src/gmxlib/nrnb.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team,
   6  * check out http://www.gromacs.org for more information.
   7  * Copyright (c) 2012, by the GROMACS development team, led by
   8  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   9  * others, as listed in the AUTHORS file in the top-level source
  10  * directory and at http://www.gromacs.org.
  11  *
  12  * GROMACS is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public License
  14  * as published by the Free Software Foundation; either version 2.1
  15  * of the License, or (at your option) any later version.
  16  *
  17  * GROMACS is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with GROMACS; if not, see
  24  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  25  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  26  *
  27  * If you want to redistribute modifications to GROMACS, please
  28  * consider that scientific software is very special. Version
  29  * control is crucial - bugs must be traceable. We will be happy to
  30  * consider code for inclusion in the official distribution, but
  31  * derived work must not be called official GROMACS. Details are found
  32  * in the README & COPYING files - if they are missing, get the
  33  * official version at http://www.gromacs.org.
  34  *
  35  * To help us fund GROMACS development, we humbly ask that you cite
  36  * the research papers on the package. Check out http://www.gromacs.org.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <string.h>
  43 #include "types/commrec.h"
  44 #include "sysstuff.h"
  45 #include "gmx_fatal.h"
  46 #include "names.h"
  47 #include "macros.h"
  48 #include "nrnb.h"
  49 #include "main.h"
  50 #include "smalloc.h"
  51 #include "copyrite.h"
  52
  53
  54
  55
  56
  57 typedef struct {
  58   const char *name;
  59   int  flop;
  60 } t_nrnb_data;
  61
  62
  63 static const t_nrnb_data nbdata[eNRNB] = {
  64     /* These are re-used for different NB kernels, since there are so many.
  65      * The actual number of flops is set dynamically.
  66      */
  67     { "NB VdW [V&F]",                    1 },
  68     { "NB VdW [F]",                      1 },
  69     { "NB Elec. [V&F]",                  1 },
  70     { "NB Elec. [F]",                    1 },
  71     { "NB Elec. [W3,V&F]",               1 },
  72     { "NB Elec. [W3,F]",                 1 },
  73     { "NB Elec. [W3-W3,V&F]",            1 },
  74     { "NB Elec. [W3-W3,F]",              1 },
  75     { "NB Elec. [W4,V&F]",               1 },
  76     { "NB Elec. [W4,F]",                 1 },
  77     { "NB Elec. [W4-W4,V&F]",            1 },
  78     { "NB Elec. [W4-W4,F]",              1 },
  79     { "NB VdW & Elec. [V&F]",            1 },
  80     { "NB VdW & Elec. [F]",              1 },
  81     { "NB VdW & Elec. [W3,V&F]",         1 },
  82     { "NB VdW & Elec. [W3,F]",           1 },
  83     { "NB VdW & Elec. [W3-W3,V&F]",      1 },
  84     { "NB VdW & Elec. [W3-W3,F]",        1 },
  85     { "NB VdW & Elec. [W4,V&F]",         1 },
  86     { "NB VdW & Elec. [W4,F]",           1 },
  87     { "NB VdW & Elec. [W4-W4,V&F]",      1 },
  88     { "NB VdW & Elec. [W4-W4,F]",        1 },
  89
  90     { "NB Generic kernel",               1 },
  91     { "NB Free energy kernel",           1 },
  92     { "NB All-vs-all",                   1 },
  93     { "NB All-vs-all, GB",               1 },
  94
  95     { "Pair Search distance check",      9 }, /* nbnxn pair dist. check */
  96     /* nbnxn kernel flops are based on inner-loops without exclusion checks.
  97      * Plain Coulomb runs through the RF kernels, except with CUDA.
  98      * invsqrt is counted as 6 flops: 1 for _mm_rsqt_ps + 5 for iteration.
  99      * The flops are equal for plain-C, x86 SIMD and CUDA, except for:
 100      * - plain-C kernel uses one flop more for Coulomb-only (F) than listed
 101      * - x86 SIMD LJ geom-comb.rule kernels (fastest) use 2 more flops
 102      * - x86 SIMD LJ LB-comb.rule kernels (fast) use 3 (8 for F+E) more flops
 103      * - GPU always does exclusions, which requires 2-4 flops, but as invsqrt
 104      *   is always counted as 6 flops, this roughly compensates.
 105      */
 106     { "NxN RF Elec. + VdW [F]",         38 }, /* nbnxn kernel LJ+RF, no ener */
 107     { "NxN RF Elec. + VdW [V&F]",       54 },
 108     { "NxN QSTab Elec. + VdW [F]",      41 }, /* nbnxn kernel LJ+tab, no en */
 109     { "NxN QSTab Elec. + VdW [V&F]",    59 },
 110     { "NxN Ewald Elec. + VdW [F]",      66 }, /* nbnxn kernel LJ+Ewald, no en */
 111     { "NxN Ewald Elec. + VdW [V&F]",   107 },
 112     { "NxN VdW [F]",                    33 }, /* nbnxn kernel LJ, no ener */
 113     { "NxN VdW [V&F]",                  43 },
 114     { "NxN RF Electrostatics [F]",      31 }, /* nbnxn kernel RF, no ener */
 115     { "NxN RF Electrostatics [V&F]",    36 },
 116     { "NxN QSTab Elec. [F]",            34 }, /* nbnxn kernel tab, no ener */
 117     { "NxN QSTab Elec. [V&F]",          41 },
 118     { "NxN Ewald Elec. [F]",            61 }, /* nbnxn kernel Ewald, no ener */
 119     { "NxN Ewald Elec. [V&F]",          84 },
 120     { "1,4 nonbonded interactions",     90 },
 121     { "Born radii (Still)",             47 },
 122     { "Born radii (HCT/OBC)",          183 },
 123     { "Born force chain rule",          15 },
 124     { "All-vs-All Still radii",          1 },
 125     { "All-vs-All HCT/OBC radii",        1 },
 126     { "All-vs-All Born chain rule",      1 },
 127     { "Calc Weights",                   36 },
 128     { "Spread Q",                        6 },
 129     { "Spread Q Bspline",                2 },
 130     { "Gather F",                      23  },
 131     { "Gather F Bspline",              6   },
 132     { "3D-FFT",                        8   },
 133     { "Convolution",                   4   },
 134     { "Solve PME",                     64  },
 135     { "NS-Pairs",                      21  },
 136     { "Reset In Box",                  3   },
 137     { "Shift-X",                       6   },
 138     { "CG-CoM",                        3   },
 139     { "Sum Forces",                    1   },
 140     { "Bonds",                         59  },
 141     { "G96Bonds",                      44  },
 142     { "FENE Bonds",                    58  },
 143     { "Tab. Bonds",                    62  },
 144     { "Restraint Potential",           86  },
 145     { "Linear Angles",                 57  },
 146     { "Angles",                        168 },
 147     { "G96Angles",                     150 },
 148     { "Quartic Angles",                160 },
 149     { "Tab. Angles",                   169 },
 150     { "Propers",                       229 },
 151     { "Impropers",                     208 },
 152     { "RB-Dihedrals",                  247 },
 153     { "Four. Dihedrals",               247 },
 154     { "Tab. Dihedrals",                227 },
 155     { "Dist. Restr.",                  200 },
 156     { "Orient. Restr.",                200 },
 157     { "Dihedral Restr.",               200 },
 158     { "Pos. Restr.",                   50  },
 159     { "Angle Restr.",                  191 },
 160     { "Angle Restr. Z",                164 },
 161     { "Morse Potent.",                 83  },
 162     { "Cubic Bonds",                   54  },
 163     { "Walls",                         31  },
 164     { "Polarization",                  59  },
 165     { "Anharmonic Polarization",       72  },
 166     { "Water Pol.",                    62  },
 167     { "Thole Pol.",                    296 },
 168     { "Virial",                        18  },
 169     { "Update",                        31  },
 170     { "Ext.ens. Update",               54  },
 171     { "Stop-CM",                       10  },
 172     { "P-Coupling",                    6   },
 173     { "Calc-Ekin",                     27  },
 174     { "Lincs",                         60  },
 175     { "Lincs-Mat",                     4   },
 176     { "Shake",                         30  },
 177     { "Constraint-V",                   8  },
 178     { "Shake-Init",                    10  },
 179     { "Constraint-Vir",                24  },
 180     { "Settle",                        323 },
 181     { "Virtual Site 2",                23  },
 182     { "Virtual Site 3",                37  },
 183     { "Virtual Site 3fd",              95  },
 184     { "Virtual Site 3fad",             176 },
 185     { "Virtual Site 3out",             87  },
 186     { "Virtual Site 4fd",              110 },
 187     { "Virtual Site 4fdn",             254 },
 188     { "Virtual Site N",                 15 },
 189     { "Mixed Generalized Born stuff",   10 }
 190 };
 191
 192
 193 void init_nrnb(t_nrnb *nrnb)
 194 {
 195   int i;
 196
 197   for(i=0; (i<eNRNB); i++)
 198     nrnb->n[i]=0.0;
 199 }
 200
 201 void cp_nrnb(t_nrnb *dest, t_nrnb *src)
 202 {
 203   int i;
 204
 205   for(i=0; (i<eNRNB); i++)
 206     dest->n[i]=src->n[i];
 207 }
 208
 209 void add_nrnb(t_nrnb *dest, t_nrnb *s1, t_nrnb *s2)
 210 {
 211   int i;
 212
 213   for(i=0; (i<eNRNB); i++)
 214     dest->n[i]=s1->n[i]+s2->n[i];
 215 }
 216
 217 void print_nrnb(FILE *out, t_nrnb *nrnb)
 218 {
 219   int i;
 220
 221   for(i=0; (i<eNRNB); i++)
 222     if (nrnb->n[i] > 0)
 223       fprintf(out," %-26s %10.0f.\n",nbdata[i].name,nrnb->n[i]);
 224 }
 225
 226 void _inc_nrnb(t_nrnb *nrnb,int enr,int inc,char *file,int line)
 227 {
 228   nrnb->n[enr]+=inc;
 229 #ifdef DEBUG_NRNB
 230   printf("nrnb %15s(%2d) incremented with %8d from file %s line %d\n",
 231           nbdata[enr].name,enr,inc,file,line);
 232 #endif
 233 }
 234
 235 void print_flop(FILE *out,t_nrnb *nrnb,double *nbfs,double *mflop)
 236 {
 237   int    i;
 238   double mni,frac,tfrac,tflop;
 239   const char   *myline = "-----------------------------------------------------------------------------";
 240
 241   *nbfs = 0.0;
 242   for(i=0; (i<eNR_NBKERNEL_ALLVSALLGB); i++) {
 243     if (strstr(nbdata[i].name,"W3-W3") != NULL)
 244       *nbfs += 9e-6*nrnb->n[i];
 245     else if (strstr(nbdata[i].name,"W3") != NULL)
 246       *nbfs += 3e-6*nrnb->n[i];
 247     else if (strstr(nbdata[i].name,"W4-W4") != NULL)
 248       *nbfs += 10e-6*nrnb->n[i];
 249     else if (strstr(nbdata[i].name,"W4") != NULL)
 250       *nbfs += 4e-6*nrnb->n[i];
 251     else
 252       *nbfs += 1e-6*nrnb->n[i];
 253   }
 254   tflop=0;
 255   for(i=0; (i<eNRNB); i++)
 256     tflop+=1e-6*nrnb->n[i]*nbdata[i].flop;
 257
 258   if (tflop == 0) {
 259     fprintf(out,"No MEGA Flopsen this time\n");
 260     return;
 261   }
 262   if (out) {
 263     fprintf(out,"\n\tM E G A - F L O P S   A C C O U N T I N G\n\n");
 264   }
 265
 266   if (out)
 267   {
 268       fprintf(out," NB=Group-cutoff nonbonded kernels    NxN=N-by-N cluster Verlet kernels\n");
 269       fprintf(out," RF=Reaction-Field  VdW=Van der Waals  QSTab=quadratic-spline table\n");
 270       fprintf(out," W3=SPC/TIP3p  W4=TIP4p (single or pairs)\n");
 271       fprintf(out," V&F=Potential and force  V=Potential only  F=Force only\n\n");
 272
 273       fprintf(out," %-32s %16s %15s  %7s\n",
 274               "Computing:","M-Number","M-Flops","% Flops");
 275       fprintf(out,"%s\n",myline);
 276   }
 277   *mflop=0.0;
 278   tfrac=0.0;
 279   for(i=0; (i<eNRNB); i++) {
 280     mni     = 1e-6*nrnb->n[i];
 281     *mflop += mni*nbdata[i].flop;
 282     frac    = 100.0*mni*nbdata[i].flop/tflop;
 283     tfrac  += frac;
 284     if (out && mni != 0)
 285       fprintf(out," %-32s %16.6f %15.3f  %6.1f\n",
 286               nbdata[i].name,mni,mni*nbdata[i].flop,frac);
 287   }
 288   if (out) {
 289     fprintf(out,"%s\n",myline);
 290     fprintf(out," %-32s %16s %15.3f  %6.1f\n",
 291             "Total","",*mflop,tfrac);
 292     fprintf(out,"%s\n\n",myline);
 293   }
 294 }
 295
 296 void print_perf(FILE *out,double nodetime,double realtime,int nprocs,
 297                 gmx_large_int_t nsteps,real delta_t,
 298                 double nbfs,double mflop,
 299                 int omp_nth_pp)
 300 {
 301   real runtime;
 302
 303   fprintf(out,"\n");
 304
 305   if (realtime > 0)
 306   {
 307     fprintf(out,"%12s %12s %12s %10s\n","","Core t (s)","Wall t (s)","(%)");
 308     fprintf(out,"%12s %12.3f %12.3f %10.1f\n","Time:",
 309             nodetime, realtime, 100.0*nodetime/realtime);
 310     /* only print day-hour-sec format if realtime is more than 30 min */
 311     if (realtime > 30*60)
 312     {
 313       fprintf(out,"%12s %12s","","");
 314       pr_difftime(out,realtime);
 315     }
 316     if (delta_t > 0)
 317     {
 318       mflop = mflop/realtime;
 319       runtime = nsteps*delta_t;
 320
 321       if (getenv("GMX_DETAILED_PERF_STATS") == NULL)
 322       {
 323           fprintf(out,"%12s %12s %12s\n",
 324                   "","(ns/day)","(hour/ns)");
 325           fprintf(out,"%12s %12.3f %12.3f\n","Performance:",
 326                   runtime*24*3.6/realtime,1000*realtime/(3600*runtime));
 327       }
 328       else
 329       {
 330         fprintf(out,"%12s %12s %12s %12s %12s\n",
 331                 "","(Mnbf/s)",(mflop > 1000) ? "(GFlops)" : "(MFlops)",
 332                 "(ns/day)","(hour/ns)");
 333         fprintf(out,"%12s %12.3f %12.3f %12.3f %12.3f\n","Performance:",
 334                 nbfs/realtime,(mflop > 1000) ? (mflop/1000) : mflop,
 335                 runtime*24*3.6/realtime,1000*realtime/(3600*runtime));
 336       }
 337     }
 338     else
 339     {
 340       if (getenv("GMX_DETAILED_PERF_STATS") == NULL)
 341       {
 342           fprintf(out,"%12s %14s\n",
 343                   "","(steps/hour)");
 344           fprintf(out,"%12s %14.1f\n","Performance:",
 345                   nsteps*3600.0/realtime);
 346       }
 347       else
 348       {
 349           fprintf(out,"%12s %12s %12s %14s\n",
 350                   "","(Mnbf/s)",(mflop > 1000) ? "(GFlops)" : "(MFlops)",
 351                   "(steps/hour)");
 352           fprintf(out,"%12s %12.3f %12.3f %14.1f\n","Performance:",
 353               nbfs/realtime,(mflop > 1000) ? (mflop/1000) : mflop,
 354               nsteps*3600.0/realtime);
 355       }
 356     }
 357   }
 358 }
 359
 360 int cost_nrnb(int enr)
 361 {
 362   return nbdata[enr].flop;
 363 }
 364
 365 const char *nrnb_str(int enr)
 366 {
 367   return nbdata[enr].name;
 368 }
 369
 370 static const int    force_index[]={
 371   eNR_BONDS,  eNR_ANGLES,  eNR_PROPER, eNR_IMPROPER,
 372   eNR_RB,     eNR_DISRES,  eNR_ORIRES, eNR_POSRES,
 373   eNR_NS,
 374 };
 375 #define NFORCE_INDEX asize(force_index)
 376
 377 static const int    constr_index[]={
 378   eNR_SHAKE,     eNR_SHAKE_RIJ, eNR_SETTLE,       eNR_UPDATE,       eNR_PCOUPL,
 379   eNR_CONSTR_VIR,eNR_CONSTR_V
 380 };
 381 #define NCONSTR_INDEX asize(constr_index)
 382
 383 static double pr_av(FILE *log,t_commrec *cr,
 384                     double fav,double ftot[],const char *title)
 385 {
 386   int    i,perc;
 387   double dperc,unb;
 388
 389   unb=0;
 390   if (fav > 0) {
 391     fav /= cr->nnodes - cr->npmenodes;
 392     fprintf(log,"\n %-26s",title);
 393     for(i=0; (i<cr->nnodes); i++) {
 394         dperc=(100.0*ftot[i])/fav;
 395         unb=max(unb,dperc);
 396         perc=dperc;
 397         fprintf(log,"%3d ",perc);
 398     }
 399     if (unb > 0) {
 400       perc=10000.0/unb;
 401       fprintf(log,"%6d%%\n\n",perc);
 402     }
 403     else
 404       fprintf(log,"\n\n");
 405   }
 406   return unb;
 407 }
 408
 409 void pr_load(FILE *log,t_commrec *cr,t_nrnb nrnb[])
 410 {
 411   int    i,j,perc;
 412   double dperc,unb,uf,us;
 413   double *ftot,fav;
 414   double *stot,sav;
 415   t_nrnb *av;
 416
 417   snew(av,1);
 418   snew(ftot,cr->nnodes);
 419   snew(stot,cr->nnodes);
 420   init_nrnb(av);
 421   for(i=0; (i<cr->nnodes); i++) {
 422       add_nrnb(av,av,&(nrnb[i]));
 423       /* Cost due to forces */
 424       for(j=0; (j<eNR_NBKERNEL_ALLVSALLGB); j++)
 425         ftot[i]+=nrnb[i].n[j]*cost_nrnb(j);
 426       for(j=0; (j<NFORCE_INDEX); j++)
 427         ftot[i]+=nrnb[i].n[force_index[j]]*cost_nrnb(force_index[j]);
 428       /* Due to shake */
 429       for(j=0; (j<NCONSTR_INDEX); j++) {
 430         stot[i]+=nrnb[i].n[constr_index[j]]*cost_nrnb(constr_index[j]);
 431       }
 432   }
 433   for(j=0; (j<eNRNB); j++)
 434     av->n[j]=av->n[j]/(double)(cr->nnodes - cr->npmenodes);
 435
 436     fprintf(log,"\nDetailed load balancing info in percentage of average\n");
 437
 438   fprintf(log," Type                 NODE:");
 439   for(i=0; (i<cr->nnodes); i++)
 440       fprintf(log,"%3d ",i);
 441   fprintf(log,"Scaling\n");
 442   fprintf(log,"---------------------------");
 443   for(i=0; (i<cr->nnodes); i++)
 444       fprintf(log,"----");
 445   fprintf(log,"-------\n");
 446
 447   for(j=0; (j<eNRNB); j++) {
 448     unb=100.0;
 449     if (av->n[j] > 0) {
 450       fprintf(log," %-26s",nrnb_str(j));
 451       for(i=0; (i<cr->nnodes); i++) {
 452           dperc=(100.0*nrnb[i].n[j])/av->n[j];
 453           unb=max(unb,dperc);
 454           perc=dperc;
 455           fprintf(log,"%3d ",perc);
 456       }
 457       if (unb > 0) {
 458         perc=10000.0/unb;
 459         fprintf(log,"%6d%%\n",perc);
 460       }
 461       else
 462         fprintf(log,"\n");
 463     }
 464   }
 465   fav=sav=0;
 466   for(i=0; (i<cr->nnodes); i++) {
 467     fav+=ftot[i];
 468     sav+=stot[i];
 469   }
 470   uf=pr_av(log,cr,fav,ftot,"Total Force");
 471   us=pr_av(log,cr,sav,stot,"Total Constr.");
 472
 473   unb=(uf*fav+us*sav)/(fav+sav);
 474   if (unb > 0) {
 475     unb=10000.0/unb;
 476     fprintf(log,"\nTotal Scaling: %.0f%% of max performance\n\n",unb);
 477   }
 478 }
 479