int main(int argc,char** argv){
plumed p;
int natoms=3;
if(plumed_installed()){
plumed_gcreate();
plumed_gcmd("setMDEngine","ACCode");
plumed_gcmd("setNatoms",&natoms);
plumed_gcmd("init",NULL);
plumed_gcmd("read",NULL);
plumed_gfinalize();
p=plumed_create();
plumed_cmd(p,"setNatoms",&natoms);
plumed_cmd(p,"init",NULL);
plumed_cmd(p,"read",NULL);
plumed_finalize(p);
}
return 0;
}
int cmain(int argc, char *argv[])
{
const char *desc[] = {
"The [TT]mdrun[tt] program is the main computational chemistry engine",
"within GROMACS. Obviously, it performs Molecular Dynamics simulations,",
"but it can also perform Stochastic Dynamics, Energy Minimization,",
"test particle insertion or (re)calculation of energies.",
"Normal mode analysis is another option. In this case [TT]mdrun[tt]",
"builds a Hessian matrix from single conformation.",
"For usual Normal Modes-like calculations, make sure that",
"the structure provided is properly energy-minimized.",
"The generated matrix can be diagonalized by [TT]g_nmeig[tt].[PAR]",
"The [TT]mdrun[tt] program reads the run input file ([TT]-s[tt])",
"and distributes the topology over nodes if needed.",
"[TT]mdrun[tt] produces at least four output files.",
"A single log file ([TT]-g[tt]) is written, unless the option",
"[TT]-seppot[tt] is used, in which case each node writes a log file.",
"The trajectory file ([TT]-o[tt]), contains coordinates, velocities and",
"optionally forces.",
"The structure file ([TT]-c[tt]) contains the coordinates and",
"velocities of the last step.",
"The energy file ([TT]-e[tt]) contains energies, the temperature,",
"pressure, etc, a lot of these things are also printed in the log file.",
"Optionally coordinates can be written to a compressed trajectory file",
"([TT]-x[tt]).[PAR]",
"The option [TT]-dhdl[tt] is only used when free energy calculation is",
"turned on.[PAR]",
"A simulation can be run in parallel using two different parallelization",
"schemes: MPI parallelization and/or OpenMP thread parallelization.",
"The MPI parallelization uses multiple processes when [TT]mdrun[tt] is",
"compiled with a normal MPI library or threads when [TT]mdrun[tt] is",
"compiled with the GROMACS built-in thread-MPI library. OpenMP threads",
"are supported when mdrun is compiled with OpenMP. Full OpenMP support",
"is only available with the Verlet cut-off scheme, with the (older)",
"group scheme only PME-only processes can use OpenMP parallelization.",
"In all cases [TT]mdrun[tt] will by default try to use all the available",
"hardware resources. With a normal MPI library only the options",
"[TT]-ntomp[tt] (with the Verlet cut-off scheme) and [TT]-ntomp_pme[tt],",
"for PME-only processes, can be used to control the number of threads.",
"With thread-MPI there are additional options [TT]-nt[tt], which sets",
"the total number of threads, and [TT]-ntmpi[tt], which sets the number",
"of thread-MPI threads.",
"Note that using combined MPI+OpenMP parallelization is almost always",
"slower than single parallelization, except at the scaling limit, where",
"especially OpenMP parallelization of PME reduces the communication cost.",
"OpenMP-only parallelization is much faster than MPI-only parallelization",
"on a single CPU(-die). Since we currently don't have proper hardware",
"topology detection, [TT]mdrun[tt] compiled with thread-MPI will only",
"automatically use OpenMP-only parallelization when you use up to 4",
"threads, up to 12 threads with Intel Nehalem/Westmere, or up to 16",
"threads with Intel Sandy Bridge or newer CPUs. Otherwise MPI-only",
"parallelization is used (except with GPUs, see below).",
"[PAR]",
"To quickly test the performance of the new Verlet cut-off scheme",
"with old [TT].tpr[tt] files, either on CPUs or CPUs+GPUs, you can use",
"the [TT]-testverlet[tt] option. This should not be used for production,",
"since it can slightly modify potentials and it will remove charge groups",
"making analysis difficult, as the [TT].tpr[tt] file will still contain",
"charge groups. For production simulations it is highly recommended",
"to specify [TT]cutoff-scheme = Verlet[tt] in the [TT].mdp[tt] file.",
"[PAR]",
"With GPUs (only supported with the Verlet cut-off scheme), the number",
"of GPUs should match the number of MPI processes or MPI threads,",
"excluding PME-only processes/threads. With thread-MPI the number",
"of MPI threads will automatically be set to the number of GPUs detected.",
"When you want to use a subset of the available GPUs, you can use",
"the [TT]-gpu_id[tt] option, where GPU id's are passed as a string,",
"e.g. 02 for using GPUs 0 and 2. When you want different GPU id's",
"on different nodes of a compute cluster, use the GMX_GPU_ID environment",
"variable instead. The format for GMX_GPU_ID is identical to ",
"[TT]-gpu_id[tt], but an environment variable can have different values",
"on different nodes of a cluster.",
"[PAR]",
"When using PME with separate PME nodes or with a GPU, the two major",
"compute tasks, the non-bonded force calculation and the PME calculation",
"run on different compute resources. If this load is not balanced,",
"some of the resources will be idle part of time. With the Verlet",
"cut-off scheme this load is automatically balanced when the PME load",
"is too high (but not when it is too low). This is done by scaling",
"the Coulomb cut-off and PME grid spacing by the same amount. In the first",
"few hundred steps different settings are tried and the fastest is chosen",
"for the rest of the simulation. This does not affect the accuracy of",
"the results, but it does affect the decomposition of the Coulomb energy",
"into particle and mesh contributions. The auto-tuning can be turned off",
"with the option [TT]-notunepme[tt].",
"[PAR]",
"[TT]mdrun[tt] pins (sets affinity of) threads to specific cores,",
"when all (logical) cores on a compute node are used by [TT]mdrun[tt],",
"even when no multi-threading is used,",
"as this usually results in significantly better performance.",
"If the queuing systems or the OpenMP library pinned threads, we honor",
"this and don't pin again, even though the layout may be sub-optimal.",
"If you want to have [TT]mdrun[tt] override an already set thread affinity",
"or pin threads when using less cores, use [TT]-pin on[tt].",
"With SMT (simultaneous multithreading), e.g. Intel Hyper-Threading,",
"there are multiple logical cores per physical core.",
"The option [TT]-pinstride[tt] sets the stride in logical cores for",
"pinning consecutive threads. Without SMT, 1 is usually the best choice.",
"With Intel Hyper-Threading 2 is best when using half or less of the",
"logical cores, 1 otherwise. The default value of 0 do exactly that:",
"it minimizes the threads per logical core, to optimize performance.",
"If you want to run multiple mdrun jobs on the same physical node,"
"you should set [TT]-pinstride[tt] to 1 when using all logical cores.",
"When running multiple mdrun (or other) simulations on the same physical",
"node, some simulations need to start pinning from a non-zero core",
"to avoid overloading cores; with [TT]-pinoffset[tt] you can specify",
"the offset in logical cores for pinning.",
"[PAR]",
"When [TT]mdrun[tt] is started using MPI with more than 1 process",
"or with thread-MPI with more than 1 thread, MPI parallelization is used.",
"By default domain decomposition is used, unless the [TT]-pd[tt]",
"option is set, which selects particle decomposition.",
"[PAR]",
"With domain decomposition, the spatial decomposition can be set",
"with option [TT]-dd[tt]. By default [TT]mdrun[tt] selects a good decomposition.",
"The user only needs to change this when the system is very inhomogeneous.",
"Dynamic load balancing is set with the option [TT]-dlb[tt],",
"which can give a significant performance improvement,",
"especially for inhomogeneous systems. The only disadvantage of",
"dynamic load balancing is that runs are no longer binary reproducible,",
"but in most cases this is not important.",
"By default the dynamic load balancing is automatically turned on",
"when the measured performance loss due to load imbalance is 5% or more.",
"At low parallelization these are the only important options",
"for domain decomposition.",
"At high parallelization the options in the next two sections",
"could be important for increasing the performace.",
"[PAR]",
"When PME is used with domain decomposition, separate nodes can",
"be assigned to do only the PME mesh calculation;",
"this is computationally more efficient starting at about 12 nodes.",
"The number of PME nodes is set with option [TT]-npme[tt],",
"this can not be more than half of the nodes.",
"By default [TT]mdrun[tt] makes a guess for the number of PME",
"nodes when the number of nodes is larger than 11 or performance wise",
"not compatible with the PME grid x dimension.",
"But the user should optimize npme. Performance statistics on this issue",
"are written at the end of the log file.",
"For good load balancing at high parallelization, the PME grid x and y",
"dimensions should be divisible by the number of PME nodes",
"(the simulation will run correctly also when this is not the case).",
"[PAR]",
"This section lists all options that affect the domain decomposition.",
"[PAR]",
"Option [TT]-rdd[tt] can be used to set the required maximum distance",
"for inter charge-group bonded interactions.",
"Communication for two-body bonded interactions below the non-bonded",
"cut-off distance always comes for free with the non-bonded communication.",
"Atoms beyond the non-bonded cut-off are only communicated when they have",
"missing bonded interactions; this means that the extra cost is minor",
"and nearly indepedent of the value of [TT]-rdd[tt].",
"With dynamic load balancing option [TT]-rdd[tt] also sets",
"the lower limit for the domain decomposition cell sizes.",
"By default [TT]-rdd[tt] is determined by [TT]mdrun[tt] based on",
"the initial coordinates. The chosen value will be a balance",
"between interaction range and communication cost.",
"[PAR]",
"When inter charge-group bonded interactions are beyond",
"the bonded cut-off distance, [TT]mdrun[tt] terminates with an error message.",
"For pair interactions and tabulated bonds",
"that do not generate exclusions, this check can be turned off",
"with the option [TT]-noddcheck[tt].",
"[PAR]",
"When constraints are present, option [TT]-rcon[tt] influences",
"the cell size limit as well.",
"Atoms connected by NC constraints, where NC is the LINCS order plus 1,",
"should not be beyond the smallest cell size. A error message is",
"generated when this happens and the user should change the decomposition",
"or decrease the LINCS order and increase the number of LINCS iterations.",
"By default [TT]mdrun[tt] estimates the minimum cell size required for P-LINCS",
"in a conservative fashion. For high parallelization it can be useful",
"to set the distance required for P-LINCS with the option [TT]-rcon[tt].",
"[PAR]",
"The [TT]-dds[tt] option sets the minimum allowed x, y and/or z scaling",
"of the cells with dynamic load balancing. [TT]mdrun[tt] will ensure that",
"the cells can scale down by at least this factor. This option is used",
"for the automated spatial decomposition (when not using [TT]-dd[tt])",
"as well as for determining the number of grid pulses, which in turn",
"sets the minimum allowed cell size. Under certain circumstances",
"the value of [TT]-dds[tt] might need to be adjusted to account for",
"high or low spatial inhomogeneity of the system.",
"[PAR]",
"The option [TT]-gcom[tt] can be used to only do global communication",
"every n steps.",
"This can improve performance for highly parallel simulations",
"where this global communication step becomes the bottleneck.",
"For a global thermostat and/or barostat the temperature",
"and/or pressure will also only be updated every [TT]-gcom[tt] steps.",
"By default it is set to the minimum of nstcalcenergy and nstlist.[PAR]",
"With [TT]-rerun[tt] an input trajectory can be given for which ",
"forces and energies will be (re)calculated. Neighbor searching will be",
"performed for every frame, unless [TT]nstlist[tt] is zero",
"(see the [TT].mdp[tt] file).[PAR]",
"ED (essential dynamics) sampling and/or additional flooding potentials",
"are switched on by using the [TT]-ei[tt] flag followed by an [TT].edi[tt]",
"file. The [TT].edi[tt] file can be produced with the [TT]make_edi[tt] tool",
"or by using options in the essdyn menu of the WHAT IF program.",
"[TT]mdrun[tt] produces a [TT].xvg[tt] output file that",
"contains projections of positions, velocities and forces onto selected",
"eigenvectors.[PAR]",
"When user-defined potential functions have been selected in the",
"[TT].mdp[tt] file the [TT]-table[tt] option is used to pass [TT]mdrun[tt]",
"a formatted table with potential functions. The file is read from",
"either the current directory or from the [TT]GMXLIB[tt] directory.",
"A number of pre-formatted tables are presented in the [TT]GMXLIB[tt] dir,",
"for 6-8, 6-9, 6-10, 6-11, 6-12 Lennard-Jones potentials with",
"normal Coulomb.",
"When pair interactions are present, a separate table for pair interaction",
"functions is read using the [TT]-tablep[tt] option.[PAR]",
"When tabulated bonded functions are present in the topology,",
"interaction functions are read using the [TT]-tableb[tt] option.",
"For each different tabulated interaction type the table file name is",
"modified in a different way: before the file extension an underscore is",
"appended, then a 'b' for bonds, an 'a' for angles or a 'd' for dihedrals",
"and finally the table number of the interaction type.[PAR]",
"The options [TT]-px[tt] and [TT]-pf[tt] are used for writing pull COM",
"coordinates and forces when pulling is selected",
"in the [TT].mdp[tt] file.[PAR]",
"With [TT]-multi[tt] or [TT]-multidir[tt], multiple systems can be ",
"simulated in parallel.",
"As many input files/directories are required as the number of systems. ",
"The [TT]-multidir[tt] option takes a list of directories (one for each ",
"system) and runs in each of them, using the input/output file names, ",
"such as specified by e.g. the [TT]-s[tt] option, relative to these ",
"directories.",
"With [TT]-multi[tt], the system number is appended to the run input ",
"and each output filename, for instance [TT]topol.tpr[tt] becomes",
"[TT]topol0.tpr[tt], [TT]topol1.tpr[tt] etc.",
"The number of nodes per system is the total number of nodes",
"divided by the number of systems.",
"One use of this option is for NMR refinement: when distance",
"or orientation restraints are present these can be ensemble averaged",
"over all the systems.[PAR]",
"With [TT]-replex[tt] replica exchange is attempted every given number",
"of steps. The number of replicas is set with the [TT]-multi[tt] or ",
"[TT]-multidir[tt] option, described above.",
"All run input files should use a different coupling temperature,",
"the order of the files is not important. The random seed is set with",
"[TT]-reseed[tt]. The velocities are scaled and neighbor searching",
"is performed after every exchange.[PAR]",
"Finally some experimental algorithms can be tested when the",
"appropriate options have been given. Currently under",
"investigation are: polarizability and X-ray bombardments.",
"[PAR]",
"The option [TT]-membed[tt] does what used to be g_membed, i.e. embed",
"a protein into a membrane. The data file should contain the options",
"that where passed to g_membed before. The [TT]-mn[tt] and [TT]-mp[tt]",
"both apply to this as well.",
"[PAR]",
"The option [TT]-pforce[tt] is useful when you suspect a simulation",
"crashes due to too large forces. With this option coordinates and",
"forces of atoms with a force larger than a certain value will",
"be printed to stderr.",
"[PAR]",
"Checkpoints containing the complete state of the system are written",
"at regular intervals (option [TT]-cpt[tt]) to the file [TT]-cpo[tt],",
"unless option [TT]-cpt[tt] is set to -1.",
"The previous checkpoint is backed up to [TT]state_prev.cpt[tt] to",
"make sure that a recent state of the system is always available,",
"even when the simulation is terminated while writing a checkpoint.",
"With [TT]-cpnum[tt] all checkpoint files are kept and appended",
"with the step number.",
"A simulation can be continued by reading the full state from file",
"with option [TT]-cpi[tt]. This option is intelligent in the way that",
"if no checkpoint file is found, Gromacs just assumes a normal run and",
"starts from the first step of the [TT].tpr[tt] file. By default the output",
"will be appending to the existing output files. The checkpoint file",
"contains checksums of all output files, such that you will never",
"loose data when some output files are modified, corrupt or removed.",
"There are three scenarios with [TT]-cpi[tt]:[PAR]",
"[TT]*[tt] no files with matching names are present: new output files are written[PAR]",
"[TT]*[tt] all files are present with names and checksums matching those stored",
"in the checkpoint file: files are appended[PAR]",
"[TT]*[tt] otherwise no files are modified and a fatal error is generated[PAR]",
"With [TT]-noappend[tt] new output files are opened and the simulation",
"part number is added to all output file names.",
"Note that in all cases the checkpoint file itself is not renamed",
"and will be overwritten, unless its name does not match",
"the [TT]-cpo[tt] option.",
"[PAR]",
"With checkpointing the output is appended to previously written",
"output files, unless [TT]-noappend[tt] is used or none of the previous",
"output files are present (except for the checkpoint file).",
"The integrity of the files to be appended is verified using checksums",
"which are stored in the checkpoint file. This ensures that output can",
"not be mixed up or corrupted due to file appending. When only some",
"of the previous output files are present, a fatal error is generated",
"and no old output files are modified and no new output files are opened.",
"The result with appending will be the same as from a single run.",
"The contents will be binary identical, unless you use a different number",
"of nodes or dynamic load balancing or the FFT library uses optimizations",
"through timing.",
"[PAR]",
"With option [TT]-maxh[tt] a simulation is terminated and a checkpoint",
"file is written at the first neighbor search step where the run time",
"exceeds [TT]-maxh[tt]*0.99 hours.",
"[PAR]",
"When [TT]mdrun[tt] receives a TERM signal, it will set nsteps to the current",
"step plus one. When [TT]mdrun[tt] receives an INT signal (e.g. when ctrl+C is",
"pressed), it will stop after the next neighbor search step ",
"(with nstlist=0 at the next step).",
"In both cases all the usual output will be written to file.",
"When running with MPI, a signal to one of the [TT]mdrun[tt] processes",
"is sufficient, this signal should not be sent to mpirun or",
"the [TT]mdrun[tt] process that is the parent of the others.",
"[PAR]",
"When [TT]mdrun[tt] is started with MPI, it does not run niced by default."
};
t_commrec *cr;
t_filenm fnm[] = {
{ efTPX, NULL, NULL, ffREAD },
{ efTRN, "-o", NULL, ffWRITE },
{ efXTC, "-x", NULL, ffOPTWR },
{ efCPT, "-cpi", NULL, ffOPTRD },
{ efCPT, "-cpo", NULL, ffOPTWR },
{ efSTO, "-c", "confout", ffWRITE },
{ efEDR, "-e", "ener", ffWRITE },
{ efLOG, "-g", "md", ffWRITE },
{ efXVG, "-dhdl", "dhdl", ffOPTWR },
{ efXVG, "-field", "field", ffOPTWR },
{ efXVG, "-table", "table", ffOPTRD },
{ efXVG, "-tabletf", "tabletf", ffOPTRD },
{ efXVG, "-tablep", "tablep", ffOPTRD },
{ efXVG, "-tableb", "table", ffOPTRD },
{ efTRX, "-rerun", "rerun", ffOPTRD },
{ efXVG, "-tpi", "tpi", ffOPTWR },
{ efXVG, "-tpid", "tpidist", ffOPTWR },
{ efEDI, "-ei", "sam", ffOPTRD },
{ efXVG, "-eo", "edsam", ffOPTWR },
{ efGCT, "-j", "wham", ffOPTRD },
{ efGCT, "-jo", "bam", ffOPTWR },
{ efXVG, "-ffout", "gct", ffOPTWR },
{ efXVG, "-devout", "deviatie", ffOPTWR },
{ efXVG, "-runav", "runaver", ffOPTWR },
{ efXVG, "-px", "pullx", ffOPTWR },
{ efXVG, "-pf", "pullf", ffOPTWR },
{ efXVG, "-ro", "rotation", ffOPTWR },
{ efLOG, "-ra", "rotangles", ffOPTWR },
{ efLOG, "-rs", "rotslabs", ffOPTWR },
{ efLOG, "-rt", "rottorque", ffOPTWR },
{ efMTX, "-mtx", "nm", ffOPTWR },
{ efNDX, "-dn", "dipole", ffOPTWR },
{ efRND, "-multidir", NULL, ffOPTRDMULT},
{ efDAT, "-plumed", "plumed", ffOPTRD }, /* PLUMED */
{ efDAT, "-membed", "membed", ffOPTRD },
{ efTOP, "-mp", "membed", ffOPTRD },
{ efNDX, "-mn", "membed", ffOPTRD }
};
#define NFILE asize(fnm)
/* Command line options ! */
gmx_bool bCart = FALSE;
gmx_bool bPPPME = FALSE;
gmx_bool bPartDec = FALSE;
gmx_bool bDDBondCheck = TRUE;
gmx_bool bDDBondComm = TRUE;
gmx_bool bTunePME = TRUE;
gmx_bool bTestVerlet = FALSE;
gmx_bool bVerbose = FALSE;
gmx_bool bCompact = TRUE;
gmx_bool bSepPot = FALSE;
gmx_bool bRerunVSite = FALSE;
gmx_bool bIonize = FALSE;
gmx_bool bConfout = TRUE;
gmx_bool bReproducible = FALSE;
int npme = -1;
int nmultisim = 0;
int nstglobalcomm = -1;
int repl_ex_nst = 0;
int repl_ex_seed = -1;
int repl_ex_nex = 0;
int nstepout = 100;
int resetstep = -1;
gmx_large_int_t nsteps = -2; /* the value -2 means that the mdp option will be used */
rvec realddxyz = {0, 0, 0};
const char *ddno_opt[ddnoNR+1] =
{ NULL, "interleave", "pp_pme", "cartesian", NULL };
const char *dddlb_opt[] =
{ NULL, "auto", "no", "yes", NULL };
const char *thread_aff_opt[threadaffNR+1] =
{ NULL, "auto", "on", "off", NULL };
const char *nbpu_opt[] =
{ NULL, "auto", "cpu", "gpu", "gpu_cpu", NULL };
real rdd = 0.0, rconstr = 0.0, dlb_scale = 0.8, pforce = -1;
char *ddcsx = NULL, *ddcsy = NULL, *ddcsz = NULL;
real cpt_period = 15.0, max_hours = -1;
gmx_bool bAppendFiles = TRUE;
gmx_bool bKeepAndNumCPT = FALSE;
gmx_bool bResetCountersHalfWay = FALSE;
output_env_t oenv = NULL;
const char *deviceOptions = "";
gmx_hw_opt_t hw_opt = {0, 0, 0, 0, threadaffSEL, 0, 0, NULL};
t_pargs pa[] = {
{ "-pd", FALSE, etBOOL, {&bPartDec},
"Use particle decompostion" },
{ "-dd", FALSE, etRVEC, {&realddxyz},
"Domain decomposition grid, 0 is optimize" },
{ "-ddorder", FALSE, etENUM, {ddno_opt},
"DD node order" },
{ "-npme", FALSE, etINT, {&npme},
"Number of separate nodes to be used for PME, -1 is guess" },
{ "-nt", FALSE, etINT, {&hw_opt.nthreads_tot},
"Total number of threads to start (0 is guess)" },
{ "-ntmpi", FALSE, etINT, {&hw_opt.nthreads_tmpi},
"Number of thread-MPI threads to start (0 is guess)" },
{ "-ntomp", FALSE, etINT, {&hw_opt.nthreads_omp},
"Number of OpenMP threads per MPI process/thread to start (0 is guess)" },
{ "-ntomp_pme", FALSE, etINT, {&hw_opt.nthreads_omp_pme},
"Number of OpenMP threads per MPI process/thread to start (0 is -ntomp)" },
{ "-pin", FALSE, etENUM, {thread_aff_opt},
"Fix threads (or processes) to specific cores" },
{ "-pinoffset", FALSE, etINT, {&hw_opt.core_pinning_offset},
"The starting logical core number for pinning to cores; used to avoid pinning threads from different mdrun instances to the same core" },
{ "-pinstride", FALSE, etINT, {&hw_opt.core_pinning_stride},
"Pinning distance in logical cores for threads, use 0 to minimize the number of threads per physical core" },
{ "-gpu_id", FALSE, etSTR, {&hw_opt.gpu_id},
"List of GPU id's to use" },
{ "-ddcheck", FALSE, etBOOL, {&bDDBondCheck},
"Check for all bonded interactions with DD" },
{ "-ddbondcomm", FALSE, etBOOL, {&bDDBondComm},
"HIDDENUse special bonded atom communication when [TT]-rdd[tt] > cut-off" },
{ "-rdd", FALSE, etREAL, {&rdd},
"The maximum distance for bonded interactions with DD (nm), 0 is determine from initial coordinates" },
{ "-rcon", FALSE, etREAL, {&rconstr},
"Maximum distance for P-LINCS (nm), 0 is estimate" },
{ "-dlb", FALSE, etENUM, {dddlb_opt},
"Dynamic load balancing (with DD)" },
{ "-dds", FALSE, etREAL, {&dlb_scale},
"Minimum allowed dlb scaling of the DD cell size" },
{ "-ddcsx", FALSE, etSTR, {&ddcsx},
"HIDDENThe DD cell sizes in x" },
{ "-ddcsy", FALSE, etSTR, {&ddcsy},
"HIDDENThe DD cell sizes in y" },
{ "-ddcsz", FALSE, etSTR, {&ddcsz},
"HIDDENThe DD cell sizes in z" },
{ "-gcom", FALSE, etINT, {&nstglobalcomm},
"Global communication frequency" },
{ "-nb", FALSE, etENUM, {&nbpu_opt},
"Calculate non-bonded interactions on" },
{ "-tunepme", FALSE, etBOOL, {&bTunePME},
"Optimize PME load between PP/PME nodes or GPU/CPU" },
{ "-testverlet", FALSE, etBOOL, {&bTestVerlet},
"Test the Verlet non-bonded scheme" },
{ "-v", FALSE, etBOOL, {&bVerbose},
"Be loud and noisy" },
{ "-compact", FALSE, etBOOL, {&bCompact},
"Write a compact log file" },
{ "-seppot", FALSE, etBOOL, {&bSepPot},
"Write separate V and dVdl terms for each interaction type and node to the log file(s)" },
{ "-pforce", FALSE, etREAL, {&pforce},
"Print all forces larger than this (kJ/mol nm)" },
{ "-reprod", FALSE, etBOOL, {&bReproducible},
"Try to avoid optimizations that affect binary reproducibility" },
{ "-cpt", FALSE, etREAL, {&cpt_period},
"Checkpoint interval (minutes)" },
{ "-cpnum", FALSE, etBOOL, {&bKeepAndNumCPT},
"Keep and number checkpoint files" },
{ "-append", FALSE, etBOOL, {&bAppendFiles},
"Append to previous output files when continuing from checkpoint instead of adding the simulation part number to all file names" },
{ "-nsteps", FALSE, etGMX_LARGE_INT, {&nsteps},
"Run this number of steps, overrides .mdp file option" },
{ "-maxh", FALSE, etREAL, {&max_hours},
"Terminate after 0.99 times this time (hours)" },
{ "-multi", FALSE, etINT, {&nmultisim},
"Do multiple simulations in parallel" },
{ "-replex", FALSE, etINT, {&repl_ex_nst},
"Attempt replica exchange periodically with this period (steps)" },
{ "-nex", FALSE, etINT, {&repl_ex_nex},
"Number of random exchanges to carry out each exchange interval (N^3 is one suggestion). -nex zero or not specified gives neighbor replica exchange." },
{ "-reseed", FALSE, etINT, {&repl_ex_seed},
"Seed for replica exchange, -1 is generate a seed" },
{ "-rerunvsite", FALSE, etBOOL, {&bRerunVSite},
"HIDDENRecalculate virtual site coordinates with [TT]-rerun[tt]" },
{ "-ionize", FALSE, etBOOL, {&bIonize},
"Do a simulation including the effect of an X-Ray bombardment on your system" },
{ "-confout", FALSE, etBOOL, {&bConfout},
"HIDDENWrite the last configuration with [TT]-c[tt] and force checkpointing at the last step" },
{ "-stepout", FALSE, etINT, {&nstepout},
"HIDDENFrequency of writing the remaining runtime" },
{ "-resetstep", FALSE, etINT, {&resetstep},
"HIDDENReset cycle counters after these many time steps" },
{ "-resethway", FALSE, etBOOL, {&bResetCountersHalfWay},
"HIDDENReset the cycle counters after half the number of steps or halfway [TT]-maxh[tt]" }
};
gmx_edsam_t ed;
unsigned long Flags, PCA_Flags;
ivec ddxyz;
int dd_node_order;
gmx_bool bAddPart;
FILE *fplog, *fpmulti;
int sim_part, sim_part_fn;
const char *part_suffix = ".part";
char suffix[STRLEN];
int rc;
char **multidir = NULL;
cr = init_par(&argc, &argv);
if (MASTER(cr))
{
CopyRight(stderr, argv[0]);
}
PCA_Flags = (PCA_CAN_SET_DEFFNM | (MASTER(cr) ? 0 : PCA_QUIET));
/* Comment this in to do fexist calls only on master
* works not with rerun or tables at the moment
* also comment out the version of init_forcerec in md.c
* with NULL instead of opt2fn
*/
/*
if (!MASTER(cr))
{
PCA_Flags |= PCA_NOT_READ_NODE;
}
*/
parse_common_args(&argc, argv, PCA_Flags, NFILE, fnm, asize(pa), pa,
asize(desc), desc, 0, NULL, &oenv);
/* we set these early because they might be used in init_multisystem()
Note that there is the potential for npme>nnodes until the number of
threads is set later on, if there's thread parallelization. That shouldn't
lead to problems. */
dd_node_order = nenum(ddno_opt);
cr->npmenodes = npme;
hw_opt.thread_affinity = nenum(thread_aff_opt);
/* now check the -multi and -multidir option */
if (opt2bSet("-multidir", NFILE, fnm))
{
int i;
if (nmultisim > 0)
{
gmx_fatal(FARGS, "mdrun -multi and -multidir options are mutually exclusive.");
}
nmultisim = opt2fns(&multidir, "-multidir", NFILE, fnm);
}
if (repl_ex_nst != 0 && nmultisim < 2)
{
gmx_fatal(FARGS, "Need at least two replicas for replica exchange (option -multi)");
}
if (repl_ex_nex < 0)
{
gmx_fatal(FARGS, "Replica exchange number of exchanges needs to be positive");
}
if (nmultisim > 1)
{
#ifndef GMX_THREAD_MPI
gmx_bool bParFn = (multidir == NULL);
init_multisystem(cr, nmultisim, multidir, NFILE, fnm, bParFn);
#else
gmx_fatal(FARGS, "mdrun -multi is not supported with the thread library.Please compile GROMACS with MPI support");
#endif
}
bAddPart = !bAppendFiles;
/* Check if there is ANY checkpoint file available */
sim_part = 1;
sim_part_fn = sim_part;
if (opt2bSet("-cpi", NFILE, fnm))
{
if (bSepPot && bAppendFiles)
{
gmx_fatal(FARGS, "Output file appending is not supported with -seppot");
}
bAppendFiles =
read_checkpoint_simulation_part(opt2fn_master("-cpi", NFILE,
fnm, cr),
&sim_part_fn, NULL, cr,
bAppendFiles, NFILE, fnm,
part_suffix, &bAddPart);
if (sim_part_fn == 0 && MULTIMASTER(cr))
{
fprintf(stdout, "No previous checkpoint file present, assuming this is a new run.\n");
}
else
{
sim_part = sim_part_fn + 1;
}
if (MULTISIM(cr) && MASTER(cr))
{
if (MULTIMASTER(cr))
{
/* Log file is not yet available, so if there's a
* problem we can only write to stderr. */
fpmulti = stderr;
}
else
{
fpmulti = NULL;
}
check_multi_int(fpmulti, cr->ms, sim_part, "simulation part", TRUE);
}
}
else
{
bAppendFiles = FALSE;
}
if (!bAppendFiles)
{
sim_part_fn = sim_part;
}
if (bAddPart)
{
/* Rename all output files (except checkpoint files) */
/* create new part name first (zero-filled) */
sprintf(suffix, "%s%04d", part_suffix, sim_part_fn);
add_suffix_to_output_names(fnm, NFILE, suffix);
if (MULTIMASTER(cr))
{
fprintf(stdout, "Checkpoint file is from part %d, new output files will be suffixed '%s'.\n", sim_part-1, suffix);
}
}
Flags = opt2bSet("-rerun", NFILE, fnm) ? MD_RERUN : 0;
Flags = Flags | (bSepPot ? MD_SEPPOT : 0);
Flags = Flags | (bIonize ? MD_IONIZE : 0);
Flags = Flags | (bPartDec ? MD_PARTDEC : 0);
Flags = Flags | (bDDBondCheck ? MD_DDBONDCHECK : 0);
Flags = Flags | (bDDBondComm ? MD_DDBONDCOMM : 0);
Flags = Flags | (bTunePME ? MD_TUNEPME : 0);
Flags = Flags | (bTestVerlet ? MD_TESTVERLET : 0);
Flags = Flags | (bConfout ? MD_CONFOUT : 0);
Flags = Flags | (bRerunVSite ? MD_RERUN_VSITE : 0);
Flags = Flags | (bReproducible ? MD_REPRODUCIBLE : 0);
Flags = Flags | (bAppendFiles ? MD_APPENDFILES : 0);
Flags = Flags | (opt2parg_bSet("-append", asize(pa), pa) ? MD_APPENDFILESSET : 0);
Flags = Flags | (bKeepAndNumCPT ? MD_KEEPANDNUMCPT : 0);
Flags = Flags | (sim_part > 1 ? MD_STARTFROMCPT : 0);
Flags = Flags | (bResetCountersHalfWay ? MD_RESETCOUNTERSHALFWAY : 0);
/* We postpone opening the log file if we are appending, so we can
first truncate the old log file and append to the correct position
there instead. */
if ((MASTER(cr) || bSepPot) && !bAppendFiles)
{
gmx_log_open(ftp2fn(efLOG, NFILE, fnm), cr,
!bSepPot, Flags & MD_APPENDFILES, &fplog);
CopyRight(fplog, argv[0]);
please_cite(fplog, "Hess2008b");
please_cite(fplog, "Spoel2005a");
please_cite(fplog, "Lindahl2001a");
please_cite(fplog, "Berendsen95a");
}
else if (!MASTER(cr) && bSepPot)
{
gmx_log_open(ftp2fn(efLOG, NFILE, fnm), cr, !bSepPot, Flags, &fplog);
}
else
{
fplog = NULL;
}
ddxyz[XX] = (int)(realddxyz[XX] + 0.5);
ddxyz[YY] = (int)(realddxyz[YY] + 0.5);
ddxyz[ZZ] = (int)(realddxyz[ZZ] + 0.5);
/* PLUMED */
plumedswitch=0;
if (opt2bSet("-plumed",NFILE,fnm)) plumedswitch=1;
if(plumedswitch){
int plumed_is_there=0;
int real_precision=sizeof(real);
real energyUnits=1.0;
real lengthUnits=1.0;
real timeUnits=1.0;
if(!plumed_installed()){
gmx_fatal(FARGS,"Plumed is not available. Check your PLUMED_KERNEL variable.");
}
plumedmain=plumed_create();
plumed_cmd(plumedmain,"setRealPrecision",&real_precision);
// this is not necessary for gromacs units:
plumed_cmd(plumedmain,"setMDEnergyUnits",&energyUnits);
plumed_cmd(plumedmain,"setMDLengthUnits",&lengthUnits);
plumed_cmd(plumedmain,"setMDTimeUnits",&timeUnits);
//
plumed_cmd(plumedmain,"setPlumedDat",ftp2fn(efDAT,NFILE,fnm));
plumedswitch=1;
}
/* END PLUMED */
rc = mdrunner(&hw_opt, fplog, cr, NFILE, fnm, oenv, bVerbose, bCompact,
nstglobalcomm, ddxyz, dd_node_order, rdd, rconstr,
dddlb_opt[0], dlb_scale, ddcsx, ddcsy, ddcsz,
nbpu_opt[0],
nsteps, nstepout, resetstep,
nmultisim, repl_ex_nst, repl_ex_nex, repl_ex_seed,
pforce, cpt_period, max_hours, deviceOptions, Flags);
/* PLUMED */
if(plumedswitch){
plumed_finalize(plumedmain);
}
/* END PLUMED */
gmx_finalize_par();
if (MULTIMASTER(cr))
{
thanx(stderr);
}
/* Log file has to be closed in mdrunner if we are appending to it
(fplog not set here) */
if (MASTER(cr) && !bAppendFiles)
{
gmx_log_close(fplog);
}
return rc;
}
void plumed_gfinalize(void){
plumed_finalize(gmain);
gmain.p=NULL;
}
void plumed_f_finalize(char*c){
plumed p;
p=plumed_f2c(c);
plumed_finalize(p);
}
int mdrunner(gmx_hw_opt_t *hw_opt,
FILE *fplog, t_commrec *cr, int nfile,
const t_filenm fnm[], const output_env_t oenv, gmx_bool bVerbose,
gmx_bool bCompact, int nstglobalcomm,
ivec ddxyz, int dd_node_order, real rdd, real rconstr,
const char *dddlb_opt, real dlb_scale,
const char *ddcsx, const char *ddcsy, const char *ddcsz,
const char *nbpu_opt, int nstlist_cmdline,
gmx_int64_t nsteps_cmdline, int nstepout, int resetstep,
int gmx_unused nmultisim, int repl_ex_nst, int repl_ex_nex,
int repl_ex_seed, real pforce, real cpt_period, real max_hours,
int imdport, unsigned long Flags)
{
gmx_bool bForceUseGPU, bTryUseGPU, bRerunMD;
t_inputrec *inputrec;
t_state *state = NULL;
matrix box;
gmx_ddbox_t ddbox = {0};
int npme_major, npme_minor;
t_nrnb *nrnb;
gmx_mtop_t *mtop = NULL;
t_mdatoms *mdatoms = NULL;
t_forcerec *fr = NULL;
t_fcdata *fcd = NULL;
real ewaldcoeff_q = 0;
real ewaldcoeff_lj = 0;
struct gmx_pme_t **pmedata = NULL;
gmx_vsite_t *vsite = NULL;
gmx_constr_t constr;
int nChargePerturbed = -1, nTypePerturbed = 0, status;
gmx_wallcycle_t wcycle;
gmx_bool bReadEkin;
gmx_walltime_accounting_t walltime_accounting = NULL;
int rc;
gmx_int64_t reset_counters;
gmx_edsam_t ed = NULL;
int nthreads_pme = 1;
int nthreads_pp = 1;
gmx_membed_t membed = NULL;
gmx_hw_info_t *hwinfo = NULL;
/* The master rank decides early on bUseGPU and broadcasts this later */
gmx_bool bUseGPU = FALSE;
/* CAUTION: threads may be started later on in this function, so
cr doesn't reflect the final parallel state right now */
snew(inputrec, 1);
snew(mtop, 1);
if (Flags & MD_APPENDFILES)
{
fplog = NULL;
}
bRerunMD = (Flags & MD_RERUN);
bForceUseGPU = (strncmp(nbpu_opt, "gpu", 3) == 0);
bTryUseGPU = (strncmp(nbpu_opt, "auto", 4) == 0) || bForceUseGPU;
/* Detect hardware, gather information. This is an operation that is
* global for this process (MPI rank). */
hwinfo = gmx_detect_hardware(fplog, cr, bTryUseGPU);
gmx_print_detected_hardware(fplog, cr, hwinfo);
if (fplog != NULL)
{
/* Print references after all software/hardware printing */
please_cite(fplog, "Abraham2015");
please_cite(fplog, "Pall2015");
please_cite(fplog, "Pronk2013");
please_cite(fplog, "Hess2008b");
please_cite(fplog, "Spoel2005a");
please_cite(fplog, "Lindahl2001a");
please_cite(fplog, "Berendsen95a");
}
snew(state, 1);
if (SIMMASTER(cr))
{
/* Read (nearly) all data required for the simulation */
read_tpx_state(ftp2fn(efTPR, nfile, fnm), inputrec, state, NULL, mtop);
if (inputrec->cutoff_scheme == ecutsVERLET)
{
/* Here the master rank decides if all ranks will use GPUs */
bUseGPU = (hwinfo->gpu_info.n_dev_compatible > 0 ||
getenv("GMX_EMULATE_GPU") != NULL);
/* TODO add GPU kernels for this and replace this check by:
* (bUseGPU && (ir->vdwtype == evdwPME &&
* ir->ljpme_combination_rule == eljpmeLB))
* update the message text and the content of nbnxn_acceleration_supported.
*/
if (bUseGPU &&
!nbnxn_gpu_acceleration_supported(fplog, cr, inputrec, bRerunMD))
{
/* Fallback message printed by nbnxn_acceleration_supported */
if (bForceUseGPU)
{
gmx_fatal(FARGS, "GPU acceleration requested, but not supported with the given input settings");
}
bUseGPU = FALSE;
}
prepare_verlet_scheme(fplog, cr,
inputrec, nstlist_cmdline, mtop, state->box,
bUseGPU);
}
else
{
if (nstlist_cmdline > 0)
{
gmx_fatal(FARGS, "Can not set nstlist with the group cut-off scheme");
}
if (hwinfo->gpu_info.n_dev_compatible > 0)
{
md_print_warn(cr, fplog,
"NOTE: GPU(s) found, but the current simulation can not use GPUs\n"
" To use a GPU, set the mdp option: cutoff-scheme = Verlet\n");
}
if (bForceUseGPU)
{
gmx_fatal(FARGS, "GPU requested, but can't be used without cutoff-scheme=Verlet");
}
#ifdef GMX_TARGET_BGQ
md_print_warn(cr, fplog,
"NOTE: There is no SIMD implementation of the group scheme kernels on\n"
" BlueGene/Q. You will observe better performance from using the\n"
" Verlet cut-off scheme.\n");
#endif
}
if (inputrec->eI == eiSD2)
{
md_print_warn(cr, fplog, "The stochastic dynamics integrator %s is deprecated, since\n"
"it is slower than integrator %s and is slightly less accurate\n"
"with constraints. Use the %s integrator.",
ei_names[inputrec->eI], ei_names[eiSD1], ei_names[eiSD1]);
}
}
/* Check and update the hardware options for internal consistency */
check_and_update_hw_opt_1(hw_opt, cr);
/* Early check for externally set process affinity. */
gmx_check_thread_affinity_set(fplog, cr,
hw_opt, hwinfo->nthreads_hw_avail, FALSE);
#ifdef GMX_THREAD_MPI
if (SIMMASTER(cr))
{
if (cr->npmenodes > 0 && hw_opt->nthreads_tmpi <= 0)
{
gmx_fatal(FARGS, "You need to explicitly specify the number of MPI threads (-ntmpi) when using separate PME ranks");
}
/* Since the master knows the cut-off scheme, update hw_opt for this.
* This is done later for normal MPI and also once more with tMPI
* for all tMPI ranks.
*/
check_and_update_hw_opt_2(hw_opt, inputrec->cutoff_scheme);
/* NOW the threads will be started: */
hw_opt->nthreads_tmpi = get_nthreads_mpi(hwinfo,
hw_opt,
inputrec, mtop,
cr, fplog, bUseGPU);
if (hw_opt->nthreads_tmpi > 1)
{
t_commrec *cr_old = cr;
/* now start the threads. */
cr = mdrunner_start_threads(hw_opt, fplog, cr_old, nfile, fnm,
oenv, bVerbose, bCompact, nstglobalcomm,
ddxyz, dd_node_order, rdd, rconstr,
dddlb_opt, dlb_scale, ddcsx, ddcsy, ddcsz,
nbpu_opt, nstlist_cmdline,
nsteps_cmdline, nstepout, resetstep, nmultisim,
repl_ex_nst, repl_ex_nex, repl_ex_seed, pforce,
cpt_period, max_hours,
Flags);
/* the main thread continues here with a new cr. We don't deallocate
the old cr because other threads may still be reading it. */
if (cr == NULL)
{
gmx_comm("Failed to spawn threads");
}
}
}
#endif
/* END OF CAUTION: cr is now reliable */
/* g_membed initialisation *
* Because we change the mtop, init_membed is called before the init_parallel *
* (in case we ever want to make it run in parallel) */
if (opt2bSet("-membed", nfile, fnm))
{
if (MASTER(cr))
{
fprintf(stderr, "Initializing membed");
}
membed = init_membed(fplog, nfile, fnm, mtop, inputrec, state, cr, &cpt_period);
}
if (PAR(cr))
{
/* now broadcast everything to the non-master nodes/threads: */
init_parallel(cr, inputrec, mtop);
/* The master rank decided on the use of GPUs,
* broadcast this information to all ranks.
*/
gmx_bcast_sim(sizeof(bUseGPU), &bUseGPU, cr);
}
if (fplog != NULL)
{
pr_inputrec(fplog, 0, "Input Parameters", inputrec, FALSE);
fprintf(fplog, "\n");
}
/* now make sure the state is initialized and propagated */
set_state_entries(state, inputrec);
/* A parallel command line option consistency check that we can
only do after any threads have started. */
if (!PAR(cr) &&
(ddxyz[XX] > 1 || ddxyz[YY] > 1 || ddxyz[ZZ] > 1 || cr->npmenodes > 0))
{
gmx_fatal(FARGS,
"The -dd or -npme option request a parallel simulation, "
#ifndef GMX_MPI
"but %s was compiled without threads or MPI enabled"
#else
#ifdef GMX_THREAD_MPI
"but the number of threads (option -nt) is 1"
#else
"but %s was not started through mpirun/mpiexec or only one rank was requested through mpirun/mpiexec"
#endif
#endif
, output_env_get_program_display_name(oenv)
);
}
if (bRerunMD &&
(EI_ENERGY_MINIMIZATION(inputrec->eI) || eiNM == inputrec->eI))
{
gmx_fatal(FARGS, "The .mdp file specified an energy mininization or normal mode algorithm, and these are not compatible with mdrun -rerun");
}
if (can_use_allvsall(inputrec, TRUE, cr, fplog) && DOMAINDECOMP(cr))
{
gmx_fatal(FARGS, "All-vs-all loops do not work with domain decomposition, use a single MPI rank");
}
if (!(EEL_PME(inputrec->coulombtype) || EVDW_PME(inputrec->vdwtype)))
{
if (cr->npmenodes > 0)
{
gmx_fatal_collective(FARGS, cr, NULL,
"PME-only ranks are requested, but the system does not use PME for electrostatics or LJ");
}
cr->npmenodes = 0;
}
if (bUseGPU && cr->npmenodes < 0)
{
/* With GPUs we don't automatically use PME-only ranks. PME ranks can
* improve performance with many threads per GPU, since our OpenMP
* scaling is bad, but it's difficult to automate the setup.
*/
cr->npmenodes = 0;
}
#ifdef GMX_FAHCORE
if (MASTER(cr))
{
fcRegisterSteps(inputrec->nsteps, inputrec->init_step);
}
#endif
/* NMR restraints must be initialized before load_checkpoint,
* since with time averaging the history is added to t_state.
* For proper consistency check we therefore need to extend
* t_state here.
* So the PME-only nodes (if present) will also initialize
* the distance restraints.
*/
snew(fcd, 1);
/* This needs to be called before read_checkpoint to extend the state */
init_disres(fplog, mtop, inputrec, cr, fcd, state, repl_ex_nst > 0);
init_orires(fplog, mtop, state->x, inputrec, cr, &(fcd->orires),
state);
if (DEFORM(*inputrec))
{
/* Store the deform reference box before reading the checkpoint */
if (SIMMASTER(cr))
{
copy_mat(state->box, box);
}
if (PAR(cr))
{
gmx_bcast(sizeof(box), box, cr);
}
/* Because we do not have the update struct available yet
* in which the reference values should be stored,
* we store them temporarily in static variables.
* This should be thread safe, since they are only written once
* and with identical values.
*/
tMPI_Thread_mutex_lock(&deform_init_box_mutex);
deform_init_init_step_tpx = inputrec->init_step;
copy_mat(box, deform_init_box_tpx);
tMPI_Thread_mutex_unlock(&deform_init_box_mutex);
}
if (opt2bSet("-cpi", nfile, fnm))
{
/* Check if checkpoint file exists before doing continuation.
* This way we can use identical input options for the first and subsequent runs...
*/
if (gmx_fexist_master(opt2fn_master("-cpi", nfile, fnm, cr), cr) )
{
load_checkpoint(opt2fn_master("-cpi", nfile, fnm, cr), &fplog,
cr, ddxyz,
inputrec, state, &bReadEkin,
(Flags & MD_APPENDFILES),
(Flags & MD_APPENDFILESSET));
if (bReadEkin)
{
Flags |= MD_READ_EKIN;
}
}
}
if (MASTER(cr) && (Flags & MD_APPENDFILES))
{
gmx_log_open(ftp2fn(efLOG, nfile, fnm), cr,
Flags, &fplog);
}
/* override nsteps with value from cmdline */
override_nsteps_cmdline(fplog, nsteps_cmdline, inputrec, cr);
if (SIMMASTER(cr))
{
copy_mat(state->box, box);
}
if (PAR(cr))
{
gmx_bcast(sizeof(box), box, cr);
}
/* Essential dynamics */
if (opt2bSet("-ei", nfile, fnm))
{
/* Open input and output files, allocate space for ED data structure */
ed = ed_open(mtop->natoms, &state->edsamstate, nfile, fnm, Flags, oenv, cr);
}
if (PAR(cr) && !(EI_TPI(inputrec->eI) ||
inputrec->eI == eiNM))
{
cr->dd = init_domain_decomposition(fplog, cr, Flags, ddxyz, rdd, rconstr,
dddlb_opt, dlb_scale,
ddcsx, ddcsy, ddcsz,
mtop, inputrec,
box, state->x,
&ddbox, &npme_major, &npme_minor);
make_dd_communicators(fplog, cr, dd_node_order);
/* Set overallocation to avoid frequent reallocation of arrays */
set_over_alloc_dd(TRUE);
}
else
{
/* PME, if used, is done on all nodes with 1D decomposition */
cr->npmenodes = 0;
cr->duty = (DUTY_PP | DUTY_PME);
npme_major = 1;
npme_minor = 1;
if (inputrec->ePBC == epbcSCREW)
{
gmx_fatal(FARGS,
"pbc=%s is only implemented with domain decomposition",
epbc_names[inputrec->ePBC]);
}
}
if (PAR(cr))
{
/* After possible communicator splitting in make_dd_communicators.
* we can set up the intra/inter node communication.
*/
gmx_setup_nodecomm(fplog, cr);
}
/* Initialize per-physical-node MPI process/thread ID and counters. */
gmx_init_intranode_counters(cr);
#ifdef GMX_MPI
if (MULTISIM(cr))
{
md_print_info(cr, fplog,
"This is simulation %d out of %d running as a composite GROMACS\n"
"multi-simulation job. Setup for this simulation:\n\n",
cr->ms->sim, cr->ms->nsim);
}
md_print_info(cr, fplog, "Using %d MPI %s\n",
cr->nnodes,
#ifdef GMX_THREAD_MPI
cr->nnodes == 1 ? "thread" : "threads"
#else
cr->nnodes == 1 ? "process" : "processes"
#endif
);
fflush(stderr);
#endif
/* Check and update hw_opt for the cut-off scheme */
check_and_update_hw_opt_2(hw_opt, inputrec->cutoff_scheme);
/* Check and update hw_opt for the number of MPI ranks */
check_and_update_hw_opt_3(hw_opt);
gmx_omp_nthreads_init(fplog, cr,
hwinfo->nthreads_hw_avail,
hw_opt->nthreads_omp,
hw_opt->nthreads_omp_pme,
(cr->duty & DUTY_PP) == 0,
inputrec->cutoff_scheme == ecutsVERLET);
#ifndef NDEBUG
if (integrator[inputrec->eI].func != do_tpi &&
inputrec->cutoff_scheme == ecutsVERLET)
{
gmx_feenableexcept();
}
#endif
if (bUseGPU)
{
/* Select GPU id's to use */
gmx_select_gpu_ids(fplog, cr, &hwinfo->gpu_info, bForceUseGPU,
&hw_opt->gpu_opt);
}
else
{
/* Ignore (potentially) manually selected GPUs */
hw_opt->gpu_opt.n_dev_use = 0;
}
/* check consistency across ranks of things like SIMD
* support and number of GPUs selected */
gmx_check_hw_runconf_consistency(fplog, hwinfo, cr, hw_opt, bUseGPU);
/* Now that we know the setup is consistent, check for efficiency */
check_resource_division_efficiency(hwinfo, hw_opt, Flags & MD_NTOMPSET,
cr, fplog);
if (DOMAINDECOMP(cr))
{
/* When we share GPUs over ranks, we need to know this for the DLB */
dd_setup_dlb_resource_sharing(cr, hwinfo, hw_opt);
}
/* getting number of PP/PME threads
PME: env variable should be read only on one node to make sure it is
identical everywhere;
*/
/* TODO nthreads_pp is only used for pinning threads.
* This is a temporary solution until we have a hw topology library.
*/
nthreads_pp = gmx_omp_nthreads_get(emntNonbonded);
nthreads_pme = gmx_omp_nthreads_get(emntPME);
wcycle = wallcycle_init(fplog, resetstep, cr, nthreads_pp, nthreads_pme);
if (PAR(cr))
{
/* Master synchronizes its value of reset_counters with all nodes
* including PME only nodes */
reset_counters = wcycle_get_reset_counters(wcycle);
gmx_bcast_sim(sizeof(reset_counters), &reset_counters, cr);
wcycle_set_reset_counters(wcycle, reset_counters);
}
snew(nrnb, 1);
if (cr->duty & DUTY_PP)
{
bcast_state(cr, state);
/* Initiate forcerecord */
fr = mk_forcerec();
fr->hwinfo = hwinfo;
fr->gpu_opt = &hw_opt->gpu_opt;
init_forcerec(fplog, oenv, fr, fcd, inputrec, mtop, cr, box,
opt2fn("-table", nfile, fnm),
opt2fn("-tabletf", nfile, fnm),
opt2fn("-tablep", nfile, fnm),
opt2fn("-tableb", nfile, fnm),
nbpu_opt,
FALSE,
pforce);
/* version for PCA_NOT_READ_NODE (see md.c) */
/*init_forcerec(fplog,fr,fcd,inputrec,mtop,cr,box,FALSE,
"nofile","nofile","nofile","nofile",FALSE,pforce);
*/
/* Initialize QM-MM */
if (fr->bQMMM)
{
init_QMMMrec(cr, mtop, inputrec, fr);
}
/* Initialize the mdatoms structure.
* mdatoms is not filled with atom data,
* as this can not be done now with domain decomposition.
*/
mdatoms = init_mdatoms(fplog, mtop, inputrec->efep != efepNO);
/* Initialize the virtual site communication */
vsite = init_vsite(mtop, cr, FALSE);
calc_shifts(box, fr->shift_vec);
/* With periodic molecules the charge groups should be whole at start up
* and the virtual sites should not be far from their proper positions.
*/
if (!inputrec->bContinuation && MASTER(cr) &&
!(inputrec->ePBC != epbcNONE && inputrec->bPeriodicMols))
{
/* Make molecules whole at start of run */
if (fr->ePBC != epbcNONE)
{
do_pbc_first_mtop(fplog, inputrec->ePBC, box, mtop, state->x);
}
if (vsite)
{
/* Correct initial vsite positions are required
* for the initial distribution in the domain decomposition
* and for the initial shell prediction.
*/
construct_vsites_mtop(vsite, mtop, state->x);
}
}
if (EEL_PME(fr->eeltype) || EVDW_PME(fr->vdwtype))
{
ewaldcoeff_q = fr->ewaldcoeff_q;
ewaldcoeff_lj = fr->ewaldcoeff_lj;
pmedata = &fr->pmedata;
}
else
{
pmedata = NULL;
}
}
else
{
/* This is a PME only node */
/* We don't need the state */
done_state(state);
ewaldcoeff_q = calc_ewaldcoeff_q(inputrec->rcoulomb, inputrec->ewald_rtol);
ewaldcoeff_lj = calc_ewaldcoeff_lj(inputrec->rvdw, inputrec->ewald_rtol_lj);
snew(pmedata, 1);
}
if (hw_opt->thread_affinity != threadaffOFF)
{
/* Before setting affinity, check whether the affinity has changed
* - which indicates that probably the OpenMP library has changed it
* since we first checked).
*/
gmx_check_thread_affinity_set(fplog, cr,
hw_opt, hwinfo->nthreads_hw_avail, TRUE);
/* Set the CPU affinity */
gmx_set_thread_affinity(fplog, cr, hw_opt, hwinfo);
}
/* Initiate PME if necessary,
* either on all nodes or on dedicated PME nodes only. */
if (EEL_PME(inputrec->coulombtype) || EVDW_PME(inputrec->vdwtype))
{
if (mdatoms)
{
nChargePerturbed = mdatoms->nChargePerturbed;
if (EVDW_PME(inputrec->vdwtype))
{
nTypePerturbed = mdatoms->nTypePerturbed;
}
}
if (cr->npmenodes > 0)
{
/* The PME only nodes need to know nChargePerturbed(FEP on Q) and nTypePerturbed(FEP on LJ)*/
gmx_bcast_sim(sizeof(nChargePerturbed), &nChargePerturbed, cr);
gmx_bcast_sim(sizeof(nTypePerturbed), &nTypePerturbed, cr);
}
if (cr->duty & DUTY_PME)
{
status = gmx_pme_init(pmedata, cr, npme_major, npme_minor, inputrec,
mtop ? mtop->natoms : 0, nChargePerturbed, nTypePerturbed,
(Flags & MD_REPRODUCIBLE), nthreads_pme);
if (status != 0)
{
gmx_fatal(FARGS, "Error %d initializing PME", status);
}
}
}
if (integrator[inputrec->eI].func == do_md)
{
/* Turn on signal handling on all nodes */
/*
* (A user signal from the PME nodes (if any)
* is communicated to the PP nodes.
*/
signal_handler_install();
}
if (cr->duty & DUTY_PP)
{
/* Assumes uniform use of the number of OpenMP threads */
walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(emntDefault));
if (inputrec->bPull)
{
/* Initialize pull code */
inputrec->pull_work =
init_pull(fplog, inputrec->pull, inputrec, nfile, fnm,
mtop, cr, oenv, inputrec->fepvals->init_lambda,
EI_DYNAMICS(inputrec->eI) && MASTER(cr), Flags);
}
if (inputrec->bRot)
{
/* Initialize enforced rotation code */
init_rot(fplog, inputrec, nfile, fnm, cr, state->x, box, mtop, oenv,
bVerbose, Flags);
}
if (inputrec->eSwapCoords != eswapNO)
{
/* Initialize ion swapping code */
init_swapcoords(fplog, bVerbose, inputrec, opt2fn_master("-swap", nfile, fnm, cr),
mtop, state->x, state->box, &state->swapstate, cr, oenv, Flags);
}
constr = init_constraints(fplog, mtop, inputrec, ed, state, cr);
if (DOMAINDECOMP(cr))
{
GMX_RELEASE_ASSERT(fr, "fr was NULL while cr->duty was DUTY_PP");
dd_init_bondeds(fplog, cr->dd, mtop, vsite, inputrec,
Flags & MD_DDBONDCHECK, fr->cginfo_mb);
set_dd_parameters(fplog, cr->dd, dlb_scale, inputrec, &ddbox);
setup_dd_grid(fplog, cr->dd);
}
/* Now do whatever the user wants us to do (how flexible...) */
integrator[inputrec->eI].func(fplog, cr, nfile, fnm,
oenv, bVerbose, bCompact,
nstglobalcomm,
vsite, constr,
nstepout, inputrec, mtop,
fcd, state,
mdatoms, nrnb, wcycle, ed, fr,
repl_ex_nst, repl_ex_nex, repl_ex_seed,
membed,
cpt_period, max_hours,
imdport,
Flags,
walltime_accounting);
if (inputrec->bPull)
{
finish_pull(inputrec->pull_work);
}
if (inputrec->bRot)
{
finish_rot(inputrec->rot);
}
}
else
{
GMX_RELEASE_ASSERT(pmedata, "pmedata was NULL while cr->duty was not DUTY_PP");
/* do PME only */
walltime_accounting = walltime_accounting_init(gmx_omp_nthreads_get(emntPME));
gmx_pmeonly(*pmedata, cr, nrnb, wcycle, walltime_accounting, ewaldcoeff_q, ewaldcoeff_lj, inputrec);
}
wallcycle_stop(wcycle, ewcRUN);
/* Finish up, write some stuff
* if rerunMD, don't write last frame again
*/
finish_run(fplog, cr,
inputrec, nrnb, wcycle, walltime_accounting,
fr ? fr->nbv : NULL,
EI_DYNAMICS(inputrec->eI) && !MULTISIM(cr));
/* Free GPU memory and context */
free_gpu_resources(fr, cr, &hwinfo->gpu_info, fr ? fr->gpu_opt : NULL);
if (opt2bSet("-membed", nfile, fnm))
{
sfree(membed);
}
gmx_hardware_info_free(hwinfo);
/* Does what it says */
print_date_and_time(fplog, cr->nodeid, "Finished mdrun", gmx_gettime());
walltime_accounting_destroy(walltime_accounting);
/* PLUMED */
if(plumedswitch){
plumed_finalize(plumedmain);
}
/* END PLUMED */
/* Close logfile already here if we were appending to it */
if (MASTER(cr) && (Flags & MD_APPENDFILES))
{
gmx_log_close(fplog);
}
rc = (int)gmx_get_stop_condition();
done_ed(&ed);
#ifdef GMX_THREAD_MPI
/* we need to join all threads. The sub-threads join when they
exit this function, but the master thread needs to be told to
wait for that. */
if (PAR(cr) && MASTER(cr))
{
tMPI_Finalize();
}
#endif
return rc;
}
