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
56 #include "checkpoint.h"
63 ddnoSEL, ddnoINTERLEAVE, ddnoPP_PME, ddnoCARTESIAN, ddnoNR
66 /* Enum for situations that can occur during log file parsing, the
67 * corresponding string entries can be found in do_the_tests() in
68 * const char* ParseLog[] */
74 eParselogResetProblem,
85 int nPMEnodes; /* number of PME-only nodes used in this test */
86 int nx, ny, nz; /* DD grid */
87 int guessPME; /* if nPMEnodes == -1, this is the guessed number of PME nodes */
88 double *Gcycles; /* This can contain more than one value if doing multiple tests */
92 float *PME_f_load; /* PME mesh/force load average*/
93 float PME_f_load_Av; /* Average average ;) ... */
94 char *mdrun_cmd_line; /* Mdrun command line used for this test */
100 int nr_inputfiles; /* The number of tpr and mdp input files */
101 gmx_large_int_t orig_sim_steps; /* Number of steps to be done in the real simulation */
102 gmx_large_int_t orig_init_step; /* Init step for the real simulation */
103 real *rcoulomb; /* The coulomb radii [0...nr_inputfiles] */
104 real *rvdw; /* The vdW radii */
105 real *rlist; /* Neighbourlist cutoff radius */
107 int *nkx, *nky, *nkz;
108 real *fsx, *fsy, *fsz; /* Fourierspacing in x,y,z dimension */
112 static void sep_line(FILE *fp)
114 fprintf(fp, "\n------------------------------------------------------------\n");
118 /* Wrapper for system calls */
119 static int gmx_system_call(char *command)
122 gmx_fatal(FARGS,"No calls to system(3) supported on this platform. Attempted to call:\n'%s'\n",command);
124 return ( system(command) );
129 /* Check if string starts with substring */
130 static gmx_bool str_starts(const char *string, const char *substring)
132 return ( strncmp(string, substring, strlen(substring)) == 0);
136 static void cleandata(t_perf *perfdata, int test_nr)
138 perfdata->Gcycles[test_nr] = 0.0;
139 perfdata->ns_per_day[test_nr] = 0.0;
140 perfdata->PME_f_load[test_nr] = 0.0;
146 static gmx_bool is_equal(real a, real b)
148 real diff, eps=1.0e-7;
153 if (diff < 0.0) diff = -diff;
162 static void finalize(const char *fn_out)
168 fp = fopen(fn_out,"r");
169 fprintf(stdout,"\n\n");
171 while( fgets(buf,STRLEN-1,fp) != NULL )
173 fprintf(stdout,"%s",buf);
176 fprintf(stdout,"\n\n");
181 enum {eFoundNothing, eFoundDDStr, eFoundAccountingStr, eFoundCycleStr};
183 static int parse_logfile(const char *logfile, const char *errfile,
184 t_perf *perfdata, int test_nr, int presteps, gmx_large_int_t cpt_steps,
188 char line[STRLEN], dumstring[STRLEN], dumstring2[STRLEN];
189 const char matchstrdd[]="Domain decomposition grid";
190 const char matchstrcr[]="resetting all time and cycle counters";
191 const char matchstrbal[]="Average PME mesh/force load:";
192 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";
193 const char errSIG[]="signal, stopping at the next";
196 float dum1,dum2,dum3;
198 gmx_large_int_t resetsteps=-1;
199 gmx_bool bFoundResetStr = FALSE;
200 gmx_bool bResetChecked = FALSE;
203 if (!gmx_fexist(logfile))
205 fprintf(stderr, "WARNING: Could not find logfile %s.\n", logfile);
206 cleandata(perfdata, test_nr);
207 return eParselogNotFound;
210 fp = fopen(logfile, "r");
211 perfdata->PME_f_load[test_nr] = -1.0;
212 perfdata->guessPME = -1;
214 iFound = eFoundNothing;
216 iFound = eFoundDDStr; /* Skip some case statements */
218 while (fgets(line, STRLEN, fp) != NULL)
220 /* Remove leading spaces */
223 /* Check for TERM and INT signals from user: */
224 if ( strstr(line, errSIG) != NULL )
227 cleandata(perfdata, test_nr);
228 return eParselogTerm;
231 /* Check whether cycle resetting worked */
232 if (presteps > 0 && !bFoundResetStr)
234 if (strstr(line, matchstrcr) != NULL)
236 sprintf(dumstring, "Step %s", gmx_large_int_pfmt);
237 sscanf(line, dumstring, &resetsteps);
238 bFoundResetStr = TRUE;
239 if (resetsteps == presteps+cpt_steps)
241 bResetChecked = TRUE;
245 sprintf(dumstring , gmx_large_int_pfmt, resetsteps);
246 sprintf(dumstring2, gmx_large_int_pfmt, presteps+cpt_steps);
247 fprintf(stderr, "WARNING: Time step counters were reset at step %s,\n"
248 " though they were supposed to be reset at step %s!\n",
249 dumstring, dumstring2);
254 /* Look for strings that appear in a certain order in the log file: */
258 /* Look for domain decomp grid and separate PME nodes: */
259 if (str_starts(line, matchstrdd))
261 sscanf(line, "Domain decomposition grid %d x %d x %d, separate PME nodes %d",
262 &(perfdata->nx), &(perfdata->ny), &(perfdata->nz), &npme);
263 if (perfdata->nPMEnodes == -1)
264 perfdata->guessPME = npme;
265 else if (perfdata->nPMEnodes != npme)
266 gmx_fatal(FARGS, "PME nodes from command line and output file are not identical");
267 iFound = eFoundDDStr;
269 /* Catch a few errors that might have occured: */
270 else if (str_starts(line, "There is no domain decomposition for"))
273 return eParselogNoDDGrid;
275 else if (str_starts(line, "reading tpx file"))
278 return eParselogTPXVersion;
280 else if (str_starts(line, "The -dd or -npme option request a parallel simulation"))
283 return eParselogNotParallel;
287 /* Look for PME mesh/force balance (not necessarily present, though) */
288 if (str_starts(line, matchstrbal))
289 sscanf(&line[strlen(matchstrbal)], "%f", &(perfdata->PME_f_load[test_nr]));
290 /* Look for matchstring */
291 if (str_starts(line, matchstring))
292 iFound = eFoundAccountingStr;
294 case eFoundAccountingStr:
295 /* Already found matchstring - look for cycle data */
296 if (str_starts(line, "Total "))
298 sscanf(line,"Total %d %lf",&procs,&(perfdata->Gcycles[test_nr]));
299 iFound = eFoundCycleStr;
303 /* Already found cycle data - look for remaining performance info and return */
304 if (str_starts(line, "Performance:"))
306 sscanf(line,"%s %f %f %f %f", dumstring, &dum1, &dum2, &(perfdata->ns_per_day[test_nr]), &dum3);
308 if (bResetChecked || presteps == 0)
311 return eParselogResetProblem;
317 /* Close the log file */
320 /* Check why there is no performance data in the log file.
321 * Did a fatal errors occur? */
322 if (gmx_fexist(errfile))
324 fp = fopen(errfile, "r");
325 while (fgets(line, STRLEN, fp) != NULL)
327 if ( str_starts(line, "Fatal error:") )
329 if (fgets(line, STRLEN, fp) != NULL)
330 fprintf(stderr, "\nWARNING: A fatal error has occured during this benchmark:\n"
333 cleandata(perfdata, test_nr);
334 return eParselogFatal;
341 fprintf(stderr, "WARNING: Could not find stderr file %s.\n", errfile);
344 /* Giving up ... we could not find out why there is no performance data in
346 fprintf(stdout, "No performance data in log file.\n");
347 cleandata(perfdata, test_nr);
349 return eParselogNoPerfData;
353 static gmx_bool analyze_data(
362 int *index_tpr, /* OUT: Nr of mdp file with best settings */
363 int *npme_optimal) /* OUT: Optimal number of PME nodes */
366 int line=0, line_win=-1;
367 int k_win=-1, i_win=-1, winPME;
368 double s=0.0; /* standard deviation */
371 char str_PME_f_load[13];
372 gmx_bool bCanUseOrigTPR;
373 gmx_bool bRefinedCoul, bRefinedVdW, bRefinedGrid;
379 fprintf(fp, "Summary of successful runs:\n");
380 fprintf(fp, "Line tpr PME nodes Gcycles Av. Std.dev. ns/day PME/f");
382 fprintf(fp, " DD grid");
387 for (k=0; k<ntprs; k++)
389 for (i=0; i<ntests; i++)
391 /* Select the right dataset: */
392 pd = &(perfdata[k][i]);
394 pd->Gcycles_Av = 0.0;
395 pd->PME_f_load_Av = 0.0;
396 pd->ns_per_day_Av = 0.0;
398 if (pd->nPMEnodes == -1)
399 sprintf(strbuf, "(%3d)", pd->guessPME);
401 sprintf(strbuf, " ");
403 /* Get the average run time of a setting */
404 for (j=0; j<nrepeats; j++)
406 pd->Gcycles_Av += pd->Gcycles[j];
407 pd->PME_f_load_Av += pd->PME_f_load[j];
409 pd->Gcycles_Av /= nrepeats;
410 pd->PME_f_load_Av /= nrepeats;
412 for (j=0; j<nrepeats; j++)
414 if (pd->ns_per_day[j] > 0.0)
415 pd->ns_per_day_Av += pd->ns_per_day[j];
418 /* Somehow the performance number was not aquired for this run,
419 * therefor set the average to some negative value: */
420 pd->ns_per_day_Av = -1.0f*nrepeats;
424 pd->ns_per_day_Av /= nrepeats;
427 if (pd->PME_f_load_Av > 0.0)
428 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load_Av);
430 sprintf(str_PME_f_load, "%s", " - ");
433 /* We assume we had a successful run if both averages are positive */
434 if (pd->Gcycles_Av > 0.0 && pd->ns_per_day_Av > 0.0)
436 /* Output statistics if repeats were done */
439 /* Calculate the standard deviation */
441 for (j=0; j<nrepeats; j++)
442 s += pow( pd->Gcycles[j] - pd->Gcycles_Av, 2 );
446 fprintf(fp, "%4d %3d %4d%s %12.3f %12.3f %12.3f %s",
447 line, k, pd->nPMEnodes, strbuf, pd->Gcycles_Av, s,
448 pd->ns_per_day_Av, str_PME_f_load);
450 fprintf(fp, " %3d %3d %3d", pd->nx, pd->ny, pd->nz);
453 /* Store the index of the best run found so far in 'winner': */
454 if ( (k_win == -1) || (pd->Gcycles_Av < perfdata[k_win][i_win].Gcycles_Av) )
466 gmx_fatal(FARGS, "None of the runs was successful! Check %s for problems.", fn);
470 winPME = perfdata[k_win][i_win].nPMEnodes;
474 /* We stuck to a fixed number of PME-only nodes */
475 sprintf(strbuf, "settings No. %d", k_win);
479 /* We have optimized the number of PME-only nodes */
481 sprintf(strbuf, "%s", "the automatic number of PME nodes");
483 sprintf(strbuf, "%d PME nodes", winPME);
485 fprintf(fp, "Best performance was achieved with %s", strbuf);
486 if ((nrepeats > 1) && (ntests > 1))
487 fprintf(fp, " (see line %d)", line_win);
490 /* Only mention settings if they were modified: */
491 bRefinedCoul = !is_equal(info->rcoulomb[k_win], info->rcoulomb[0]);
492 bRefinedVdW = !is_equal(info->rvdw[k_win] , info->rvdw[0] );
493 bRefinedGrid = !(info->nkx[k_win] == info->nkx[0] &&
494 info->nky[k_win] == info->nky[0] &&
495 info->nkz[k_win] == info->nkz[0]);
497 if (bRefinedCoul || bRefinedVdW || bRefinedGrid)
499 fprintf(fp, "Optimized PME settings:\n");
500 bCanUseOrigTPR = FALSE;
504 bCanUseOrigTPR = TRUE;
508 fprintf(fp, " New Coulomb radius: %f nm (was %f nm)\n", info->rcoulomb[k_win], info->rcoulomb[0]);
511 fprintf(fp, " New Van der Waals radius: %f nm (was %f nm)\n", info->rvdw[k_win], info->rvdw[0]);
514 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],
515 info->nkx[0], info->nky[0], info->nkz[0]);
517 if (bCanUseOrigTPR && ntprs > 1)
518 fprintf(fp, "and original PME settings.\n");
522 /* Return the index of the mdp file that showed the highest performance
523 * and the optimal number of PME nodes */
525 *npme_optimal = winPME;
527 return bCanUseOrigTPR;
531 /* Get the commands we need to set up the runs from environment variables */
532 static void get_program_paths(gmx_bool bThreads, char *cmd_mpirun[], char cmd_np[],
533 char *cmd_mdrun[], int repeats)
540 const char def_mpirun[] = "mpirun";
541 const char def_mdrun[] = "mdrun";
542 const char filename[] = "benchtest.log";
543 const char match_mpi[] = "NNODES=";
544 const char match_mdrun[]= "Program: ";
545 const char empty_mpirun[] = "";
546 gmx_bool bMdrun = FALSE;
547 gmx_bool bMPI = FALSE;
550 /* Get the commands we need to set up the runs from environment variables */
553 if ( (cp = getenv("MPIRUN")) != NULL)
554 *cmd_mpirun = strdup(cp);
556 *cmd_mpirun = strdup(def_mpirun);
560 *cmd_mpirun = strdup(empty_mpirun);
563 if ( (cp = getenv("MDRUN" )) != NULL )
564 *cmd_mdrun = strdup(cp);
566 *cmd_mdrun = strdup(def_mdrun);
569 /* If no simulations have to be performed, we are done here */
573 /* Run a small test to see whether mpirun + mdrun work */
574 fprintf(stdout, "Making sure that mdrun can be executed. ");
577 snew(command, strlen(*cmd_mdrun) + strlen(cmd_np) + strlen(filename) + 50);
578 sprintf(command, "%s%s-version -maxh 0.001 1> %s 2>&1", *cmd_mdrun, cmd_np, filename);
582 snew(command, strlen(*cmd_mpirun) + strlen(cmd_np) + strlen(*cmd_mdrun) + strlen(filename) + 50);
583 sprintf(command, "%s%s%s -version -maxh 0.001 1> %s 2>&1", *cmd_mpirun, cmd_np, *cmd_mdrun, filename);
585 fprintf(stdout, "Trying '%s' ... ", command);
586 make_backup(filename);
587 gmx_system_call(command);
589 /* Check if we find the characteristic string in the output: */
590 if (!gmx_fexist(filename))
591 gmx_fatal(FARGS, "Output from test run could not be found.");
593 fp = fopen(filename, "r");
594 /* We need to scan the whole output file, since sometimes the queuing system
595 * also writes stuff to stdout/err */
598 cp2=fgets(line, STRLEN, fp);
601 if ( str_starts(line, match_mdrun) )
603 if ( str_starts(line, match_mpi) )
613 gmx_fatal(FARGS, "Need a threaded version of mdrun. This one\n"
615 "seems to have been compiled with MPI instead.",
623 gmx_fatal(FARGS, "Need an MPI-enabled version of mdrun. This one\n"
625 "seems to have been compiled without MPI support.",
632 gmx_fatal(FARGS, "Cannot execute mdrun. Please check %s for problems!",
636 fprintf(stdout, "passed.\n");
644 static void launch_simulation(
645 gmx_bool bLaunch, /* Should the simulation be launched? */
646 FILE *fp, /* General log file */
647 gmx_bool bThreads, /* whether to use threads */
648 char *cmd_mpirun, /* Command for mpirun */
649 char *cmd_np, /* Switch for -np or -nt or empty */
650 char *cmd_mdrun, /* Command for mdrun */
651 char *args_for_mdrun, /* Arguments for mdrun */
652 const char *simulation_tpr, /* This tpr will be simulated */
653 int nnodes, /* Number of nodes to run on */
654 int nPMEnodes) /* Number of PME nodes to use */
659 /* Make enough space for the system call command,
660 * (100 extra chars for -npme ... etc. options should suffice): */
661 snew(command, strlen(cmd_mpirun)+strlen(cmd_mdrun)+strlen(cmd_np)+strlen(args_for_mdrun)+strlen(simulation_tpr)+100);
663 /* Note that the -passall options requires args_for_mdrun to be at the end
664 * of the command line string */
667 sprintf(command, "%s%s-npme %d -s %s %s",
668 cmd_mdrun, cmd_np, nPMEnodes, simulation_tpr, args_for_mdrun);
672 sprintf(command, "%s%s%s -npme %d -s %s %s",
673 cmd_mpirun, cmd_np, cmd_mdrun, nPMEnodes, simulation_tpr, args_for_mdrun);
676 fprintf(fp, "%s this command line to launch the simulation:\n\n%s", bLaunch? "Using":"Please use", command);
680 /* Now the real thing! */
683 fprintf(stdout, "\nLaunching simulation with best parameters now.\nExecuting '%s'", command);
686 gmx_system_call(command);
692 static void modify_PMEsettings(
693 gmx_large_int_t simsteps, /* Set this value as number of time steps */
694 gmx_large_int_t init_step, /* Set this value as init_step */
695 const char *fn_best_tpr, /* tpr file with the best performance */
696 const char *fn_sim_tpr) /* name of tpr file to be launched */
704 read_tpx_state(fn_best_tpr,ir,&state,NULL,&mtop);
706 /* Reset nsteps and init_step to the value of the input .tpr file */
707 ir->nsteps = simsteps;
708 ir->init_step = init_step;
710 /* Write the tpr file which will be launched */
711 sprintf(buf, "Writing optimized simulation file %s with nsteps=%s.\n", fn_sim_tpr, gmx_large_int_pfmt);
712 fprintf(stdout,buf,ir->nsteps);
714 write_tpx_state(fn_sim_tpr,ir,&state,&mtop);
722 int nkx, nky, nkz; /* Fourier grid */
723 real fac; /* actual scaling factor */
724 real fs; /* Fourierspacing */
728 static void copy_grid(t_pmegrid *ingrid, t_pmegrid *outgrid)
730 outgrid->nkx = ingrid->nkx;
731 outgrid->nky = ingrid->nky;
732 outgrid->nkz = ingrid->nkz;
733 outgrid->fac = ingrid->fac;
734 outgrid->fs = ingrid->fs;
737 /* Removes entry 'index' from the t_pmegrid list */
738 static void remove_from_list(
739 t_pmegrid gridlist[],
740 int *nlist, /* Length of the list */
741 int index) /* Index to remove from the list */
746 for (j = index; j < (*nlist - 1); j++)
748 copy_grid(&gridlist[j+1], &gridlist[j]);
754 /* Returns the index of the least necessary grid in the list.
756 * This is the one where the following conditions hold for the scaling factor:
757 * 1. this grid has the smallest distance to its neighboring grid (distance
758 * measured by fac) -> this condition is true for two grids at the same time
759 * 2. this grid (of the two) has the smaller distance to the other neighboring
762 * The extreme grids (the ones with the smallest and largest
763 * scaling factor) are never thrown away.
765 static int get_throwaway_index(
766 t_pmegrid gridlist[],
767 int nlist) /* Length of the list */
770 real dist,mindist,d_left,d_right;
773 /* Find the two factors with the smallest mutual distance */
774 mindist = GMX_FLOAT_MAX;
775 for (i = 1; i < nlist; i++)
777 dist = fabs(gridlist[i].fac - gridlist[i-1].fac);
784 /* index and index-1 have the smallest mutual distance */
787 /* Never return the first index, i.e. index == 0 */
792 d_left = fabs(gridlist[index-1].fac - gridlist[index-2].fac);
793 d_right = fabs(gridlist[index+1].fac - gridlist[index ].fac);
795 /* Return the left index if its distance to its left neighbor is shorter
796 * than the distance of the right index to its right neighbor */
797 if (d_left < d_right)
800 /* Never return the last index */
801 if (index == nlist-1)
808 static void make_grid_list(
809 real fmin, /* minimum scaling factor (downscaling fac) */
810 real fmax, /* maximum scaling factor (upscaling fac) */
811 int *ntprs, /* Number of tpr files to test */
812 matrix box, /* The box */
813 t_pmegrid *griduse[], /* List of grids that have to be tested */
814 real fs) /* Requested fourierspacing at a scaling factor
817 real req_fac,act_fac=0,act_fs,eps;
819 int i,jmin=0,d,ngridall=0;
820 int nkx=0,nky=0,nkz=0;
821 int nkx_old=0,nky_old=0,nkz_old=0;
823 int gridalloc,excess;
826 /* Determine length of triclinic box vectors */
831 box_size[d] += box[d][i]*box[d][i];
832 box_size[d] = sqrt(box_size[d]);
836 snew(gridall, gridalloc);
838 fprintf(stdout, "Possible PME grid settings (apart from input file settings):\n");
839 fprintf(stdout, " nkx nky nkz max spacing scaling factor\n");
841 /* eps should provide a fine enough spacing not to miss any grid */
843 eps = (fmax-fmin)/(100.0*(*ntprs - 1));
845 eps = 1.0/max( (*griduse)[0].nkz, max( (*griduse)[0].nkx, (*griduse)[0].nky ) );
847 for (req_fac = fmin; act_fac < fmax; req_fac += eps)
852 calc_grid(NULL,box,fs*req_fac,&nkx,&nky,&nkz);
853 act_fs = max(box_size[XX]/nkx,max(box_size[YY]/nky,box_size[ZZ]/nkz));
855 if ( ! ( nkx==nkx_old && nky==nky_old && nkz==nkz_old ) /* Exclude if grid is already in list */
856 && ! ( nkx==(*griduse)[0].nkx && nky==(*griduse)[0].nky && nkz==(*griduse)[0].nkz ) ) /* Exclude input file grid */
858 /* We found a new grid that will do */
862 gridall[ngridall].nkx = nkx;
863 gridall[ngridall].nky = nky;
864 gridall[ngridall].nkz = nkz;
865 gridall[ngridall].fac = act_fac;
866 gridall[ngridall].fs = act_fs;
867 fprintf(stdout, "%5d%5d%5d %12f %12f\n",nkx,nky,nkz,act_fs,act_fac);
869 if (ngridall >= gridalloc)
872 srenew(gridall, gridalloc);
877 /* excess is the number of grids we have to get rid of */
878 excess = ngridall+1 - *ntprs;
880 /* If we found less grids than tpr files were requested, simply test all
881 * the grid there are ... */
884 fprintf(stdout, "NOTE: You requested %d tpr files, but apart from the input file grid only the\n"
885 " above listed %d suitable PME grids were found. Will test all suitable settings.\n",
893 /* We can only keep the input tpr file plus one extra tpr file.
894 * We make that choice depending on the values of -upfac and -downfac */
897 /* Keep the one with the -downfac as scaling factor. This is already
898 * stored in gridall[0] */
903 /* Keep the one with -upfac as scaling factor */
904 copy_grid(&(gridall[ngridall-1]), &(gridall[0]));
910 /* From the grid list throw away the unnecessary ones (keep the extremes) */
911 for (i = 0; i < excess; i++)
913 /* Get the index of the least necessary grid from the list ... */
914 jmin = get_throwaway_index(gridall, ngridall);
915 /* ... and remove the grid from the list */
916 remove_from_list(gridall, &ngridall, jmin);
921 /* The remaining list contains the grids we want to test */
922 for (i=1; i < *ntprs; i++)
923 copy_grid(&(gridall[i-1]), &((*griduse)[i]));
929 #define EPME_SWITCHED(e) ((e) == eelPMESWITCH || (e) == eelPMEUSERSWITCH)
931 /* Make additional TPR files with more computational load for the
932 * direct space processors: */
933 static void make_benchmark_tprs(
934 const char *fn_sim_tpr, /* READ : User-provided tpr file */
935 char *fn_bench_tprs[], /* WRITE: Names of benchmark tpr files */
936 gmx_large_int_t benchsteps, /* Number of time steps for benchmark runs */
937 gmx_large_int_t statesteps, /* Step counter in checkpoint file */
938 real upfac, /* Scale rcoulomb inbetween downfac and upfac */
940 int *ntprs, /* No. of TPRs to write, each with a different
941 rcoulomb and fourierspacing. If not enough
942 grids are found, ntprs is reduced accordingly */
943 real fourierspacing, /* Basic fourierspacing from tpr input file */
944 t_inputinfo *info, /* Contains information about mdp file options */
945 FILE *fp) /* Write the output here */
951 real nlist_buffer; /* Thickness of the buffer regions for PME-switch potentials: */
954 gmx_bool bNote = FALSE;
955 t_pmegrid *pmegrid=NULL; /* Grid settings for the PME grids to test */
958 sprintf(buf, "Making benchmark tpr file%s with %s time step%s",
959 *ntprs > 1? "s":"", gmx_large_int_pfmt, benchsteps>1?"s":"");
960 fprintf(stdout, buf, benchsteps);
963 sprintf(buf, " (adding %s steps from checkpoint file)", gmx_large_int_pfmt);
964 fprintf(stdout, buf, statesteps);
965 benchsteps += statesteps;
967 fprintf(stdout, ".\n");
971 read_tpx_state(fn_sim_tpr,ir,&state,NULL,&mtop);
973 /* Check if some kind of PME was chosen */
974 if (EEL_PME(ir->coulombtype) == FALSE)
975 gmx_fatal(FARGS, "Can only do optimizations for simulations with %s electrostatics.",
978 /* Check if rcoulomb == rlist, which is necessary for plain PME. */
979 if ( (eelPME == ir->coulombtype) && !(ir->rcoulomb == ir->rlist) )
981 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to rlist (%f).",
982 EELTYPE(eelPME), ir->rcoulomb, ir->rlist);
984 /* For other PME types, rcoulomb is allowed to be smaller than rlist */
985 else if (ir->rcoulomb > ir->rlist)
987 gmx_fatal(FARGS, "%s requires rcoulomb (%f) to be equal to or smaller than rlist (%f)",
988 EELTYPE(ir->coulombtype), ir->rcoulomb, ir->rlist);
991 /* Reduce the number of steps for the benchmarks */
992 info->orig_sim_steps = ir->nsteps;
993 ir->nsteps = benchsteps;
994 /* We must not use init_step from the input tpr file for the benchmarks */
995 info->orig_init_step = ir->init_step;
998 /* For PME-switch potentials, keep the radial distance of the buffer region */
999 nlist_buffer = ir->rlist - ir->rcoulomb;
1001 /* Determine length of triclinic box vectors */
1002 for(d=0; d<DIM; d++)
1006 box_size[d] += state.box[d][i]*state.box[d][i];
1007 box_size[d] = sqrt(box_size[d]);
1010 /* Reconstruct fourierspacing per dimension from the number of grid points and box size */
1011 info->fsx[0] = box_size[XX]/ir->nkx;
1012 info->fsy[0] = box_size[YY]/ir->nky;
1013 info->fsz[0] = box_size[ZZ]/ir->nkz;
1015 /* Put the input grid as first entry into the grid list */
1016 snew(pmegrid, *ntprs);
1017 pmegrid[0].fac = 1.00;
1018 pmegrid[0].nkx = ir->nkx;
1019 pmegrid[0].nky = ir->nky;
1020 pmegrid[0].nkz = ir->nkz;
1021 pmegrid[0].fs = max(box_size[ZZ]/ir->nkz, max(box_size[XX]/ir->nkx, box_size[YY]/ir->nky));
1023 /* If no value for the fourierspacing was provided on the command line, we
1024 * use the reconstruction from the tpr file */
1025 if (fourierspacing <= 0)
1027 fourierspacing = pmegrid[0].fs;
1030 fprintf(stdout, "Calculating PME grid points on the basis of a fourierspacing of %f nm\n", fourierspacing);
1032 /* Print information about settings of which some are potentially modified: */
1033 fprintf(fp, " Coulomb type : %s\n", EELTYPE(ir->coulombtype));
1034 fprintf(fp, " Grid spacing x y z : %f %f %f\n",
1035 box_size[XX]/ir->nkx, box_size[YY]/ir->nky, box_size[ZZ]/ir->nkz);
1036 fprintf(fp, " Van der Waals type : %s\n", EVDWTYPE(ir->vdwtype));
1037 if (EVDW_SWITCHED(ir->vdwtype))
1038 fprintf(fp, " rvdw_switch : %f nm\n", ir->rvdw_switch);
1039 if (EPME_SWITCHED(ir->coulombtype))
1040 fprintf(fp, " rlist : %f nm\n", ir->rlist);
1041 if (ir->rlistlong != max_cutoff(ir->rvdw,ir->rcoulomb))
1042 fprintf(fp, " rlistlong : %f nm\n", ir->rlistlong);
1044 /* Print a descriptive line about the tpr settings tested */
1045 fprintf(fp, "\nWill try these real/reciprocal workload settings:\n");
1046 fprintf(fp, " No. scaling rcoulomb");
1047 fprintf(fp, " nkx nky nkz");
1048 fprintf(fp, " spacing");
1049 if (evdwCUT == ir->vdwtype)
1050 fprintf(fp, " rvdw");
1051 if (EPME_SWITCHED(ir->coulombtype))
1052 fprintf(fp, " rlist");
1053 if ( ir->rlistlong != max_cutoff(ir->rlist,max_cutoff(ir->rvdw,ir->rcoulomb)) )
1054 fprintf(fp, " rlistlong");
1055 fprintf(fp, " tpr file\n");
1058 /* Get useful PME grids if requested, the actual number of entries is
1059 * returned in npmegrid */
1061 make_grid_list(downfac, upfac, ntprs, state.box, &pmegrid, fourierspacing);
1063 /* Loop to create the requested number of tpr input files */
1064 for (j = 0; j < *ntprs; j++)
1066 /* The first one is the provided tpr file, just need to modify the number
1067 * of steps, so skip the following block */
1070 /* Scale the Coulomb radius */
1071 ir->rcoulomb = info->rcoulomb[0]*pmegrid[j].fac;
1073 /* Adjust other radii since various conditions neet to be fulfilled */
1074 if (eelPME == ir->coulombtype)
1076 /* plain PME, rcoulomb must be equal to rlist */
1077 ir->rlist = ir->rcoulomb;
1081 /* rlist must be >= rcoulomb, we keep the size of the buffer region */
1082 ir->rlist = ir->rcoulomb + nlist_buffer;
1085 if (evdwCUT == ir->vdwtype)
1087 /* For vdw cutoff, rvdw >= rlist */
1088 ir->rvdw = max(info->rvdw[0], ir->rlist);
1091 ir->rlistlong = max_cutoff(ir->rlist,max_cutoff(ir->rvdw,ir->rcoulomb));
1093 /* Copy the optimal grid dimensions to ir */
1094 ir->nkx = pmegrid[j].nkx;
1095 ir->nky = pmegrid[j].nky;
1096 ir->nkz = pmegrid[j].nkz;
1098 } /* end of "if (j != 0)" */
1100 /* for j==0: Save the original settings
1101 * for j >0: Save modified radii and fourier grids */
1102 info->rcoulomb[j] = ir->rcoulomb;
1103 info->rvdw[j] = ir->rvdw;
1104 info->nkx[j] = ir->nkx;
1105 info->nky[j] = ir->nky;
1106 info->nkz[j] = ir->nkz;
1107 info->rlist[j] = ir->rlist;
1108 info->rlistlong[j] = ir->rlistlong;
1109 info->fsx[j] = pmegrid[j].fs;
1110 info->fsy[j] = pmegrid[j].fs;
1111 info->fsz[j] = pmegrid[j].fs;
1113 /* Write the benchmark tpr file */
1114 strncpy(fn_bench_tprs[j],fn_sim_tpr,strlen(fn_sim_tpr)-strlen(".tpr"));
1115 sprintf(buf, "_bench%.2d.tpr", j);
1116 strcat(fn_bench_tprs[j], buf);
1117 fprintf(stdout,"Writing benchmark tpr %s with nsteps=", fn_bench_tprs[j]);
1118 fprintf(stdout, gmx_large_int_pfmt, ir->nsteps);
1120 fprintf(stdout,", scaling factor %f\n", pmegrid[j].fac);
1122 fprintf(stdout,", unmodified settings\n");
1124 write_tpx_state(fn_bench_tprs[j],ir,&state,&mtop);
1126 /* Write information about modified tpr settings to log file */
1128 fprintf(fp, "%4d%10s%10f", j, "-input-", ir->rcoulomb);
1130 fprintf(fp, "%4d%10f%10f", j, pmegrid[j].fac, ir->rcoulomb);
1131 fprintf(fp, "%5d%5d%5d", ir->nkx, ir->nky, ir->nkz);
1132 fprintf(fp, " %9f ", info->fsx[j]);
1133 if (evdwCUT == ir->vdwtype)
1134 fprintf(fp, "%10f", ir->rvdw);
1135 if (EPME_SWITCHED(ir->coulombtype))
1136 fprintf(fp, "%10f", ir->rlist);
1137 if ( info->rlistlong[0] != max_cutoff(info->rlist[0],max_cutoff(info->rvdw[0],info->rcoulomb[0])) )
1138 fprintf(fp, "%10f", ir->rlistlong);
1139 fprintf(fp, " %-14s\n",fn_bench_tprs[j]);
1141 /* Make it clear to the user that some additional settings were modified */
1142 if ( !is_equal(ir->rvdw , info->rvdw[0])
1143 || !is_equal(ir->rlistlong, info->rlistlong[0]) )
1149 fprintf(fp, "\nNote that in addition to rcoulomb and the fourier grid\n"
1150 "also other input settings were changed (see table above).\n"
1151 "Please check if the modified settings are appropriate.\n");
1158 /* Rename the files we want to keep to some meaningful filename and
1159 * delete the rest */
1160 static void cleanup(const t_filenm *fnm, int nfile, int k, int nnodes,
1161 int nPMEnodes, int nr, gmx_bool bKeepStderr)
1163 char numstring[STRLEN];
1164 char newfilename[STRLEN];
1165 const char *fn=NULL;
1170 fprintf(stdout, "Cleaning up, deleting benchmark temp files ...\n");
1172 for (i=0; i<nfile; i++)
1174 opt = (char *)fnm[i].opt;
1175 if ( strcmp(opt, "-p") == 0 )
1177 /* do nothing; keep this file */
1180 else if (strcmp(opt, "-bg") == 0)
1182 /* Give the log file a nice name so one can later see which parameters were used */
1183 numstring[0] = '\0';
1185 sprintf(numstring, "_%d", nr);
1186 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", opt2fn("-bg",nfile,fnm), k, nnodes, nPMEnodes, numstring);
1187 if (gmx_fexist(opt2fn("-bg",nfile,fnm)))
1189 fprintf(stdout, "renaming log file to %s\n", newfilename);
1190 make_backup(newfilename);
1191 rename(opt2fn("-bg",nfile,fnm), newfilename);
1194 else if (strcmp(opt, "-err") == 0)
1196 /* This file contains the output of stderr. We want to keep it in
1197 * cases where there have been problems. */
1198 fn = opt2fn(opt, nfile, fnm);
1199 numstring[0] = '\0';
1201 sprintf(numstring, "_%d", nr);
1202 sprintf(newfilename, "%s_no%d_np%d_npme%d%s", fn, k, nnodes, nPMEnodes, numstring);
1207 fprintf(stdout, "Saving stderr output in %s\n", newfilename);
1208 make_backup(newfilename);
1209 rename(fn, newfilename);
1213 fprintf(stdout, "Deleting %s\n", fn);
1218 /* Delete the files which are created for each benchmark run: (options -b*) */
1219 else if ( (0 == strncmp(opt, "-b", 2)) && (opt2bSet(opt,nfile,fnm) || !is_optional(&fnm[i])) )
1221 fn = opt2fn(opt, nfile, fnm);
1224 fprintf(stdout, "Deleting %s\n", fn);
1232 /* Returns the largest common factor of n1 and n2 */
1233 static int largest_common_factor(int n1, int n2)
1238 for (factor=nmax; factor > 0; factor--)
1240 if ( 0==(n1 % factor) && 0==(n2 % factor) )
1245 return 0; /* one for the compiler */
1248 enum {eNpmeAuto, eNpmeAll, eNpmeReduced, eNpmeSubset, eNpmeNr};
1250 /* Create a list of numbers of PME nodes to test */
1251 static void make_npme_list(
1252 const char *npmevalues_opt, /* Make a complete list with all
1253 * possibilities or a short list that keeps only
1254 * reasonable numbers of PME nodes */
1255 int *nentries, /* Number of entries we put in the nPMEnodes list */
1256 int *nPMEnodes[], /* Each entry contains the value for -npme */
1257 int nnodes, /* Total number of nodes to do the tests on */
1258 int minPMEnodes, /* Minimum number of PME nodes */
1259 int maxPMEnodes) /* Maximum number of PME nodes */
1262 int min_factor=1; /* We request that npp and npme have this minimal
1263 * largest common factor (depends on npp) */
1264 int nlistmax; /* Max. list size */
1265 int nlist; /* Actual number of entries in list */
1269 /* Do we need to check all possible values for -npme or is a reduced list enough? */
1270 if ( 0 == strcmp(npmevalues_opt, "all") )
1274 else if ( 0 == strcmp(npmevalues_opt, "subset") )
1276 eNPME = eNpmeSubset;
1278 else if ( 0 == strcmp(npmevalues_opt, "auto") )
1282 else if (nnodes < 128)
1283 eNPME = eNpmeReduced;
1285 eNPME = eNpmeSubset;
1289 gmx_fatal(FARGS, "Unknown option for -npme in make_npme_list");
1292 /* Calculate how many entries we could possibly have (in case of -npme all) */
1295 nlistmax = maxPMEnodes - minPMEnodes + 3;
1296 if (0 == minPMEnodes)
1302 /* Now make the actual list which is at most of size nlist */
1303 snew(*nPMEnodes, nlistmax);
1304 nlist = 0; /* start counting again, now the real entries in the list */
1305 for (i = 0; i < nlistmax - 2; i++)
1307 npme = maxPMEnodes - i;
1318 /* For 2d PME we want a common largest factor of at least the cube
1319 * root of the number of PP nodes */
1320 min_factor = (int) pow(npp, 1.0/3.0);
1323 gmx_fatal(FARGS, "Unknown option for eNPME in make_npme_list");
1326 if (largest_common_factor(npp, npme) >= min_factor)
1328 (*nPMEnodes)[nlist] = npme;
1332 /* We always test 0 PME nodes and the automatic number */
1333 *nentries = nlist + 2;
1334 (*nPMEnodes)[nlist ] = 0;
1335 (*nPMEnodes)[nlist+1] = -1;
1337 fprintf(stderr, "Will try the following %d different values for -npme:\n", *nentries);
1338 for (i=0; i<*nentries-1; i++)
1339 fprintf(stderr, "%d, ", (*nPMEnodes)[i]);
1340 fprintf(stderr, "and %d (auto).\n", (*nPMEnodes)[*nentries-1]);
1344 /* Allocate memory to store the performance data */
1345 static void init_perfdata(t_perf *perfdata[], int ntprs, int datasets, int repeats)
1350 for (k=0; k<ntprs; k++)
1352 snew(perfdata[k], datasets);
1353 for (i=0; i<datasets; i++)
1355 for (j=0; j<repeats; j++)
1357 snew(perfdata[k][i].Gcycles , repeats);
1358 snew(perfdata[k][i].ns_per_day, repeats);
1359 snew(perfdata[k][i].PME_f_load, repeats);
1366 static void do_the_tests(
1367 FILE *fp, /* General g_tune_pme output file */
1368 char **tpr_names, /* Filenames of the input files to test */
1369 int maxPMEnodes, /* Max fraction of nodes to use for PME */
1370 int minPMEnodes, /* Min fraction of nodes to use for PME */
1371 int npme_fixed, /* If >= -1, test fixed number of PME
1373 const char *npmevalues_opt, /* Which -npme values should be tested */
1374 t_perf **perfdata, /* Here the performace data is stored */
1375 int *pmeentries, /* Entries in the nPMEnodes list */
1376 int repeats, /* Repeat each test this often */
1377 int nnodes, /* Total number of nodes = nPP + nPME */
1378 int nr_tprs, /* Total number of tpr files to test */
1379 gmx_bool bThreads, /* Threads or MPI? */
1380 char *cmd_mpirun, /* mpirun command string */
1381 char *cmd_np, /* "-np", "-n", whatever mpirun needs */
1382 char *cmd_mdrun, /* mdrun command string */
1383 char *cmd_args_bench, /* arguments for mdrun in a string */
1384 const t_filenm *fnm, /* List of filenames from command line */
1385 int nfile, /* Number of files specified on the cmdl. */
1386 int sim_part, /* For checkpointing */
1387 int presteps, /* DLB equilibration steps, is checked */
1388 gmx_large_int_t cpt_steps) /* Time step counter in the checkpoint */
1390 int i,nr,k,ret,count=0,totaltests;
1391 int *nPMEnodes=NULL;
1394 char *command, *cmd_stub;
1396 gmx_bool bResetProblem=FALSE;
1399 /* This string array corresponds to the eParselog enum type at the start
1401 const char* ParseLog[] = {"OK.",
1402 "Logfile not found!",
1403 "No timings, logfile truncated?",
1404 "Run was terminated.",
1405 "Counters were not reset properly.",
1406 "No DD grid found for these settings.",
1407 "TPX version conflict!",
1408 "mdrun was not started in parallel!",
1409 "A fatal error occured!" };
1410 char str_PME_f_load[13];
1413 /* Allocate space for the mdrun command line. 100 extra characters should
1414 be more than enough for the -npme etcetera arguments */
1415 cmdline_length = strlen(cmd_mpirun)
1418 + strlen(cmd_args_bench)
1419 + strlen(tpr_names[0]) + 100;
1420 snew(command , cmdline_length);
1421 snew(cmd_stub, cmdline_length);
1423 /* Construct the part of the command line that stays the same for all tests: */
1426 sprintf(cmd_stub, "%s%s", cmd_mdrun, cmd_np);
1430 sprintf(cmd_stub, "%s%s%s ", cmd_mpirun, cmd_np, cmd_mdrun);
1433 /* Create a list of numbers of PME nodes to test */
1434 if (npme_fixed < -1)
1436 make_npme_list(npmevalues_opt, pmeentries, &nPMEnodes,
1437 nnodes, minPMEnodes, maxPMEnodes);
1443 nPMEnodes[0] = npme_fixed;
1444 fprintf(stderr, "Will use a fixed number of %d PME-only nodes.\n", nPMEnodes[0]);
1449 fprintf(fp, "\nNo benchmarks done since number of repeats (-r) is 0.\n");
1451 finalize(opt2fn("-p", nfile, fnm));
1455 /* Allocate one dataset for each tpr input file: */
1456 init_perfdata(perfdata, nr_tprs, *pmeentries, repeats);
1458 /*****************************************/
1459 /* Main loop over all tpr files to test: */
1460 /*****************************************/
1461 totaltests = nr_tprs*(*pmeentries)*repeats;
1462 for (k=0; k<nr_tprs;k++)
1464 fprintf(fp, "\nIndividual timings for input file %d (%s):\n", k, tpr_names[k]);
1465 fprintf(fp, "PME nodes Gcycles ns/day PME/f Remark\n");
1466 /* Loop over various numbers of PME nodes: */
1467 for (i = 0; i < *pmeentries; i++)
1469 pd = &perfdata[k][i];
1471 /* Loop over the repeats for each scenario: */
1472 for (nr = 0; nr < repeats; nr++)
1474 pd->nPMEnodes = nPMEnodes[i];
1476 /* Add -npme and -s to the command line and save it. Note that
1477 * the -passall (if set) options requires cmd_args_bench to be
1478 * at the end of the command line string */
1479 snew(pd->mdrun_cmd_line, cmdline_length);
1480 sprintf(pd->mdrun_cmd_line, "%s-npme %d -s %s %s",
1481 cmd_stub, pd->nPMEnodes, tpr_names[k], cmd_args_bench);
1483 /* Do a benchmark simulation: */
1485 sprintf(buf, ", pass %d/%d", nr+1, repeats);
1488 fprintf(stdout, "\n=== Progress %2.0f%%, tpr %d/%d, run %d/%d%s:\n",
1489 (100.0*count)/totaltests,
1490 k+1, nr_tprs, i+1, *pmeentries, buf);
1491 make_backup(opt2fn("-err",nfile,fnm));
1492 sprintf(command, "%s 1> /dev/null 2>%s", pd->mdrun_cmd_line, opt2fn("-err",nfile,fnm));
1493 fprintf(stdout, "%s\n", pd->mdrun_cmd_line);
1494 gmx_system_call(command);
1496 /* Collect the performance data from the log file; also check stderr
1497 * for fatal errors */
1498 ret = parse_logfile(opt2fn("-bg",nfile,fnm), opt2fn("-err",nfile,fnm),
1499 pd, nr, presteps, cpt_steps, nnodes);
1500 if ((presteps > 0) && (ret == eParselogResetProblem))
1501 bResetProblem = TRUE;
1503 if (-1 == pd->nPMEnodes)
1504 sprintf(buf, "(%3d)", pd->guessPME);
1509 if (pd->PME_f_load[nr] > 0.0)
1510 sprintf(str_PME_f_load, "%12.3f", pd->PME_f_load[nr]);
1512 sprintf(str_PME_f_load, "%s", " - ");
1514 /* Write the data we got to disk */
1515 fprintf(fp, "%4d%s %12.3f %12.3f %s %s", pd->nPMEnodes,
1516 buf, pd->Gcycles[nr], pd->ns_per_day[nr], str_PME_f_load, ParseLog[ret]);
1517 if (! (ret==eParselogOK || ret==eParselogNoDDGrid || ret==eParselogNotFound) )
1518 fprintf(fp, " Check %s file for problems.", ret==eParselogFatal? "err":"log");
1523 /* Do some cleaning up and delete the files we do not need any more */
1524 cleanup(fnm, nfile, k, nnodes, pd->nPMEnodes, nr, ret==eParselogFatal);
1526 /* If the first run with this number of processors already failed, do not try again: */
1527 if (pd->Gcycles[0] <= 0.0 && repeats > 1)
1529 fprintf(stdout, "Skipping remaining passes of unsuccessful setting, see log file for details.\n");
1530 count += repeats-(nr+1);
1533 } /* end of repeats loop */
1534 } /* end of -npme loop */
1535 } /* end of tpr file loop */
1539 fprintf(fp, "WARNING: The cycle and time step counters could not be reset properly. ");
1547 static void check_input(
1553 real maxPMEfraction,
1554 real minPMEfraction,
1556 real fourierspacing,
1557 gmx_large_int_t bench_nsteps,
1558 const t_filenm *fnm,
1565 /* Make sure the input file exists */
1566 if (!gmx_fexist(opt2fn("-s",nfile,fnm)))
1567 gmx_fatal(FARGS, "File %s not found.", opt2fn("-s",nfile,fnm));
1569 /* Make sure that the checkpoint file is not overwritten during benchmarking */
1570 if ( (0 == strcmp(opt2fn("-cpi",nfile,fnm), opt2fn("-bcpo",nfile,fnm)) ) && (sim_part > 1) )
1571 gmx_fatal(FARGS, "Checkpoint input (-cpi) and benchmark checkpoint output (-bcpo) files must not be identical.\n"
1572 "The checkpoint input file must not be overwritten during the benchmarks.\n");
1574 /* Make sure that repeats is >= 0 (if == 0, only write tpr files) */
1576 gmx_fatal(FARGS, "Number of repeats < 0!");
1578 /* Check number of nodes */
1580 gmx_fatal(FARGS, "Number of nodes/threads must be a positive integer.");
1582 /* Automatically choose -ntpr if not set */
1589 fprintf(stderr, "Will test %d tpr file%s.\n", *ntprs, *ntprs==1?"":"s");
1594 fprintf(stderr, "Note: Choose ntpr>1 to shift PME load between real and reciprocal space.\n");
1597 if ( is_equal(*downfac,*upfac) && (*ntprs > 2) && !(fourierspacing > 0.0))
1599 fprintf(stderr, "WARNING: Resetting -ntpr to 2 since both scaling factors are the same.\n"
1600 "Please choose upfac unequal to downfac to test various PME grid settings\n");
1604 /* Check whether max and min fraction are within required values */
1605 if (maxPMEfraction > 0.5 || maxPMEfraction < 0)
1606 gmx_fatal(FARGS, "-max must be between 0 and 0.5");
1607 if (minPMEfraction > 0.5 || minPMEfraction < 0)
1608 gmx_fatal(FARGS, "-min must be between 0 and 0.5");
1609 if (maxPMEfraction < minPMEfraction)
1610 gmx_fatal(FARGS, "-max must be larger or equal to -min");
1612 /* Check whether the number of steps - if it was set - has a reasonable value */
1613 if (bench_nsteps < 0)
1614 gmx_fatal(FARGS, "Number of steps must be positive.");
1616 if (bench_nsteps > 10000 || bench_nsteps < 100)
1618 fprintf(stderr, "WARNING: steps=");
1619 fprintf(stderr, gmx_large_int_pfmt, bench_nsteps);
1620 fprintf(stderr, ". Are you sure you want to perform so %s steps for each benchmark?\n", (bench_nsteps < 100)? "few" : "many");
1625 gmx_fatal(FARGS, "Cannot have a negative number of presteps.\n");
1628 if (*upfac <= 0.0 || *downfac <= 0.0 || *downfac > *upfac)
1629 gmx_fatal(FARGS, "Both scaling factors must be larger than zero and upper\n"
1630 "scaling limit (%f) must be larger than lower limit (%f).",
1633 if (*downfac < 0.75 || *upfac > 1.25)
1634 fprintf(stderr, "WARNING: Applying extreme scaling factor. I hope you know what you are doing.\n");
1636 if (fourierspacing < 0)
1637 gmx_fatal(FARGS, "Please choose a positive value for fourierspacing.");
1639 /* Make shure that the scaling factor options are compatible with the number of tprs */
1640 if ( (*ntprs < 3) && ( opt2parg_bSet("-upfac",npargs,pa) || opt2parg_bSet("-downfac",npargs,pa) ) )
1642 if (opt2parg_bSet("-upfac",npargs,pa) && opt2parg_bSet("-downfac",npargs,pa) && !is_equal(*upfac,*downfac))
1644 fprintf(stderr, "NOTE: Specify -ntpr > 2 for both scaling factors to take effect.\n"
1645 "(upfac=%f, downfac=%f)\n", *upfac, *downfac);
1647 if (opt2parg_bSet("-upfac",npargs,pa))
1649 if (opt2parg_bSet("-downfac",npargs,pa))
1652 if ( (2 == *ntprs) && (opt2parg_bSet("-downfac",npargs,pa)) && !is_equal(*downfac, 1.0))
1657 /* If a fixed number of PME nodes is set we do rcoulomb and PME gird tuning
1658 * only. We need to check whether the requested number of PME-only nodes
1660 if (npme_fixed > -1)
1662 /* No more than 50% of all nodes can be assigned as PME-only nodes. */
1663 if (2*npme_fixed > nnodes)
1665 gmx_fatal(FARGS, "Cannot have more than %d PME-only nodes for a total of %d nodes (you chose %d).\n",
1666 nnodes/2, nnodes, npme_fixed);
1668 if ((npme_fixed > 0) && (5*npme_fixed < nnodes))
1670 fprintf(stderr, "WARNING: Only %g percent of the nodes are assigned as PME-only nodes.\n",
1671 100.0*((real)npme_fixed / (real)nnodes));
1673 if (opt2parg_bSet("-min",npargs,pa) || opt2parg_bSet("-max",npargs,pa))
1675 fprintf(stderr, "NOTE: The -min, -max, and -npme options have no effect when a\n"
1676 " fixed number of PME-only nodes is requested with -fix.\n");
1682 /* Returns TRUE when "opt" is a switch for g_tune_pme itself */
1683 static gmx_bool is_main_switch(char *opt)
1685 if ( (0 == strcmp(opt,"-s" ))
1686 || (0 == strcmp(opt,"-p" ))
1687 || (0 == strcmp(opt,"-launch" ))
1688 || (0 == strcmp(opt,"-r" ))
1689 || (0 == strcmp(opt,"-ntpr" ))
1690 || (0 == strcmp(opt,"-max" ))
1691 || (0 == strcmp(opt,"-min" ))
1692 || (0 == strcmp(opt,"-upfac" ))
1693 || (0 == strcmp(opt,"-downfac" ))
1694 || (0 == strcmp(opt,"-fix" ))
1695 || (0 == strcmp(opt,"-four" ))
1696 || (0 == strcmp(opt,"-steps" ))
1697 || (0 == strcmp(opt,"-simsteps" ))
1698 || (0 == strcmp(opt,"-resetstep"))
1699 || (0 == strcmp(opt,"-so" ))
1700 || (0 == strcmp(opt,"-npstring" ))
1701 || (0 == strcmp(opt,"-npme" ))
1702 || (0 == strcmp(opt,"-err" ))
1703 || (0 == strcmp(opt,"-passall" )) )
1710 /* Returns TRUE when "opt" is needed at launch time */
1711 static gmx_bool is_launch_option(char *opt, gmx_bool bSet)
1720 /* Returns TRUE when "opt" is needed at launch time */
1721 static gmx_bool is_launch_file(char *opt, gmx_bool bSet)
1723 /* We need all options that were set on the command line
1724 * and that do not start with -b */
1725 if (0 == strncmp(opt,"-b", 2))
1735 /* Returns TRUE when "opt" gives an option needed for the benchmarks runs */
1736 static gmx_bool is_bench_option(char *opt, gmx_bool bSet)
1738 /* If option is set, we might need it for the benchmarks.
1739 * This includes -cpi */
1742 if ( (0 == strcmp(opt, "-append" ))
1743 || (0 == strcmp(opt, "-maxh" ))
1744 || (0 == strcmp(opt, "-deffnm" ))
1745 || (0 == strcmp(opt, "-resethway"))
1746 || (0 == strcmp(opt, "-cpnum" )) )
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 /* All options starting with "-b" are for _b_enchmark files exclusively */
1760 if (0 == strncmp(opt,"-b", 2))
1762 if (!bOptional || bSet)
1772 if (bSet) /* These are additional input files like -cpi -ei */
1780 /* Adds 'buf' to 'str' */
1781 static void add_to_string(char **str, char *buf)
1786 len = strlen(*str) + strlen(buf) + 1;
1792 /* Create the command line for the benchmark as well as for the real run */
1793 static void create_command_line_snippets(
1800 const char *procstring, /* How to pass the number of processors to $MPIRUN */
1801 char *cmd_np[], /* Actual command line snippet, e.g. '-np <N>' */
1802 char *cmd_args_bench[], /* command line arguments for benchmark runs */
1803 char *cmd_args_launch[], /* command line arguments for simulation run */
1804 char extra_args[]) /* Add this to the end of the command line */
1810 #define BUFLENGTH 255
1811 char buf[BUFLENGTH];
1812 char strbuf[BUFLENGTH];
1813 char strbuf2[BUFLENGTH];
1817 np_or_nt=strdup("-nt");
1819 np_or_nt=strdup("-np");
1821 /* strlen needs at least '\0' as a string: */
1822 snew(*cmd_args_bench ,1);
1823 snew(*cmd_args_launch,1);
1824 *cmd_args_launch[0]='\0';
1825 *cmd_args_bench[0] ='\0';
1828 /*******************************************/
1829 /* 1. Process other command line arguments */
1830 /*******************************************/
1831 for (i=0; i<npargs; i++)
1833 /* What command line switch are we currently processing: */
1834 opt = (char *)pa[i].option;
1835 /* Skip options not meant for mdrun */
1836 if (!is_main_switch(opt))
1838 /* Boolean arguments need to be generated in the -[no]argname format */
1839 if (pa[i].type == etBOOL)
1841 sprintf(strbuf,"-%s%s ",*pa[i].u.b ? "" : "no",opt+1);
1845 /* Print it to a string buffer, strip away trailing whitespaces that pa_val also returns: */
1846 sprintf(strbuf2, "%s", pa_val(&pa[i],buf,BUFLENGTH));
1848 sprintf(strbuf, "%s %s ", opt, strbuf2);
1851 /* We need the -np (or -nt) switch in a separate buffer - whether or not it was set! */
1852 if (0 == strcmp(opt,np_or_nt))
1854 if (strcmp(procstring, "none")==0 && !bThreads)
1856 /* Omit -np <N> entirely */
1858 sprintf(*cmd_np, " ");
1862 /* This is the normal case with -np <N> */
1863 snew(*cmd_np, strlen(procstring)+strlen(strbuf2)+4);
1864 sprintf(*cmd_np, " %s %s ", bThreads? "-nt" : procstring, strbuf2);
1869 if (is_bench_option(opt,pa[i].bSet))
1870 add_to_string(cmd_args_bench, strbuf);
1872 if (is_launch_option(opt,pa[i].bSet))
1873 add_to_string(cmd_args_launch, strbuf);
1879 /* Add equilibration steps to benchmark options */
1880 sprintf(strbuf, "-resetstep %d ", presteps);
1881 add_to_string(cmd_args_bench, strbuf);
1884 /********************/
1885 /* 2. Process files */
1886 /********************/
1887 for (i=0; i<nfile; i++)
1889 opt = (char *)fnm[i].opt;
1890 name = opt2fn(opt,nfile,fnm);
1892 /* Strbuf contains the options, now let's sort out where we need that */
1893 sprintf(strbuf, "%s %s ", opt, name);
1895 /* Skip options not meant for mdrun */
1896 if (!is_main_switch(opt))
1899 if ( is_bench_file(opt, opt2bSet(opt,nfile,fnm), is_optional(&fnm[i]), is_output(&fnm[i])) )
1901 /* All options starting with -b* need the 'b' removed,
1902 * therefore overwrite strbuf */
1903 if (0 == strncmp(opt, "-b", 2))
1904 sprintf(strbuf, "-%s %s ", &opt[2], name);
1906 add_to_string(cmd_args_bench, strbuf);
1909 if ( is_launch_file(opt,opt2bSet(opt,nfile,fnm)) )
1910 add_to_string(cmd_args_launch, strbuf);
1914 add_to_string(cmd_args_bench , extra_args);
1915 add_to_string(cmd_args_launch, extra_args);
1920 /* Set option opt */
1921 static void setopt(const char *opt,int nfile,t_filenm fnm[])
1925 for(i=0; (i<nfile); i++)
1926 if (strcmp(opt,fnm[i].opt)==0)
1927 fnm[i].flag |= ffSET;
1931 /* This routine checks for output files that get triggered by a tpr option.
1932 * These output files are marked as set to allow for proper cleanup after
1933 * each tuning run. */
1934 static void get_tpr_outfiles(int nfile, t_filenm fnm[])
1936 gmx_bool bPull; /* Is pulling requested in .tpr file? */
1937 gmx_bool bTpi; /* Is test particle insertion requested? */
1938 gmx_bool bFree; /* Is a free energy simulation requested? */
1939 gmx_bool bNM; /* Is a normal mode analysis requested? */
1945 /* Check tpr file for options that trigger extra output files */
1946 read_tpx_state(opt2fn("-s",nfile,fnm),&ir,&state,NULL,&mtop);
1947 bPull = (epullNO != ir.ePull);
1948 bFree = (efepNO != ir.efep );
1949 bNM = (eiNM == ir.eI );
1950 bTpi = EI_TPI(ir.eI);
1952 /* Set these output files on the tuning command-line */
1955 setopt("-pf" , nfile, fnm);
1956 setopt("-px" , nfile, fnm);
1960 setopt("-dhdl", nfile, fnm);
1964 setopt("-tpi" , nfile, fnm);
1965 setopt("-tpid", nfile, fnm);
1969 setopt("-mtx" , nfile, fnm);
1974 static void couple_files_options(int nfile, t_filenm fnm[])
1977 gmx_bool bSet,bBench;
1982 for (i=0; i<nfile; i++)
1984 opt = (char *)fnm[i].opt;
1985 bSet = ((fnm[i].flag & ffSET) != 0);
1986 bBench = (0 == strncmp(opt,"-b", 2));
1988 /* Check optional files */
1989 /* If e.g. -eo is set, then -beo also needs to be set */
1990 if (is_optional(&fnm[i]) && bSet && !bBench)
1992 sprintf(buf, "-b%s", &opt[1]);
1993 setopt(buf,nfile,fnm);
1995 /* If -beo is set, then -eo also needs to be! */
1996 if (is_optional(&fnm[i]) && bSet && bBench)
1998 sprintf(buf, "-%s", &opt[2]);
1999 setopt(buf,nfile,fnm);
2005 static double gettime()
2007 #ifdef HAVE_GETTIMEOFDAY
2011 gettimeofday(&t,NULL);
2013 seconds = (double) t.tv_sec + 1e-6*(double)t.tv_usec;
2019 seconds = time(NULL);
2026 #define BENCHSTEPS (1000)
2028 int gmx_tune_pme(int argc,char *argv[])
2030 const char *desc[] = {
2031 "For a given number [TT]-np[tt] or [TT]-nt[tt] of processors/threads, this program systematically",
2032 "times [TT]mdrun[tt] with various numbers of PME-only nodes and determines",
2033 "which setting is fastest. It will also test whether performance can",
2034 "be enhanced by shifting load from the reciprocal to the real space",
2035 "part of the Ewald sum. ",
2036 "Simply pass your [TT].tpr[tt] file to [TT]g_tune_pme[tt] together with other options",
2037 "for [TT]mdrun[tt] as needed.[PAR]",
2038 "Which executables are used can be set in the environment variables",
2039 "MPIRUN and MDRUN. If these are not present, 'mpirun' and 'mdrun'",
2040 "will be used as defaults. Note that for certain MPI frameworks you",
2041 "need to provide a machine- or hostfile. This can also be passed",
2042 "via the MPIRUN variable, e.g.[PAR]",
2043 "[TT]export MPIRUN=\"/usr/local/mpirun -machinefile hosts\"[tt][PAR]",
2044 "Please call [TT]g_tune_pme[tt] with the normal options you would pass to",
2045 "[TT]mdrun[tt] and add [TT]-np[tt] for the number of processors to perform the",
2046 "tests on, or [TT]-nt[tt] for the number of threads. You can also add [TT]-r[tt]",
2047 "to repeat each test several times to get better statistics. [PAR]",
2048 "[TT]g_tune_pme[tt] can test various real space / reciprocal space workloads",
2049 "for you. With [TT]-ntpr[tt] you control how many extra [TT].tpr[tt] files will be",
2050 "written with enlarged cutoffs and smaller fourier grids respectively.",
2051 "Typically, the first test (number 0) will be with the settings from the input",
2052 "[TT].tpr[tt] file; the last test (number [TT]ntpr[tt]) will have cutoffs multiplied",
2053 "by (and at the same time fourier grid dimensions divided by) the scaling",
2054 "factor [TT]-fac[tt] (default 1.2). The remaining [TT].tpr[tt] files will have about ",
2055 "equally-spaced values in between these extremes. [BB]Note[bb] that you can set [TT]-ntpr[tt] to 1",
2056 "if you just want to find the optimal number of PME-only nodes; in that case",
2057 "your input [TT].tpr[tt] file will remain unchanged.[PAR]",
2058 "For the benchmark runs, the default of 1000 time steps should suffice for most",
2059 "MD systems. The dynamic load balancing needs about 100 time steps",
2060 "to adapt to local load imbalances, therefore the time step counters",
2061 "are by default reset after 100 steps. For large systems",
2062 "(>1M atoms) you may have to set [TT]-resetstep[tt] to a higher value.",
2063 "From the 'DD' load imbalance entries in the md.log output file you",
2064 "can tell after how many steps the load is sufficiently balanced. Example call:[PAR]"
2065 "[TT]g_tune_pme -np 64 -s protein.tpr -launch[tt][PAR]",
2066 "After calling [TT]mdrun[tt] several times, detailed performance information",
2067 "is available in the output file [TT]perf.out.[tt] ",
2068 "[BB]Note[bb] that during the benchmarks, a couple of temporary files are written",
2069 "(options [TT]-b[tt]*), these will be automatically deleted after each test.[PAR]",
2070 "If you want the simulation to be started automatically with the",
2071 "optimized parameters, use the command line option [TT]-launch[tt].[PAR]",
2076 int pmeentries=0; /* How many values for -npme do we actually test for each tpr file */
2077 real maxPMEfraction=0.50;
2078 real minPMEfraction=0.25;
2079 int maxPMEnodes, minPMEnodes;
2080 int npme_fixed=-2; /* If >= -1, use only this number
2081 * of PME-only nodes */
2082 real downfac=1.0,upfac=1.2;
2084 real fs=0.0; /* 0 indicates: not set by the user */
2085 gmx_large_int_t bench_nsteps=BENCHSTEPS;
2086 gmx_large_int_t new_sim_nsteps=-1; /* -1 indicates: not set by the user */
2087 gmx_large_int_t cpt_steps=0; /* Step counter in .cpt input file */
2088 int presteps=100; /* Do a full cycle reset after presteps steps */
2089 gmx_bool bOverwrite=FALSE, bKeepTPR;
2090 gmx_bool bLaunch=FALSE;
2091 gmx_bool bPassAll=FALSE;
2092 char *ExtraArgs=NULL;
2093 char **tpr_names=NULL;
2094 const char *simulation_tpr=NULL;
2095 int best_npme, best_tpr;
2096 int sim_part = 1; /* For benchmarks with checkpoint files */
2098 /* Default program names if nothing else is found */
2099 char *cmd_mpirun=NULL, *cmd_mdrun=NULL;
2100 char *cmd_args_bench, *cmd_args_launch;
2103 t_perf **perfdata=NULL;
2109 /* Print out how long the tuning took */
2112 static t_filenm fnm[] = {
2114 { efOUT, "-p", "perf", ffWRITE },
2115 { efLOG, "-err", "errors", ffWRITE },
2116 { efTPX, "-so", "tuned", ffWRITE },
2118 { efTPX, NULL, NULL, ffREAD },
2119 { efTRN, "-o", NULL, ffWRITE },
2120 { efXTC, "-x", NULL, ffOPTWR },
2121 { efCPT, "-cpi", NULL, ffOPTRD },
2122 { efCPT, "-cpo", NULL, ffOPTWR },
2123 { efSTO, "-c", "confout", ffWRITE },
2124 { efEDR, "-e", "ener", ffWRITE },
2125 { efLOG, "-g", "md", ffWRITE },
2126 { efXVG, "-dhdl", "dhdl", ffOPTWR },
2127 { efXVG, "-field", "field", ffOPTWR },
2128 { efXVG, "-table", "table", ffOPTRD },
2129 { efXVG, "-tablep", "tablep", ffOPTRD },
2130 { efXVG, "-tableb", "table", ffOPTRD },
2131 { efTRX, "-rerun", "rerun", ffOPTRD },
2132 { efXVG, "-tpi", "tpi", ffOPTWR },
2133 { efXVG, "-tpid", "tpidist", ffOPTWR },
2134 { efEDI, "-ei", "sam", ffOPTRD },
2135 { efEDO, "-eo", "sam", ffOPTWR },
2136 { efGCT, "-j", "wham", ffOPTRD },
2137 { efGCT, "-jo", "bam", ffOPTWR },
2138 { efXVG, "-ffout", "gct", ffOPTWR },
2139 { efXVG, "-devout", "deviatie", ffOPTWR },
2140 { efXVG, "-runav", "runaver", ffOPTWR },
2141 { efXVG, "-px", "pullx", ffOPTWR },
2142 { efXVG, "-pf", "pullf", ffOPTWR },
2143 { efMTX, "-mtx", "nm", ffOPTWR },
2144 { efNDX, "-dn", "dipole", ffOPTWR },
2145 /* Output files that are deleted after each benchmark run */
2146 { efTRN, "-bo", "bench", ffWRITE },
2147 { efXTC, "-bx", "bench", ffWRITE },
2148 { efCPT, "-bcpo", "bench", ffWRITE },
2149 { efSTO, "-bc", "bench", ffWRITE },
2150 { efEDR, "-be", "bench", ffWRITE },
2151 { efLOG, "-bg", "bench", ffWRITE },
2152 { efEDO, "-beo", "bench", ffOPTWR },
2153 { efXVG, "-bdhdl", "benchdhdl",ffOPTWR },
2154 { efXVG, "-bfield", "benchfld" ,ffOPTWR },
2155 { efXVG, "-btpi", "benchtpi", ffOPTWR },
2156 { efXVG, "-btpid", "benchtpid",ffOPTWR },
2157 { efGCT, "-bjo", "bench", ffOPTWR },
2158 { efXVG, "-bffout", "benchgct", ffOPTWR },
2159 { efXVG, "-bdevout","benchdev", ffOPTWR },
2160 { efXVG, "-brunav", "benchrnav",ffOPTWR },
2161 { efXVG, "-bpx", "benchpx", ffOPTWR },
2162 { efXVG, "-bpf", "benchpf", ffOPTWR },
2163 { efMTX, "-bmtx", "benchn", ffOPTWR },
2164 { efNDX, "-bdn", "bench", ffOPTWR }
2167 /* Command line options of mdrun */
2168 gmx_bool bDDBondCheck = TRUE;
2169 gmx_bool bDDBondComm = TRUE;
2170 gmx_bool bVerbose = FALSE;
2171 gmx_bool bCompact = TRUE;
2172 gmx_bool bSepPot = FALSE;
2173 gmx_bool bRerunVSite = FALSE;
2174 gmx_bool bIonize = FALSE;
2175 gmx_bool bConfout = TRUE;
2176 gmx_bool bReproducible = FALSE;
2177 gmx_bool bThreads = FALSE;
2180 int nstglobalcomm=-1;
2182 int repl_ex_seed=-1;
2186 const char *ddno_opt[ddnoNR+1] =
2187 { NULL, "interleave", "pp_pme", "cartesian", NULL };
2188 const char *dddlb_opt[] =
2189 { NULL, "auto", "no", "yes", NULL };
2190 const char *procstring[] =
2191 { NULL, "-np", "-n", "none", NULL };
2192 const char *npmevalues_opt[] =
2193 { NULL, "auto", "all", "subset", NULL };
2194 real rdd=0.0,rconstr=0.0,dlb_scale=0.8,pforce=-1;
2195 char *ddcsx=NULL,*ddcsy=NULL,*ddcsz=NULL;
2197 #define STD_CPT_PERIOD (15.0)
2198 real cpt_period=STD_CPT_PERIOD,max_hours=-1;
2199 gmx_bool bAppendFiles=TRUE;
2200 gmx_bool bKeepAndNumCPT=FALSE;
2201 gmx_bool bResetCountersHalfWay=FALSE;
2202 output_env_t oenv=NULL;
2205 /***********************/
2206 /* g_tune_pme options: */
2207 /***********************/
2208 { "-np", FALSE, etINT, {&nnodes},
2209 "Number of nodes to run the tests on (must be > 2 for separate PME nodes)" },
2210 { "-npstring", FALSE, etENUM, {procstring},
2211 "Specify the number of processors to [TT]$MPIRUN[tt] using this string"},
2212 { "-passall", FALSE, etBOOL, {&bPassAll},
2213 "HIDDENPut arguments unknown to [TT]mdrun[tt] at the end of the command line. Can be used for debugging purposes. "},
2214 { "-nt", FALSE, etINT, {&nthreads},
2215 "Number of threads to run the tests on (turns MPI & mpirun off)"},
2216 { "-r", FALSE, etINT, {&repeats},
2217 "Repeat each test this often" },
2218 { "-max", FALSE, etREAL, {&maxPMEfraction},
2219 "Max fraction of PME nodes to test with" },
2220 { "-min", FALSE, etREAL, {&minPMEfraction},
2221 "Min fraction of PME nodes to test with" },
2222 { "-npme", FALSE, etENUM, {npmevalues_opt},
2223 "Benchmark all possible values for [TT]-npme[tt] or just the subset that is expected to perform well"},
2224 { "-fix", FALSE, etINT, {&npme_fixed},
2225 "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."},
2226 { "-upfac", FALSE, etREAL, {&upfac},
2227 "Upper limit for rcoulomb scaling factor (Note that rcoulomb upscaling results in fourier grid downscaling)" },
2228 { "-downfac", FALSE, etREAL, {&downfac},
2229 "Lower limit for rcoulomb scaling factor" },
2230 { "-ntpr", FALSE, etINT, {&ntprs},
2231 "Number of [TT].tpr[tt] files to benchmark. Create this many files with scaling factors ranging from 1.0 to fac. If < 1, automatically choose the number of [TT].tpr[tt] files to test" },
2232 { "-four", FALSE, etREAL, {&fs},
2233 "Use this fourierspacing value instead of the grid found in the [TT].tpr[tt] input file. (Spacing applies to a scaling factor of 1.0 if multiple [TT].tpr[tt] files are written)" },
2234 { "-steps", FALSE, etGMX_LARGE_INT, {&bench_nsteps},
2235 "Take timings for this many steps in the benchmark runs" },
2236 { "-resetstep",FALSE, etINT, {&presteps},
2237 "Let dlb equilibrate this many steps before timings are taken (reset cycle counters after this many steps)" },
2238 { "-simsteps", FALSE, etGMX_LARGE_INT, {&new_sim_nsteps},
2239 "If non-negative, perform this many steps in the real run (overwrites nsteps from [TT].tpr[tt], add [TT].cpt[tt] steps)" },
2240 { "-launch", FALSE, etBOOL, {&bLaunch},
2241 "Launch the real simulation after optimization" },
2242 /******************/
2243 /* mdrun options: */
2244 /******************/
2245 { "-deffnm", FALSE, etSTR, {&deffnm},
2246 "Set the default filename for all file options at launch time" },
2247 { "-ddorder", FALSE, etENUM, {ddno_opt},
2249 { "-ddcheck", FALSE, etBOOL, {&bDDBondCheck},
2250 "Check for all bonded interactions with DD" },
2251 { "-ddbondcomm",FALSE, etBOOL, {&bDDBondComm},
2252 "HIDDENUse special bonded atom communication when [TT]-rdd[tt] > cut-off" },
2253 { "-rdd", FALSE, etREAL, {&rdd},
2254 "The maximum distance for bonded interactions with DD (nm), 0 is determine from initial coordinates" },
2255 { "-rcon", FALSE, etREAL, {&rconstr},
2256 "Maximum distance for P-LINCS (nm), 0 is estimate" },
2257 { "-dlb", FALSE, etENUM, {dddlb_opt},
2258 "Dynamic load balancing (with DD)" },
2259 { "-dds", FALSE, etREAL, {&dlb_scale},
2260 "Minimum allowed dlb scaling of the DD cell size" },
2261 { "-ddcsx", FALSE, etSTR, {&ddcsx},
2262 "HIDDENThe DD cell sizes in x" },
2263 { "-ddcsy", FALSE, etSTR, {&ddcsy},
2264 "HIDDENThe DD cell sizes in y" },
2265 { "-ddcsz", FALSE, etSTR, {&ddcsz},
2266 "HIDDENThe DD cell sizes in z" },
2267 { "-gcom", FALSE, etINT, {&nstglobalcomm},
2268 "Global communication frequency" },
2269 { "-v", FALSE, etBOOL, {&bVerbose},
2270 "Be loud and noisy" },
2271 { "-compact", FALSE, etBOOL, {&bCompact},
2272 "Write a compact log file" },
2273 { "-seppot", FALSE, etBOOL, {&bSepPot},
2274 "Write separate V and dVdl terms for each interaction type and node to the log file(s)" },
2275 { "-pforce", FALSE, etREAL, {&pforce},
2276 "Print all forces larger than this (kJ/mol nm)" },
2277 { "-reprod", FALSE, etBOOL, {&bReproducible},
2278 "Try to avoid optimizations that affect binary reproducibility" },
2279 { "-cpt", FALSE, etREAL, {&cpt_period},
2280 "Checkpoint interval (minutes)" },
2281 { "-cpnum", FALSE, etBOOL, {&bKeepAndNumCPT},
2282 "Keep and number checkpoint files" },
2283 { "-append", FALSE, etBOOL, {&bAppendFiles},
2284 "Append to previous output files when continuing from checkpoint instead of adding the simulation part number to all file names (for launch only)" },
2285 { "-maxh", FALSE, etREAL, {&max_hours},
2286 "Terminate after 0.99 times this time (hours)" },
2287 { "-multi", FALSE, etINT, {&nmultisim},
2288 "Do multiple simulations in parallel" },
2289 { "-replex", FALSE, etINT, {&repl_ex_nst},
2290 "Attempt replica exchange periodically with this period (steps)" },
2291 { "-reseed", FALSE, etINT, {&repl_ex_seed},
2292 "Seed for replica exchange, -1 is generate a seed" },
2293 { "-rerunvsite", FALSE, etBOOL, {&bRerunVSite},
2294 "HIDDENRecalculate virtual site coordinates with [TT]-rerun[tt]" },
2295 { "-ionize", FALSE, etBOOL, {&bIonize},
2296 "Do a simulation including the effect of an X-ray bombardment on your system" },
2297 { "-confout", FALSE, etBOOL, {&bConfout},
2298 "HIDDENWrite the last configuration with [TT]-c[tt] and force checkpointing at the last step" },
2299 { "-stepout", FALSE, etINT, {&nstepout},
2300 "HIDDENFrequency of writing the remaining runtime" },
2301 { "-resethway", FALSE, etBOOL, {&bResetCountersHalfWay},
2302 "HIDDENReset the cycle counters after half the number of steps or halfway [TT]-maxh[tt] (launch only)" }
2306 #define NFILE asize(fnm)
2308 CopyRight(stderr,argv[0]);
2310 seconds = gettime();
2312 parse_common_args(&argc,argv,PCA_NOEXIT_ON_ARGS,
2313 NFILE,fnm,asize(pa),pa,asize(desc),desc,
2316 /* Store the remaining unparsed command line entries in a string */
2318 ExtraArgs[0] = '\0';
2319 for (i=1; i<argc; i++) /* argc will now be 1 if everything was understood */
2321 add_to_string(&ExtraArgs, argv[i]);
2322 add_to_string(&ExtraArgs, " ");
2324 if ( !bPassAll && (ExtraArgs[0] != '\0') )
2326 fprintf(stderr, "\nWARNING: The following arguments you provided have no effect:\n"
2328 "Use the -passall option to force them to appear on the command lines\n"
2329 "for the benchmark simulations%s.\n\n",
2330 ExtraArgs, bLaunch? " and at launch time" : "");
2333 if (opt2parg_bSet("-nt",asize(pa),pa))
2336 if (opt2parg_bSet("-npstring",asize(pa),pa))
2337 fprintf(stderr, "WARNING: -npstring has no effect when using threads.\n");
2340 gmx_fatal(FARGS, "Can't run multi-threaded MPI simulation yet!");
2341 /* and now we just set this; a bit of an ugly hack*/
2344 /* tpr-triggered output files */
2345 get_tpr_outfiles(NFILE,fnm);
2347 /* Automatically set -beo options if -eo is set etc. */
2348 couple_files_options(NFILE,fnm);
2350 /* Construct the command line arguments for benchmark runs
2351 * as well as for the simulation run
2353 create_command_line_snippets(bThreads,presteps,NFILE,fnm,asize(pa),pa,procstring[0],
2354 &cmd_np, &cmd_args_bench, &cmd_args_launch,
2355 bPassAll? ExtraArgs : (char *)"");
2357 /* Read in checkpoint file if requested */
2359 if(opt2bSet("-cpi",NFILE,fnm))
2362 cr->duty=DUTY_PP; /* makes the following routine happy */
2363 read_checkpoint_simulation_part(opt2fn("-cpi",NFILE,fnm),
2364 &sim_part,&cpt_steps,cr,
2365 FALSE,NFILE,fnm,NULL,NULL);
2368 /* sim_part will now be 1 if no checkpoint file was found */
2370 gmx_fatal(FARGS, "Checkpoint file %s not found!", opt2fn("-cpi",
2375 /* Open performance output file and write header info */
2376 fp = ffopen(opt2fn("-p",NFILE,fnm),"w");
2378 /* Make a quick consistency check of command line parameters */
2379 check_input(nnodes, repeats, &ntprs, &upfac, &downfac,
2380 maxPMEfraction, minPMEfraction, npme_fixed,
2381 fs, bench_nsteps, fnm, NFILE, sim_part, presteps,
2384 /* Determine the maximum and minimum number of PME nodes to test,
2385 * the actual list of settings is build in do_the_tests(). */
2386 if ((nnodes > 2) && (npme_fixed < -1))
2388 maxPMEnodes = floor(maxPMEfraction*nnodes);
2389 minPMEnodes = max(floor(minPMEfraction*nnodes), 0);
2390 fprintf(stdout, "Will try runs with %d ", minPMEnodes);
2391 if (maxPMEnodes != minPMEnodes)
2392 fprintf(stdout, "- %d ", maxPMEnodes);
2393 fprintf(stdout, "PME-only nodes.\n Note that the automatic number of PME-only nodes and no separate PME nodes are always tested.\n");
2401 /* Get the commands we need to set up the runs from environment variables */
2402 get_program_paths(bThreads, &cmd_mpirun, cmd_np, &cmd_mdrun, repeats);
2404 /* Print some header info to file */
2406 fprintf(fp, "\n P E R F O R M A N C E R E S U L T S\n");
2408 fprintf(fp, "%s for Gromacs %s\n", ShortProgram(),GromacsVersion());
2411 fprintf(fp, "Number of nodes : %d\n", nnodes);
2412 fprintf(fp, "The mpirun command is : %s\n", cmd_mpirun);
2413 if ( strcmp(procstring[0], "none") != 0)
2414 fprintf(fp, "Passing # of nodes via : %s\n", procstring[0]);
2416 fprintf(fp, "Not setting number of nodes in system call\n");
2419 fprintf(fp, "Number of threads : %d\n", nnodes);
2421 fprintf(fp, "The mdrun command is : %s\n", cmd_mdrun);
2422 fprintf(fp, "mdrun args benchmarks : %s\n", cmd_args_bench);
2423 fprintf(fp, "Benchmark steps : ");
2424 fprintf(fp, gmx_large_int_pfmt, bench_nsteps);
2426 fprintf(fp, "dlb equilibration steps : %d\n", presteps);
2429 fprintf(fp, "Checkpoint time step : ");
2430 fprintf(fp, gmx_large_int_pfmt, cpt_steps);
2434 fprintf(fp, "mdrun args at launchtime: %s\n", cmd_args_launch);
2435 if (!bPassAll && ExtraArgs[0] != '\0')
2436 fprintf(fp, "Unused arguments : %s\n", ExtraArgs);
2437 if (new_sim_nsteps >= 0)
2440 fprintf(stderr, "Note: Simulation input file %s will have ", opt2fn("-so",NFILE,fnm));
2441 fprintf(stderr, gmx_large_int_pfmt, new_sim_nsteps+cpt_steps);
2442 fprintf(stderr, " steps.\n");
2443 fprintf(fp, "Simulation steps : ");
2444 fprintf(fp, gmx_large_int_pfmt, new_sim_nsteps);
2448 fprintf(fp, "Repeats for each test : %d\n", repeats);
2452 fprintf(fp, "Requested grid spacing : %f (This will be the grid spacing at a scaling factor of 1.0)\n", fs);
2455 if (npme_fixed >= -1)
2457 fprintf(fp, "Fixing -npme at : %d\n", npme_fixed);
2460 fprintf(fp, "Input file : %s\n", opt2fn("-s",NFILE,fnm));
2462 /* Allocate memory for the inputinfo struct: */
2464 info->nr_inputfiles = ntprs;
2465 for (i=0; i<ntprs; i++)
2467 snew(info->rcoulomb , ntprs);
2468 snew(info->rvdw , ntprs);
2469 snew(info->rlist , ntprs);
2470 snew(info->rlistlong, ntprs);
2471 snew(info->nkx , ntprs);
2472 snew(info->nky , ntprs);
2473 snew(info->nkz , ntprs);
2474 snew(info->fsx , ntprs);
2475 snew(info->fsy , ntprs);
2476 snew(info->fsz , ntprs);
2478 /* Make alternative tpr files to test: */
2479 snew(tpr_names, ntprs);
2480 for (i=0; i<ntprs; i++)
2481 snew(tpr_names[i], STRLEN);
2483 /* It can be that ntprs is reduced by make_benchmark_tprs if not enough
2484 * different grids could be found. */
2485 make_benchmark_tprs(opt2fn("-s",NFILE,fnm), tpr_names, bench_nsteps+presteps,
2486 cpt_steps, upfac, downfac, &ntprs, fs, info, fp);
2489 /********************************************************************************/
2490 /* Main loop over all scenarios we need to test: tpr files, PME nodes, repeats */
2491 /********************************************************************************/
2492 snew(perfdata, ntprs);
2493 do_the_tests(fp, tpr_names, maxPMEnodes, minPMEnodes, npme_fixed, npmevalues_opt[0], perfdata, &pmeentries,
2494 repeats, nnodes, ntprs, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2495 cmd_args_bench, fnm, NFILE, sim_part, presteps, cpt_steps);
2497 fprintf(fp, "\nTuning took%8.1f minutes.\n", (gettime()-seconds)/60.0);
2499 /* Analyse the results and give a suggestion for optimal settings: */
2500 bKeepTPR = analyze_data(fp, opt2fn("-p", NFILE, fnm), perfdata, nnodes, ntprs, pmeentries,
2501 repeats, info, &best_tpr, &best_npme);
2503 /* Take the best-performing tpr file and enlarge nsteps to original value */
2504 if ( bKeepTPR && !bOverwrite && !(fs > 0.0) )
2506 simulation_tpr = opt2fn("-s",NFILE,fnm);
2510 simulation_tpr = opt2fn("-so",NFILE,fnm);
2511 modify_PMEsettings(bOverwrite? (new_sim_nsteps+cpt_steps) : info->orig_sim_steps,
2512 info->orig_init_step, tpr_names[best_tpr], simulation_tpr);
2515 /* Now start the real simulation if the user requested it ... */
2516 launch_simulation(bLaunch, fp, bThreads, cmd_mpirun, cmd_np, cmd_mdrun,
2517 cmd_args_launch, simulation_tpr, nnodes, best_npme);
2520 /* ... or simply print the performance results to screen: */
2522 finalize(opt2fn("-p", NFILE, fnm));