
 137  *
 138  * \param[in]     pairlistSet   Pairlists with local or non-local interactions to compute
 139  * \param[in]     kernelSetup   The non-bonded kernel setup
 140  * \param[in,out] nbat          The atomdata for the interactions
 141  * \param[in]     ic            Non-bonded interaction constants
 142  * \param[in]     shiftVectors  The PBC shift vectors
 143  * \param[in]     forceFlags    Flags that tell what to compute
 144  * \param[in]     clearF        Enum that tells if to clear the force output buffer
 145  * \param[out]    fshift        Shift force output buffer
 146  * \param[out]    vCoulomb      Output buffer for Coulomb energies
 147  * \param[out]    vVdw          Output buffer for Van der Waals energies
 148  */
 149 static void
 150 nbnxn_kernel_cpu(const PairlistSet              &pairlistSet,
 151                  const Nbnxm::KernelSetup       &kernelSetup,
 152                  nbnxn_atomdata_t               *nbat,
 153                  const interaction_const_t      &ic,
 154                  rvec                           *shiftVectors,
 155                  int                             forceFlags,
 156                  int                             clearF,
 157                  real                           *fshift,
 158                  real                           *vCoulomb,
 159                  real                           *vVdw)
 160 {
 161
 162     int                      coulkt;
 163     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 164     {
 165         coulkt = coulktRF;
 166     }
 167     else
 168     {
 169         if (kernelSetup.ewaldExclusionType == Nbnxm::EwaldExclusionType::Table)
 170         {
 171             if (ic.rcoulomb == ic.rvdw)
 172             {
 173                 coulkt = coulktTAB;
 174             }
 175             else
 176             {
 177                 coulkt = coulktTAB_TWIN;
 178             }
 179         }
 180         else
 181         {
 182             if (ic.rcoulomb == ic.rvdw)
 183             {
 184                 coulkt = coulktEWALD;
 185             }
 186             else
 187             {
 188                 coulkt = coulktEWALD_TWIN;
 189             }
 190         }
 191     }
 192
 193     const nbnxn_atomdata_t::Params &nbatParams = nbat->params();
 194
 195     int vdwkt = 0;
 196     if (ic.vdwtype == evdwCUT)
 197     {
 198         switch (ic.vdw_modifier)
 199         {
 200             case eintmodNONE:
 201             case eintmodPOTSHIFT:
 202                 switch (nbatParams.comb_rule)
 203                 {
 204                     case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
 205                     case ljcrLB:   vdwkt = vdwktLJCUT_COMBLB;   break;
 206                     case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
 207                     default:
 208                         GMX_RELEASE_ASSERT(false, "Unknown combination rule");
 209                 }
 210                 break;
 211             case eintmodFORCESWITCH:
 212                 vdwkt = vdwktLJFORCESWITCH;
 213                 break;
 214             case eintmodPOTSWITCH:
 215                 vdwkt = vdwktLJPOTSWITCH;
 216                 break;
 217             default:
 218                 GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction modifier");
 219         }
 220     }
 221     else if (ic.vdwtype == evdwPME)
 222     {
 223         if (ic.ljpme_comb_rule == eljpmeGEOM)
 224         {
 225             vdwkt = vdwktLJEWALDCOMBGEOM;
 226         }
 227         else
 228         {
 229             vdwkt = vdwktLJEWALDCOMBLB;
 230             /* At setup we (should have) selected the C reference kernel */
 231             GMX_RELEASE_ASSERT(kernelSetup.kernelType == Nbnxm::KernelType::Cpu4x4_PlainC, "Only the C reference nbnxn SIMD kernel supports LJ-PME with LB combination rules");
 232         }
 233     }
 234     else
 235     {
 236         GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction type");
 237     }
 238
 239     gmx::ArrayRef<const NbnxnPairlistCpu> pairlists = pairlistSet.cpuLists();
 240
 241     int gmx_unused                        nthreads = gmx_omp_nthreads_get(emntNonbonded);
 242 #pragma omp parallel for schedule(static) num_threads(nthreads)
 243     for (int nb = 0; nb < pairlists.ssize(); nb++)
 244     {
 245         // Presently, the kernels do not call C++ code that can throw,
 246         // so no need for a try/catch pair in this OpenMP region.
 247         nbnxn_atomdata_output_t *out = &nbat->out[nb];
 248
 249         if (clearF == enbvClearFYes)
 250         {
 251             clear_f(nbat, nb, out->f.data());
 252         }
 253
 254         real *fshift_p;
 255         if ((forceFlags & GMX_FORCE_VIRIAL) && pairlists.ssize() == 1)
 256         {
 257             fshift_p = fshift;
 258         }
 259         else
 260         {
 261             fshift_p = out->fshift.data();
 262
 263             if (clearF == enbvClearFYes)
 264             {
 265                 clear_fshift(fshift_p);
 266             }
 267         }
 268
 269         // TODO: Change to reference
 270         const NbnxnPairlistCpu *pairlist = &pairlists[nb];
 271
 272         if (!(forceFlags & GMX_FORCE_ENERGY))
 273         {
 274             /* Don't calculate energies */
 275             switch (kernelSetup.kernelType)
 276             {
 277                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 278                     nbnxn_kernel_noener_ref[coulkt][vdwkt](pairlist, nbat,
 279                                                            &ic,
 280                                                            shiftVectors,
 281                                                            out->f.data(),
 282                                                            fshift_p);
 283                     break;
 284 #ifdef GMX_NBNXN_SIMD_2XNN
 285                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 286                     nbnxm_kernel_noener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 287                                                                  &ic,
 288                                                                  shiftVectors,
 289                                                                  out->f.data(),
 290                                                                  fshift_p);
 291                     break;
 292 #endif
 293 #ifdef GMX_NBNXN_SIMD_4XN
 294                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 295                     nbnxm_kernel_noener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 296                                                                 &ic,
 297                                                                 shiftVectors,
 298                                                                 out->f.data(),
 299                                                                 fshift_p);
 300                     break;
 301 #endif
 302                 default:
 303                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 304             }
 305         }
 306         else if (out->Vvdw.size() == 1)
 307         {
 308             /* A single energy group (pair) */
 309             out->Vvdw[0] = 0;
 310             out->Vc[0]   = 0;
 311
 312             switch (kernelSetup.kernelType)
 313             {
 314                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 315                     nbnxn_kernel_ener_ref[coulkt][vdwkt](pairlist, nbat,
 316                                                          &ic,
 317                                                          shiftVectors,
 318                                                          out->f.data(),
 319                                                          fshift_p,
 320                                                          out->Vvdw.data(),
 321                                                          out->Vc.data());
 322                     break;
 323 #ifdef GMX_NBNXN_SIMD_2XNN
 324                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 325                     nbnxm_kernel_ener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 326                                                                &ic,
 327                                                                shiftVectors,
 328                                                                out->f.data(),
 329                                                                fshift_p,
 330                                                                out->Vvdw.data(),
 331                                                                out->Vc.data());
 332                     break;
 333 #endif
 334 #ifdef GMX_NBNXN_SIMD_4XN
 335                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 336                     nbnxm_kernel_ener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 337                                                               &ic,
 338                                                               shiftVectors,
 339                                                               out->f.data(),
 340                                                               fshift_p,
 341                                                               out->Vvdw.data(),
 342                                                               out->Vc.data());
 343                     break;
 344 #endif
 345                 default:
 346                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 347             }
 348         }
 349         else
 350         {
 351             /* Calculate energy group contributions */
 352             clearGroupEnergies(out);
 353
 354             int unrollj = 0;
 355
 356             switch (kernelSetup.kernelType)
 357             {
 358                 case Nbnxm::KernelType::Cpu4x4_PlainC:
 359                     unrollj = c_nbnxnCpuIClusterSize;
 360                     nbnxn_kernel_energrp_ref[coulkt][vdwkt](pairlist, nbat,
 361                                                             &ic,
 362                                                             shiftVectors,
 363                                                             out->f.data(),
 364                                                             fshift_p,
 365                                                             out->Vvdw.data(),
 366                                                             out->Vc.data());
 367                     break;
 368 #ifdef GMX_NBNXN_SIMD_2XNN
 369                 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 370                     unrollj = GMX_SIMD_REAL_WIDTH/2;
 371                     nbnxm_kernel_energrp_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
 372                                                                   &ic,
 373                                                                   shiftVectors,
 374                                                                   out->f.data(),
 375                                                                   fshift_p,
 376                                                                   out->VSvdw.data(),
 377                                                                   out->VSc.data());
 378                     break;
 379 #endif
 380 #ifdef GMX_NBNXN_SIMD_4XN
 381                 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 382                     unrollj = GMX_SIMD_REAL_WIDTH;
 383                     nbnxm_kernel_energrp_simd_4xm[coulkt][vdwkt](pairlist, nbat,
 384                                                                  &ic,
 385                                                                  shiftVectors,
 386                                                                  out->f.data(),
 387                                                                  fshift_p,
 388                                                                  out->VSvdw.data(),
 389                                                                  out->VSc.data());
 390                     break;
 391 #endif
 392                 default:
 393                     GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
 394             }
 395
 396             if (kernelSetup.kernelType != Nbnxm::KernelType::Cpu4x4_PlainC)
 397             {
 398                 switch (unrollj)
 399                 {
 400                     case 2:
 401                         reduceGroupEnergySimdBuffers<2>(nbatParams.nenergrp,
 402                                                         nbatParams.neg_2log,
 403                                                         out);
 404                         break;
 405                     case 4:
 406                         reduceGroupEnergySimdBuffers<4>(nbatParams.nenergrp,
 407                                                         nbatParams.neg_2log,
 408                                                         out);
 409                         break;
 410                     case 8:
 411                         reduceGroupEnergySimdBuffers<8>(nbatParams.nenergrp,
 412                                                         nbatParams.neg_2log,
 413                                                         out);
 414                         break;
 415                     default:
 416                         GMX_RELEASE_ASSERT(false, "Unsupported j-unroll size");
 417                 }
 418             }
 419         }
 420     }
 421
 422     if (forceFlags & GMX_FORCE_ENERGY)
 423     {
 424         reduce_energies_over_lists(nbat, pairlists.ssize(), vVdw, vCoulomb);
 425     }
 426 }
 427
 428 static void accountFlops(t_nrnb                           *nrnb,
 429                          const PairlistSet                &pairlistSet,
 430                          const nonbonded_verlet_t         &nbv,
 431                          const interaction_const_t        &ic,
 432                          const int                         forceFlags)
 433 {
 434     const bool usingGpuKernels = nbv.useGpu();
 435
 436     int        enr_nbnxn_kernel_ljc;
 437     if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
 438     {
 439         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_RF;
 440     }
 441     else if ((!usingGpuKernels && nbv.kernelSetup().ewaldExclusionType == Nbnxm::EwaldExclusionType::Analytical) ||
 442              (usingGpuKernels && Nbnxm::gpu_is_kernel_ewald_analytical(nbv.gpu_nbv)))
 443     {
 444         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_EWALD;
 445     }
 446     else
 447     {
 448         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_TAB;
 449     }
 450     int enr_nbnxn_kernel_lj = eNR_NBNXN_LJ;
 451     if (forceFlags & GMX_FORCE_ENERGY)
 452     {
 453         /* In eNR_??? the nbnxn F+E kernels are always the F kernel + 1 */
 454         enr_nbnxn_kernel_ljc += 1;
 455         enr_nbnxn_kernel_lj  += 1;
 456     }
 457
 458     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc,
 459              pairlistSet.natpair_ljq_);
 460     inc_nrnb(nrnb, enr_nbnxn_kernel_lj,
 461              pairlistSet.natpair_lj_);
 462     /* The Coulomb-only kernels are offset -eNR_NBNXN_LJ_RF+eNR_NBNXN_RF */
 463     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc-eNR_NBNXN_LJ_RF+eNR_NBNXN_RF,
 464              pairlistSet.natpair_q_);
 465
 466     const bool calcEnergy = ((forceFlags & GMX_FORCE_ENERGY) != 0);
 467     if (ic.vdw_modifier == eintmodFORCESWITCH)
 468     {
 469         /* We add up the switch cost separately */
 470         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_FSW + (calcEnergy ? 1 : 0),
 471                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 472     }
 473     if (ic.vdw_modifier == eintmodPOTSWITCH)
 474     {
 475         /* We add up the switch cost separately */
 476         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_PSW + (calcEnergy ? 1 : 0),
 477                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 478     }
 479     if (ic.vdwtype == evdwPME)
 480     {
 481         /* We add up the LJ Ewald cost separately */
 482         inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_EWALD + (calcEnergy ? 1 : 0),
 483                  pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
 484     }
 485 }
 486
 487 void
 488 nonbonded_verlet_t::dispatchNonbondedKernel(Nbnxm::InteractionLocality iLocality,
 489                                             const interaction_const_t &ic,
 490                                             int                        forceFlags,
 491                                             int                        clearF,
 492                                             t_forcerec                *fr,
 493                                             gmx_enerdata_t            *enerd,
 494                                             t_nrnb                    *nrnb)
 495 {
 496     const PairlistSet &pairlistSet = pairlistSets().pairlistSet(iLocality);
 497
 498     switch (kernelSetup().kernelType)
 499     {
 500         case Nbnxm::KernelType::Cpu4x4_PlainC:
 501         case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
 502         case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
 503             nbnxn_kernel_cpu(pairlistSet,
 504                              kernelSetup(),
 505                              nbat.get(),
 506                              ic,
 507                              fr->shift_vec,
 508                              forceFlags,
 509                              clearF,
 510                              fr->fshift[0],
 511                              enerd->grpp.ener[egCOULSR],
 512                              fr->bBHAM ?
 513                              enerd->grpp.ener[egBHAMSR] :
 514                              enerd->grpp.ener[egLJSR]);
 515             break;
 516
 517         case Nbnxm::KernelType::Gpu8x8x8:
 518             Nbnxm::gpu_launch_kernel(gpu_nbv, forceFlags, iLocality);
 519             break;
 520
 521         case Nbnxm::KernelType::Cpu8x8x8_PlainC:
 522             nbnxn_kernel_gpu_ref(pairlistSet.gpuList(),
 523                                  nbat.get(), &ic,
 524                                  fr->shift_vec,
 525                                  forceFlags,
 526                                  clearF,
 527                                  nbat->out[0].f,
 528                                  fr->fshift[0],
 529                                  enerd->grpp.ener[egCOULSR],
 530                                  fr->bBHAM ?
 531                                  enerd->grpp.ener[egBHAMSR] :
 532                                  enerd->grpp.ener[egLJSR]);
 533             break;
 534
 535         default:
 536             GMX_RELEASE_ASSERT(false, "Invalid nonbonded kernel type passed!");
 537
 538     }
 539
 540     accountFlops(nrnb, pairlistSet, *this, ic, forceFlags);
 541 }
 542
 543 void
 544 nonbonded_verlet_t::dispatchFreeEnergyKernel(Nbnxm::InteractionLocality  iLocality,
 545                                              t_forcerec                 *fr,
 546                                              rvec                        x[],
 547                                              rvec                        f[],
 548                                              const t_mdatoms            &mdatoms,
 549                                              t_lambda                   *fepvals,
 550                                              real                       *lambda,
 551                                              gmx_enerdata_t             *enerd,
 552                                              const int                   forceFlags,
 553                                              t_nrnb                     *nrnb)
 554 {
 555     const gmx::ArrayRef<t_nblist const * const > nbl_fep = pairlistSets().pairlistSet(iLocality).fepLists();
 556
 557     /* When the first list is empty, all are empty and there is nothing to do */
 558     if (!pairlistSets().params().haveFep || nbl_fep[0]->nrj == 0)
 559     {
 560         return;
 561     }
 562
 563     int donb_flags = 0;
 564     /* Add short-range interactions */
 565     donb_flags |= GMX_NONBONDED_DO_SR;
 566
 567     /* Currently all group scheme kernels always calculate (shift-)forces */
 568     if (forceFlags & GMX_FORCE_FORCES)
 569     {
 570         donb_flags |= GMX_NONBONDED_DO_FORCE;
 571     }
 572     if (forceFlags & GMX_FORCE_VIRIAL)
 573     {
 574         donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 575     }
 576     if (forceFlags & GMX_FORCE_ENERGY)
 577     {
 578         donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 579     }
 580
 581     nb_kernel_data_t kernel_data;
 582     real             dvdl_nb[efptNR] = { 0 };
 583     kernel_data.flags  = donb_flags;
 584     kernel_data.lambda = lambda;
 585     kernel_data.dvdl   = dvdl_nb;
 586
 587     kernel_data.energygrp_elec = enerd->grpp.ener[egCOULSR];
 588     kernel_data.energygrp_vdw  = enerd->grpp.ener[egLJSR];
 589
 590     GMX_ASSERT(gmx_omp_nthreads_get(emntNonbonded) == nbl_fep.ssize(), "Number of lists should be same as number of NB threads");
 591
 592 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 593     for (int th = 0; th < nbl_fep.ssize(); th++)
 594     {
 595         try
 596         {
 597             gmx_nb_free_energy_kernel(nbl_fep[th],
 598                                       x, f, fr, &mdatoms, &kernel_data, nrnb);
 599         }
 600         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 601     }
 602
 603     if (fepvals->sc_alpha != 0)
 604     {
 605         enerd->dvdl_nonlin[efptVDW]  += dvdl_nb[efptVDW];
 606         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 607     }
 608     else
 609     {
 610         enerd->dvdl_lin[efptVDW]  += dvdl_nb[efptVDW];
 611         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 612     }
 613
 614     /* If we do foreign lambda and we have soft-core interactions
 615      * we have to recalculate the (non-linear) energies contributions.
 616      */
 617     if (fepvals->n_lambda > 0 && (forceFlags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 618     {
 619         real lam_i[efptNR];
 620         kernel_data.flags          = (donb_flags & ~(GMX_NONBONDED_DO_FORCE | GMX_NONBONDED_DO_SHIFTFORCE)) | GMX_NONBONDED_DO_FOREIGNLAMBDA;
 621         kernel_data.lambda         = lam_i;
 622         kernel_data.energygrp_elec = enerd->foreign_grpp.ener[egCOULSR];
 623         kernel_data.energygrp_vdw  = enerd->foreign_grpp.ener[egLJSR];
 624         /* Note that we add to kernel_data.dvdl, but ignore the result */
 625
 626         for (int i = 0; i < enerd->n_lambda; i++)
 627         {
 628             for (int j = 0; j < efptNR; j++)
 629             {
 630                 lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 631             }
 632             reset_foreign_enerdata(enerd);
 633 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
 634             for (int th = 0; th < nbl_fep.ssize(); th++)
 635             {
 636                 try
 637                 {
 638                     gmx_nb_free_energy_kernel(nbl_fep[th],
 639                                               x, f, fr, &mdatoms, &kernel_data, nrnb);
 640                 }
 641                 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
 642             }
 643
 644             sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 645             enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 646         }
 647     }
 648 }