2 * This file is part of the GROMACS molecular simulation package.
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.
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.
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.
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.
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.
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.
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/gmx_omp_nthreads.h"
46 #include "gromacs/mdtypes/enerdata.h"
47 #include "gromacs/mdtypes/forceoutput.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/mdtypes/simulation_workload.h"
53 #include "gromacs/nbnxm/gpu_data_mgmt.h"
54 #include "gromacs/nbnxm/nbnxm.h"
55 #include "gromacs/simd/simd.h"
56 #include "gromacs/timing/wallcycle.h"
57 #include "gromacs/utility/gmxassert.h"
58 #include "gromacs/utility/real.h"
60 #include "kernel_common.h"
61 #include "nbnxm_simd.h"
62 #include "pairlistset.h"
63 #include "pairlistsets.h"
64 #include "kernels_reference/kernel_gpu_ref.h"
65 #define INCLUDE_KERNELFUNCTION_TABLES
66 #include "kernels_reference/kernel_ref.h"
67 #ifdef GMX_NBNXN_SIMD_2XNN
68 #include "kernels_simd_2xmm/kernels.h"
70 #ifdef GMX_NBNXN_SIMD_4XN
71 #include "kernels_simd_4xm/kernels.h"
73 #undef INCLUDE_FUNCTION_TABLES
75 /*! \brief Clears the energy group output buffers
77 * \param[in,out] out nbnxn kernel output struct
79 static void clearGroupEnergies(nbnxn_atomdata_output_t *out)
81 std::fill(out->Vvdw.begin(), out->Vvdw.end(), 0.0_real);
82 std::fill(out->Vc.begin(), out->Vc.end(), 0.0_real);
83 std::fill(out->VSvdw.begin(), out->VSvdw.end(), 0.0_real);
84 std::fill(out->VSc.begin(), out->VSc.end(), 0.0_real);
87 /*! \brief Reduce the group-pair energy buffers produced by a SIMD kernel
88 * to single terms in the output buffers.
90 * The SIMD kernels produce a large number of energy buffer in SIMD registers
91 * to avoid scattered reads and writes.
93 * \tparam unrollj The unroll size for j-particles in the SIMD kernel
94 * \param[in] numGroups The number of energy groups
95 * \param[in] numGroups_2log Log2 of numGroups, rounded up
96 * \param[in,out] out Struct with energy buffers
98 template <int unrollj> static void
99 reduceGroupEnergySimdBuffers(int numGroups,
101 nbnxn_atomdata_output_t *out)
103 const int unrollj_half = unrollj/2;
104 /* Energies are stored in SIMD registers with size 2^numGroups_2log */
105 const int numGroupsStorage = (1 << numGroups_2log);
107 const real * gmx_restrict vVdwSimd = out->VSvdw.data();
108 const real * gmx_restrict vCoulombSimd = out->VSc.data();
109 real * gmx_restrict vVdw = out->Vvdw.data();
110 real * gmx_restrict vCoulomb = out->Vc.data();
112 /* The size of the SIMD energy group buffer array is:
113 * numGroups*numGroups*numGroupsStorage*unrollj_half*simd_width
115 for (int i = 0; i < numGroups; i++)
117 for (int j1 = 0; j1 < numGroups; j1++)
119 for (int j0 = 0; j0 < numGroups; j0++)
121 int c = ((i*numGroups + j1)*numGroupsStorage + j0)*unrollj_half*unrollj;
122 for (int s = 0; s < unrollj_half; s++)
124 vVdw [i*numGroups + j0] += vVdwSimd [c + 0];
125 vVdw [i*numGroups + j1] += vVdwSimd [c + 1];
126 vCoulomb[i*numGroups + j0] += vCoulombSimd[c + 0];
127 vCoulomb[i*numGroups + j1] += vCoulombSimd[c + 1];
135 /*! \brief Dispatches the non-bonded N versus M atom cluster CPU kernels.
137 * OpenMP parallelization is performed within this function.
138 * Energy reduction, but not force and shift force reduction, is performed
139 * within this function.
141 * \param[in] pairlistSet Pairlists with local or non-local interactions to compute
142 * \param[in] kernelSetup The non-bonded kernel setup
143 * \param[in,out] nbat The atomdata for the interactions
144 * \param[in] ic Non-bonded interaction constants
145 * \param[in] shiftVectors The PBC shift vectors
146 * \param[in] stepWork Flags that tell what to compute
147 * \param[in] clearF Enum that tells if to clear the force output buffer
148 * \param[out] vCoulomb Output buffer for Coulomb energies
149 * \param[out] vVdw Output buffer for Van der Waals energies
150 * \param[in] wcycle Pointer to cycle counting data structure.
153 nbnxn_kernel_cpu(const PairlistSet &pairlistSet,
154 const Nbnxm::KernelSetup &kernelSetup,
155 nbnxn_atomdata_t *nbat,
156 const interaction_const_t &ic,
158 const gmx::StepWorkload &stepWork,
162 gmx_wallcycle *wcycle)
166 if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
172 if (kernelSetup.ewaldExclusionType == Nbnxm::EwaldExclusionType::Table)
174 if (ic.rcoulomb == ic.rvdw)
180 coulkt = coulktTAB_TWIN;
185 if (ic.rcoulomb == ic.rvdw)
187 coulkt = coulktEWALD;
191 coulkt = coulktEWALD_TWIN;
196 const nbnxn_atomdata_t::Params &nbatParams = nbat->params();
199 if (ic.vdwtype == evdwCUT)
201 switch (ic.vdw_modifier)
204 case eintmodPOTSHIFT:
205 switch (nbatParams.comb_rule)
207 case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
208 case ljcrLB: vdwkt = vdwktLJCUT_COMBLB; break;
209 case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
211 GMX_RELEASE_ASSERT(false, "Unknown combination rule");
214 case eintmodFORCESWITCH:
215 vdwkt = vdwktLJFORCESWITCH;
217 case eintmodPOTSWITCH:
218 vdwkt = vdwktLJPOTSWITCH;
221 GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction modifier");
224 else if (ic.vdwtype == evdwPME)
226 if (ic.ljpme_comb_rule == eljpmeGEOM)
228 vdwkt = vdwktLJEWALDCOMBGEOM;
232 vdwkt = vdwktLJEWALDCOMBLB;
233 /* At setup we (should have) selected the C reference kernel */
234 GMX_RELEASE_ASSERT(kernelSetup.kernelType == Nbnxm::KernelType::Cpu4x4_PlainC, "Only the C reference nbnxn SIMD kernel supports LJ-PME with LB combination rules");
239 GMX_RELEASE_ASSERT(false, "Unsupported VdW interaction type");
242 gmx::ArrayRef<const NbnxnPairlistCpu> pairlists = pairlistSet.cpuLists();
244 int gmx_unused nthreads = gmx_omp_nthreads_get(emntNonbonded);
245 wallcycle_sub_start(wcycle, ewcsNONBONDED_CLEAR);
246 #pragma omp parallel for schedule(static) num_threads(nthreads)
247 for (gmx::index nb = 0; nb < pairlists.ssize(); nb++)
249 // Presently, the kernels do not call C++ code that can throw,
250 // so no need for a try/catch pair in this OpenMP region.
251 nbnxn_atomdata_output_t *out = &nbat->out[nb];
253 if (clearF == enbvClearFYes)
255 clearForceBuffer(nbat, nb);
257 clear_fshift(out->fshift.data());
262 wallcycle_sub_stop(wcycle, ewcsNONBONDED_CLEAR);
263 wallcycle_sub_start(wcycle, ewcsNONBONDED_KERNEL);
266 // TODO: Change to reference
267 const NbnxnPairlistCpu *pairlist = &pairlists[nb];
269 if (!stepWork.computeEnergy)
271 /* Don't calculate energies */
272 switch (kernelSetup.kernelType)
274 case Nbnxm::KernelType::Cpu4x4_PlainC:
275 nbnxn_kernel_noener_ref[coulkt][vdwkt](pairlist, nbat,
280 #ifdef GMX_NBNXN_SIMD_2XNN
281 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
282 nbnxm_kernel_noener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
288 #ifdef GMX_NBNXN_SIMD_4XN
289 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
290 nbnxm_kernel_noener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
297 GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
300 else if (out->Vvdw.size() == 1)
302 /* A single energy group (pair) */
306 switch (kernelSetup.kernelType)
308 case Nbnxm::KernelType::Cpu4x4_PlainC:
309 nbnxn_kernel_ener_ref[coulkt][vdwkt](pairlist, nbat,
314 #ifdef GMX_NBNXN_SIMD_2XNN
315 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
316 nbnxm_kernel_ener_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
322 #ifdef GMX_NBNXN_SIMD_4XN
323 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
324 nbnxm_kernel_ener_simd_4xm[coulkt][vdwkt](pairlist, nbat,
331 GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
336 /* Calculate energy group contributions */
337 clearGroupEnergies(out);
341 switch (kernelSetup.kernelType)
343 case Nbnxm::KernelType::Cpu4x4_PlainC:
344 unrollj = c_nbnxnCpuIClusterSize;
345 nbnxn_kernel_energrp_ref[coulkt][vdwkt](pairlist, nbat,
350 #ifdef GMX_NBNXN_SIMD_2XNN
351 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
352 unrollj = GMX_SIMD_REAL_WIDTH/2;
353 nbnxm_kernel_energrp_simd_2xmm[coulkt][vdwkt](pairlist, nbat,
359 #ifdef GMX_NBNXN_SIMD_4XN
360 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
361 unrollj = GMX_SIMD_REAL_WIDTH;
362 nbnxm_kernel_energrp_simd_4xm[coulkt][vdwkt](pairlist, nbat,
369 GMX_RELEASE_ASSERT(false, "Unsupported kernel architecture");
372 if (kernelSetup.kernelType != Nbnxm::KernelType::Cpu4x4_PlainC)
377 reduceGroupEnergySimdBuffers<2>(nbatParams.nenergrp,
382 reduceGroupEnergySimdBuffers<4>(nbatParams.nenergrp,
387 reduceGroupEnergySimdBuffers<8>(nbatParams.nenergrp,
392 GMX_RELEASE_ASSERT(false, "Unsupported j-unroll size");
397 wallcycle_sub_stop(wcycle, ewcsNONBONDED_KERNEL);
399 if (stepWork.computeEnergy)
401 reduce_energies_over_lists(nbat, pairlists.ssize(), vVdw, vCoulomb);
405 static void accountFlops(t_nrnb *nrnb,
406 const PairlistSet &pairlistSet,
407 const nonbonded_verlet_t &nbv,
408 const interaction_const_t &ic,
409 const gmx::StepWorkload &stepWork)
411 const bool usingGpuKernels = nbv.useGpu();
413 int enr_nbnxn_kernel_ljc;
414 if (EEL_RF(ic.eeltype) || ic.eeltype == eelCUT)
416 enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_RF;
418 else if ((!usingGpuKernels && nbv.kernelSetup().ewaldExclusionType == Nbnxm::EwaldExclusionType::Analytical) ||
419 (usingGpuKernels && Nbnxm::gpu_is_kernel_ewald_analytical(nbv.gpu_nbv)))
421 enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_EWALD;
425 enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_TAB;
427 int enr_nbnxn_kernel_lj = eNR_NBNXN_LJ;
428 if (stepWork.computeEnergy)
430 /* In eNR_??? the nbnxn F+E kernels are always the F kernel + 1 */
431 enr_nbnxn_kernel_ljc += 1;
432 enr_nbnxn_kernel_lj += 1;
435 inc_nrnb(nrnb, enr_nbnxn_kernel_ljc,
436 pairlistSet.natpair_ljq_);
437 inc_nrnb(nrnb, enr_nbnxn_kernel_lj,
438 pairlistSet.natpair_lj_);
439 /* The Coulomb-only kernels are offset -eNR_NBNXN_LJ_RF+eNR_NBNXN_RF */
440 inc_nrnb(nrnb, enr_nbnxn_kernel_ljc-eNR_NBNXN_LJ_RF+eNR_NBNXN_RF,
441 pairlistSet.natpair_q_);
443 if (ic.vdw_modifier == eintmodFORCESWITCH)
445 /* We add up the switch cost separately */
446 inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_FSW + (stepWork.computeEnergy ? 1 : 0),
447 pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
449 if (ic.vdw_modifier == eintmodPOTSWITCH)
451 /* We add up the switch cost separately */
452 inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_PSW + (stepWork.computeEnergy ? 1 : 0),
453 pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
455 if (ic.vdwtype == evdwPME)
457 /* We add up the LJ Ewald cost separately */
458 inc_nrnb(nrnb, eNR_NBNXN_ADD_LJ_EWALD + (stepWork.computeEnergy ? 1 : 0),
459 pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_);
464 nonbonded_verlet_t::dispatchNonbondedKernel(gmx::InteractionLocality iLocality,
465 const interaction_const_t &ic,
466 const gmx::StepWorkload &stepWork,
468 const t_forcerec &fr,
469 gmx_enerdata_t *enerd,
472 const PairlistSet &pairlistSet = pairlistSets().pairlistSet(iLocality);
474 switch (kernelSetup().kernelType)
476 case Nbnxm::KernelType::Cpu4x4_PlainC:
477 case Nbnxm::KernelType::Cpu4xN_Simd_4xN:
478 case Nbnxm::KernelType::Cpu4xN_Simd_2xNN:
479 nbnxn_kernel_cpu(pairlistSet,
486 enerd->grpp.ener[egCOULSR].data(),
488 enerd->grpp.ener[egBHAMSR].data() :
489 enerd->grpp.ener[egLJSR].data(),
493 case Nbnxm::KernelType::Gpu8x8x8:
494 Nbnxm::gpu_launch_kernel(gpu_nbv, stepWork, iLocality);
497 case Nbnxm::KernelType::Cpu8x8x8_PlainC:
498 nbnxn_kernel_gpu_ref(pairlistSet.gpuList(),
504 nbat->out[0].fshift.data(),
505 enerd->grpp.ener[egCOULSR].data(),
507 enerd->grpp.ener[egBHAMSR].data() :
508 enerd->grpp.ener[egLJSR].data());
512 GMX_RELEASE_ASSERT(false, "Invalid nonbonded kernel type passed!");
516 accountFlops(nrnb, pairlistSet, *this, ic, stepWork);
520 nonbonded_verlet_t::dispatchFreeEnergyKernel(gmx::InteractionLocality iLocality,
521 const t_forcerec *fr,
523 gmx::ForceWithShiftForces *forceWithShiftForces,
524 const t_mdatoms &mdatoms,
527 gmx_enerdata_t *enerd,
528 const gmx::StepWorkload &stepWork,
531 const auto nbl_fep = pairlistSets().pairlistSet(iLocality).fepLists();
533 /* When the first list is empty, all are empty and there is nothing to do */
534 if (!pairlistSets().params().haveFep || nbl_fep[0]->nrj == 0)
540 /* Add short-range interactions */
541 donb_flags |= GMX_NONBONDED_DO_SR;
543 if (stepWork.computeForces)
545 donb_flags |= GMX_NONBONDED_DO_FORCE;
547 if (stepWork.computeVirial)
549 donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
551 if (stepWork.computeEnergy)
553 donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
556 nb_kernel_data_t kernel_data;
557 real dvdl_nb[efptNR] = { 0 };
558 kernel_data.flags = donb_flags;
559 kernel_data.lambda = lambda;
560 kernel_data.dvdl = dvdl_nb;
562 kernel_data.energygrp_elec = enerd->grpp.ener[egCOULSR].data();
563 kernel_data.energygrp_vdw = enerd->grpp.ener[egLJSR].data();
565 GMX_ASSERT(gmx_omp_nthreads_get(emntNonbonded) == nbl_fep.ssize(), "Number of lists should be same as number of NB threads");
567 wallcycle_sub_start(wcycle_, ewcsNONBONDED_FEP);
568 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
569 for (gmx::index th = 0; th < nbl_fep.ssize(); th++)
573 gmx_nb_free_energy_kernel(nbl_fep[th].get(),
574 x, forceWithShiftForces,
575 fr, &mdatoms, &kernel_data, nrnb);
577 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
580 if (fepvals->sc_alpha != 0)
582 enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
583 enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
587 enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
588 enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
591 /* If we do foreign lambda and we have soft-core interactions
592 * we have to recalculate the (non-linear) energies contributions.
594 if (fepvals->n_lambda > 0 && stepWork.computeDhdl && fepvals->sc_alpha != 0)
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 */
603 for (size_t i = 0; i < enerd->enerpart_lambda.size(); i++)
605 for (int j = 0; j < efptNR; j++)
607 lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
609 reset_foreign_enerdata(enerd);
610 #pragma omp parallel for schedule(static) num_threads(nbl_fep.ssize())
611 for (gmx::index th = 0; th < nbl_fep.ssize(); th++)
615 gmx_nb_free_energy_kernel(nbl_fep[th].get(),
616 x, forceWithShiftForces,
617 fr, &mdatoms, &kernel_data, nrnb);
619 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
622 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
623 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
626 wallcycle_sub_stop(wcycle_, ewcsNONBONDED_FEP);