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