src/gromacs/nbnxm/benchmark/bench_setup.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2019,2020,2021, 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 /*! \internal \file
  37  * \brief
  38  * This file defines functions for setting up kernel benchmarks
  39  *
  40  * \author Berk Hess <hess@kth.se>
  41  * \ingroup module_nbnxm
  42  */
  43
  44 #include "gmxpre.h"
  45
  46 #include "bench_setup.h"
  47
  48 #include <optional>
  49
  50 #include "gromacs/gmxlib/nrnb.h"
  51 #include "gromacs/math/units.h"
  52 #include "gromacs/math/vectypes.h"
  53 #include "gromacs/mdlib/dispersioncorrection.h"
  54 #include "gromacs/mdlib/force_flags.h"
  55 #include "gromacs/mdlib/forcerec.h"
  56 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  57 #include "gromacs/mdtypes/enerdata.h"
  58 #include "gromacs/mdtypes/forcerec.h"
  59 #include "gromacs/mdtypes/interaction_const.h"
  60 #include "gromacs/mdtypes/mdatom.h"
  61 #include "gromacs/mdtypes/simulation_workload.h"
  62 #include "gromacs/nbnxm/atomdata.h"
  63 #include "gromacs/nbnxm/gridset.h"
  64 #include "gromacs/nbnxm/nbnxm.h"
  65 #include "gromacs/nbnxm/nbnxm_simd.h"
  66 #include "gromacs/nbnxm/pairlistset.h"
  67 #include "gromacs/nbnxm/pairlistsets.h"
  68 #include "gromacs/nbnxm/pairsearch.h"
  69 #include "gromacs/pbcutil/ishift.h"
  70 #include "gromacs/pbcutil/pbc.h"
  71 #include "gromacs/simd/simd.h"
  72 #include "gromacs/timing/cyclecounter.h"
  73 #include "gromacs/utility/enumerationhelpers.h"
  74 #include "gromacs/utility/fatalerror.h"
  75 #include "gromacs/utility/logger.h"
  76
  77 #include "bench_system.h"
  78
  79 namespace Nbnxm
  80 {
  81
  82 /*! \brief Checks the kernel setup
  83  *
  84  * Returns an error string when the kernel is not available.
  85  */
  86 static std::optional<std::string> checkKernelSetup(const KernelBenchOptions& options)
  87 {
  88     GMX_RELEASE_ASSERT(options.nbnxmSimd < BenchMarkKernels::Count
  89                                && options.nbnxmSimd != BenchMarkKernels::SimdAuto,
  90                        "Need a valid kernel SIMD type");
  91
  92     // Check SIMD support
  93     if ((options.nbnxmSimd != BenchMarkKernels::SimdNo && !GMX_SIMD)
  94 #ifndef GMX_NBNXN_SIMD_4XN
  95         || options.nbnxmSimd == BenchMarkKernels::Simd4XM
  96 #endif
  97 #ifndef GMX_NBNXN_SIMD_2XNN
  98         || options.nbnxmSimd == BenchMarkKernels::Simd2XMM
  99 #endif
 100     )
 101     {
 102         return "the requested SIMD kernel was not set up at configuration time";
 103     }
 104
 105     if (options.reportTime && (0 > gmx_cycles_calibrate(1.0)))
 106     {
 107         return "the -time option is not supported on this system";
 108     }
 109
 110     return {};
 111 }
 112
 113 //! Helper to translate between the different enumeration values.
 114 static KernelType translateBenchmarkEnum(const BenchMarkKernels& kernel)
 115 {
 116     int kernelInt = static_cast<int>(kernel);
 117     return static_cast<KernelType>(kernelInt);
 118 }
 119
 120 /*! \brief Returns the kernel setup
 121  */
 122 static KernelSetup getKernelSetup(const KernelBenchOptions& options)
 123 {
 124     auto messageWhenInvalid = checkKernelSetup(options);
 125     GMX_RELEASE_ASSERT(!messageWhenInvalid, "Need valid options");
 126
 127     KernelSetup kernelSetup;
 128
 129     // The int enum options.nbnxnSimd is set up to match KernelType + 1
 130     kernelSetup.kernelType = translateBenchmarkEnum(options.nbnxmSimd);
 131     // The plain-C kernel does not support analytical ewald correction
 132     if (kernelSetup.kernelType == KernelType::Cpu4x4_PlainC)
 133     {
 134         kernelSetup.ewaldExclusionType = EwaldExclusionType::Table;
 135     }
 136     else
 137     {
 138         kernelSetup.ewaldExclusionType = options.useTabulatedEwaldCorr ? EwaldExclusionType::Table
 139                                                                        : EwaldExclusionType::Analytical;
 140     }
 141
 142     return kernelSetup;
 143 }
 144
 145 //! Return an interaction constants struct with members used in the benchmark set appropriately
 146 static interaction_const_t setupInteractionConst(const KernelBenchOptions& options)
 147
 148 {
 149     interaction_const_t ic;
 150
 151     ic.vdwtype      = VanDerWaalsType::Cut;
 152     ic.vdw_modifier = InteractionModifiers::PotShift;
 153     ic.rvdw         = options.pairlistCutoff;
 154
 155     ic.eeltype = (options.coulombType == BenchMarkCoulomb::Pme ? CoulombInteractionType::Pme
 156                                                                : CoulombInteractionType::RF);
 157     ic.coulomb_modifier = InteractionModifiers::PotShift;
 158     ic.rcoulomb         = options.pairlistCutoff;
 159
 160     // Reaction-field with reactionFieldPermitivity=inf
 161     // TODO: Replace by calc_rffac() after refactoring that
 162     ic.reactionFieldCoefficient = 0.5 * std::pow(ic.rcoulomb, -3);
 163     ic.reactionFieldShift = 1 / ic.rcoulomb + ic.reactionFieldCoefficient * ic.rcoulomb * ic.rcoulomb;
 164
 165     if (EEL_PME_EWALD(ic.eeltype))
 166     {
 167         // Ewald coefficients, we ignore the potential shift
 168         GMX_RELEASE_ASSERT(options.ewaldcoeff_q > 0, "Ewald coefficient should be > 0");
 169         ic.ewaldcoeff_q       = options.ewaldcoeff_q;
 170         ic.coulombEwaldTables = std::make_unique<EwaldCorrectionTables>();
 171         init_interaction_const_tables(nullptr, &ic, 0, 0);
 172     }
 173
 174     return ic;
 175 }
 176
 177 //! Sets up and returns a Nbnxm object for the given benchmark options and system
 178 static std::unique_ptr<nonbonded_verlet_t> setupNbnxmForBenchInstance(const KernelBenchOptions& options,
 179                                                                       const gmx::BenchmarkSystem& system)
 180 {
 181     const auto pinPolicy  = (options.useGpu ? gmx::PinningPolicy::PinnedIfSupported
 182                                            : gmx::PinningPolicy::CannotBePinned);
 183     const int  numThreads = options.numThreads;
 184     // Note: the options and Nbnxm combination rule enums values should match
 185     const int combinationRule = static_cast<int>(options.ljCombinationRule);
 186
 187     auto messageWhenInvalid = checkKernelSetup(options);
 188     if (messageWhenInvalid)
 189     {
 190         gmx_fatal(FARGS, "Requested kernel is unavailable because %s.", messageWhenInvalid->c_str());
 191     }
 192     Nbnxm::KernelSetup kernelSetup = getKernelSetup(options);
 193
 194     PairlistParams pairlistParams(kernelSetup.kernelType, false, options.pairlistCutoff, false);
 195
 196     GridSet gridSet(
 197             PbcType::Xyz, false, nullptr, nullptr, pairlistParams.pairlistType, false, numThreads, pinPolicy);
 198
 199     auto pairlistSets = std::make_unique<PairlistSets>(pairlistParams, false, 0);
 200
 201     auto pairSearch = std::make_unique<PairSearch>(
 202             PbcType::Xyz, false, nullptr, nullptr, pairlistParams.pairlistType, false, numThreads, pinPolicy);
 203
 204     auto atomData = std::make_unique<nbnxn_atomdata_t>(pinPolicy,
 205                                                        gmx::MDLogger(),
 206                                                        kernelSetup.kernelType,
 207                                                        combinationRule,
 208                                                        system.numAtomTypes,
 209                                                        system.nonbondedParameters,
 210                                                        1,
 211                                                        numThreads);
 212
 213     // Put everything together
 214     auto nbv = std::make_unique<nonbonded_verlet_t>(
 215             std::move(pairlistSets), std::move(pairSearch), std::move(atomData), kernelSetup, nullptr, nullptr);
 216
 217     t_nrnb nrnb;
 218
 219     GMX_RELEASE_ASSERT(!TRICLINIC(system.box), "Only rectangular unit-cells are supported here");
 220     const rvec lowerCorner = { 0, 0, 0 };
 221     const rvec upperCorner = { system.box[XX][XX], system.box[YY][YY], system.box[ZZ][ZZ] };
 222
 223     gmx::ArrayRef<const int> atomInfo;
 224     if (options.useHalfLJOptimization)
 225     {
 226         atomInfo = system.atomInfoOxygenVdw;
 227     }
 228     else
 229     {
 230         atomInfo = system.atomInfoAllVdw;
 231     }
 232
 233     const real atomDensity = system.coordinates.size() / det(system.box);
 234
 235     nbnxn_put_on_grid(nbv.get(),
 236                       system.box,
 237                       0,
 238                       lowerCorner,
 239                       upperCorner,
 240                       nullptr,
 241                       { 0, int(system.coordinates.size()) },
 242                       atomDensity,
 243                       atomInfo,
 244                       system.coordinates,
 245                       0,
 246                       nullptr);
 247
 248     nbv->constructPairlist(gmx::InteractionLocality::Local, system.excls, 0, &nrnb);
 249
 250     nbv->setAtomProperties(system.atomTypes, system.charges, atomInfo);
 251
 252     return nbv;
 253 }
 254
 255 //! Add the options instance to the list for all requested kernel SIMD types
 256 static void expandSimdOptionAndPushBack(const KernelBenchOptions&        options,
 257                                         std::vector<KernelBenchOptions>* optionsList)
 258 {
 259     if (options.nbnxmSimd == BenchMarkKernels::SimdAuto)
 260     {
 261         bool addedInstance = false;
 262 #ifdef GMX_NBNXN_SIMD_4XN
 263         optionsList->push_back(options);
 264         optionsList->back().nbnxmSimd = BenchMarkKernels::Simd4XM;
 265         addedInstance                 = true;
 266 #endif
 267 #ifdef GMX_NBNXN_SIMD_2XNN
 268         optionsList->push_back(options);
 269         optionsList->back().nbnxmSimd = BenchMarkKernels::Simd2XMM;
 270         addedInstance                 = true;
 271 #endif
 272         if (!addedInstance)
 273         {
 274             optionsList->push_back(options);
 275             optionsList->back().nbnxmSimd = BenchMarkKernels::SimdNo;
 276         }
 277     }
 278     else
 279     {
 280         optionsList->push_back(options);
 281     }
 282 }
 283
 284 //! Sets up and runs the requested benchmark instance and prints the results
 285 //
 286 // When \p doWarmup is true runs the warmup iterations instead
 287 // of the normal ones and does not print any results
 288 static void setupAndRunInstance(const gmx::BenchmarkSystem& system,
 289                                 const KernelBenchOptions&   options,
 290                                 const bool                  doWarmup)
 291 {
 292     // Generate an, accurate, estimate of the number of non-zero pair interactions
 293     const real atomDensity = system.coordinates.size() / det(system.box);
 294     const real numPairsWithinCutoff =
 295             atomDensity * 4.0 / 3.0 * M_PI * std::pow(options.pairlistCutoff, 3);
 296     const real numUsefulPairs = system.coordinates.size() * 0.5 * (numPairsWithinCutoff + 1);
 297
 298     std::unique_ptr<nonbonded_verlet_t> nbv = setupNbnxmForBenchInstance(options, system);
 299
 300     // We set the interaction cut-off to the pairlist cut-off
 301     interaction_const_t ic = setupInteractionConst(options);
 302
 303     t_nrnb nrnb = { 0 };
 304
 305     gmx_enerdata_t enerd(1, 0);
 306
 307     gmx::StepWorkload stepWork;
 308     stepWork.computeForces = true;
 309     if (options.computeVirialAndEnergy)
 310     {
 311         stepWork.computeVirial = true;
 312         stepWork.computeEnergy = true;
 313     }
 314
 315     const gmx::EnumerationArray<BenchMarkKernels, std::string> kernelNames = {
 316         "auto", "no", "4xM", "2xMM"
 317     };
 318
 319     const gmx::EnumerationArray<BenchMarkCombRule, std::string> combruleNames = { "geom.",
 320                                                                                   "LB",
 321                                                                                   "none" };
 322
 323     if (!doWarmup)
 324     {
 325         fprintf(stdout,
 326                 "%-7s %-4s %-5s %-4s ",
 327                 options.coulombType == BenchMarkCoulomb::Pme ? "Ewald" : "RF",
 328                 options.useHalfLJOptimization ? "half" : "all",
 329                 combruleNames[options.ljCombinationRule].c_str(),
 330                 kernelNames[options.nbnxmSimd].c_str());
 331         if (!options.outputFile.empty())
 332         {
 333             fprintf(system.csv,
 334                     "\"%d\",\"%zu\",\"%g\",\"%d\",\"%d\",\"%s\",\"%s\",\"%s\",\"%s\",\"%s\",\"%"
 335                     "s\",",
 336 #if GMX_SIMD
 337                     (options.nbnxmSimd != BenchMarkKernels::SimdNo) ? GMX_SIMD_REAL_WIDTH : 0,
 338 #else
 339                     0,
 340 #endif
 341                     system.coordinates.size(),
 342                     options.pairlistCutoff,
 343                     options.numThreads,
 344                     options.numIterations,
 345                     options.computeVirialAndEnergy ? "yes" : "no",
 346                     (options.coulombType != BenchMarkCoulomb::ReactionField)
 347                             ? ((options.nbnxmSimd == BenchMarkKernels::SimdNo || options.useTabulatedEwaldCorr)
 348                                        ? "table"
 349                                        : "analytical")
 350                             : "",
 351                     options.coulombType == BenchMarkCoulomb::Pme ? "Ewald" : "RF",
 352                     options.useHalfLJOptimization ? "half" : "all",
 353                     combruleNames[options.ljCombinationRule].c_str(),
 354                     kernelNames[options.nbnxmSimd].c_str());
 355         }
 356     }
 357
 358     // Run pre-iteration to avoid cache misses
 359     for (int iter = 0; iter < options.numPreIterations; iter++)
 360     {
 361         nbv->dispatchNonbondedKernel(
 362                 gmx::InteractionLocality::Local,
 363                 ic,
 364                 stepWork,
 365                 enbvClearFYes,
 366                 system.forceRec.shift_vec,
 367                 enerd.grpp.energyGroupPairTerms[system.forceRec.haveBuckingham ? NonBondedEnergyTerms::BuckinghamSR
 368                                                                                : NonBondedEnergyTerms::LJSR],
 369                 enerd.grpp.energyGroupPairTerms[NonBondedEnergyTerms::CoulombSR],
 370                 &nrnb);
 371     }
 372
 373     const int numIterations = (doWarmup ? options.numWarmupIterations : options.numIterations);
 374     const PairlistSet& pairlistSet = nbv->pairlistSets().pairlistSet(gmx::InteractionLocality::Local);
 375     const gmx::index numPairs = pairlistSet.natpair_ljq_ + pairlistSet.natpair_lj_ + pairlistSet.natpair_q_;
 376     gmx_cycles_t cycles = gmx_cycles_read();
 377     for (int iter = 0; iter < numIterations; iter++)
 378     {
 379         // Run the kernel without force clearing
 380         nbv->dispatchNonbondedKernel(
 381                 gmx::InteractionLocality::Local,
 382                 ic,
 383                 stepWork,
 384                 enbvClearFNo,
 385                 system.forceRec.shift_vec,
 386                 enerd.grpp.energyGroupPairTerms[system.forceRec.haveBuckingham ? NonBondedEnergyTerms::BuckinghamSR
 387                                                                                : NonBondedEnergyTerms::LJSR],
 388                 enerd.grpp.energyGroupPairTerms[NonBondedEnergyTerms::CoulombSR],
 389                 &nrnb);
 390     }
 391     cycles = gmx_cycles_read() - cycles;
 392     if (!doWarmup)
 393     {
 394         if (options.reportTime)
 395         {
 396             const double uSec = static_cast<double>(cycles) * gmx_cycles_calibrate(1.0) * 1.e6;
 397             if (options.cyclesPerPair)
 398             {
 399                 fprintf(stdout,
 400                         "%13.2f %13.3f %10.3f %10.3f\n",
 401                         uSec,
 402                         uSec / options.numIterations,
 403                         uSec / (options.numIterations * numPairs),
 404                         uSec / (options.numIterations * numUsefulPairs));
 405                 if (!options.outputFile.empty())
 406                 {
 407                     fprintf(system.csv,
 408                             "\"%.3f\",\"%.4f\",\"%.4f\",\"%.4f\"\n",
 409                             uSec,
 410                             uSec / options.numIterations,
 411                             uSec / (options.numIterations * numPairs),
 412                             uSec / (options.numIterations * numUsefulPairs));
 413                 }
 414             }
 415             else
 416             {
 417                 fprintf(stdout,
 418                         "%13.2f %13.3f %10.3f %10.3f\n",
 419                         uSec,
 420                         uSec / options.numIterations,
 421                         options.numIterations * numPairs / uSec,
 422                         options.numIterations * numUsefulPairs / uSec);
 423                 if (!options.outputFile.empty())
 424                 {
 425                     fprintf(system.csv,
 426                             "\"%.3f\",\"%.4f\",\"%.4f\",\"%.4f\"\n",
 427                             uSec,
 428                             uSec / options.numIterations,
 429                             options.numIterations * numPairs / uSec,
 430                             options.numIterations * numUsefulPairs / uSec);
 431                 }
 432             }
 433         }
 434         else
 435         {
 436             const double dCycles = static_cast<double>(cycles);
 437             if (options.cyclesPerPair)
 438             {
 439                 fprintf(stdout,
 440                         "%10.3f %10.4f %8.4f %8.4f\n",
 441                         cycles * 1e-6,
 442                         dCycles / options.numIterations * 1e-6,
 443                         dCycles / (options.numIterations * numPairs),
 444                         dCycles / (options.numIterations * numUsefulPairs));
 445             }
 446             else
 447             {
 448                 fprintf(stdout,
 449                         "%10.3f %10.4f %8.4f %8.4f\n",
 450                         dCycles * 1e-6,
 451                         dCycles / options.numIterations * 1e-6,
 452                         options.numIterations * numPairs / dCycles,
 453                         options.numIterations * numUsefulPairs / dCycles);
 454             }
 455         }
 456     }
 457 }
 458
 459 void bench(const int sizeFactor, const KernelBenchOptions& options)
 460 {
 461     // We don't want to call gmx_omp_nthreads_init(), so we init what we need
 462     gmx_omp_nthreads_set(ModuleMultiThread::Pairsearch, options.numThreads);
 463     gmx_omp_nthreads_set(ModuleMultiThread::Nonbonded, options.numThreads);
 464
 465     const gmx::BenchmarkSystem system(sizeFactor, options.outputFile);
 466
 467     real minBoxSize = norm(system.box[XX]);
 468     for (int dim = YY; dim < DIM; dim++)
 469     {
 470         minBoxSize = std::min(minBoxSize, norm(system.box[dim]));
 471     }
 472     if (options.pairlistCutoff > 0.5 * minBoxSize)
 473     {
 474         gmx_fatal(FARGS, "The cut-off should be shorter than half the box size");
 475     }
 476
 477     std::vector<KernelBenchOptions> optionsList;
 478     if (options.doAll)
 479     {
 480         KernelBenchOptions                        opt = options;
 481         gmx::EnumerationWrapper<BenchMarkCoulomb> coulombIter;
 482         for (auto coulombType : coulombIter)
 483         {
 484             opt.coulombType = coulombType;
 485             for (int halfLJ = 0; halfLJ <= 1; halfLJ++)
 486             {
 487                 opt.useHalfLJOptimization = (halfLJ == 1);
 488
 489                 gmx::EnumerationWrapper<BenchMarkCombRule> combRuleIter;
 490                 for (auto combRule : combRuleIter)
 491                 {
 492                     opt.ljCombinationRule = combRule;
 493
 494                     expandSimdOptionAndPushBack(opt, &optionsList);
 495                 }
 496             }
 497         }
 498     }
 499     else
 500     {
 501         expandSimdOptionAndPushBack(options, &optionsList);
 502     }
 503     GMX_RELEASE_ASSERT(!optionsList.empty(), "Expect at least on benchmark setup");
 504
 505 #if GMX_SIMD
 506     if (options.nbnxmSimd != BenchMarkKernels::SimdNo)
 507     {
 508         fprintf(stdout, "SIMD width:           %d\n", GMX_SIMD_REAL_WIDTH);
 509     }
 510 #endif
 511     fprintf(stdout, "System size:          %zu atoms\n", system.coordinates.size());
 512     fprintf(stdout, "Cut-off radius:       %g nm\n", options.pairlistCutoff);
 513     fprintf(stdout, "Number of threads:    %d\n", options.numThreads);
 514     fprintf(stdout, "Number of iterations: %d\n", options.numIterations);
 515     fprintf(stdout, "Compute energies:     %s\n", options.computeVirialAndEnergy ? "yes" : "no");
 516     if (options.coulombType != BenchMarkCoulomb::ReactionField)
 517     {
 518         fprintf(stdout,
 519                 "Ewald excl. corr.:    %s\n",
 520                 options.nbnxmSimd == BenchMarkKernels::SimdNo || options.useTabulatedEwaldCorr
 521                         ? "table"
 522                         : "analytical");
 523     }
 524     printf("\n");
 525
 526     if (options.numWarmupIterations > 0)
 527     {
 528         setupAndRunInstance(system, optionsList[0], true);
 529     }
 530
 531     if (options.reportTime)
 532     {
 533         fprintf(stdout,
 534                 "Coulomb LJ   comb. SIMD       usec         usec/it.        %s\n",
 535                 options.cyclesPerPair ? "usec/pair" : "pairs/usec");
 536         if (!options.outputFile.empty())
 537         {
 538             fprintf(system.csv,
 539                     "\"width\",\"atoms\",\"cut-off radius\",\"threads\",\"iter\",\"compute "
 540                     "energy\",\"Ewald excl. "
 541                     "corr.\",\"Coulomb\",\"LJ\",\"comb\",\"SIMD\",\"usec\",\"usec/it\",\"total "
 542                     "pairs/usec\",\"useful pairs/usec\"\n");
 543         }
 544         fprintf(stdout,
 545                 "                                                        total      useful\n");
 546     }
 547     else
 548     {
 549         fprintf(stdout,
 550                 "Coulomb LJ   comb. SIMD    Mcycles  Mcycles/it.   %s\n",
 551                 options.cyclesPerPair ? "cycles/pair" : "pairs/cycle");
 552         if (!options.outputFile.empty())
 553         {
 554             fprintf(system.csv,
 555                     "\"width\",\"atoms\",\"cut-off radius\",\"threads\",\"iter\",\"compute "
 556                     "energy\",\"Ewald excl. "
 557                     "corr.\",\"Coulomb\",\"LJ\",\"comb\",\"SIMD\",\"Mcycles\",\"Mcycles/"
 558                     "it\",\"total "
 559                     "total cycles/pair\",\"total cycles per useful pair\"\n");
 560         }
 561         fprintf(stdout, "                                                total    useful\n");
 562     }
 563
 564     for (const auto& optionsInstance : optionsList)
 565     {
 566         setupAndRunInstance(system, optionsInstance, false);
 567     }
 568
 569     if (!options.outputFile.empty())
 570     {
 571         fclose(system.csv);
 572     }
 573 }
 574
 575 } // namespace Nbnxm