2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2008, The GROMACS development team,
6 * check out http://www.gromacs.org for more information.
7 * Copyright (c) 2012,2013, by the GROMACS development team, led by
8 * David van der Spoel, Berk Hess, Erik Lindahl, and including many
9 * others, as listed in the AUTHORS file in the top-level source
10 * directory and at http://www.gromacs.org.
12 * GROMACS is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public License
14 * as published by the Free Software Foundation; either version 2.1
15 * of the License, or (at your option) any later version.
17 * GROMACS is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with GROMACS; if not, see
24 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
25 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
27 * If you want to redistribute modifications to GROMACS, please
28 * consider that scientific software is very special. Version
29 * control is crucial - bugs must be traceable. We will be happy to
30 * consider code for inclusion in the official distribution, but
31 * derived work must not be called official GROMACS. Details are found
32 * in the README & COPYING files - if they are missing, get the
33 * official version at http://www.gromacs.org.
35 * To help us fund GROMACS development, we humbly ask that you cite
36 * the research papers on the package. Check out http://www.gromacs.org.
44 #ifdef HAVE_SYS_TIME_H
60 #include "checkpoint.h"
67 /* Enum for situations that can occur during log file parsing, the
68 * corresponding string entries can be found in do_the_tests() in
69 * const char* ParseLog[] */
75 eParselogResetProblem,
86 int nPMEnodes; /* number of PME-only nodes used in this test */
87 int nx, ny, nz; /* DD grid */
88 int guessPME; /* if nPMEnodes == -1, this is the guessed number of PME nodes */
89 double *Gcycles; /* This can contain more than one value if doing multiple tests */
93 float *PME_f_load; /* PME mesh/force load average*/
94 float PME_f_load_Av; /* Average average ;) ... */
95 char *mdrun_cmd_line; /* Mdrun command line used for this test */
101 int nr_inputfiles; /* The number of tpr and mdp input files */
102 gmx_large_int_t orig_sim_steps; /* Number of steps to be done in the real simulation */
103 gmx_large_int_t orig_init_step; /* Init step for the real simulation */
104 real *rcoulomb; /* The coulomb radii [0...nr_inputfiles] */
105 real *rvdw; /* The vdW radii */
106 real *rlist; /* Neighbourlist cutoff radius */
108 int *nkx, *nky, *nkz;
109 real *fsx, *fsy, *fsz; /* Fourierspacing in x,y,z dimension */
113 static void sep_line(FILE *fp)
115 fprintf(fp, "\n------------------------------------------------------------\n");
119 /* Wrapper for system calls */
120 static int gmx_system_call(char *command)
123 gmx_fatal(FARGS, "No calls to system(3) supported on this platform. Attempted to call:\n'%s'\n", command);
125 return ( system(command) );
130 /* Check if string starts with substring */
131 static gmx_bool str_starts(const char *string, const char *substring)
133 return ( strncmp(string, substring, strlen(substring)) == 0);
137 static void cleandata(t_perf *perfdata, int test_nr)
139 perfdata->Gcycles[test_nr] = 0.0;
140 perfdata->ns_per_day[test_nr] = 0.0;
141 perfdata->PME_f_load[test_nr] = 0.0;
147 static gmx_bool is_equal(real a, real b)
149 real diff, eps = 1.0e-7;
170 static void finalize(const char *fn_out)
176 fp = fopen(fn_out, "r");
177 fprintf(stdout, "\n\n");
179 while (fgets(buf, STRLEN-1, fp) != NULL)
181 fprintf(stdout, "%s", buf);
184 fprintf(stdout, "\n\n");
189 eFoundNothing, eFoundDDStr, eFoundAccountingStr, eFoundCycleStr
192 static int parse_logfile(const char *logfile, const char *errfile,
193 t_perf *perfdata, int test_nr, int presteps, gmx_large_int_t cpt_steps,
197 char line[STRLEN], dumstring[STRLEN], dumstring2[STRLEN];
198 const char matchstrdd[] = "Domain decomposition grid";
199 const char matchstrcr[] = "resetting all time and cycle counters";
200 const char matchstrbal[] = "Average PME mesh/force load:";
201 const char matchstring[] = "R E A L C Y C L E A N D T I M E A C C O U N T I N G";
202 const char errSIG[] = "signal, stopping at the next";
205 float dum1, dum2, dum3, dum4;
208 gmx_large_int_t resetsteps = -1;
209 gmx_bool bFoundResetStr = FALSE;
210 gmx_bool bResetChecked = FALSE;
213 if (!gmx_fexist(logfile))
215 fprintf(stderr, "WARNING: Could not find logfile %s.\n", logfile);
216 cleandata(perfdata, test_nr);
217 return eParselogNotFound;
220 fp = fopen(logfile, "r");
221 perfdata->PME_f_load[test_nr] = -1.0;
222 perfdata->guessPME = -1;
224 iFound = eFoundNothing;
227 iFound = eFoundDDStr; /* Skip some case statements */
230 while (fgets(line, STRLEN, fp) != NULL)
232 /* Remove leading spaces */
235 /* Check for TERM and INT signals from user: */
236 if (strstr(line, errSIG) != NULL)
239 cleandata(perfdata, test_nr);
240 return eParselogTerm;
243 /* Check whether cycle resetting worked */
244 if (presteps > 0 && !bFoundResetStr)
246 if (strstr(line, matchstrcr) != NULL)
248 sprintf(dumstring, "step %s", gmx_large_int_pfmt);
249 sscanf(line, dumstring, &resetsteps);
250 bFoundResetStr = TRUE;
251 if (resetsteps == presteps+cpt_steps)
253 bResetChecked = TRUE;
257 sprintf(dumstring, gmx_large_int_pfmt, resetsteps);
258 sprintf(dumstring2, gmx_large_int_pfmt, presteps+cpt_steps);
259 fprintf(stderr, "WARNING: Time step counters were reset at step %s,\n"
260 " though they were supposed to be reset at step %s!\n",
261 dumstring, dumstring2);
266 /* Look for strings that appear in a certain order in the log file: */
270 /* Look for domain decomp grid and separate PME nodes: */
271 if (str_starts(line, matchstrdd))
273 sscanf(line, "Domain decomposition grid %d x %d x %d, separate PME nodes %d",
274 &(perfdata->nx), &(perfdata->ny), &(perfdata->nz), &npme);
275 if (perfdata->nPMEnodes == -1)
277 perfdata->guessPME = npme;
279 else if (perfdata->nPMEnodes != npme)
281 gmx_fatal(FARGS, "PME nodes from command line and output file are not identical");
283 iFound = eFoundDDStr;
285 /* Catch a few errors that might have occured: */
286 else if (str_starts(line, "There is no domain decomposition for"))
289 return eParselogNoDDGrid;
291 else if (str_starts(line, "reading tpx file"))
294 return eParselogTPXVersion;
296 else if (str_starts(line, "The -dd or -npme option request a parallel simulation"))
299 return eParselogNotParallel;
303 /* Look for PME mesh/force balance (not necessarily present, though) */
304 if (str_starts(line, matchstrbal))
306 sscanf(&line[strlen(matchstrbal)], "%f", &(perfdata->PME_f_load[test_nr]));
308 /* Look for matchstring */
309 if (str_starts(line, matchstring))
311 iFound = eFoundAccountingStr;
314 case eFoundAccountingStr:
315 /* Already found matchstring - look for cycle data */
316 if (str_starts(line, "Total "))
318 sscanf(line, "Total %d %lf", &procs, &(perfdata->Gcycles[test_nr]));
319 iFound = eFoundCycleStr;
323 /* Already found cycle data - look for remaining performance info and return */
324 if (str_starts(line, "Performance:"))
326 ndum = sscanf(line, "%s %f %f %f %f", dumstring, &dum1, &dum2, &dum3, &dum4);
327 /* (ns/day) is the second last entry, depending on whether GMX_DETAILED_PERF_STATS was set in print_perf(), nrnb.c */
328 perfdata->ns_per_day[test_nr] = (ndum == 5) ? dum3 : dum1;
330 if (bResetChecked || presteps == 0)
336 return eParselogResetProblem;
343 /* Close the log file */
346 /* Check why there is no performance data in the log file.
347 * Did a fatal errors occur? */
348 if (gmx_fexist(errfile))
350 fp = fopen(errfile, "r");
351 while (fgets(line, STRLEN, fp) != NULL)
353 if (str_starts(line, "Fatal error:") )
355 if (fgets(line, STRLEN, fp) != NULL)
357 fprintf(stderr, "\nWARNING: An error occured during this benchmark:\n"
361 cleandata(perfdata, test_nr);
362 return eParselogFatal;
369 fprintf(stderr, "WARNING: Could not find stderr file %s.\n", errfile);
372 /* Giving up ... we could not find out why there is no performance data in
374 fprintf(stdout, "No performance data in log file.\n");
375 cleandata(perfdata, test_nr);
377 return eParselogNoPerfData;
381 static gmx_bool analyze_data(
390 int *index_tpr, /* OUT: Nr of mdp file with best settings */
391 int *npme_optimal) /* OUT: Optimal number of PME nodes */
394 int line = 0, line_win = -1;
395 int k_win = -1, i_win = -1, winPME;
396 double s = 0.0; /* standard deviation */
399 char str_PME_f_load[13];
400 gmx_bool bCanUseOrigTPR;
401 gmx_bool bRefinedCoul, bRefinedVdW, bRefinedGrid;
407 fprintf(fp, "Summary of successful runs:\n");
408 fprintf(fp, "Line tpr PME nodes Gcycles Av. Std.dev. ns/day PME/f");
411 fprintf(fp, " DD grid");
417 for (k = 0; k < ntprs; k++)
419 for (i = 0; i < ntests; i++)
421 /* Select the right dataset: */
422 pd = &(perfdata[k][i]);
424 pd->Gcycles_Av = 0.0;
425 pd->PME_f_load_Av = 0.0;
426 pd->ns_per_day_Av = 0.0;
428 if (pd->nPMEnodes == -1)
430 sprintf(strbuf, "(%3d)", pd->guessPME);
434 sprintf(strbuf, " ");
437 /* Get the average run time of a setting */
438 for (j = 0; j < nrepeats; j++)
440 pd->Gcycles_Av += pd->Gcycles[j];
441 pd->PME_f_load_Av += pd->PME_f_load[j];
443 pd->Gcycles_Av /= nrepeats;
444 pd->PME_f_load_Av /= nrepeats;
446 for (j = 0; j < nrepeats; j++)
448 if (pd->ns_per_day[j] > 0.0)
450 pd->ns_per_day_Av += pd->ns_per_day[j];
454 /* Somehow the performance number was not aquired for this run,
455 * therefor set the average to some negative value: */
456 pd->ns_per_day_Av = -1.0f*nrepeats;
460 pd->ns_per_day_Av /= nrepeats;
463 if (pd->PME_f_load_Av > 0.0)
465 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load_Av);
469 sprintf(str_PME_f_load, "%s", " - ");
473 /* We assume we had a successful run if both averages are positive */
474 if (pd->Gcycles_Av > 0.0 && pd->ns_per_day_Av > 0.0)
476 /* Output statistics if repeats were done */
479 /* Calculate the standard deviation */
481 for (j = 0; j < nrepeats; j++)
483 s += pow( pd->Gcycles[j] - pd->Gcycles_Av, 2 );
488 fprintf(fp, "%4d %3d %4d%s %12.3f %12.3f %12.3f %s",
489 line, k, pd->nPMEnodes, strbuf, pd->Gcycles_Av, s,
490 pd->ns_per_day_Av, str_PME_f_load);
493 fprintf(fp, " %3d %3d %3d", pd->nx, pd->ny, pd->nz);
497 /* Store the index of the best run found so far in 'winner': */
498 if ( (k_win == -1) || (pd->Gcycles_Av < perfdata[k_win][i_win].Gcycles_Av) )
511 gmx_fatal(FARGS, "None of the runs was successful! Check %s for problems.", fn);
516 winPME = perfdata[k_win][i_win].nPMEnodes;
520 /* We stuck to a fixed number of PME-only nodes */
521 sprintf(strbuf, "settings No. %d", k_win);
525 /* We have optimized the number of PME-only nodes */
528 sprintf(strbuf, "%s", "the automatic number of PME nodes");
532 sprintf(strbuf, "%d PME nodes", winPME);
535 fprintf(fp, "Best performance was achieved with %s", strbuf);
536 if ((nrepeats > 1) && (ntests > 1))
538 fprintf(fp, " (see line %d)", line_win);
542 /* Only mention settings if they were modified: */
543 bRefinedCoul = !is_equal(info->rcoulomb[k_win], info->rcoulomb[0]);
544 bRefinedVdW = !is_equal(info->rvdw[k_win], info->rvdw[0] );
545 bRefinedGrid = !(info->nkx[k_win] == info->nkx[0] &&
546 info->nky[k_win] == info->nky[0] &&
547 info->nkz[k_win] == info->nkz[0]);
549 if (bRefinedCoul || bRefinedVdW || bRefinedGrid)
551 fprintf(fp, "Optimized PME settings:\n");
552 bCanUseOrigTPR = FALSE;
556 bCanUseOrigTPR = TRUE;
561 fprintf(fp, " New Coulomb radius: %f nm (was %f nm)\n", info->rcoulomb[k_win], info->rcoulomb[0]);
566 fprintf(fp, " New Van der Waals radius: %f nm (was %f nm)\n", info->rvdw[k_win], info->rvdw[0]);
571 fprintf(fp, " New Fourier grid xyz: %d %d %d (was %d %d %d)\n", info->nkx[k_win], info->nky[k_win], info->nkz[k_win],
572 info->nkx[0], info->nky[0], info->nkz[0]);
575 if (bCanUseOrigTPR && ntprs > 1)
577 fprintf(fp, "and original PME settings.\n");
582 /* Return the index of the mdp file that showed the highest performance
583 * and the optimal number of PME nodes */
585 *npme_optimal = winPME;
587 return bCanUseOrigTPR;
591 /* Get the commands we need to set up the runs from environment variables */
592 static void get_program_paths(gmx_bool bThreads, char *cmd_mpirun[], char cmd_np[],
593 char *cmd_mdrun[], int repeats)
595 char *command = NULL;
600 const char def_mpirun[] = "mpirun";
601 const char def_mdrun[] = "mdrun";
602 const char filename[] = "benchtest.log";
604 /* This string should always be identical to the one in copyrite.c,
605 * gmx_print_version_info() in the defined(GMX_MPI) section */
606 const char match_mpi[] = "MPI library: MPI";
607 const char match_mdrun[] = "Program: ";
608 const char empty_mpirun[] = "";
609 gmx_bool bMdrun = FALSE;
610 gmx_bool bMPI = FALSE;
613 /* Get the commands we need to set up the runs from environment variables */
616 if ( (cp = getenv("MPIRUN")) != NULL)
618 *cmd_mpirun = strdup(cp);
622 *cmd_mpirun = strdup(def_mpirun);
627 *cmd_mpirun = strdup(empty_mpirun);
630 if ( (cp = getenv("MDRUN" )) != NULL)
632 *cmd_mdrun = strdup(cp);
636 *cmd_mdrun = strdup(def_mdrun);
640 /* If no simulations have to be performed, we are done here */
646 /* Run a small test to see whether mpirun + mdrun work */
647 fprintf(stdout, "Making sure that mdrun can be executed. ");
650 snew(command, strlen(*cmd_mdrun) + strlen(cmd_np) + strlen(filename) + 50);
651 sprintf(command, "%s%s-version -maxh 0.001 1> %s 2>&1", *cmd_mdrun, cmd_np, filename);
655 snew(command, strlen(*cmd_mpirun) + strlen(cmd_np) + strlen(*cmd_mdrun) + strlen(filename) + 50);
656 sprintf(command, "%s%s%s -version -maxh 0.001 1> %s 2>&1", *cmd_mpirun, cmd_np, *cmd_mdrun, filename);
658 fprintf(stdout, "Trying '%s' ... ", command);
659 make_backup(filename);
660 gmx_system_call(command);
662 /* Check if we find the characteristic string in the output: */
663 if (!gmx_fexist(filename))
665 gmx_fatal(FARGS, "Output from test run could not be found.");
668 fp = fopen(filename, "r");
669 /* We need to scan the whole output file, since sometimes the queuing system
670 * also writes stuff to stdout/err */
673 cp2 = fgets(line, STRLEN, fp);
676 if (str_starts(line, match_mdrun) )
680 if (str_starts(line, match_mpi) )
692 gmx_fatal(FARGS, "Need a threaded version of mdrun. This one\n"
694 "seems to have been compiled with MPI instead.",
702 gmx_fatal(FARGS, "Need an MPI-enabled version of mdrun. This one\n"
704 "seems to have been compiled without MPI support.",
711 gmx_fatal(FARGS, "Cannot execute mdrun. Please check %s for problems!",
715 fprintf(stdout, "passed.\n");
723 static void launch_simulation(
724 gmx_bool bLaunch, /* Should the simulation be launched? */
725 FILE *fp, /* General log file */
726 gmx_bool bThreads, /* whether to use threads */
727 char *cmd_mpirun, /* Command for mpirun */
728 char *cmd_np, /* Switch for -np or -ntmpi or empty */
729 char *cmd_mdrun, /* Command for mdrun */
730 char *args_for_mdrun, /* Arguments for mdrun */
731 const char *simulation_tpr, /* This tpr will be simulated */
732 int nPMEnodes) /* Number of PME nodes to use */
737 /* Make enough space for the system call command,
738 * (100 extra chars for -npme ... etc. options should suffice): */
739 snew(command, strlen(cmd_mpirun)+strlen(cmd_mdrun)+strlen(cmd_np)+strlen(args_for_mdrun)+strlen(simulation_tpr)+100);
741 /* Note that the -passall options requires args_for_mdrun to be at the end
742 * of the command line string */
745 sprintf(command, "%s%s-npme %d -s %s %s",
746 cmd_mdrun, cmd_np, nPMEnodes, simulation_tpr, args_for_mdrun);
750 sprintf(command, "%s%s%s -npme %d -s %s %s",
751 cmd_mpirun, cmd_np, cmd_mdrun, nPMEnodes, simulation_tpr, args_for_mdrun);
754 fprintf(fp, "%s this command line to launch the simulation:\n\n%s", bLaunch ? "Using" : "Please use", command);
758 /* Now the real thing! */
761 fprintf(stdout, "\nLaunching simulation with best parameters now.\nExecuting '%s'", command);
764 gmx_system_call(command);
769 static void modify_PMEsettings(
770 gmx_large_int_t simsteps, /* Set this value as number of time steps */
771 gmx_large_int_t init_step, /* Set this value as init_step */
772 const char *fn_best_tpr, /* tpr file with the best performance */
773 const char *fn_sim_tpr) /* name of tpr file to be launched */
781 read_tpx_state(fn_best_tpr, ir, &state, NULL, &mtop);
783 /* Reset nsteps and init_step to the value of the input .tpr file */
784 ir->nsteps = simsteps;
785 ir->init_step = init_step;
787 /* Write the tpr file which will be launched */
788 sprintf(buf, "Writing optimized simulation file %s with nsteps=%s.\n", fn_sim_tpr, gmx_large_int_pfmt);
789 fprintf(stdout, buf, ir->nsteps);
791 write_tpx_state(fn_sim_tpr, ir, &state, &mtop);
797 #define EPME_SWITCHED(e) ((e) == eelPMESWITCH || (e) == eelPMEUSERSWITCH)
799 /* Make additional TPR files with more computational load for the
800 * direct space processors: */
801 static void make_benchmark_tprs(
802 const char *fn_sim_tpr, /* READ : User-provided tpr file */
803 char *fn_bench_tprs[], /* WRITE: Names of benchmark tpr files */
804 gmx_large_int_t benchsteps, /* Number of time steps for benchmark runs */
805 gmx_large_int_t statesteps, /* Step counter in checkpoint file */
806 real rmin, /* Minimal Coulomb radius */
807 real rmax, /* Maximal Coulomb radius */
808 real bScaleRvdw, /* Scale rvdw along with rcoulomb */
809 int *ntprs, /* No. of TPRs to write, each with a different
810 rcoulomb and fourierspacing */
811 t_inputinfo *info, /* Contains information about mdp file options */
812 FILE *fp) /* Write the output here */
818 real nlist_buffer; /* Thickness of the buffer regions for PME-switch potentials */
821 gmx_bool bNote = FALSE;
822 real add; /* Add this to rcoul for the next test */
823 real fac = 1.0; /* Scaling factor for Coulomb radius */
824 real fourierspacing; /* Basic fourierspacing from tpr */
827 sprintf(buf, "Making benchmark tpr file%s with %s time step%s",
828 *ntprs > 1 ? "s" : "", gmx_large_int_pfmt, benchsteps > 1 ? "s" : "");
829 fprintf(stdout, buf, benchsteps);
832 sprintf(buf, " (adding %s steps from checkpoint file)", gmx_large_int_pfmt);
833 fprintf(stdout, buf, statesteps);
834 benchsteps += statesteps;
836 fprintf(stdout, ".\n");
840 read_tpx_state(fn_sim_tpr, ir, &state, NULL, &mtop);
842 /* Check if some kind of PME was chosen */
843 if (EEL_PME(ir->coulombtype) == FALSE)
845 gmx_fatal(FARGS, "Can only do optimizations for simulations with %s electrostatics.",
849 /* Check if rcoulomb == rlist, which is necessary for plain PME. */
850 if ( (ir->cutoff_scheme != ecutsVERLET) &&
851 (eelPME == ir->coulombtype) && !(ir->rcoulomb == ir->rlist))
853 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to rlist (%f).",
854 EELTYPE(eelPME), ir->rcoulomb, ir->rlist);
856 /* For other PME types, rcoulomb is allowed to be smaller than rlist */
857 else if (ir->rcoulomb > ir->rlist)
859 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to or smaller than rlist (%f)",
860 EELTYPE(ir->coulombtype), ir->rcoulomb, ir->rlist);
863 if (bScaleRvdw && ir->rvdw != ir->rcoulomb)
865 fprintf(stdout, "NOTE: input rvdw != rcoulomb, will not scale rvdw\n");
869 /* Reduce the number of steps for the benchmarks */
870 info->orig_sim_steps = ir->nsteps;
871 ir->nsteps = benchsteps;
872 /* We must not use init_step from the input tpr file for the benchmarks */
873 info->orig_init_step = ir->init_step;
876 /* For PME-switch potentials, keep the radial distance of the buffer region */
877 nlist_buffer = ir->rlist - ir->rcoulomb;
879 /* Determine length of triclinic box vectors */
880 for (d = 0; d < DIM; d++)
883 for (i = 0; i < DIM; i++)
885 box_size[d] += state.box[d][i]*state.box[d][i];
887 box_size[d] = sqrt(box_size[d]);
890 if (ir->fourier_spacing > 0)
892 info->fsx[0] = ir->fourier_spacing;
893 info->fsy[0] = ir->fourier_spacing;
894 info->fsz[0] = ir->fourier_spacing;
898 /* Reconstruct fourierspacing per dimension from the number of grid points and box size */
899 info->fsx[0] = box_size[XX]/ir->nkx;
900 info->fsy[0] = box_size[YY]/ir->nky;
901 info->fsz[0] = box_size[ZZ]/ir->nkz;
904 /* If no value for the fourierspacing was provided on the command line, we
905 * use the reconstruction from the tpr file */
906 if (ir->fourier_spacing > 0)
908 /* Use the spacing from the tpr */
909 fourierspacing = ir->fourier_spacing;
913 /* Use the maximum observed spacing */
914 fourierspacing = max(max(info->fsx[0], info->fsy[0]), info->fsz[0]);
917 fprintf(stdout, "Calculating PME grid points on the basis of a fourierspacing of %f nm\n", fourierspacing);
919 /* For performance comparisons the number of particles is useful to have */
920 fprintf(fp, " Number of particles : %d\n", mtop.natoms);
922 /* Print information about settings of which some are potentially modified: */
923 fprintf(fp, " Coulomb type : %s\n", EELTYPE(ir->coulombtype));
924 fprintf(fp, " Grid spacing x y z : %f %f %f\n",
925 box_size[XX]/ir->nkx, box_size[YY]/ir->nky, box_size[ZZ]/ir->nkz);
926 fprintf(fp, " Van der Waals type : %s\n", EVDWTYPE(ir->vdwtype));
927 if (EVDW_SWITCHED(ir->vdwtype))
929 fprintf(fp, " rvdw_switch : %f nm\n", ir->rvdw_switch);
931 if (EPME_SWITCHED(ir->coulombtype))
933 fprintf(fp, " rlist : %f nm\n", ir->rlist);
935 if (ir->rlistlong != max_cutoff(ir->rvdw, ir->rcoulomb))
937 fprintf(fp, " rlistlong : %f nm\n", ir->rlistlong);
940 /* Print a descriptive line about the tpr settings tested */
941 fprintf(fp, "\nWill try these real/reciprocal workload settings:\n");
942 fprintf(fp, " No. scaling rcoulomb");
943 fprintf(fp, " nkx nky nkz");
944 fprintf(fp, " spacing");
945 if (evdwCUT == ir->vdwtype)
947 fprintf(fp, " rvdw");
949 if (EPME_SWITCHED(ir->coulombtype))
951 fprintf(fp, " rlist");
953 if (ir->rlistlong != max_cutoff(ir->rlist, max_cutoff(ir->rvdw, ir->rcoulomb)) )
955 fprintf(fp, " rlistlong");
957 fprintf(fp, " tpr file\n");
959 /* Loop to create the requested number of tpr input files */
960 for (j = 0; j < *ntprs; j++)
962 /* The first .tpr is the provided one, just need to modify nsteps,
963 * so skip the following block */
966 /* Determine which Coulomb radii rc to use in the benchmarks */
967 add = (rmax-rmin)/(*ntprs-1);
968 if (is_equal(rmin, info->rcoulomb[0]))
970 ir->rcoulomb = rmin + j*add;
972 else if (is_equal(rmax, info->rcoulomb[0]))
974 ir->rcoulomb = rmin + (j-1)*add;
978 /* rmin != rcoul != rmax, ergo test between rmin and rmax */
979 add = (rmax-rmin)/(*ntprs-2);
980 ir->rcoulomb = rmin + (j-1)*add;
983 /* Determine the scaling factor fac */
984 fac = ir->rcoulomb/info->rcoulomb[0];
986 /* Scale the Fourier grid spacing */
987 ir->nkx = ir->nky = ir->nkz = 0;
988 calc_grid(NULL, state.box, fourierspacing*fac, &ir->nkx, &ir->nky, &ir->nkz);
990 /* Adjust other radii since various conditions neet to be fulfilled */
991 if (eelPME == ir->coulombtype)
993 /* plain PME, rcoulomb must be equal to rlist */
994 ir->rlist = ir->rcoulomb;
998 /* rlist must be >= rcoulomb, we keep the size of the buffer region */
999 ir->rlist = ir->rcoulomb + nlist_buffer;
1002 if (bScaleRvdw && evdwCUT == ir->vdwtype)
1004 /* For vdw cutoff, rvdw >= rlist */
1005 ir->rvdw = max(info->rvdw[0], ir->rlist);
1008 ir->rlistlong = max_cutoff(ir->rlist, max_cutoff(ir->rvdw, ir->rcoulomb));
1010 } /* end of "if (j != 0)" */
1012 /* for j==0: Save the original settings
1013 * for j >0: Save modified radii and Fourier grids */
1014 info->rcoulomb[j] = ir->rcoulomb;
1015 info->rvdw[j] = ir->rvdw;
1016 info->nkx[j] = ir->nkx;
1017 info->nky[j] = ir->nky;
1018 info->nkz[j] = ir->nkz;
1019 info->rlist[j] = ir->rlist;
1020 info->rlistlong[j] = ir->rlistlong;
1021 info->fsx[j] = fac*fourierspacing;
1022 info->fsy[j] = fac*fourierspacing;
1023 info->fsz[j] = fac*fourierspacing;
1025 /* Write the benchmark tpr file */
1026 strncpy(fn_bench_tprs[j], fn_sim_tpr, strlen(fn_sim_tpr)-strlen(".tpr"));
1027 sprintf(buf, "_bench%.2d.tpr", j);
1028 strcat(fn_bench_tprs[j], buf);
1029 fprintf(stdout, "Writing benchmark tpr %s with nsteps=", fn_bench_tprs[j]);
1030 fprintf(stdout, gmx_large_int_pfmt, ir->nsteps);
1033 fprintf(stdout, ", scaling factor %f\n", fac);
1037 fprintf(stdout, ", unmodified settings\n");
1040 write_tpx_state(fn_bench_tprs[j], ir, &state, &mtop);
1042 /* Write information about modified tpr settings to log file */
1043 fprintf(fp, "%4d%10f%10f", j, fac, ir->rcoulomb);
1044 fprintf(fp, "%5d%5d%5d", ir->nkx, ir->nky, ir->nkz);
1045 fprintf(fp, " %9f ", info->fsx[j]);
1046 if (evdwCUT == ir->vdwtype)
1048 fprintf(fp, "%10f", ir->rvdw);
1050 if (EPME_SWITCHED(ir->coulombtype))
1052 fprintf(fp, "%10f", ir->rlist);
1054 if (info->rlistlong[0] != max_cutoff(info->rlist[0], max_cutoff(info->rvdw[0], info->rcoulomb[0])) )
1056 fprintf(fp, "%10f", ir->rlistlong);
1058 fprintf(fp, " %-14s\n", fn_bench_tprs[j]);
1060 /* Make it clear to the user that some additional settings were modified */
1061 if (!is_equal(ir->rvdw, info->rvdw[0])
1062 || !is_equal(ir->rlistlong, info->rlistlong[0]) )
1069 fprintf(fp, "\nNote that in addition to the Coulomb radius and the Fourier grid\n"
1070 "other input settings were also changed (see table above).\n"
1071 "Please check if the modified settings are appropriate.\n");
1079 /* Rename the files we want to keep to some meaningful filename and
1080 * delete the rest */
1081 static void cleanup(const t_filenm *fnm, int nfile, int k, int nnodes,
1082 int nPMEnodes, int nr, gmx_bool bKeepStderr)
1084 char numstring[STRLEN];
1085 char newfilename[STRLEN];
1086 const char *fn = NULL;
1091 fprintf(stdout, "Cleaning up, deleting benchmark temp files ...\n");
1093 for (i = 0; i < nfile; i++)
1095 opt = (char *)fnm[i].opt;
1096 if (strcmp(opt, "-p") == 0)
1098 /* do nothing; keep this file */
1101 else if (strcmp(opt, "-bg") == 0)
1103 /* Give the log file a nice name so one can later see which parameters were used */
1104 numstring[0] = '\0';
1107 sprintf(numstring, "_%d", nr);
1109 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", opt2fn("-bg", nfile, fnm), k, nnodes, nPMEnodes, numstring);
1110 if (gmx_fexist(opt2fn("-bg", nfile, fnm)))
1112 fprintf(stdout, "renaming log file to %s\n", newfilename);
1113 make_backup(newfilename);
1114 rename(opt2fn("-bg", nfile, fnm), newfilename);
1117 else if (strcmp(opt, "-err") == 0)
1119 /* This file contains the output of stderr. We want to keep it in
1120 * cases where there have been problems. */
1121 fn = opt2fn(opt, nfile, fnm);
1122 numstring[0] = '\0';
1125 sprintf(numstring, "_%d", nr);
1127 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", fn, k, nnodes, nPMEnodes, numstring);
1132 fprintf(stdout, "Saving stderr output in %s\n", newfilename);
1133 make_backup(newfilename);
1134 rename(fn, newfilename);
1138 fprintf(stdout, "Deleting %s\n", fn);
1143 /* Delete the files which are created for each benchmark run: (options -b*) */
1144 else if ( (0 == strncmp(opt, "-b", 2)) && (opt2bSet(opt, nfile, fnm) || !is_optional(&fnm[i])) )
1146 fn = opt2fn(opt, nfile, fnm);
1149 fprintf(stdout, "Deleting %s\n", fn);
1157 /* Returns the largest common factor of n1 and n2 */
1158 static int largest_common_factor(int n1, int n2)
1163 for (factor = nmax; factor > 0; factor--)
1165 if (0 == (n1 % factor) && 0 == (n2 % factor) )
1170 return 0; /* one for the compiler */
1174 eNpmeAuto, eNpmeAll, eNpmeReduced, eNpmeSubset, eNpmeNr
1177 /* Create a list of numbers of PME nodes to test */
1178 static void make_npme_list(
1179 const char *npmevalues_opt, /* Make a complete list with all
1180 * possibilities or a short list that keeps only
1181 * reasonable numbers of PME nodes */
1182 int *nentries, /* Number of entries we put in the nPMEnodes list */
1183 int *nPMEnodes[], /* Each entry contains the value for -npme */
1184 int nnodes, /* Total number of nodes to do the tests on */
1185 int minPMEnodes, /* Minimum number of PME nodes */
1186 int maxPMEnodes) /* Maximum number of PME nodes */
1189 int min_factor = 1; /* We request that npp and npme have this minimal
1190 * largest common factor (depends on npp) */
1191 int nlistmax; /* Max. list size */
1192 int nlist; /* Actual number of entries in list */
1196 /* Do we need to check all possible values for -npme or is a reduced list enough? */
1197 if (0 == strcmp(npmevalues_opt, "all") )
1201 else if (0 == strcmp(npmevalues_opt, "subset") )
1203 eNPME = eNpmeSubset;
1205 else /* "auto" or "range" */
1211 else if (nnodes < 128)
1213 eNPME = eNpmeReduced;
1217 eNPME = eNpmeSubset;
1221 /* Calculate how many entries we could possibly have (in case of -npme all) */
1224 nlistmax = maxPMEnodes - minPMEnodes + 3;
1225 if (0 == minPMEnodes)
1235 /* Now make the actual list which is at most of size nlist */
1236 snew(*nPMEnodes, nlistmax);
1237 nlist = 0; /* start counting again, now the real entries in the list */
1238 for (i = 0; i < nlistmax - 2; i++)
1240 npme = maxPMEnodes - i;
1251 /* For 2d PME we want a common largest factor of at least the cube
1252 * root of the number of PP nodes */
1253 min_factor = (int) pow(npp, 1.0/3.0);
1256 gmx_fatal(FARGS, "Unknown option for eNPME in make_npme_list");
1259 if (largest_common_factor(npp, npme) >= min_factor)
1261 (*nPMEnodes)[nlist] = npme;
1265 /* We always test 0 PME nodes and the automatic number */
1266 *nentries = nlist + 2;
1267 (*nPMEnodes)[nlist ] = 0;
1268 (*nPMEnodes)[nlist+1] = -1;
1270 fprintf(stderr, "Will try the following %d different values for -npme:\n", *nentries);
1271 for (i = 0; i < *nentries-1; i++)
1273 fprintf(stderr, "%d, ", (*nPMEnodes)[i]);
1275 fprintf(stderr, "and %d (auto).\n", (*nPMEnodes)[*nentries-1]);
1279 /* Allocate memory to store the performance data */
1280 static void init_perfdata(t_perf *perfdata[], int ntprs, int datasets, int repeats)
1285 for (k = 0; k < ntprs; k++)
1287 snew(perfdata[k], datasets);
1288 for (i = 0; i < datasets; i++)
1290 for (j = 0; j < repeats; j++)
1292 snew(perfdata[k][i].Gcycles, repeats);
1293 snew(perfdata[k][i].ns_per_day, repeats);
1294 snew(perfdata[k][i].PME_f_load, repeats);
1301 /* Check for errors on mdrun -h */
1302 static void make_sure_it_runs(char *mdrun_cmd_line, int length, FILE *fp)
1304 char *command, *msg;
1308 snew(command, length + 15);
1309 snew(msg, length + 500);
1311 fprintf(stdout, "Making shure the benchmarks can be executed ...\n");
1312 sprintf(command, "%s-h -quiet", mdrun_cmd_line);
1313 ret = gmx_system_call(command);
1317 /* To prevent confusion, do not again issue a gmx_fatal here since we already
1318 * get the error message from mdrun itself */
1319 sprintf(msg, "Cannot run the benchmark simulations! Please check the error message of\n"
1320 "mdrun for the source of the problem. Did you provide a command line\n"
1321 "argument that neither g_tune_pme nor mdrun understands? Offending command:\n"
1322 "\n%s\n\n", command);
1324 fprintf(stderr, "%s", msg);
1326 fprintf(fp, "%s", msg);
1336 static void do_the_tests(
1337 FILE *fp, /* General g_tune_pme output file */
1338 char **tpr_names, /* Filenames of the input files to test */
1339 int maxPMEnodes, /* Max fraction of nodes to use for PME */
1340 int minPMEnodes, /* Min fraction of nodes to use for PME */
1341 int npme_fixed, /* If >= -1, test fixed number of PME
1343 const char *npmevalues_opt, /* Which -npme values should be tested */
1344 t_perf **perfdata, /* Here the performace data is stored */
1345 int *pmeentries, /* Entries in the nPMEnodes list */
1346 int repeats, /* Repeat each test this often */
1347 int nnodes, /* Total number of nodes = nPP + nPME */
1348 int nr_tprs, /* Total number of tpr files to test */
1349 gmx_bool bThreads, /* Threads or MPI? */
1350 char *cmd_mpirun, /* mpirun command string */
1351 char *cmd_np, /* "-np", "-n", whatever mpirun needs */
1352 char *cmd_mdrun, /* mdrun command string */
1353 char *cmd_args_bench, /* arguments for mdrun in a string */
1354 const t_filenm *fnm, /* List of filenames from command line */
1355 int nfile, /* Number of files specified on the cmdl. */
1356 int presteps, /* DLB equilibration steps, is checked */
1357 gmx_large_int_t cpt_steps) /* Time step counter in the checkpoint */
1359 int i, nr, k, ret, count = 0, totaltests;
1360 int *nPMEnodes = NULL;
1363 char *command, *cmd_stub;
1365 gmx_bool bResetProblem = FALSE;
1366 gmx_bool bFirst = TRUE;
1369 /* This string array corresponds to the eParselog enum type at the start
1371 const char* ParseLog[] = {
1373 "Logfile not found!",
1374 "No timings, logfile truncated?",
1375 "Run was terminated.",
1376 "Counters were not reset properly.",
1377 "No DD grid found for these settings.",
1378 "TPX version conflict!",
1379 "mdrun was not started in parallel!",
1382 char str_PME_f_load[13];
1385 /* Allocate space for the mdrun command line. 100 extra characters should
1386 be more than enough for the -npme etcetera arguments */
1387 cmdline_length = strlen(cmd_mpirun)
1390 + strlen(cmd_args_bench)
1391 + strlen(tpr_names[0]) + 100;
1392 snew(command, cmdline_length);
1393 snew(cmd_stub, cmdline_length);
1395 /* Construct the part of the command line that stays the same for all tests: */
1398 sprintf(cmd_stub, "%s%s", cmd_mdrun, cmd_np);
1402 sprintf(cmd_stub, "%s%s%s ", cmd_mpirun, cmd_np, cmd_mdrun);
1405 /* Create a list of numbers of PME nodes to test */
1406 if (npme_fixed < -1)
1408 make_npme_list(npmevalues_opt, pmeentries, &nPMEnodes,
1409 nnodes, minPMEnodes, maxPMEnodes);
1415 nPMEnodes[0] = npme_fixed;
1416 fprintf(stderr, "Will use a fixed number of %d PME-only nodes.\n", nPMEnodes[0]);
1421 fprintf(fp, "\nNo benchmarks done since number of repeats (-r) is 0.\n");
1423 finalize(opt2fn("-p", nfile, fnm));
1427 /* Allocate one dataset for each tpr input file: */
1428 init_perfdata(perfdata, nr_tprs, *pmeentries, repeats);
1430 /*****************************************/
1431 /* Main loop over all tpr files to test: */
1432 /*****************************************/
1433 totaltests = nr_tprs*(*pmeentries)*repeats;
1434 for (k = 0; k < nr_tprs; k++)
1436 fprintf(fp, "\nIndividual timings for input file %d (%s):\n", k, tpr_names[k]);
1437 fprintf(fp, "PME nodes Gcycles ns/day PME/f Remark\n");
1438 /* Loop over various numbers of PME nodes: */
1439 for (i = 0; i < *pmeentries; i++)
1441 pd = &perfdata[k][i];
1443 /* Loop over the repeats for each scenario: */
1444 for (nr = 0; nr < repeats; nr++)
1446 pd->nPMEnodes = nPMEnodes[i];
1448 /* Add -npme and -s to the command line and save it. Note that
1449 * the -passall (if set) options requires cmd_args_bench to be
1450 * at the end of the command line string */
1451 snew(pd->mdrun_cmd_line, cmdline_length);
1452 sprintf(pd->mdrun_cmd_line, "%s-npme %d -s %s %s",
1453 cmd_stub, pd->nPMEnodes, tpr_names[k], cmd_args_bench);
1455 /* To prevent that all benchmarks fail due to a show-stopper argument
1456 * on the mdrun command line, we make a quick check with mdrun -h first */
1459 make_sure_it_runs(pd->mdrun_cmd_line, cmdline_length, fp);
1463 /* Do a benchmark simulation: */
1466 sprintf(buf, ", pass %d/%d", nr+1, repeats);
1472 fprintf(stdout, "\n=== Progress %2.0f%%, tpr %d/%d, run %d/%d%s:\n",
1473 (100.0*count)/totaltests,
1474 k+1, nr_tprs, i+1, *pmeentries, buf);
1475 make_backup(opt2fn("-err", nfile, fnm));
1476 sprintf(command, "%s 1> /dev/null 2>%s", pd->mdrun_cmd_line, opt2fn("-err", nfile, fnm));
1477 fprintf(stdout, "%s\n", pd->mdrun_cmd_line);
1478 gmx_system_call(command);
1480 /* Collect the performance data from the log file; also check stderr
1481 * for fatal errors */
1482 ret = parse_logfile(opt2fn("-bg", nfile, fnm), opt2fn("-err", nfile, fnm),
1483 pd, nr, presteps, cpt_steps, nnodes);
1484 if ((presteps > 0) && (ret == eParselogResetProblem))
1486 bResetProblem = TRUE;
1489 if (-1 == pd->nPMEnodes)
1491 sprintf(buf, "(%3d)", pd->guessPME);
1499 if (pd->PME_f_load[nr] > 0.0)
1501 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load[nr]);
1505 sprintf(str_PME_f_load, "%s", " - ");
1508 /* Write the data we got to disk */
1509 fprintf(fp, "%4d%s %12.3f %12.3f %s %s", pd->nPMEnodes,
1510 buf, pd->Gcycles[nr], pd->ns_per_day[nr], str_PME_f_load, ParseLog[ret]);
1511 if (!(ret == eParselogOK || ret == eParselogNoDDGrid || ret == eParselogNotFound) )
1513 fprintf(fp, " Check %s file for problems.", ret == eParselogFatal ? "err" : "log");
1519 /* Do some cleaning up and delete the files we do not need any more */
1520 cleanup(fnm, nfile, k, nnodes, pd->nPMEnodes, nr, ret == eParselogFatal);
1522 /* If the first run with this number of processors already failed, do not try again: */
1523 if (pd->Gcycles[0] <= 0.0 && repeats > 1)
1525 fprintf(stdout, "Skipping remaining passes of unsuccessful setting, see log file for details.\n");
1526 count += repeats-(nr+1);
1529 } /* end of repeats loop */
1530 } /* end of -npme loop */
1531 } /* end of tpr file loop */
1536 fprintf(fp, "WARNING: The cycle and time step counters could not be reset properly. ");
1544 static void check_input(
1551 real maxPMEfraction,
1552 real minPMEfraction,
1554 gmx_large_int_t bench_nsteps,
1555 const t_filenm *fnm,
1565 /* Make sure the input file exists */
1566 if (!gmx_fexist(opt2fn("-s", nfile, fnm)))
1568 gmx_fatal(FARGS, "File %s not found.", opt2fn("-s", nfile, fnm));
1571 /* Make sure that the checkpoint file is not overwritten during benchmarking */
1572 if ( (0 == strcmp(opt2fn("-cpi", nfile, fnm), opt2fn("-bcpo", nfile, fnm)) ) && (sim_part > 1) )
1574 gmx_fatal(FARGS, "Checkpoint input (-cpi) and benchmark checkpoint output (-bcpo) files must not be identical.\n"
1575 "The checkpoint input file must not be overwritten during the benchmarks.\n");
1578 /* Make sure that repeats is >= 0 (if == 0, only write tpr files) */
1581 gmx_fatal(FARGS, "Number of repeats < 0!");
1584 /* Check number of nodes */
1587 gmx_fatal(FARGS, "Number of nodes/threads must be a positive integer.");
1590 /* Automatically choose -ntpr if not set */
1600 /* Set a reasonable scaling factor for rcoulomb */
1603 *rmax = rcoulomb * 1.2;
1606 fprintf(stderr, "Will test %d tpr file%s.\n", *ntprs, *ntprs == 1 ? "" : "s");
1612 fprintf(stderr, "Note: Choose ntpr>1 to shift PME load between real and reciprocal space.\n");
1616 /* Make shure that rmin <= rcoulomb <= rmax */
1625 if (!(*rmin <= *rmax) )
1627 gmx_fatal(FARGS, "Please choose the Coulomb radii such that rmin <= rmax.\n"
1628 "rmin = %g, rmax = %g, actual rcoul from .tpr file = %g\n", *rmin, *rmax, rcoulomb);
1630 /* Add test scenarios if rmin or rmax were set */
1633 if (!is_equal(*rmin, rcoulomb) && (*ntprs == 1) )
1636 fprintf(stderr, "NOTE: Setting -rmin to %g changed -ntpr to %d\n",
1639 if (!is_equal(*rmax, rcoulomb) && (*ntprs == 1) )
1642 fprintf(stderr, "NOTE: Setting -rmax to %g changed -ntpr to %d\n",
1647 /* If one of rmin, rmax is set, we need 2 tpr files at minimum */
1648 if (!is_equal(*rmax, rcoulomb) || !is_equal(*rmin, rcoulomb) )
1650 *ntprs = max(*ntprs, 2);
1653 /* If both rmin, rmax are set, we need 3 tpr files at minimum */
1654 if (!is_equal(*rmax, rcoulomb) && !is_equal(*rmin, rcoulomb) )
1656 *ntprs = max(*ntprs, 3);
1661 fprintf(stderr, "NOTE: Your rmin, rmax setting changed -ntpr to %d\n", *ntprs);
1666 if (is_equal(*rmin, rcoulomb) && is_equal(rcoulomb, *rmax)) /* We have just a single rc */
1668 fprintf(stderr, "WARNING: Resetting -ntpr to 1 since no Coulomb radius scaling is requested.\n"
1669 "Please set rmin < rmax to test Coulomb radii in the [rmin, rmax] interval\n"
1670 "with correspondingly adjusted PME grid settings\n");
1675 /* Check whether max and min fraction are within required values */
1676 if (maxPMEfraction > 0.5 || maxPMEfraction < 0)
1678 gmx_fatal(FARGS, "-max must be between 0 and 0.5");
1680 if (minPMEfraction > 0.5 || minPMEfraction < 0)
1682 gmx_fatal(FARGS, "-min must be between 0 and 0.5");
1684 if (maxPMEfraction < minPMEfraction)
1686 gmx_fatal(FARGS, "-max must be larger or equal to -min");
1689 /* Check whether the number of steps - if it was set - has a reasonable value */
1690 if (bench_nsteps < 0)
1692 gmx_fatal(FARGS, "Number of steps must be positive.");
1695 if (bench_nsteps > 10000 || bench_nsteps < 100)
1697 fprintf(stderr, "WARNING: steps=");
1698 fprintf(stderr, gmx_large_int_pfmt, bench_nsteps);
1699 fprintf(stderr, ". Are you sure you want to perform so %s steps for each benchmark?\n", (bench_nsteps < 100) ? "few" : "many");
1704 gmx_fatal(FARGS, "Cannot have a negative number of presteps.\n");
1707 /* Check for rcoulomb scaling if more than one .tpr file is tested */
1710 if (*rmin/rcoulomb < 0.75 || *rmax/rcoulomb > 1.25)
1712 fprintf(stderr, "WARNING: Applying extreme scaling factor. I hope you know what you are doing.\n");
1716 /* If a fixed number of PME nodes is set we do rcoulomb and PME gird tuning
1717 * only. We need to check whether the requested number of PME-only nodes
1719 if (npme_fixed > -1)
1721 /* No more than 50% of all nodes can be assigned as PME-only nodes. */
1722 if (2*npme_fixed > nnodes)
1724 gmx_fatal(FARGS, "Cannot have more than %d PME-only nodes for a total of %d nodes (you chose %d).\n",
1725 nnodes/2, nnodes, npme_fixed);
1727 if ((npme_fixed > 0) && (5*npme_fixed < nnodes))
1729 fprintf(stderr, "WARNING: Only %g percent of the nodes are assigned as PME-only nodes.\n",
1730 100.0*((real)npme_fixed / (real)nnodes));
1732 if (opt2parg_bSet("-min", npargs, pa) || opt2parg_bSet("-max", npargs, pa))
1734 fprintf(stderr, "NOTE: The -min, -max, and -npme options have no effect when a\n"
1735 " fixed number of PME-only nodes is requested with -fix.\n");
1741 /* Returns TRUE when "opt" is needed at launch time */
1742 static gmx_bool is_launch_file(char *opt, gmx_bool bSet)
1744 /* Apart from the input .tpr and the output log files we need all options that
1745 * were set on the command line and that do not start with -b */
1746 if (0 == strncmp(opt, "-b" , 2) || 0 == strncmp(opt, "-s", 2)
1747 || 0 == strncmp(opt, "-err", 4) || 0 == strncmp(opt, "-p", 2) )
1756 /* Returns TRUE when "opt" defines a file which is needed for the benchmarks runs */
1757 static gmx_bool is_bench_file(char *opt, gmx_bool bSet, gmx_bool bOptional, gmx_bool bIsOutput)
1759 /* Apart from the input .tpr, all files starting with "-b" are for
1760 * _b_enchmark files exclusively */
1761 if (0 == strncmp(opt, "-s", 2))
1766 if (0 == strncmp(opt, "-b", 2) || 0 == strncmp(opt, "-s", 2))
1768 if (!bOptional || bSet)
1785 if (bSet) /* These are additional input files like -cpi -ei */
1798 /* Adds 'buf' to 'str' */
1799 static void add_to_string(char **str, char *buf)
1804 len = strlen(*str) + strlen(buf) + 1;
1810 /* Create the command line for the benchmark as well as for the real run */
1811 static void create_command_line_snippets(
1812 gmx_bool bAppendFiles,
1813 gmx_bool bKeepAndNumCPT,
1814 gmx_bool bResetHWay,
1818 char *cmd_args_bench[], /* command line arguments for benchmark runs */
1819 char *cmd_args_launch[], /* command line arguments for simulation run */
1820 char extra_args[]) /* Add this to the end of the command line */
1825 char strbuf[STRLEN];
1828 /* strlen needs at least '\0' as a string: */
1829 snew(*cmd_args_bench, 1);
1830 snew(*cmd_args_launch, 1);
1831 *cmd_args_launch[0] = '\0';
1832 *cmd_args_bench[0] = '\0';
1835 /*******************************************/
1836 /* 1. Process other command line arguments */
1837 /*******************************************/
1840 /* Add equilibration steps to benchmark options */
1841 sprintf(strbuf, "-resetstep %d ", presteps);
1842 add_to_string(cmd_args_bench, strbuf);
1844 /* These switches take effect only at launch time */
1845 if (FALSE == bAppendFiles)
1847 add_to_string(cmd_args_launch, "-noappend ");
1851 add_to_string(cmd_args_launch, "-cpnum ");
1855 add_to_string(cmd_args_launch, "-resethway ");
1858 /********************/
1859 /* 2. Process files */
1860 /********************/
1861 for (i = 0; i < nfile; i++)
1863 opt = (char *)fnm[i].opt;
1864 name = opt2fn(opt, nfile, fnm);
1866 /* Strbuf contains the options, now let's sort out where we need that */
1867 sprintf(strbuf, "%s %s ", opt, name);
1869 if (is_bench_file(opt, opt2bSet(opt, nfile, fnm), is_optional(&fnm[i]), is_output(&fnm[i])) )
1871 /* All options starting with -b* need the 'b' removed,
1872 * therefore overwrite strbuf */
1873 if (0 == strncmp(opt, "-b", 2))
1875 sprintf(strbuf, "-%s %s ", &opt[2], name);
1878 add_to_string(cmd_args_bench, strbuf);
1881 if (is_launch_file(opt, opt2bSet(opt, nfile, fnm)) )
1883 add_to_string(cmd_args_launch, strbuf);
1887 add_to_string(cmd_args_bench, extra_args);
1888 add_to_string(cmd_args_launch, extra_args);
1892 /* Set option opt */
1893 static void setopt(const char *opt, int nfile, t_filenm fnm[])
1897 for (i = 0; (i < nfile); i++)
1899 if (strcmp(opt, fnm[i].opt) == 0)
1901 fnm[i].flag |= ffSET;
1907 /* This routine inspects the tpr file and ...
1908 * 1. checks for output files that get triggered by a tpr option. These output
1909 * files are marked as 'set' to allow for a proper cleanup after each
1911 * 2. returns the PME:PP load ratio
1912 * 3. returns rcoulomb from the tpr */
1913 static float inspect_tpr(int nfile, t_filenm fnm[], real *rcoulomb)
1915 gmx_bool bPull; /* Is pulling requested in .tpr file? */
1916 gmx_bool bTpi; /* Is test particle insertion requested? */
1917 gmx_bool bFree; /* Is a free energy simulation requested? */
1918 gmx_bool bNM; /* Is a normal mode analysis requested? */
1924 /* Check tpr file for options that trigger extra output files */
1925 read_tpx_state(opt2fn("-s", nfile, fnm), &ir, &state, NULL, &mtop);
1926 bPull = (epullNO != ir.ePull);
1927 bFree = (efepNO != ir.efep );
1928 bNM = (eiNM == ir.eI );
1929 bTpi = EI_TPI(ir.eI);
1931 /* Set these output files on the tuning command-line */
1934 setopt("-pf", nfile, fnm);
1935 setopt("-px", nfile, fnm);
1939 setopt("-dhdl", nfile, fnm);
1943 setopt("-tpi", nfile, fnm);
1944 setopt("-tpid", nfile, fnm);
1948 setopt("-mtx", nfile, fnm);
1951 *rcoulomb = ir.rcoulomb;
1953 /* Return the estimate for the number of PME nodes */
1954 return pme_load_estimate(&mtop, &ir, state.box);
1958 static void couple_files_options(int nfile, t_filenm fnm[])
1961 gmx_bool bSet, bBench;
1966 for (i = 0; i < nfile; i++)
1968 opt = (char *)fnm[i].opt;
1969 bSet = ((fnm[i].flag & ffSET) != 0);
1970 bBench = (0 == strncmp(opt, "-b", 2));
1972 /* Check optional files */
1973 /* If e.g. -eo is set, then -beo also needs to be set */
1974 if (is_optional(&fnm[i]) && bSet && !bBench)
1976 sprintf(buf, "-b%s", &opt[1]);
1977 setopt(buf, nfile, fnm);
1979 /* If -beo is set, then -eo also needs to be! */
1980 if (is_optional(&fnm[i]) && bSet && bBench)
1982 sprintf(buf, "-%s", &opt[2]);
1983 setopt(buf, nfile, fnm);
1989 static double gettime()
1991 #ifdef HAVE_GETTIMEOFDAY
1995 gettimeofday(&t, NULL);
1997 seconds = (double) t.tv_sec + 1e-6*(double)t.tv_usec;
2003 seconds = time(NULL);
2010 #define BENCHSTEPS (1000)
2012 int gmx_tune_pme(int argc, char *argv[])
2014 const char *desc[] = {
2015 "For a given number [TT]-np[tt] or [TT]-ntmpi[tt] of processors/threads, this program systematically",
2016 "times [TT]mdrun[tt] with various numbers of PME-only nodes and determines",
2017 "which setting is fastest. It will also test whether performance can",
2018 "be enhanced by shifting load from the reciprocal to the real space",
2019 "part of the Ewald sum. ",
2020 "Simply pass your [TT].tpr[tt] file to [TT]g_tune_pme[tt] together with other options",
2021 "for [TT]mdrun[tt] as needed.[PAR]",
2022 "Which executables are used can be set in the environment variables",
2023 "MPIRUN and MDRUN. If these are not present, 'mpirun' and 'mdrun'",
2024 "will be used as defaults. Note that for certain MPI frameworks you",
2025 "need to provide a machine- or hostfile. This can also be passed",
2026 "via the MPIRUN variable, e.g.[PAR]",
2027 "[TT]export MPIRUN=\"/usr/local/mpirun -machinefile hosts\"[tt][PAR]",
2028 "Please call [TT]g_tune_pme[tt] with the normal options you would pass to",
2029 "[TT]mdrun[tt] and add [TT]-np[tt] for the number of processors to perform the",
2030 "tests on, or [TT]-ntmpi[tt] for the number of threads. You can also add [TT]-r[tt]",
2031 "to repeat each test several times to get better statistics. [PAR]",
2032 "[TT]g_tune_pme[tt] can test various real space / reciprocal space workloads",
2033 "for you. With [TT]-ntpr[tt] you control how many extra [TT].tpr[tt] files will be",
2034 "written with enlarged cutoffs and smaller Fourier grids respectively.",
2035 "Typically, the first test (number 0) will be with the settings from the input",
2036 "[TT].tpr[tt] file; the last test (number [TT]ntpr[tt]) will have the Coulomb cutoff",
2037 "specified by [TT]-rmax[tt] with a somwhat smaller PME grid at the same time. ",
2038 "In this last test, the Fourier spacing is multiplied with [TT]rmax[tt]/rcoulomb. ",
2039 "The remaining [TT].tpr[tt] files will have equally-spaced Coulomb radii (and Fourier "
2040 "spacings) between these extremes. [BB]Note[bb] that you can set [TT]-ntpr[tt] to 1",
2041 "if you just seek the optimal number of PME-only nodes; in that case",
2042 "your input [TT].tpr[tt] file will remain unchanged.[PAR]",
2043 "For the benchmark runs, the default of 1000 time steps should suffice for most",
2044 "MD systems. The dynamic load balancing needs about 100 time steps",
2045 "to adapt to local load imbalances, therefore the time step counters",
2046 "are by default reset after 100 steps. For large systems (>1M atoms), as well as ",
2047 "for a higher accuarcy of the measurements, you should set [TT]-resetstep[tt] to a higher value.",
2048 "From the 'DD' load imbalance entries in the md.log output file you",
2049 "can tell after how many steps the load is sufficiently balanced. Example call:[PAR]"
2050 "[TT]g_tune_pme -np 64 -s protein.tpr -launch[tt][PAR]",
2051 "After calling [TT]mdrun[tt] several times, detailed performance information",
2052 "is available in the output file [TT]perf.out.[tt] ",
2053 "[BB]Note[bb] that during the benchmarks, a couple of temporary files are written",
2054 "(options [TT]-b[tt]*), these will be automatically deleted after each test.[PAR]",
2055 "If you want the simulation to be started automatically with the",
2056 "optimized parameters, use the command line option [TT]-launch[tt].[PAR]",
2061 int pmeentries = 0; /* How many values for -npme do we actually test for each tpr file */
2062 real maxPMEfraction = 0.50;
2063 real minPMEfraction = 0.25;
2064 int maxPMEnodes, minPMEnodes;
2065 float guessPMEratio; /* guessed PME:PP ratio based on the tpr file */
2066 float guessPMEnodes;
2067 int npme_fixed = -2; /* If >= -1, use only this number
2068 * of PME-only nodes */
2070 real rmin = 0.0, rmax = 0.0; /* min and max value for rcoulomb if scaling is requested */
2071 real rcoulomb = -1.0; /* Coulomb radius as set in .tpr file */
2072 gmx_bool bScaleRvdw = TRUE;
2073 gmx_large_int_t bench_nsteps = BENCHSTEPS;
2074 gmx_large_int_t new_sim_nsteps = -1; /* -1 indicates: not set by the user */
2075 gmx_large_int_t cpt_steps = 0; /* Step counter in .cpt input file */
2076 int presteps = 100; /* Do a full cycle reset after presteps steps */
2077 gmx_bool bOverwrite = FALSE, bKeepTPR;
2078 gmx_bool bLaunch = FALSE;
2079 char *ExtraArgs = NULL;
2080 char **tpr_names = NULL;
2081 const char *simulation_tpr = NULL;
2082 int best_npme, best_tpr;
2083 int sim_part = 1; /* For benchmarks with checkpoint files */
2086 /* Default program names if nothing else is found */
2087 char *cmd_mpirun = NULL, *cmd_mdrun = NULL;
2088 char *cmd_args_bench, *cmd_args_launch;
2089 char *cmd_np = NULL;
2091 t_perf **perfdata = NULL;
2097 /* Print out how long the tuning took */
2100 static t_filenm fnm[] = {
2102 { efOUT, "-p", "perf", ffWRITE },
2103 { efLOG, "-err", "bencherr", ffWRITE },
2104 { efTPX, "-so", "tuned", ffWRITE },
2106 { efTPX, NULL, NULL, ffREAD },
2107 { efTRN, "-o", NULL, ffWRITE },
2108 { efXTC, "-x", NULL, ffOPTWR },
2109 { efCPT, "-cpi", NULL, ffOPTRD },
2110 { efCPT, "-cpo", NULL, ffOPTWR },
2111 { efSTO, "-c", "confout", ffWRITE },
2112 { efEDR, "-e", "ener", ffWRITE },
2113 { efLOG, "-g", "md", ffWRITE },
2114 { efXVG, "-dhdl", "dhdl", ffOPTWR },
2115 { efXVG, "-field", "field", ffOPTWR },
2116 { efXVG, "-table", "table", ffOPTRD },
2117 { efXVG, "-tabletf", "tabletf", ffOPTRD },
2118 { efXVG, "-tablep", "tablep", ffOPTRD },
2119 { efXVG, "-tableb", "table", ffOPTRD },
2120 { efTRX, "-rerun", "rerun", ffOPTRD },
2121 { efXVG, "-tpi", "tpi", ffOPTWR },
2122 { efXVG, "-tpid", "tpidist", ffOPTWR },
2123 { efEDI, "-ei", "sam", ffOPTRD },
2124 { efXVG, "-eo", "edsam", ffOPTWR },
2125 { efGCT, "-j", "wham", ffOPTRD },
2126 { efGCT, "-jo", "bam", ffOPTWR },
2127 { efXVG, "-ffout", "gct", ffOPTWR },
2128 { efXVG, "-devout", "deviatie", ffOPTWR },
2129 { efXVG, "-runav", "runaver", ffOPTWR },
2130 { efXVG, "-px", "pullx", ffOPTWR },
2131 { efXVG, "-pf", "pullf", ffOPTWR },
2132 { efXVG, "-ro", "rotation", ffOPTWR },
2133 { efLOG, "-ra", "rotangles", ffOPTWR },
2134 { efLOG, "-rs", "rotslabs", ffOPTWR },
2135 { efLOG, "-rt", "rottorque", ffOPTWR },
2136 { efMTX, "-mtx", "nm", ffOPTWR },
2137 { efNDX, "-dn", "dipole", ffOPTWR },
2138 /* Output files that are deleted after each benchmark run */
2139 { efTRN, "-bo", "bench", ffWRITE },
2140 { efXTC, "-bx", "bench", ffWRITE },
2141 { efCPT, "-bcpo", "bench", ffWRITE },
2142 { efSTO, "-bc", "bench", ffWRITE },
2143 { efEDR, "-be", "bench", ffWRITE },
2144 { efLOG, "-bg", "bench", ffWRITE },
2145 { efXVG, "-beo", "benchedo", ffOPTWR },
2146 { efXVG, "-bdhdl", "benchdhdl", ffOPTWR },
2147 { efXVG, "-bfield", "benchfld", ffOPTWR },
2148 { efXVG, "-btpi", "benchtpi", ffOPTWR },
2149 { efXVG, "-btpid", "benchtpid", ffOPTWR },
2150 { efGCT, "-bjo", "bench", ffOPTWR },
2151 { efXVG, "-bffout", "benchgct", ffOPTWR },
2152 { efXVG, "-bdevout", "benchdev", ffOPTWR },
2153 { efXVG, "-brunav", "benchrnav", ffOPTWR },
2154 { efXVG, "-bpx", "benchpx", ffOPTWR },
2155 { efXVG, "-bpf", "benchpf", ffOPTWR },
2156 { efXVG, "-bro", "benchrot", ffOPTWR },
2157 { efLOG, "-bra", "benchrota", ffOPTWR },
2158 { efLOG, "-brs", "benchrots", ffOPTWR },
2159 { efLOG, "-brt", "benchrott", ffOPTWR },
2160 { efMTX, "-bmtx", "benchn", ffOPTWR },
2161 { efNDX, "-bdn", "bench", ffOPTWR }
2164 gmx_bool bThreads = FALSE;
2168 const char *procstring[] =
2169 { NULL, "-np", "-n", "none", NULL };
2170 const char *npmevalues_opt[] =
2171 { NULL, "auto", "all", "subset", NULL };
2173 gmx_bool bAppendFiles = TRUE;
2174 gmx_bool bKeepAndNumCPT = FALSE;
2175 gmx_bool bResetCountersHalfWay = FALSE;
2176 gmx_bool bBenchmark = TRUE;
2178 output_env_t oenv = NULL;
2181 /***********************/
2182 /* g_tune_pme options: */
2183 /***********************/
2184 { "-np", FALSE, etINT, {&nnodes},
2185 "Number of nodes to run the tests on (must be > 2 for separate PME nodes)" },
2186 { "-npstring", FALSE, etENUM, {procstring},
2187 "Specify the number of processors to [TT]$MPIRUN[tt] using this string"},
2188 { "-ntmpi", FALSE, etINT, {&nthreads},
2189 "Number of MPI-threads to run the tests on (turns MPI & mpirun off)"},
2190 { "-r", FALSE, etINT, {&repeats},
2191 "Repeat each test this often" },
2192 { "-max", FALSE, etREAL, {&maxPMEfraction},
2193 "Max fraction of PME nodes to test with" },
2194 { "-min", FALSE, etREAL, {&minPMEfraction},
2195 "Min fraction of PME nodes to test with" },
2196 { "-npme", FALSE, etENUM, {npmevalues_opt},
2197 "Within -min and -max, benchmark all possible values for [TT]-npme[tt], or just a reasonable subset. "
2198 "Auto neglects -min and -max and chooses reasonable values around a guess for npme derived from the .tpr"},
2199 { "-fix", FALSE, etINT, {&npme_fixed},
2200 "If >= -1, do not vary the number of PME-only nodes, instead use this fixed value and only vary rcoulomb and the PME grid spacing."},
2201 { "-rmax", FALSE, etREAL, {&rmax},
2202 "If >0, maximal rcoulomb for -ntpr>1 (rcoulomb upscaling results in fourier grid downscaling)" },
2203 { "-rmin", FALSE, etREAL, {&rmin},
2204 "If >0, minimal rcoulomb for -ntpr>1" },
2205 { "-scalevdw", FALSE, etBOOL, {&bScaleRvdw},
2206 "Scale rvdw along with rcoulomb"},
2207 { "-ntpr", FALSE, etINT, {&ntprs},
2208 "Number of [TT].tpr[tt] files to benchmark. Create this many files with different rcoulomb scaling factors depending on -rmin and -rmax. "
2209 "If < 1, automatically choose the number of [TT].tpr[tt] files to test" },
2210 { "-steps", FALSE, etGMX_LARGE_INT, {&bench_nsteps},
2211 "Take timings for this many steps in the benchmark runs" },
2212 { "-resetstep", FALSE, etINT, {&presteps},
2213 "Let dlb equilibrate this many steps before timings are taken (reset cycle counters after this many steps)" },
2214 { "-simsteps", FALSE, etGMX_LARGE_INT, {&new_sim_nsteps},
2215 "If non-negative, perform this many steps in the real run (overwrites nsteps from [TT].tpr[tt], add [TT].cpt[tt] steps)" },
2216 { "-launch", FALSE, etBOOL, {&bLaunch},
2217 "Launch the real simulation after optimization" },
2218 { "-bench", FALSE, etBOOL, {&bBenchmark},
2219 "Run the benchmarks or just create the input [TT].tpr[tt] files?" },
2220 /******************/
2221 /* mdrun options: */
2222 /******************/
2223 /* We let g_tune_pme parse and understand these options, because we need to
2224 * prevent that they appear on the mdrun command line for the benchmarks */
2225 { "-append", FALSE, etBOOL, {&bAppendFiles},
2226 "Append to previous output files when continuing from checkpoint instead of adding the simulation part number to all file names (for launch only)" },
2227 { "-cpnum", FALSE, etBOOL, {&bKeepAndNumCPT},
2228 "Keep and number checkpoint files (launch only)" },
2229 { "-resethway", FALSE, etBOOL, {&bResetCountersHalfWay},
2230 "HIDDENReset the cycle counters after half the number of steps or halfway [TT]-maxh[tt] (launch only)" }
2234 #define NFILE asize(fnm)
2236 CopyRight(stderr, argv[0]);
2238 seconds = gettime();
2240 parse_common_args(&argc, argv, PCA_NOEXIT_ON_ARGS,
2241 NFILE, fnm, asize(pa), pa, asize(desc), desc,
2244 /* Store the remaining unparsed command line entries in a string which
2245 * is then attached to the mdrun command line */
2247 ExtraArgs[0] = '\0';
2248 for (i = 1; i < argc; i++) /* argc will now be 1 if everything was understood */
2250 add_to_string(&ExtraArgs, argv[i]);
2251 add_to_string(&ExtraArgs, " ");
2254 if (opt2parg_bSet("-ntmpi", asize(pa), pa))
2257 if (opt2parg_bSet("-npstring", asize(pa), pa))
2259 fprintf(stderr, "WARNING: -npstring has no effect when using threads.\n");
2264 gmx_fatal(FARGS, "Can't run multi-threaded MPI simulation yet!");
2266 /* and now we just set this; a bit of an ugly hack*/
2269 /* Check for PME:PP ratio and whether tpr triggers additional output files */
2270 guessPMEratio = inspect_tpr(NFILE, fnm, &rcoulomb);
2272 /* Automatically set -beo options if -eo is set etc. */
2273 couple_files_options(NFILE, fnm);
2275 /* Construct the command line arguments for benchmark runs
2276 * as well as for the simulation run */
2279 sprintf(bbuf, " -ntmpi %d ", nthreads);
2283 sprintf(bbuf, " -np %d ", nnodes);
2288 create_command_line_snippets(bAppendFiles, bKeepAndNumCPT, bResetCountersHalfWay, presteps,
2289 NFILE, fnm, &cmd_args_bench, &cmd_args_launch, ExtraArgs);
2291 /* Read in checkpoint file if requested */
2293 if (opt2bSet("-cpi", NFILE, fnm))
2296 cr->duty = DUTY_PP; /* makes the following routine happy */
2297 read_checkpoint_simulation_part(opt2fn("-cpi", NFILE, fnm),
2298 &sim_part, &cpt_steps, cr,
2299 FALSE, NFILE, fnm, NULL, NULL);
2302 /* sim_part will now be 1 if no checkpoint file was found */
2305 gmx_fatal(FARGS, "Checkpoint file %s not found!", opt2fn("-cpi", NFILE, fnm));
2309 /* Open performance output file and write header info */
2310 fp = ffopen(opt2fn("-p", NFILE, fnm), "w");
2312 /* Make a quick consistency check of command line parameters */
2313 check_input(nnodes, repeats, &ntprs, &rmin, rcoulomb, &rmax,
2314 maxPMEfraction, minPMEfraction, npme_fixed,
2315 bench_nsteps, fnm, NFILE, sim_part, presteps,
2318 /* Determine the maximum and minimum number of PME nodes to test,
2319 * the actual list of settings is build in do_the_tests(). */
2320 if ((nnodes > 2) && (npme_fixed < -1))
2322 if (0 == strcmp(npmevalues_opt[0], "auto"))
2324 /* Determine the npme range automatically based on the PME:PP load guess */
2325 if (guessPMEratio > 1.0)
2327 /* More PME than PP work, probably we do not need separate PME nodes at all! */
2328 maxPMEnodes = nnodes/2;
2329 minPMEnodes = nnodes/2;
2333 /* PME : PP load is in the range 0..1, let's test around the guess */
2334 guessPMEnodes = nnodes/(1.0 + 1.0/guessPMEratio);
2335 minPMEnodes = floor(0.7*guessPMEnodes);
2336 maxPMEnodes = ceil(1.6*guessPMEnodes);
2337 maxPMEnodes = min(maxPMEnodes, nnodes/2);
2342 /* Determine the npme range based on user input */
2343 maxPMEnodes = floor(maxPMEfraction*nnodes);
2344 minPMEnodes = max(floor(minPMEfraction*nnodes), 0);
2345 fprintf(stdout, "Will try runs with %d ", minPMEnodes);
2346 if (maxPMEnodes != minPMEnodes)
2348 fprintf(stdout, "- %d ", maxPMEnodes);
2350 fprintf(stdout, "PME-only nodes.\n Note that the automatic number of PME-only nodes and no separate PME nodes are always tested.\n");
2359 /* Get the commands we need to set up the runs from environment variables */
2360 get_program_paths(bThreads, &cmd_mpirun, cmd_np, &cmd_mdrun, repeats);
2362 /* Print some header info to file */
2364 fprintf(fp, "\n P E R F O R M A N C E R E S U L T S\n");
2366 fprintf(fp, "%s for Gromacs %s\n", ShortProgram(), GromacsVersion());
2369 fprintf(fp, "Number of nodes : %d\n", nnodes);
2370 fprintf(fp, "The mpirun command is : %s\n", cmd_mpirun);
2371 if (strcmp(procstring[0], "none") != 0)
2373 fprintf(fp, "Passing # of nodes via : %s\n", procstring[0]);
2377 fprintf(fp, "Not setting number of nodes in system call\n");
2382 fprintf(fp, "Number of threads : %d\n", nnodes);
2385 fprintf(fp, "The mdrun command is : %s\n", cmd_mdrun);
2386 fprintf(fp, "mdrun args benchmarks : %s\n", cmd_args_bench);
2387 fprintf(fp, "Benchmark steps : ");
2388 fprintf(fp, gmx_large_int_pfmt, bench_nsteps);
2390 fprintf(fp, "dlb equilibration steps : %d\n", presteps);
2393 fprintf(fp, "Checkpoint time step : ");
2394 fprintf(fp, gmx_large_int_pfmt, cpt_steps);
2397 fprintf(fp, "mdrun args at launchtime: %s\n", cmd_args_launch);
2399 if (new_sim_nsteps >= 0)
2402 fprintf(stderr, "Note: Simulation input file %s will have ", opt2fn("-so", NFILE, fnm));
2403 fprintf(stderr, gmx_large_int_pfmt, new_sim_nsteps+cpt_steps);
2404 fprintf(stderr, " steps.\n");
2405 fprintf(fp, "Simulation steps : ");
2406 fprintf(fp, gmx_large_int_pfmt, new_sim_nsteps);
2411 fprintf(fp, "Repeats for each test : %d\n", repeats);
2414 if (npme_fixed >= -1)
2416 fprintf(fp, "Fixing -npme at : %d\n", npme_fixed);
2419 fprintf(fp, "Input file : %s\n", opt2fn("-s", NFILE, fnm));
2420 fprintf(fp, " PME/PP load estimate : %g\n", guessPMEratio);
2422 /* Allocate memory for the inputinfo struct: */
2424 info->nr_inputfiles = ntprs;
2425 for (i = 0; i < ntprs; i++)
2427 snew(info->rcoulomb, ntprs);
2428 snew(info->rvdw, ntprs);
2429 snew(info->rlist, ntprs);
2430 snew(info->rlistlong, ntprs);
2431 snew(info->nkx, ntprs);
2432 snew(info->nky, ntprs);
2433 snew(info->nkz, ntprs);
2434 snew(info->fsx, ntprs);
2435 snew(info->fsy, ntprs);
2436 snew(info->fsz, ntprs);
2438 /* Make alternative tpr files to test: */
2439 snew(tpr_names, ntprs);
2440 for (i = 0; i < ntprs; i++)
2442 snew(tpr_names[i], STRLEN);
2445 /* It can be that ntprs is reduced by make_benchmark_tprs if not enough
2446 * different grids could be found. */
2447 make_benchmark_tprs(opt2fn("-s", NFILE, fnm), tpr_names, bench_nsteps+presteps,
2448 cpt_steps, rmin, rmax, bScaleRvdw, &ntprs, info, fp);
2450 /********************************************************************************/
2451 /* Main loop over all scenarios we need to test: tpr files, PME nodes, repeats */
2452 /********************************************************************************/
2453 snew(perfdata, ntprs);
2456 do_the_tests(fp, tpr_names, maxPMEnodes, minPMEnodes, npme_fixed, npmevalues_opt[0], perfdata, &pmeentries,
2457 repeats, nnodes, ntprs, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2458 cmd_args_bench, fnm, NFILE, presteps, cpt_steps);
2460 fprintf(fp, "\nTuning took%8.1f minutes.\n", (gettime()-seconds)/60.0);
2462 /* Analyse the results and give a suggestion for optimal settings: */
2463 bKeepTPR = analyze_data(fp, opt2fn("-p", NFILE, fnm), perfdata, nnodes, ntprs, pmeentries,
2464 repeats, info, &best_tpr, &best_npme);
2466 /* Take the best-performing tpr file and enlarge nsteps to original value */
2467 if (bKeepTPR && !bOverwrite)
2469 simulation_tpr = opt2fn("-s", NFILE, fnm);
2473 simulation_tpr = opt2fn("-so", NFILE, fnm);
2474 modify_PMEsettings(bOverwrite ? (new_sim_nsteps+cpt_steps) : info->orig_sim_steps,
2475 info->orig_init_step, tpr_names[best_tpr], simulation_tpr);
2478 /* Let's get rid of the temporary benchmark input files */
2479 for (i = 0; i < ntprs; i++)
2481 fprintf(stdout, "Deleting temporary benchmark input file %s\n", tpr_names[i]);
2482 remove(tpr_names[i]);
2485 /* Now start the real simulation if the user requested it ... */
2486 launch_simulation(bLaunch, fp, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2487 cmd_args_launch, simulation_tpr, best_npme);
2491 /* ... or simply print the performance results to screen: */
2494 finalize(opt2fn("-p", NFILE, fnm));