3 * This source code is part of
7 * GROningen MAchine for Chemical Simulations
9 * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
10 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
11 * Copyright (c) 2001-2008, The GROMACS development team,
12 * check out http://www.gromacs.org for more information.
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version 2
17 * of the License, or (at your option) any later version.
19 * If you want to redistribute modifications, please consider that
20 * scientific software is very special. Version control is crucial -
21 * bugs must be traceable. We will be happy to consider code for
22 * inclusion in the official distribution, but derived work must not
23 * be called official GROMACS. Details are found in the README & COPYING
24 * files - if they are missing, get the official version at www.gromacs.org.
26 * To help us fund GROMACS development, we humbly ask that you cite
27 * the papers on the package - you can find them in the top README file.
29 * For more info, check our website at http://www.gromacs.org
32 * Gallium Rubidium Oxygen Manganese Argon Carbon Silicon
40 #ifdef HAVE_SYS_TIME_H
46 #include "gromacs/commandline/pargs.h"
51 #include "gromacs/fileio/tpxio.h"
55 #include "checkpoint.h"
60 #include "gromacs/timing/walltime_accounting.h"
63 /* Enum for situations that can occur during log file parsing, the
64 * corresponding string entries can be found in do_the_tests() in
65 * const char* ParseLog[] */
71 eParselogResetProblem,
75 eParselogLargePrimeFactor,
83 int nPMEnodes; /* number of PME-only nodes used in this test */
84 int nx, ny, nz; /* DD grid */
85 int guessPME; /* if nPMEnodes == -1, this is the guessed number of PME nodes */
86 double *Gcycles; /* This can contain more than one value if doing multiple tests */
90 float *PME_f_load; /* PME mesh/force load average*/
91 float PME_f_load_Av; /* Average average ;) ... */
92 char *mdrun_cmd_line; /* Mdrun command line used for this test */
98 int nr_inputfiles; /* The number of tpr and mdp input files */
99 gmx_large_int_t orig_sim_steps; /* Number of steps to be done in the real simulation */
100 gmx_large_int_t orig_init_step; /* Init step for the real simulation */
101 real *rcoulomb; /* The coulomb radii [0...nr_inputfiles] */
102 real *rvdw; /* The vdW radii */
103 real *rlist; /* Neighbourlist cutoff radius */
105 int *nkx, *nky, *nkz;
106 real *fsx, *fsy, *fsz; /* Fourierspacing in x,y,z dimension */
110 static void sep_line(FILE *fp)
112 fprintf(fp, "\n------------------------------------------------------------\n");
116 /* Wrapper for system calls */
117 static int gmx_system_call(char *command)
120 gmx_fatal(FARGS, "No calls to system(3) supported on this platform. Attempted to call:\n'%s'\n", command);
122 return ( system(command) );
127 /* Check if string starts with substring */
128 static gmx_bool str_starts(const char *string, const char *substring)
130 return ( strncmp(string, substring, strlen(substring)) == 0);
134 static void cleandata(t_perf *perfdata, int test_nr)
136 perfdata->Gcycles[test_nr] = 0.0;
137 perfdata->ns_per_day[test_nr] = 0.0;
138 perfdata->PME_f_load[test_nr] = 0.0;
144 static gmx_bool is_equal(real a, real b)
146 real diff, eps = 1.0e-7;
167 static void finalize(const char *fn_out)
173 fp = fopen(fn_out, "r");
174 fprintf(stdout, "\n\n");
176 while (fgets(buf, STRLEN-1, fp) != NULL)
178 fprintf(stdout, "%s", buf);
181 fprintf(stdout, "\n\n");
186 eFoundNothing, eFoundDDStr, eFoundAccountingStr, eFoundCycleStr
189 static int parse_logfile(const char *logfile, const char *errfile,
190 t_perf *perfdata, int test_nr, int presteps, gmx_large_int_t cpt_steps,
194 char line[STRLEN], dumstring[STRLEN], dumstring2[STRLEN];
195 const char matchstrdd[] = "Domain decomposition grid";
196 const char matchstrcr[] = "resetting all time and cycle counters";
197 const char matchstrbal[] = "Average PME mesh/force load:";
198 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";
199 const char errSIG[] = "signal, stopping at the next";
202 float dum1, dum2, dum3, dum4;
205 gmx_large_int_t resetsteps = -1;
206 gmx_bool bFoundResetStr = FALSE;
207 gmx_bool bResetChecked = FALSE;
210 if (!gmx_fexist(logfile))
212 fprintf(stderr, "WARNING: Could not find logfile %s.\n", logfile);
213 cleandata(perfdata, test_nr);
214 return eParselogNotFound;
217 fp = fopen(logfile, "r");
218 perfdata->PME_f_load[test_nr] = -1.0;
219 perfdata->guessPME = -1;
221 iFound = eFoundNothing;
224 iFound = eFoundDDStr; /* Skip some case statements */
227 while (fgets(line, STRLEN, fp) != NULL)
229 /* Remove leading spaces */
232 /* Check for TERM and INT signals from user: */
233 if (strstr(line, errSIG) != NULL)
236 cleandata(perfdata, test_nr);
237 return eParselogTerm;
240 /* Check whether cycle resetting worked */
241 if (presteps > 0 && !bFoundResetStr)
243 if (strstr(line, matchstrcr) != NULL)
245 sprintf(dumstring, "step %s", gmx_large_int_pfmt);
246 sscanf(line, dumstring, &resetsteps);
247 bFoundResetStr = TRUE;
248 if (resetsteps == presteps+cpt_steps)
250 bResetChecked = TRUE;
254 sprintf(dumstring, gmx_large_int_pfmt, resetsteps);
255 sprintf(dumstring2, gmx_large_int_pfmt, presteps+cpt_steps);
256 fprintf(stderr, "WARNING: Time step counters were reset at step %s,\n"
257 " though they were supposed to be reset at step %s!\n",
258 dumstring, dumstring2);
263 /* Look for strings that appear in a certain order in the log file: */
267 /* Look for domain decomp grid and separate PME nodes: */
268 if (str_starts(line, matchstrdd))
270 sscanf(line, "Domain decomposition grid %d x %d x %d, separate PME nodes %d",
271 &(perfdata->nx), &(perfdata->ny), &(perfdata->nz), &npme);
272 if (perfdata->nPMEnodes == -1)
274 perfdata->guessPME = npme;
276 else if (perfdata->nPMEnodes != npme)
278 gmx_fatal(FARGS, "PME nodes from command line and output file are not identical");
280 iFound = eFoundDDStr;
282 /* Catch a few errors that might have occured: */
283 else if (str_starts(line, "There is no domain decomposition for"))
286 return eParselogNoDDGrid;
288 else if (str_starts(line, "The number of nodes you selected"))
291 return eParselogLargePrimeFactor;
293 else if (str_starts(line, "reading tpx file"))
296 return eParselogTPXVersion;
298 else if (str_starts(line, "The -dd or -npme option request a parallel simulation"))
301 return eParselogNotParallel;
305 /* Look for PME mesh/force balance (not necessarily present, though) */
306 if (str_starts(line, matchstrbal))
308 sscanf(&line[strlen(matchstrbal)], "%f", &(perfdata->PME_f_load[test_nr]));
310 /* Look for matchstring */
311 if (str_starts(line, matchstring))
313 iFound = eFoundAccountingStr;
316 case eFoundAccountingStr:
317 /* Already found matchstring - look for cycle data */
318 if (str_starts(line, "Total "))
320 sscanf(line, "Total %d %lf", &procs, &(perfdata->Gcycles[test_nr]));
321 iFound = eFoundCycleStr;
325 /* Already found cycle data - look for remaining performance info and return */
326 if (str_starts(line, "Performance:"))
328 ndum = sscanf(line, "%s %f %f %f %f", dumstring, &dum1, &dum2, &dum3, &dum4);
329 /* (ns/day) is the second last entry, depending on whether GMX_DETAILED_PERF_STATS was set in print_perf(), nrnb.c */
330 perfdata->ns_per_day[test_nr] = (ndum == 5) ? dum3 : dum1;
332 if (bResetChecked || presteps == 0)
338 return eParselogResetProblem;
345 /* Close the log file */
348 /* Check why there is no performance data in the log file.
349 * Did a fatal errors occur? */
350 if (gmx_fexist(errfile))
352 fp = fopen(errfile, "r");
353 while (fgets(line, STRLEN, fp) != NULL)
355 if (str_starts(line, "Fatal error:") )
357 if (fgets(line, STRLEN, fp) != NULL)
359 fprintf(stderr, "\nWARNING: An error occured during this benchmark:\n"
363 cleandata(perfdata, test_nr);
364 return eParselogFatal;
371 fprintf(stderr, "WARNING: Could not find stderr file %s.\n", errfile);
374 /* Giving up ... we could not find out why there is no performance data in
376 fprintf(stdout, "No performance data in log file.\n");
377 cleandata(perfdata, test_nr);
379 return eParselogNoPerfData;
383 static gmx_bool analyze_data(
392 int *index_tpr, /* OUT: Nr of mdp file with best settings */
393 int *npme_optimal) /* OUT: Optimal number of PME nodes */
396 int line = 0, line_win = -1;
397 int k_win = -1, i_win = -1, winPME;
398 double s = 0.0; /* standard deviation */
401 char str_PME_f_load[13];
402 gmx_bool bCanUseOrigTPR;
403 gmx_bool bRefinedCoul, bRefinedVdW, bRefinedGrid;
409 fprintf(fp, "Summary of successful runs:\n");
410 fprintf(fp, "Line tpr PME nodes Gcycles Av. Std.dev. ns/day PME/f");
413 fprintf(fp, " DD grid");
419 for (k = 0; k < ntprs; k++)
421 for (i = 0; i < ntests; i++)
423 /* Select the right dataset: */
424 pd = &(perfdata[k][i]);
426 pd->Gcycles_Av = 0.0;
427 pd->PME_f_load_Av = 0.0;
428 pd->ns_per_day_Av = 0.0;
430 if (pd->nPMEnodes == -1)
432 sprintf(strbuf, "(%3d)", pd->guessPME);
436 sprintf(strbuf, " ");
439 /* Get the average run time of a setting */
440 for (j = 0; j < nrepeats; j++)
442 pd->Gcycles_Av += pd->Gcycles[j];
443 pd->PME_f_load_Av += pd->PME_f_load[j];
445 pd->Gcycles_Av /= nrepeats;
446 pd->PME_f_load_Av /= nrepeats;
448 for (j = 0; j < nrepeats; j++)
450 if (pd->ns_per_day[j] > 0.0)
452 pd->ns_per_day_Av += pd->ns_per_day[j];
456 /* Somehow the performance number was not aquired for this run,
457 * therefor set the average to some negative value: */
458 pd->ns_per_day_Av = -1.0f*nrepeats;
462 pd->ns_per_day_Av /= nrepeats;
465 if (pd->PME_f_load_Av > 0.0)
467 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load_Av);
471 sprintf(str_PME_f_load, "%s", " - ");
475 /* We assume we had a successful run if both averages are positive */
476 if (pd->Gcycles_Av > 0.0 && pd->ns_per_day_Av > 0.0)
478 /* Output statistics if repeats were done */
481 /* Calculate the standard deviation */
483 for (j = 0; j < nrepeats; j++)
485 s += pow( pd->Gcycles[j] - pd->Gcycles_Av, 2 );
490 fprintf(fp, "%4d %3d %4d%s %12.3f %12.3f %12.3f %s",
491 line, k, pd->nPMEnodes, strbuf, pd->Gcycles_Av, s,
492 pd->ns_per_day_Av, str_PME_f_load);
495 fprintf(fp, " %3d %3d %3d", pd->nx, pd->ny, pd->nz);
499 /* Store the index of the best run found so far in 'winner': */
500 if ( (k_win == -1) || (pd->Gcycles_Av < perfdata[k_win][i_win].Gcycles_Av) )
513 gmx_fatal(FARGS, "None of the runs was successful! Check %s for problems.", fn);
518 winPME = perfdata[k_win][i_win].nPMEnodes;
522 /* We stuck to a fixed number of PME-only nodes */
523 sprintf(strbuf, "settings No. %d", k_win);
527 /* We have optimized the number of PME-only nodes */
530 sprintf(strbuf, "%s", "the automatic number of PME nodes");
534 sprintf(strbuf, "%d PME nodes", winPME);
537 fprintf(fp, "Best performance was achieved with %s", strbuf);
538 if ((nrepeats > 1) && (ntests > 1))
540 fprintf(fp, " (see line %d)", line_win);
544 /* Only mention settings if they were modified: */
545 bRefinedCoul = !is_equal(info->rcoulomb[k_win], info->rcoulomb[0]);
546 bRefinedVdW = !is_equal(info->rvdw[k_win], info->rvdw[0] );
547 bRefinedGrid = !(info->nkx[k_win] == info->nkx[0] &&
548 info->nky[k_win] == info->nky[0] &&
549 info->nkz[k_win] == info->nkz[0]);
551 if (bRefinedCoul || bRefinedVdW || bRefinedGrid)
553 fprintf(fp, "Optimized PME settings:\n");
554 bCanUseOrigTPR = FALSE;
558 bCanUseOrigTPR = TRUE;
563 fprintf(fp, " New Coulomb radius: %f nm (was %f nm)\n", info->rcoulomb[k_win], info->rcoulomb[0]);
568 fprintf(fp, " New Van der Waals radius: %f nm (was %f nm)\n", info->rvdw[k_win], info->rvdw[0]);
573 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],
574 info->nkx[0], info->nky[0], info->nkz[0]);
577 if (bCanUseOrigTPR && ntprs > 1)
579 fprintf(fp, "and original PME settings.\n");
584 /* Return the index of the mdp file that showed the highest performance
585 * and the optimal number of PME nodes */
587 *npme_optimal = winPME;
589 return bCanUseOrigTPR;
593 /* Get the commands we need to set up the runs from environment variables */
594 static void get_program_paths(gmx_bool bThreads, char *cmd_mpirun[], char *cmd_mdrun[])
598 const char def_mpirun[] = "mpirun";
599 const char def_mdrun[] = "mdrun";
601 const char empty_mpirun[] = "";
603 /* Get the commands we need to set up the runs from environment variables */
606 if ( (cp = getenv("MPIRUN")) != NULL)
608 *cmd_mpirun = strdup(cp);
612 *cmd_mpirun = strdup(def_mpirun);
617 *cmd_mpirun = strdup(empty_mpirun);
620 if ( (cp = getenv("MDRUN" )) != NULL)
622 *cmd_mdrun = strdup(cp);
626 *cmd_mdrun = strdup(def_mdrun);
630 /* Check that the commands will run mdrun (perhaps via mpirun) by
631 * running a very quick test simulation. Requires MPI environment to
632 * be available if applicable. */
633 static void check_mdrun_works(gmx_bool bThreads,
634 const char *cmd_mpirun,
636 const char *cmd_mdrun)
638 char *command = NULL;
642 const char filename[] = "benchtest.log";
644 /* This string should always be identical to the one in copyrite.c,
645 * gmx_print_version_info() in the defined(GMX_MPI) section */
646 const char match_mpi[] = "MPI library: MPI";
647 const char match_mdrun[] = "Program: ";
648 gmx_bool bMdrun = FALSE;
649 gmx_bool bMPI = FALSE;
651 /* Run a small test to see whether mpirun + mdrun work */
652 fprintf(stdout, "Making sure that mdrun can be executed. ");
655 snew(command, strlen(cmd_mdrun) + strlen(cmd_np) + strlen(filename) + 50);
656 sprintf(command, "%s%s-version -maxh 0.001 1> %s 2>&1", cmd_mdrun, cmd_np, filename);
660 snew(command, strlen(cmd_mpirun) + strlen(cmd_np) + strlen(cmd_mdrun) + strlen(filename) + 50);
661 sprintf(command, "%s%s%s -version -maxh 0.001 1> %s 2>&1", cmd_mpirun, cmd_np, cmd_mdrun, filename);
663 fprintf(stdout, "Trying '%s' ... ", command);
664 make_backup(filename);
665 gmx_system_call(command);
667 /* Check if we find the characteristic string in the output: */
668 if (!gmx_fexist(filename))
670 gmx_fatal(FARGS, "Output from test run could not be found.");
673 fp = fopen(filename, "r");
674 /* We need to scan the whole output file, since sometimes the queuing system
675 * also writes stuff to stdout/err */
678 cp = fgets(line, STRLEN, fp);
681 if (str_starts(line, match_mdrun) )
685 if (str_starts(line, match_mpi) )
697 gmx_fatal(FARGS, "Need a threaded version of mdrun. This one\n"
699 "seems to have been compiled with MPI instead.",
707 gmx_fatal(FARGS, "Need an MPI-enabled version of mdrun. This one\n"
709 "seems to have been compiled without MPI support.",
716 gmx_fatal(FARGS, "Cannot execute mdrun. Please check %s for problems!",
720 fprintf(stdout, "passed.\n");
728 static void launch_simulation(
729 gmx_bool bLaunch, /* Should the simulation be launched? */
730 FILE *fp, /* General log file */
731 gmx_bool bThreads, /* whether to use threads */
732 char *cmd_mpirun, /* Command for mpirun */
733 char *cmd_np, /* Switch for -np or -ntmpi or empty */
734 char *cmd_mdrun, /* Command for mdrun */
735 char *args_for_mdrun, /* Arguments for mdrun */
736 const char *simulation_tpr, /* This tpr will be simulated */
737 int nPMEnodes) /* Number of PME nodes to use */
742 /* Make enough space for the system call command,
743 * (100 extra chars for -npme ... etc. options should suffice): */
744 snew(command, strlen(cmd_mpirun)+strlen(cmd_mdrun)+strlen(cmd_np)+strlen(args_for_mdrun)+strlen(simulation_tpr)+100);
746 /* Note that the -passall options requires args_for_mdrun to be at the end
747 * of the command line string */
750 sprintf(command, "%s%s-npme %d -s %s %s",
751 cmd_mdrun, cmd_np, nPMEnodes, simulation_tpr, args_for_mdrun);
755 sprintf(command, "%s%s%s -npme %d -s %s %s",
756 cmd_mpirun, cmd_np, cmd_mdrun, nPMEnodes, simulation_tpr, args_for_mdrun);
759 fprintf(fp, "%s this command line to launch the simulation:\n\n%s", bLaunch ? "Using" : "Please use", command);
763 /* Now the real thing! */
766 fprintf(stdout, "\nLaunching simulation with best parameters now.\nExecuting '%s'", command);
769 gmx_system_call(command);
774 static void modify_PMEsettings(
775 gmx_large_int_t simsteps, /* Set this value as number of time steps */
776 gmx_large_int_t init_step, /* Set this value as init_step */
777 const char *fn_best_tpr, /* tpr file with the best performance */
778 const char *fn_sim_tpr) /* name of tpr file to be launched */
786 read_tpx_state(fn_best_tpr, ir, &state, NULL, &mtop);
788 /* Reset nsteps and init_step to the value of the input .tpr file */
789 ir->nsteps = simsteps;
790 ir->init_step = init_step;
792 /* Write the tpr file which will be launched */
793 sprintf(buf, "Writing optimized simulation file %s with nsteps=%s.\n", fn_sim_tpr, gmx_large_int_pfmt);
794 fprintf(stdout, buf, ir->nsteps);
796 write_tpx_state(fn_sim_tpr, ir, &state, &mtop);
802 #define EPME_SWITCHED(e) ((e) == eelPMESWITCH || (e) == eelPMEUSERSWITCH)
804 /* Make additional TPR files with more computational load for the
805 * direct space processors: */
806 static void make_benchmark_tprs(
807 const char *fn_sim_tpr, /* READ : User-provided tpr file */
808 char *fn_bench_tprs[], /* WRITE: Names of benchmark tpr files */
809 gmx_large_int_t benchsteps, /* Number of time steps for benchmark runs */
810 gmx_large_int_t statesteps, /* Step counter in checkpoint file */
811 real rmin, /* Minimal Coulomb radius */
812 real rmax, /* Maximal Coulomb radius */
813 real bScaleRvdw, /* Scale rvdw along with rcoulomb */
814 int *ntprs, /* No. of TPRs to write, each with a different
815 rcoulomb and fourierspacing */
816 t_inputinfo *info, /* Contains information about mdp file options */
817 FILE *fp) /* Write the output here */
823 real nlist_buffer; /* Thickness of the buffer regions for PME-switch potentials */
826 gmx_bool bNote = FALSE;
827 real add; /* Add this to rcoul for the next test */
828 real fac = 1.0; /* Scaling factor for Coulomb radius */
829 real fourierspacing; /* Basic fourierspacing from tpr */
832 sprintf(buf, "Making benchmark tpr file%s with %s time step%s",
833 *ntprs > 1 ? "s" : "", gmx_large_int_pfmt, benchsteps > 1 ? "s" : "");
834 fprintf(stdout, buf, benchsteps);
837 sprintf(buf, " (adding %s steps from checkpoint file)", gmx_large_int_pfmt);
838 fprintf(stdout, buf, statesteps);
839 benchsteps += statesteps;
841 fprintf(stdout, ".\n");
845 read_tpx_state(fn_sim_tpr, ir, &state, NULL, &mtop);
847 /* Check if some kind of PME was chosen */
848 if (EEL_PME(ir->coulombtype) == FALSE)
850 gmx_fatal(FARGS, "Can only do optimizations for simulations with %s electrostatics.",
854 /* Check if rcoulomb == rlist, which is necessary for plain PME. */
855 if ( (ir->cutoff_scheme != ecutsVERLET) &&
856 (eelPME == ir->coulombtype) && !(ir->rcoulomb == ir->rlist))
858 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to rlist (%f).",
859 EELTYPE(eelPME), ir->rcoulomb, ir->rlist);
861 /* For other PME types, rcoulomb is allowed to be smaller than rlist */
862 else if (ir->rcoulomb > ir->rlist)
864 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to or smaller than rlist (%f)",
865 EELTYPE(ir->coulombtype), ir->rcoulomb, ir->rlist);
868 if (bScaleRvdw && ir->rvdw != ir->rcoulomb)
870 fprintf(stdout, "NOTE: input rvdw != rcoulomb, will not scale rvdw\n");
874 /* Reduce the number of steps for the benchmarks */
875 info->orig_sim_steps = ir->nsteps;
876 ir->nsteps = benchsteps;
877 /* We must not use init_step from the input tpr file for the benchmarks */
878 info->orig_init_step = ir->init_step;
881 /* For PME-switch potentials, keep the radial distance of the buffer region */
882 nlist_buffer = ir->rlist - ir->rcoulomb;
884 /* Determine length of triclinic box vectors */
885 for (d = 0; d < DIM; d++)
888 for (i = 0; i < DIM; i++)
890 box_size[d] += state.box[d][i]*state.box[d][i];
892 box_size[d] = sqrt(box_size[d]);
895 if (ir->fourier_spacing > 0)
897 info->fsx[0] = ir->fourier_spacing;
898 info->fsy[0] = ir->fourier_spacing;
899 info->fsz[0] = ir->fourier_spacing;
903 /* Reconstruct fourierspacing per dimension from the number of grid points and box size */
904 info->fsx[0] = box_size[XX]/ir->nkx;
905 info->fsy[0] = box_size[YY]/ir->nky;
906 info->fsz[0] = box_size[ZZ]/ir->nkz;
909 /* If no value for the fourierspacing was provided on the command line, we
910 * use the reconstruction from the tpr file */
911 if (ir->fourier_spacing > 0)
913 /* Use the spacing from the tpr */
914 fourierspacing = ir->fourier_spacing;
918 /* Use the maximum observed spacing */
919 fourierspacing = max(max(info->fsx[0], info->fsy[0]), info->fsz[0]);
922 fprintf(stdout, "Calculating PME grid points on the basis of a fourierspacing of %f nm\n", fourierspacing);
924 /* For performance comparisons the number of particles is useful to have */
925 fprintf(fp, " Number of particles : %d\n", mtop.natoms);
927 /* Print information about settings of which some are potentially modified: */
928 fprintf(fp, " Coulomb type : %s\n", EELTYPE(ir->coulombtype));
929 fprintf(fp, " Grid spacing x y z : %f %f %f\n",
930 box_size[XX]/ir->nkx, box_size[YY]/ir->nky, box_size[ZZ]/ir->nkz);
931 fprintf(fp, " Van der Waals type : %s\n", EVDWTYPE(ir->vdwtype));
932 if (EVDW_SWITCHED(ir->vdwtype))
934 fprintf(fp, " rvdw_switch : %f nm\n", ir->rvdw_switch);
936 if (EPME_SWITCHED(ir->coulombtype))
938 fprintf(fp, " rlist : %f nm\n", ir->rlist);
940 if (ir->rlistlong != max_cutoff(ir->rvdw, ir->rcoulomb))
942 fprintf(fp, " rlistlong : %f nm\n", ir->rlistlong);
945 /* Print a descriptive line about the tpr settings tested */
946 fprintf(fp, "\nWill try these real/reciprocal workload settings:\n");
947 fprintf(fp, " No. scaling rcoulomb");
948 fprintf(fp, " nkx nky nkz");
949 fprintf(fp, " spacing");
950 if (evdwCUT == ir->vdwtype)
952 fprintf(fp, " rvdw");
954 if (EPME_SWITCHED(ir->coulombtype))
956 fprintf(fp, " rlist");
958 if (ir->rlistlong != max_cutoff(ir->rlist, max_cutoff(ir->rvdw, ir->rcoulomb)) )
960 fprintf(fp, " rlistlong");
962 fprintf(fp, " tpr file\n");
964 /* Loop to create the requested number of tpr input files */
965 for (j = 0; j < *ntprs; j++)
967 /* The first .tpr is the provided one, just need to modify nsteps,
968 * so skip the following block */
971 /* Determine which Coulomb radii rc to use in the benchmarks */
972 add = (rmax-rmin)/(*ntprs-1);
973 if (is_equal(rmin, info->rcoulomb[0]))
975 ir->rcoulomb = rmin + j*add;
977 else if (is_equal(rmax, info->rcoulomb[0]))
979 ir->rcoulomb = rmin + (j-1)*add;
983 /* rmin != rcoul != rmax, ergo test between rmin and rmax */
984 add = (rmax-rmin)/(*ntprs-2);
985 ir->rcoulomb = rmin + (j-1)*add;
988 /* Determine the scaling factor fac */
989 fac = ir->rcoulomb/info->rcoulomb[0];
991 /* Scale the Fourier grid spacing */
992 ir->nkx = ir->nky = ir->nkz = 0;
993 calc_grid(NULL, state.box, fourierspacing*fac, &ir->nkx, &ir->nky, &ir->nkz);
995 /* Adjust other radii since various conditions neet to be fulfilled */
996 if (eelPME == ir->coulombtype)
998 /* plain PME, rcoulomb must be equal to rlist */
999 ir->rlist = ir->rcoulomb;
1003 /* rlist must be >= rcoulomb, we keep the size of the buffer region */
1004 ir->rlist = ir->rcoulomb + nlist_buffer;
1007 if (bScaleRvdw && evdwCUT == ir->vdwtype)
1009 /* For vdw cutoff, rvdw >= rlist */
1010 ir->rvdw = max(info->rvdw[0], ir->rlist);
1013 ir->rlistlong = max_cutoff(ir->rlist, max_cutoff(ir->rvdw, ir->rcoulomb));
1015 } /* end of "if (j != 0)" */
1017 /* for j==0: Save the original settings
1018 * for j >0: Save modified radii and Fourier grids */
1019 info->rcoulomb[j] = ir->rcoulomb;
1020 info->rvdw[j] = ir->rvdw;
1021 info->nkx[j] = ir->nkx;
1022 info->nky[j] = ir->nky;
1023 info->nkz[j] = ir->nkz;
1024 info->rlist[j] = ir->rlist;
1025 info->rlistlong[j] = ir->rlistlong;
1026 info->fsx[j] = fac*fourierspacing;
1027 info->fsy[j] = fac*fourierspacing;
1028 info->fsz[j] = fac*fourierspacing;
1030 /* Write the benchmark tpr file */
1031 strncpy(fn_bench_tprs[j], fn_sim_tpr, strlen(fn_sim_tpr)-strlen(".tpr"));
1032 sprintf(buf, "_bench%.2d.tpr", j);
1033 strcat(fn_bench_tprs[j], buf);
1034 fprintf(stdout, "Writing benchmark tpr %s with nsteps=", fn_bench_tprs[j]);
1035 fprintf(stdout, gmx_large_int_pfmt, ir->nsteps);
1038 fprintf(stdout, ", scaling factor %f\n", fac);
1042 fprintf(stdout, ", unmodified settings\n");
1045 write_tpx_state(fn_bench_tprs[j], ir, &state, &mtop);
1047 /* Write information about modified tpr settings to log file */
1048 fprintf(fp, "%4d%10f%10f", j, fac, ir->rcoulomb);
1049 fprintf(fp, "%5d%5d%5d", ir->nkx, ir->nky, ir->nkz);
1050 fprintf(fp, " %9f ", info->fsx[j]);
1051 if (evdwCUT == ir->vdwtype)
1053 fprintf(fp, "%10f", ir->rvdw);
1055 if (EPME_SWITCHED(ir->coulombtype))
1057 fprintf(fp, "%10f", ir->rlist);
1059 if (info->rlistlong[0] != max_cutoff(info->rlist[0], max_cutoff(info->rvdw[0], info->rcoulomb[0])) )
1061 fprintf(fp, "%10f", ir->rlistlong);
1063 fprintf(fp, " %-14s\n", fn_bench_tprs[j]);
1065 /* Make it clear to the user that some additional settings were modified */
1066 if (!is_equal(ir->rvdw, info->rvdw[0])
1067 || !is_equal(ir->rlistlong, info->rlistlong[0]) )
1074 fprintf(fp, "\nNote that in addition to the Coulomb radius and the Fourier grid\n"
1075 "other input settings were also changed (see table above).\n"
1076 "Please check if the modified settings are appropriate.\n");
1084 /* Rename the files we want to keep to some meaningful filename and
1085 * delete the rest */
1086 static void cleanup(const t_filenm *fnm, int nfile, int k, int nnodes,
1087 int nPMEnodes, int nr, gmx_bool bKeepStderr)
1089 char numstring[STRLEN];
1090 char newfilename[STRLEN];
1091 const char *fn = NULL;
1096 fprintf(stdout, "Cleaning up, deleting benchmark temp files ...\n");
1098 for (i = 0; i < nfile; i++)
1100 opt = (char *)fnm[i].opt;
1101 if (strcmp(opt, "-p") == 0)
1103 /* do nothing; keep this file */
1106 else if (strcmp(opt, "-bg") == 0)
1108 /* Give the log file a nice name so one can later see which parameters were used */
1109 numstring[0] = '\0';
1112 sprintf(numstring, "_%d", nr);
1114 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", opt2fn("-bg", nfile, fnm), k, nnodes, nPMEnodes, numstring);
1115 if (gmx_fexist(opt2fn("-bg", nfile, fnm)))
1117 fprintf(stdout, "renaming log file to %s\n", newfilename);
1118 make_backup(newfilename);
1119 rename(opt2fn("-bg", nfile, fnm), newfilename);
1122 else if (strcmp(opt, "-err") == 0)
1124 /* This file contains the output of stderr. We want to keep it in
1125 * cases where there have been problems. */
1126 fn = opt2fn(opt, nfile, fnm);
1127 numstring[0] = '\0';
1130 sprintf(numstring, "_%d", nr);
1132 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", fn, k, nnodes, nPMEnodes, numstring);
1137 fprintf(stdout, "Saving stderr output in %s\n", newfilename);
1138 make_backup(newfilename);
1139 rename(fn, newfilename);
1143 fprintf(stdout, "Deleting %s\n", fn);
1148 /* Delete the files which are created for each benchmark run: (options -b*) */
1149 else if ( (0 == strncmp(opt, "-b", 2)) && (opt2bSet(opt, nfile, fnm) || !is_optional(&fnm[i])) )
1151 fn = opt2fn(opt, nfile, fnm);
1154 fprintf(stdout, "Deleting %s\n", fn);
1162 /* Returns the largest common factor of n1 and n2 */
1163 static int largest_common_factor(int n1, int n2)
1168 for (factor = nmax; factor > 0; factor--)
1170 if (0 == (n1 % factor) && 0 == (n2 % factor) )
1175 return 0; /* one for the compiler */
1179 eNpmeAuto, eNpmeAll, eNpmeReduced, eNpmeSubset, eNpmeNr
1182 /* Create a list of numbers of PME nodes to test */
1183 static void make_npme_list(
1184 const char *npmevalues_opt, /* Make a complete list with all
1185 * possibilities or a short list that keeps only
1186 * reasonable numbers of PME nodes */
1187 int *nentries, /* Number of entries we put in the nPMEnodes list */
1188 int *nPMEnodes[], /* Each entry contains the value for -npme */
1189 int nnodes, /* Total number of nodes to do the tests on */
1190 int minPMEnodes, /* Minimum number of PME nodes */
1191 int maxPMEnodes) /* Maximum number of PME nodes */
1194 int min_factor = 1; /* We request that npp and npme have this minimal
1195 * largest common factor (depends on npp) */
1196 int nlistmax; /* Max. list size */
1197 int nlist; /* Actual number of entries in list */
1201 /* Do we need to check all possible values for -npme or is a reduced list enough? */
1202 if (0 == strcmp(npmevalues_opt, "all") )
1206 else if (0 == strcmp(npmevalues_opt, "subset") )
1208 eNPME = eNpmeSubset;
1210 else /* "auto" or "range" */
1216 else if (nnodes < 128)
1218 eNPME = eNpmeReduced;
1222 eNPME = eNpmeSubset;
1226 /* Calculate how many entries we could possibly have (in case of -npme all) */
1229 nlistmax = maxPMEnodes - minPMEnodes + 3;
1230 if (0 == minPMEnodes)
1240 /* Now make the actual list which is at most of size nlist */
1241 snew(*nPMEnodes, nlistmax);
1242 nlist = 0; /* start counting again, now the real entries in the list */
1243 for (i = 0; i < nlistmax - 2; i++)
1245 npme = maxPMEnodes - i;
1256 /* For 2d PME we want a common largest factor of at least the cube
1257 * root of the number of PP nodes */
1258 min_factor = (int) pow(npp, 1.0/3.0);
1261 gmx_fatal(FARGS, "Unknown option for eNPME in make_npme_list");
1264 if (largest_common_factor(npp, npme) >= min_factor)
1266 (*nPMEnodes)[nlist] = npme;
1270 /* We always test 0 PME nodes and the automatic number */
1271 *nentries = nlist + 2;
1272 (*nPMEnodes)[nlist ] = 0;
1273 (*nPMEnodes)[nlist+1] = -1;
1275 fprintf(stderr, "Will try the following %d different values for -npme:\n", *nentries);
1276 for (i = 0; i < *nentries-1; i++)
1278 fprintf(stderr, "%d, ", (*nPMEnodes)[i]);
1280 fprintf(stderr, "and %d (auto).\n", (*nPMEnodes)[*nentries-1]);
1284 /* Allocate memory to store the performance data */
1285 static void init_perfdata(t_perf *perfdata[], int ntprs, int datasets, int repeats)
1290 for (k = 0; k < ntprs; k++)
1292 snew(perfdata[k], datasets);
1293 for (i = 0; i < datasets; i++)
1295 for (j = 0; j < repeats; j++)
1297 snew(perfdata[k][i].Gcycles, repeats);
1298 snew(perfdata[k][i].ns_per_day, repeats);
1299 snew(perfdata[k][i].PME_f_load, repeats);
1306 /* Check for errors on mdrun -h */
1307 static void make_sure_it_runs(char *mdrun_cmd_line, int length, FILE *fp)
1309 char *command, *msg;
1313 snew(command, length + 15);
1314 snew(msg, length + 500);
1316 fprintf(stdout, "Making sure the benchmarks can be executed ...\n");
1317 /* FIXME: mdrun -h no longer actually does anything useful.
1318 * It unconditionally prints the help, ignoring all other options. */
1319 sprintf(command, "%s-h -quiet", mdrun_cmd_line);
1320 ret = gmx_system_call(command);
1324 /* To prevent confusion, do not again issue a gmx_fatal here since we already
1325 * get the error message from mdrun itself */
1326 sprintf(msg, "Cannot run the benchmark simulations! Please check the error message of\n"
1327 "mdrun for the source of the problem. Did you provide a command line\n"
1328 "argument that neither g_tune_pme nor mdrun understands? Offending command:\n"
1329 "\n%s\n\n", command);
1331 fprintf(stderr, "%s", msg);
1333 fprintf(fp, "%s", msg);
1343 static void do_the_tests(
1344 FILE *fp, /* General g_tune_pme output file */
1345 char **tpr_names, /* Filenames of the input files to test */
1346 int maxPMEnodes, /* Max fraction of nodes to use for PME */
1347 int minPMEnodes, /* Min fraction of nodes to use for PME */
1348 int npme_fixed, /* If >= -1, test fixed number of PME
1350 const char *npmevalues_opt, /* Which -npme values should be tested */
1351 t_perf **perfdata, /* Here the performace data is stored */
1352 int *pmeentries, /* Entries in the nPMEnodes list */
1353 int repeats, /* Repeat each test this often */
1354 int nnodes, /* Total number of nodes = nPP + nPME */
1355 int nr_tprs, /* Total number of tpr files to test */
1356 gmx_bool bThreads, /* Threads or MPI? */
1357 char *cmd_mpirun, /* mpirun command string */
1358 char *cmd_np, /* "-np", "-n", whatever mpirun needs */
1359 char *cmd_mdrun, /* mdrun command string */
1360 char *cmd_args_bench, /* arguments for mdrun in a string */
1361 const t_filenm *fnm, /* List of filenames from command line */
1362 int nfile, /* Number of files specified on the cmdl. */
1363 int presteps, /* DLB equilibration steps, is checked */
1364 gmx_large_int_t cpt_steps) /* Time step counter in the checkpoint */
1366 int i, nr, k, ret, count = 0, totaltests;
1367 int *nPMEnodes = NULL;
1370 char *command, *cmd_stub;
1372 gmx_bool bResetProblem = FALSE;
1373 gmx_bool bFirst = TRUE;
1376 /* This string array corresponds to the eParselog enum type at the start
1378 const char* ParseLog[] = {
1380 "Logfile not found!",
1381 "No timings, logfile truncated?",
1382 "Run was terminated.",
1383 "Counters were not reset properly.",
1384 "No DD grid found for these settings.",
1385 "TPX version conflict!",
1386 "mdrun was not started in parallel!",
1387 "Number of PP nodes has a prime factor that is too large.",
1390 char str_PME_f_load[13];
1393 /* Allocate space for the mdrun command line. 100 extra characters should
1394 be more than enough for the -npme etcetera arguments */
1395 cmdline_length = strlen(cmd_mpirun)
1398 + strlen(cmd_args_bench)
1399 + strlen(tpr_names[0]) + 100;
1400 snew(command, cmdline_length);
1401 snew(cmd_stub, cmdline_length);
1403 /* Construct the part of the command line that stays the same for all tests: */
1406 sprintf(cmd_stub, "%s%s", cmd_mdrun, cmd_np);
1410 sprintf(cmd_stub, "%s%s%s ", cmd_mpirun, cmd_np, cmd_mdrun);
1413 /* Create a list of numbers of PME nodes to test */
1414 if (npme_fixed < -1)
1416 make_npme_list(npmevalues_opt, pmeentries, &nPMEnodes,
1417 nnodes, minPMEnodes, maxPMEnodes);
1423 nPMEnodes[0] = npme_fixed;
1424 fprintf(stderr, "Will use a fixed number of %d PME-only nodes.\n", nPMEnodes[0]);
1429 fprintf(fp, "\nNo benchmarks done since number of repeats (-r) is 0.\n");
1431 finalize(opt2fn("-p", nfile, fnm));
1435 /* Allocate one dataset for each tpr input file: */
1436 init_perfdata(perfdata, nr_tprs, *pmeentries, repeats);
1438 /*****************************************/
1439 /* Main loop over all tpr files to test: */
1440 /*****************************************/
1441 totaltests = nr_tprs*(*pmeentries)*repeats;
1442 for (k = 0; k < nr_tprs; k++)
1444 fprintf(fp, "\nIndividual timings for input file %d (%s):\n", k, tpr_names[k]);
1445 fprintf(fp, "PME nodes Gcycles ns/day PME/f Remark\n");
1446 /* Loop over various numbers of PME nodes: */
1447 for (i = 0; i < *pmeentries; i++)
1449 pd = &perfdata[k][i];
1451 /* Loop over the repeats for each scenario: */
1452 for (nr = 0; nr < repeats; nr++)
1454 pd->nPMEnodes = nPMEnodes[i];
1456 /* Add -npme and -s to the command line and save it. Note that
1457 * the -passall (if set) options requires cmd_args_bench to be
1458 * at the end of the command line string */
1459 snew(pd->mdrun_cmd_line, cmdline_length);
1460 sprintf(pd->mdrun_cmd_line, "%s-npme %d -s %s %s",
1461 cmd_stub, pd->nPMEnodes, tpr_names[k], cmd_args_bench);
1463 /* To prevent that all benchmarks fail due to a show-stopper argument
1464 * on the mdrun command line, we make a quick check with mdrun -h first */
1467 make_sure_it_runs(pd->mdrun_cmd_line, cmdline_length, fp);
1471 /* Do a benchmark simulation: */
1474 sprintf(buf, ", pass %d/%d", nr+1, repeats);
1480 fprintf(stdout, "\n=== Progress %2.0f%%, tpr %d/%d, run %d/%d%s:\n",
1481 (100.0*count)/totaltests,
1482 k+1, nr_tprs, i+1, *pmeentries, buf);
1483 make_backup(opt2fn("-err", nfile, fnm));
1484 sprintf(command, "%s 1> /dev/null 2>%s", pd->mdrun_cmd_line, opt2fn("-err", nfile, fnm));
1485 fprintf(stdout, "%s\n", pd->mdrun_cmd_line);
1486 gmx_system_call(command);
1488 /* Collect the performance data from the log file; also check stderr
1489 * for fatal errors */
1490 ret = parse_logfile(opt2fn("-bg", nfile, fnm), opt2fn("-err", nfile, fnm),
1491 pd, nr, presteps, cpt_steps, nnodes);
1492 if ((presteps > 0) && (ret == eParselogResetProblem))
1494 bResetProblem = TRUE;
1497 if (-1 == pd->nPMEnodes)
1499 sprintf(buf, "(%3d)", pd->guessPME);
1507 if (pd->PME_f_load[nr] > 0.0)
1509 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load[nr]);
1513 sprintf(str_PME_f_load, "%s", " - ");
1516 /* Write the data we got to disk */
1517 fprintf(fp, "%4d%s %12.3f %12.3f %s %s", pd->nPMEnodes,
1518 buf, pd->Gcycles[nr], pd->ns_per_day[nr], str_PME_f_load, ParseLog[ret]);
1519 if (!(ret == eParselogOK || ret == eParselogNoDDGrid || ret == eParselogNotFound) )
1521 fprintf(fp, " Check %s file for problems.", ret == eParselogFatal ? "err" : "log");
1527 /* Do some cleaning up and delete the files we do not need any more */
1528 cleanup(fnm, nfile, k, nnodes, pd->nPMEnodes, nr, ret == eParselogFatal);
1530 /* If the first run with this number of processors already failed, do not try again: */
1531 if (pd->Gcycles[0] <= 0.0 && repeats > 1)
1533 fprintf(stdout, "Skipping remaining passes of unsuccessful setting, see log file for details.\n");
1534 count += repeats-(nr+1);
1537 } /* end of repeats loop */
1538 } /* end of -npme loop */
1539 } /* end of tpr file loop */
1544 fprintf(fp, "WARNING: The cycle and time step counters could not be reset properly. ");
1552 static void check_input(
1559 real maxPMEfraction,
1560 real minPMEfraction,
1562 gmx_large_int_t bench_nsteps,
1563 const t_filenm *fnm,
1573 /* Make sure the input file exists */
1574 if (!gmx_fexist(opt2fn("-s", nfile, fnm)))
1576 gmx_fatal(FARGS, "File %s not found.", opt2fn("-s", nfile, fnm));
1579 /* Make sure that the checkpoint file is not overwritten during benchmarking */
1580 if ( (0 == strcmp(opt2fn("-cpi", nfile, fnm), opt2fn("-bcpo", nfile, fnm)) ) && (sim_part > 1) )
1582 gmx_fatal(FARGS, "Checkpoint input (-cpi) and benchmark checkpoint output (-bcpo) files must not be identical.\n"
1583 "The checkpoint input file must not be overwritten during the benchmarks.\n");
1586 /* Make sure that repeats is >= 0 (if == 0, only write tpr files) */
1589 gmx_fatal(FARGS, "Number of repeats < 0!");
1592 /* Check number of nodes */
1595 gmx_fatal(FARGS, "Number of nodes/threads must be a positive integer.");
1598 /* Automatically choose -ntpr if not set */
1608 /* Set a reasonable scaling factor for rcoulomb */
1611 *rmax = rcoulomb * 1.2;
1614 fprintf(stderr, "Will test %d tpr file%s.\n", *ntprs, *ntprs == 1 ? "" : "s");
1620 fprintf(stderr, "Note: Choose ntpr>1 to shift PME load between real and reciprocal space.\n");
1624 /* Make shure that rmin <= rcoulomb <= rmax */
1633 if (!(*rmin <= *rmax) )
1635 gmx_fatal(FARGS, "Please choose the Coulomb radii such that rmin <= rmax.\n"
1636 "rmin = %g, rmax = %g, actual rcoul from .tpr file = %g\n", *rmin, *rmax, rcoulomb);
1638 /* Add test scenarios if rmin or rmax were set */
1641 if (!is_equal(*rmin, rcoulomb) && (*ntprs == 1) )
1644 fprintf(stderr, "NOTE: Setting -rmin to %g changed -ntpr to %d\n",
1647 if (!is_equal(*rmax, rcoulomb) && (*ntprs == 1) )
1650 fprintf(stderr, "NOTE: Setting -rmax to %g changed -ntpr to %d\n",
1655 /* If one of rmin, rmax is set, we need 2 tpr files at minimum */
1656 if (!is_equal(*rmax, rcoulomb) || !is_equal(*rmin, rcoulomb) )
1658 *ntprs = max(*ntprs, 2);
1661 /* If both rmin, rmax are set, we need 3 tpr files at minimum */
1662 if (!is_equal(*rmax, rcoulomb) && !is_equal(*rmin, rcoulomb) )
1664 *ntprs = max(*ntprs, 3);
1669 fprintf(stderr, "NOTE: Your rmin, rmax setting changed -ntpr to %d\n", *ntprs);
1674 if (is_equal(*rmin, rcoulomb) && is_equal(rcoulomb, *rmax)) /* We have just a single rc */
1676 fprintf(stderr, "WARNING: Resetting -ntpr to 1 since no Coulomb radius scaling is requested.\n"
1677 "Please set rmin < rmax to test Coulomb radii in the [rmin, rmax] interval\n"
1678 "with correspondingly adjusted PME grid settings\n");
1683 /* Check whether max and min fraction are within required values */
1684 if (maxPMEfraction > 0.5 || maxPMEfraction < 0)
1686 gmx_fatal(FARGS, "-max must be between 0 and 0.5");
1688 if (minPMEfraction > 0.5 || minPMEfraction < 0)
1690 gmx_fatal(FARGS, "-min must be between 0 and 0.5");
1692 if (maxPMEfraction < minPMEfraction)
1694 gmx_fatal(FARGS, "-max must be larger or equal to -min");
1697 /* Check whether the number of steps - if it was set - has a reasonable value */
1698 if (bench_nsteps < 0)
1700 gmx_fatal(FARGS, "Number of steps must be positive.");
1703 if (bench_nsteps > 10000 || bench_nsteps < 100)
1705 fprintf(stderr, "WARNING: steps=");
1706 fprintf(stderr, gmx_large_int_pfmt, bench_nsteps);
1707 fprintf(stderr, ". Are you sure you want to perform so %s steps for each benchmark?\n", (bench_nsteps < 100) ? "few" : "many");
1712 gmx_fatal(FARGS, "Cannot have a negative number of presteps.\n");
1715 /* Check for rcoulomb scaling if more than one .tpr file is tested */
1718 if (*rmin/rcoulomb < 0.75 || *rmax/rcoulomb > 1.25)
1720 fprintf(stderr, "WARNING: Applying extreme scaling factor. I hope you know what you are doing.\n");
1724 /* If a fixed number of PME nodes is set we do rcoulomb and PME gird tuning
1725 * only. We need to check whether the requested number of PME-only nodes
1727 if (npme_fixed > -1)
1729 /* No more than 50% of all nodes can be assigned as PME-only nodes. */
1730 if (2*npme_fixed > nnodes)
1732 gmx_fatal(FARGS, "Cannot have more than %d PME-only nodes for a total of %d nodes (you chose %d).\n",
1733 nnodes/2, nnodes, npme_fixed);
1735 if ((npme_fixed > 0) && (5*npme_fixed < nnodes))
1737 fprintf(stderr, "WARNING: Only %g percent of the nodes are assigned as PME-only nodes.\n",
1738 100.0*((real)npme_fixed / (real)nnodes));
1740 if (opt2parg_bSet("-min", npargs, pa) || opt2parg_bSet("-max", npargs, pa))
1742 fprintf(stderr, "NOTE: The -min, -max, and -npme options have no effect when a\n"
1743 " fixed number of PME-only nodes is requested with -fix.\n");
1749 /* Returns TRUE when "opt" is needed at launch time */
1750 static gmx_bool is_launch_file(char *opt, gmx_bool bSet)
1752 /* Apart from the input .tpr and the output log files we need all options that
1753 * were set on the command line and that do not start with -b */
1754 if (0 == strncmp(opt, "-b", 2) || 0 == strncmp(opt, "-s", 2)
1755 || 0 == strncmp(opt, "-err", 4) || 0 == strncmp(opt, "-p", 2) )
1764 /* Returns TRUE when "opt" defines a file which is needed for the benchmarks runs */
1765 static gmx_bool is_bench_file(char *opt, gmx_bool bSet, gmx_bool bOptional, gmx_bool bIsOutput)
1767 /* Apart from the input .tpr, all files starting with "-b" are for
1768 * _b_enchmark files exclusively */
1769 if (0 == strncmp(opt, "-s", 2))
1774 if (0 == strncmp(opt, "-b", 2) || 0 == strncmp(opt, "-s", 2))
1776 if (!bOptional || bSet)
1793 if (bSet) /* These are additional input files like -cpi -ei */
1806 /* Adds 'buf' to 'str' */
1807 static void add_to_string(char **str, char *buf)
1812 len = strlen(*str) + strlen(buf) + 1;
1818 /* Create the command line for the benchmark as well as for the real run */
1819 static void create_command_line_snippets(
1820 gmx_bool bAppendFiles,
1821 gmx_bool bKeepAndNumCPT,
1822 gmx_bool bResetHWay,
1826 char *cmd_args_bench[], /* command line arguments for benchmark runs */
1827 char *cmd_args_launch[], /* command line arguments for simulation run */
1828 char extra_args[]) /* Add this to the end of the command line */
1833 char strbuf[STRLEN];
1836 /* strlen needs at least '\0' as a string: */
1837 snew(*cmd_args_bench, 1);
1838 snew(*cmd_args_launch, 1);
1839 *cmd_args_launch[0] = '\0';
1840 *cmd_args_bench[0] = '\0';
1843 /*******************************************/
1844 /* 1. Process other command line arguments */
1845 /*******************************************/
1848 /* Add equilibration steps to benchmark options */
1849 sprintf(strbuf, "-resetstep %d ", presteps);
1850 add_to_string(cmd_args_bench, strbuf);
1852 /* These switches take effect only at launch time */
1853 if (FALSE == bAppendFiles)
1855 add_to_string(cmd_args_launch, "-noappend ");
1859 add_to_string(cmd_args_launch, "-cpnum ");
1863 add_to_string(cmd_args_launch, "-resethway ");
1866 /********************/
1867 /* 2. Process files */
1868 /********************/
1869 for (i = 0; i < nfile; i++)
1871 opt = (char *)fnm[i].opt;
1872 name = opt2fn(opt, nfile, fnm);
1874 /* Strbuf contains the options, now let's sort out where we need that */
1875 sprintf(strbuf, "%s %s ", opt, name);
1877 if (is_bench_file(opt, opt2bSet(opt, nfile, fnm), is_optional(&fnm[i]), is_output(&fnm[i])) )
1879 /* All options starting with -b* need the 'b' removed,
1880 * therefore overwrite strbuf */
1881 if (0 == strncmp(opt, "-b", 2))
1883 sprintf(strbuf, "-%s %s ", &opt[2], name);
1886 add_to_string(cmd_args_bench, strbuf);
1889 if (is_launch_file(opt, opt2bSet(opt, nfile, fnm)) )
1891 add_to_string(cmd_args_launch, strbuf);
1895 add_to_string(cmd_args_bench, extra_args);
1896 add_to_string(cmd_args_launch, extra_args);
1900 /* Set option opt */
1901 static void setopt(const char *opt, int nfile, t_filenm fnm[])
1905 for (i = 0; (i < nfile); i++)
1907 if (strcmp(opt, fnm[i].opt) == 0)
1909 fnm[i].flag |= ffSET;
1915 /* This routine inspects the tpr file and ...
1916 * 1. checks for output files that get triggered by a tpr option. These output
1917 * files are marked as 'set' to allow for a proper cleanup after each
1919 * 2. returns the PME:PP load ratio
1920 * 3. returns rcoulomb from the tpr */
1921 static float inspect_tpr(int nfile, t_filenm fnm[], real *rcoulomb)
1923 gmx_bool bPull; /* Is pulling requested in .tpr file? */
1924 gmx_bool bTpi; /* Is test particle insertion requested? */
1925 gmx_bool bFree; /* Is a free energy simulation requested? */
1926 gmx_bool bNM; /* Is a normal mode analysis requested? */
1932 /* Check tpr file for options that trigger extra output files */
1933 read_tpx_state(opt2fn("-s", nfile, fnm), &ir, &state, NULL, &mtop);
1934 bPull = (epullNO != ir.ePull);
1935 bFree = (efepNO != ir.efep );
1936 bNM = (eiNM == ir.eI );
1937 bTpi = EI_TPI(ir.eI);
1939 /* Set these output files on the tuning command-line */
1942 setopt("-pf", nfile, fnm);
1943 setopt("-px", nfile, fnm);
1947 setopt("-dhdl", nfile, fnm);
1951 setopt("-tpi", nfile, fnm);
1952 setopt("-tpid", nfile, fnm);
1956 setopt("-mtx", nfile, fnm);
1959 *rcoulomb = ir.rcoulomb;
1961 /* Return the estimate for the number of PME nodes */
1962 return pme_load_estimate(&mtop, &ir, state.box);
1966 static void couple_files_options(int nfile, t_filenm fnm[])
1969 gmx_bool bSet, bBench;
1974 for (i = 0; i < nfile; i++)
1976 opt = (char *)fnm[i].opt;
1977 bSet = ((fnm[i].flag & ffSET) != 0);
1978 bBench = (0 == strncmp(opt, "-b", 2));
1980 /* Check optional files */
1981 /* If e.g. -eo is set, then -beo also needs to be set */
1982 if (is_optional(&fnm[i]) && bSet && !bBench)
1984 sprintf(buf, "-b%s", &opt[1]);
1985 setopt(buf, nfile, fnm);
1987 /* If -beo is set, then -eo also needs to be! */
1988 if (is_optional(&fnm[i]) && bSet && bBench)
1990 sprintf(buf, "-%s", &opt[2]);
1991 setopt(buf, nfile, fnm);
1997 #define BENCHSTEPS (1000)
1999 int gmx_tune_pme(int argc, char *argv[])
2001 const char *desc[] = {
2002 "For a given number [TT]-np[tt] or [TT]-ntmpi[tt] of processors/threads, [THISMODULE] systematically",
2003 "times [gmx-mdrun] with various numbers of PME-only nodes and determines",
2004 "which setting is fastest. It will also test whether performance can",
2005 "be enhanced by shifting load from the reciprocal to the real space",
2006 "part of the Ewald sum. ",
2007 "Simply pass your [TT].tpr[tt] file to [THISMODULE] together with other options",
2008 "for [gmx-mdrun] as needed.[PAR]",
2009 "Which executables are used can be set in the environment variables",
2010 "MPIRUN and MDRUN. If these are not present, 'mpirun' and 'mdrun'",
2011 "will be used as defaults. Note that for certain MPI frameworks you",
2012 "need to provide a machine- or hostfile. This can also be passed",
2013 "via the MPIRUN variable, e.g.[PAR]",
2014 "[TT]export MPIRUN=\"/usr/local/mpirun -machinefile hosts\"[tt][PAR]",
2015 "Please call [THISMODULE] with the normal options you would pass to",
2016 "[gmx-mdrun] and add [TT]-np[tt] for the number of processors to perform the",
2017 "tests on, or [TT]-ntmpi[tt] for the number of threads. You can also add [TT]-r[tt]",
2018 "to repeat each test several times to get better statistics. [PAR]",
2019 "[THISMODULE] can test various real space / reciprocal space workloads",
2020 "for you. With [TT]-ntpr[tt] you control how many extra [TT].tpr[tt] files will be",
2021 "written with enlarged cutoffs and smaller Fourier grids respectively.",
2022 "Typically, the first test (number 0) will be with the settings from the input",
2023 "[TT].tpr[tt] file; the last test (number [TT]ntpr[tt]) will have the Coulomb cutoff",
2024 "specified by [TT]-rmax[tt] with a somwhat smaller PME grid at the same time. ",
2025 "In this last test, the Fourier spacing is multiplied with [TT]rmax[tt]/rcoulomb. ",
2026 "The remaining [TT].tpr[tt] files will have equally-spaced Coulomb radii (and Fourier "
2027 "spacings) between these extremes. [BB]Note[bb] that you can set [TT]-ntpr[tt] to 1",
2028 "if you just seek the optimal number of PME-only nodes; in that case",
2029 "your input [TT].tpr[tt] file will remain unchanged.[PAR]",
2030 "For the benchmark runs, the default of 1000 time steps should suffice for most",
2031 "MD systems. The dynamic load balancing needs about 100 time steps",
2032 "to adapt to local load imbalances, therefore the time step counters",
2033 "are by default reset after 100 steps. For large systems (>1M atoms), as well as ",
2034 "for a higher accuarcy of the measurements, you should set [TT]-resetstep[tt] to a higher value.",
2035 "From the 'DD' load imbalance entries in the md.log output file you",
2036 "can tell after how many steps the load is sufficiently balanced. Example call:[PAR]"
2037 "[TT]gmx tune_pme -np 64 -s protein.tpr -launch[tt][PAR]",
2038 "After calling [gmx-mdrun] several times, detailed performance information",
2039 "is available in the output file [TT]perf.out[tt].",
2040 "[BB]Note[bb] that during the benchmarks, a couple of temporary files are written",
2041 "(options [TT]-b[tt]*), these will be automatically deleted after each test.[PAR]",
2042 "If you want the simulation to be started automatically with the",
2043 "optimized parameters, use the command line option [TT]-launch[tt].[PAR]",
2048 int pmeentries = 0; /* How many values for -npme do we actually test for each tpr file */
2049 real maxPMEfraction = 0.50;
2050 real minPMEfraction = 0.25;
2051 int maxPMEnodes, minPMEnodes;
2052 float guessPMEratio; /* guessed PME:PP ratio based on the tpr file */
2053 float guessPMEnodes;
2054 int npme_fixed = -2; /* If >= -1, use only this number
2055 * of PME-only nodes */
2057 real rmin = 0.0, rmax = 0.0; /* min and max value for rcoulomb if scaling is requested */
2058 real rcoulomb = -1.0; /* Coulomb radius as set in .tpr file */
2059 gmx_bool bScaleRvdw = TRUE;
2060 gmx_large_int_t bench_nsteps = BENCHSTEPS;
2061 gmx_large_int_t new_sim_nsteps = -1; /* -1 indicates: not set by the user */
2062 gmx_large_int_t cpt_steps = 0; /* Step counter in .cpt input file */
2063 int presteps = 100; /* Do a full cycle reset after presteps steps */
2064 gmx_bool bOverwrite = FALSE, bKeepTPR;
2065 gmx_bool bLaunch = FALSE;
2066 char *ExtraArgs = NULL;
2067 char **tpr_names = NULL;
2068 const char *simulation_tpr = NULL;
2069 int best_npme, best_tpr;
2070 int sim_part = 1; /* For benchmarks with checkpoint files */
2073 /* Default program names if nothing else is found */
2074 char *cmd_mpirun = NULL, *cmd_mdrun = NULL;
2075 char *cmd_args_bench, *cmd_args_launch;
2076 char *cmd_np = NULL;
2078 t_perf **perfdata = NULL;
2084 /* Print out how long the tuning took */
2087 static t_filenm fnm[] = {
2089 { efOUT, "-p", "perf", ffWRITE },
2090 { efLOG, "-err", "bencherr", ffWRITE },
2091 { efTPX, "-so", "tuned", ffWRITE },
2093 { efTPX, NULL, NULL, ffREAD },
2094 { efTRN, "-o", NULL, ffWRITE },
2095 { efXTC, "-x", NULL, ffOPTWR },
2096 { efCPT, "-cpi", NULL, ffOPTRD },
2097 { efCPT, "-cpo", NULL, ffOPTWR },
2098 { efSTO, "-c", "confout", ffWRITE },
2099 { efEDR, "-e", "ener", ffWRITE },
2100 { efLOG, "-g", "md", ffWRITE },
2101 { efXVG, "-dhdl", "dhdl", ffOPTWR },
2102 { efXVG, "-field", "field", ffOPTWR },
2103 { efXVG, "-table", "table", ffOPTRD },
2104 { efXVG, "-tabletf", "tabletf", ffOPTRD },
2105 { efXVG, "-tablep", "tablep", ffOPTRD },
2106 { efXVG, "-tableb", "table", ffOPTRD },
2107 { efTRX, "-rerun", "rerun", ffOPTRD },
2108 { efXVG, "-tpi", "tpi", ffOPTWR },
2109 { efXVG, "-tpid", "tpidist", ffOPTWR },
2110 { efEDI, "-ei", "sam", ffOPTRD },
2111 { efXVG, "-eo", "edsam", ffOPTWR },
2112 { efXVG, "-devout", "deviatie", ffOPTWR },
2113 { efXVG, "-runav", "runaver", ffOPTWR },
2114 { efXVG, "-px", "pullx", ffOPTWR },
2115 { efXVG, "-pf", "pullf", ffOPTWR },
2116 { efXVG, "-ro", "rotation", ffOPTWR },
2117 { efLOG, "-ra", "rotangles", ffOPTWR },
2118 { efLOG, "-rs", "rotslabs", ffOPTWR },
2119 { efLOG, "-rt", "rottorque", ffOPTWR },
2120 { efMTX, "-mtx", "nm", ffOPTWR },
2121 { efNDX, "-dn", "dipole", ffOPTWR },
2122 /* Output files that are deleted after each benchmark run */
2123 { efTRN, "-bo", "bench", ffWRITE },
2124 { efXTC, "-bx", "bench", ffWRITE },
2125 { efCPT, "-bcpo", "bench", ffWRITE },
2126 { efSTO, "-bc", "bench", ffWRITE },
2127 { efEDR, "-be", "bench", ffWRITE },
2128 { efLOG, "-bg", "bench", ffWRITE },
2129 { efXVG, "-beo", "benchedo", ffOPTWR },
2130 { efXVG, "-bdhdl", "benchdhdl", ffOPTWR },
2131 { efXVG, "-bfield", "benchfld", ffOPTWR },
2132 { efXVG, "-btpi", "benchtpi", ffOPTWR },
2133 { efXVG, "-btpid", "benchtpid", ffOPTWR },
2134 { efXVG, "-bdevout", "benchdev", ffOPTWR },
2135 { efXVG, "-brunav", "benchrnav", ffOPTWR },
2136 { efXVG, "-bpx", "benchpx", ffOPTWR },
2137 { efXVG, "-bpf", "benchpf", ffOPTWR },
2138 { efXVG, "-bro", "benchrot", ffOPTWR },
2139 { efLOG, "-bra", "benchrota", ffOPTWR },
2140 { efLOG, "-brs", "benchrots", ffOPTWR },
2141 { efLOG, "-brt", "benchrott", ffOPTWR },
2142 { efMTX, "-bmtx", "benchn", ffOPTWR },
2143 { efNDX, "-bdn", "bench", ffOPTWR }
2146 gmx_bool bThreads = FALSE;
2150 const char *procstring[] =
2151 { NULL, "-np", "-n", "none", NULL };
2152 const char *npmevalues_opt[] =
2153 { NULL, "auto", "all", "subset", NULL };
2155 gmx_bool bAppendFiles = TRUE;
2156 gmx_bool bKeepAndNumCPT = FALSE;
2157 gmx_bool bResetCountersHalfWay = FALSE;
2158 gmx_bool bBenchmark = TRUE;
2160 output_env_t oenv = NULL;
2163 /***********************/
2164 /* g_tune_pme options: */
2165 /***********************/
2166 { "-np", FALSE, etINT, {&nnodes},
2167 "Number of nodes to run the tests on (must be > 2 for separate PME nodes)" },
2168 { "-npstring", FALSE, etENUM, {procstring},
2169 "Specify the number of processors to [TT]$MPIRUN[tt] using this string"},
2170 { "-ntmpi", FALSE, etINT, {&nthreads},
2171 "Number of MPI-threads to run the tests on (turns MPI & mpirun off)"},
2172 { "-r", FALSE, etINT, {&repeats},
2173 "Repeat each test this often" },
2174 { "-max", FALSE, etREAL, {&maxPMEfraction},
2175 "Max fraction of PME nodes to test with" },
2176 { "-min", FALSE, etREAL, {&minPMEfraction},
2177 "Min fraction of PME nodes to test with" },
2178 { "-npme", FALSE, etENUM, {npmevalues_opt},
2179 "Within -min and -max, benchmark all possible values for [TT]-npme[tt], or just a reasonable subset. "
2180 "Auto neglects -min and -max and chooses reasonable values around a guess for npme derived from the .tpr"},
2181 { "-fix", FALSE, etINT, {&npme_fixed},
2182 "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."},
2183 { "-rmax", FALSE, etREAL, {&rmax},
2184 "If >0, maximal rcoulomb for -ntpr>1 (rcoulomb upscaling results in fourier grid downscaling)" },
2185 { "-rmin", FALSE, etREAL, {&rmin},
2186 "If >0, minimal rcoulomb for -ntpr>1" },
2187 { "-scalevdw", FALSE, etBOOL, {&bScaleRvdw},
2188 "Scale rvdw along with rcoulomb"},
2189 { "-ntpr", FALSE, etINT, {&ntprs},
2190 "Number of [TT].tpr[tt] files to benchmark. Create this many files with different rcoulomb scaling factors depending on -rmin and -rmax. "
2191 "If < 1, automatically choose the number of [TT].tpr[tt] files to test" },
2192 { "-steps", FALSE, etGMX_LARGE_INT, {&bench_nsteps},
2193 "Take timings for this many steps in the benchmark runs" },
2194 { "-resetstep", FALSE, etINT, {&presteps},
2195 "Let dlb equilibrate this many steps before timings are taken (reset cycle counters after this many steps)" },
2196 { "-simsteps", FALSE, etGMX_LARGE_INT, {&new_sim_nsteps},
2197 "If non-negative, perform this many steps in the real run (overwrites nsteps from [TT].tpr[tt], add [TT].cpt[tt] steps)" },
2198 { "-launch", FALSE, etBOOL, {&bLaunch},
2199 "Launch the real simulation after optimization" },
2200 { "-bench", FALSE, etBOOL, {&bBenchmark},
2201 "Run the benchmarks or just create the input [TT].tpr[tt] files?" },
2202 /******************/
2203 /* mdrun options: */
2204 /******************/
2205 /* We let g_tune_pme parse and understand these options, because we need to
2206 * prevent that they appear on the mdrun command line for the benchmarks */
2207 { "-append", FALSE, etBOOL, {&bAppendFiles},
2208 "Append to previous output files when continuing from checkpoint instead of adding the simulation part number to all file names (for launch only)" },
2209 { "-cpnum", FALSE, etBOOL, {&bKeepAndNumCPT},
2210 "Keep and number checkpoint files (launch only)" },
2211 { "-resethway", FALSE, etBOOL, {&bResetCountersHalfWay},
2212 "HIDDENReset the cycle counters after half the number of steps or halfway [TT]-maxh[tt] (launch only)" }
2215 #define NFILE asize(fnm)
2217 seconds = gmx_gettime();
2219 if (!parse_common_args(&argc, argv, PCA_NOEXIT_ON_ARGS,
2220 NFILE, fnm, asize(pa), pa, asize(desc), desc,
2226 /* Store the remaining unparsed command line entries in a string which
2227 * is then attached to the mdrun command line */
2229 ExtraArgs[0] = '\0';
2230 for (i = 1; i < argc; i++) /* argc will now be 1 if everything was understood */
2232 add_to_string(&ExtraArgs, argv[i]);
2233 add_to_string(&ExtraArgs, " ");
2236 if (opt2parg_bSet("-ntmpi", asize(pa), pa))
2239 if (opt2parg_bSet("-npstring", asize(pa), pa))
2241 fprintf(stderr, "WARNING: -npstring has no effect when using threads.\n");
2246 gmx_fatal(FARGS, "Can't run multi-threaded MPI simulation yet!");
2248 /* and now we just set this; a bit of an ugly hack*/
2251 /* Check for PME:PP ratio and whether tpr triggers additional output files */
2252 guessPMEratio = inspect_tpr(NFILE, fnm, &rcoulomb);
2254 /* Automatically set -beo options if -eo is set etc. */
2255 couple_files_options(NFILE, fnm);
2257 /* Construct the command line arguments for benchmark runs
2258 * as well as for the simulation run */
2261 sprintf(bbuf, " -ntmpi %d ", nthreads);
2265 /* This string will be used for MPI runs and will appear after the
2266 * mpirun command. */
2267 if (strcmp(procstring[0], "none") != 0)
2269 sprintf(bbuf, " %s %d ", procstring[0], nnodes);
2279 create_command_line_snippets(bAppendFiles, bKeepAndNumCPT, bResetCountersHalfWay, presteps,
2280 NFILE, fnm, &cmd_args_bench, &cmd_args_launch, ExtraArgs);
2282 /* Read in checkpoint file if requested */
2284 if (opt2bSet("-cpi", NFILE, fnm))
2287 cr->duty = DUTY_PP; /* makes the following routine happy */
2288 read_checkpoint_simulation_part(opt2fn("-cpi", NFILE, fnm),
2289 &sim_part, &cpt_steps, cr,
2290 FALSE, NFILE, fnm, NULL, NULL);
2293 /* sim_part will now be 1 if no checkpoint file was found */
2296 gmx_fatal(FARGS, "Checkpoint file %s not found!", opt2fn("-cpi", NFILE, fnm));
2300 /* Open performance output file and write header info */
2301 fp = ffopen(opt2fn("-p", NFILE, fnm), "w");
2303 /* Make a quick consistency check of command line parameters */
2304 check_input(nnodes, repeats, &ntprs, &rmin, rcoulomb, &rmax,
2305 maxPMEfraction, minPMEfraction, npme_fixed,
2306 bench_nsteps, fnm, NFILE, sim_part, presteps,
2309 /* Determine the maximum and minimum number of PME nodes to test,
2310 * the actual list of settings is build in do_the_tests(). */
2311 if ((nnodes > 2) && (npme_fixed < -1))
2313 if (0 == strcmp(npmevalues_opt[0], "auto"))
2315 /* Determine the npme range automatically based on the PME:PP load guess */
2316 if (guessPMEratio > 1.0)
2318 /* More PME than PP work, probably we do not need separate PME nodes at all! */
2319 maxPMEnodes = nnodes/2;
2320 minPMEnodes = nnodes/2;
2324 /* PME : PP load is in the range 0..1, let's test around the guess */
2325 guessPMEnodes = nnodes/(1.0 + 1.0/guessPMEratio);
2326 minPMEnodes = floor(0.7*guessPMEnodes);
2327 maxPMEnodes = ceil(1.6*guessPMEnodes);
2328 maxPMEnodes = min(maxPMEnodes, nnodes/2);
2333 /* Determine the npme range based on user input */
2334 maxPMEnodes = floor(maxPMEfraction*nnodes);
2335 minPMEnodes = max(floor(minPMEfraction*nnodes), 0);
2336 fprintf(stdout, "Will try runs with %d ", minPMEnodes);
2337 if (maxPMEnodes != minPMEnodes)
2339 fprintf(stdout, "- %d ", maxPMEnodes);
2341 fprintf(stdout, "PME-only nodes.\n Note that the automatic number of PME-only nodes and no separate PME nodes are always tested.\n");
2350 /* Get the commands we need to set up the runs from environment variables */
2351 get_program_paths(bThreads, &cmd_mpirun, &cmd_mdrun);
2352 if (bBenchmark && repeats > 0)
2354 check_mdrun_works(bThreads, cmd_mpirun, cmd_np, cmd_mdrun);
2357 /* Print some header info to file */
2359 fprintf(fp, "\n P E R F O R M A N C E R E S U L T S\n");
2361 fprintf(fp, "%s for Gromacs %s\n", ShortProgram(), GromacsVersion());
2364 fprintf(fp, "Number of nodes : %d\n", nnodes);
2365 fprintf(fp, "The mpirun command is : %s\n", cmd_mpirun);
2366 if (strcmp(procstring[0], "none") != 0)
2368 fprintf(fp, "Passing # of nodes via : %s\n", procstring[0]);
2372 fprintf(fp, "Not setting number of nodes in system call\n");
2377 fprintf(fp, "Number of threads : %d\n", nnodes);
2380 fprintf(fp, "The mdrun command is : %s\n", cmd_mdrun);
2381 fprintf(fp, "mdrun args benchmarks : %s\n", cmd_args_bench);
2382 fprintf(fp, "Benchmark steps : ");
2383 fprintf(fp, gmx_large_int_pfmt, bench_nsteps);
2385 fprintf(fp, "dlb equilibration steps : %d\n", presteps);
2388 fprintf(fp, "Checkpoint time step : ");
2389 fprintf(fp, gmx_large_int_pfmt, cpt_steps);
2392 fprintf(fp, "mdrun args at launchtime: %s\n", cmd_args_launch);
2394 if (new_sim_nsteps >= 0)
2397 fprintf(stderr, "Note: Simulation input file %s will have ", opt2fn("-so", NFILE, fnm));
2398 fprintf(stderr, gmx_large_int_pfmt, new_sim_nsteps+cpt_steps);
2399 fprintf(stderr, " steps.\n");
2400 fprintf(fp, "Simulation steps : ");
2401 fprintf(fp, gmx_large_int_pfmt, new_sim_nsteps);
2406 fprintf(fp, "Repeats for each test : %d\n", repeats);
2409 if (npme_fixed >= -1)
2411 fprintf(fp, "Fixing -npme at : %d\n", npme_fixed);
2414 fprintf(fp, "Input file : %s\n", opt2fn("-s", NFILE, fnm));
2415 fprintf(fp, " PME/PP load estimate : %g\n", guessPMEratio);
2417 /* Allocate memory for the inputinfo struct: */
2419 info->nr_inputfiles = ntprs;
2420 for (i = 0; i < ntprs; i++)
2422 snew(info->rcoulomb, ntprs);
2423 snew(info->rvdw, ntprs);
2424 snew(info->rlist, ntprs);
2425 snew(info->rlistlong, ntprs);
2426 snew(info->nkx, ntprs);
2427 snew(info->nky, ntprs);
2428 snew(info->nkz, ntprs);
2429 snew(info->fsx, ntprs);
2430 snew(info->fsy, ntprs);
2431 snew(info->fsz, ntprs);
2433 /* Make alternative tpr files to test: */
2434 snew(tpr_names, ntprs);
2435 for (i = 0; i < ntprs; i++)
2437 snew(tpr_names[i], STRLEN);
2440 /* It can be that ntprs is reduced by make_benchmark_tprs if not enough
2441 * different grids could be found. */
2442 make_benchmark_tprs(opt2fn("-s", NFILE, fnm), tpr_names, bench_nsteps+presteps,
2443 cpt_steps, rmin, rmax, bScaleRvdw, &ntprs, info, fp);
2445 /********************************************************************************/
2446 /* Main loop over all scenarios we need to test: tpr files, PME nodes, repeats */
2447 /********************************************************************************/
2448 snew(perfdata, ntprs);
2451 do_the_tests(fp, tpr_names, maxPMEnodes, minPMEnodes, npme_fixed, npmevalues_opt[0], perfdata, &pmeentries,
2452 repeats, nnodes, ntprs, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2453 cmd_args_bench, fnm, NFILE, presteps, cpt_steps);
2455 fprintf(fp, "\nTuning took%8.1f minutes.\n", (gmx_gettime()-seconds)/60.0);
2457 /* Analyse the results and give a suggestion for optimal settings: */
2458 bKeepTPR = analyze_data(fp, opt2fn("-p", NFILE, fnm), perfdata, nnodes, ntprs, pmeentries,
2459 repeats, info, &best_tpr, &best_npme);
2461 /* Take the best-performing tpr file and enlarge nsteps to original value */
2462 if (bKeepTPR && !bOverwrite)
2464 simulation_tpr = opt2fn("-s", NFILE, fnm);
2468 simulation_tpr = opt2fn("-so", NFILE, fnm);
2469 modify_PMEsettings(bOverwrite ? (new_sim_nsteps+cpt_steps) : info->orig_sim_steps,
2470 info->orig_init_step, tpr_names[best_tpr], simulation_tpr);
2473 /* Let's get rid of the temporary benchmark input files */
2474 for (i = 0; i < ntprs; i++)
2476 fprintf(stdout, "Deleting temporary benchmark input file %s\n", tpr_names[i]);
2477 remove(tpr_names[i]);
2480 /* Now start the real simulation if the user requested it ... */
2481 launch_simulation(bLaunch, fp, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2482 cmd_args_launch, simulation_tpr, best_npme);
2486 /* ... or simply print the performance results to screen: */
2489 finalize(opt2fn("-p", NFILE, fnm));