Documentation

From OxDNA
Revision as of 14:42, 13 April 2012 by Romano (talk | contribs) (New page: BLA BLA Documentation introduction ==MAKE OPTIONS== To compile oxDNA you have to change the first rows in the makefile to match your CUDA installation and to specify the architecture (64...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

BLA BLA Documentation introduction

MAKE OPTIONS

To compile oxDNA you have to change the first rows in the makefile to match your CUDA installation and to specify the architecture (64 or 32 bit) of your machine.

dbg=1 oxDNA will be compiled with debug flags (both for nvcc and gcc). The resulting executable will be put in the Debug directory.

g=1 oxDNA will be compiled with both debug and optimization flags. The resulting executable will be put in the Release directory.

arch=old oxDNA will be compiled with -arch=sm_13, i.e. for the old (1.x) architecture.

v=1 verbose output from nvcc (about ptx).

cuda=1 oxDNA will be compiled with CUDA support (who could have told?).

intel=1 oxDNA will be compile without CUDA support using the Intel icpc compiler. The resulting executable will be named oxDNA_intel.

USAGE

oxDNA input_file

OUTPUT FILE

The energy.dat (default name, can be changed in the configuration file) has this layout for MD: time potential_energy kinetic_energy total_energy hydrogen_bonding_energy

The energy.dat (default name, can be changed in the configuration file) has this layout for MC: time potential_energy hydrogen_bonding_energy acc_trasl acc_rot

Mind that potential, kinetic and total energies are divided by the number of particles whereas the hydrogen bonding energy is not.

Configurations are saved in the trajectory file.

INPUT FILE OPTIONS

As always in UNIX environments, everything is case sensitive. The options are in the form key = value. There can be arbitrary spaces before and after both key and value. Line with a leading # will be treated as comments. In this part | (pipe) is the separator between the different values that can be used to specify a value for the key. Keys between [ and ] are optional, the value after the equal sign is the default value.

GENERIC PROGRAM OPTIONS: The options listed here define the generic behavior of the entire program.

sim_type MD|MC|CMMC MD = Molecular Dynamics, MC = Monte Carlo, CMMC = Cluster Move Monte Carlo

backend CUDA|CPU

backend_precision float|double if backend == CPU. float|double|mixed if backend == CUDA. With mixed precision the most important calculations will be performed in double precision while everything is stored in single precision.

[CUDA_device=0] used if backend == CUDA. the GPU where the simulation will be run on.

[CUDA_list=no] no|verlet|bin_verlet used if backend == CUDA. no means no lists at all. O(N^2) algorithm. verlet means usual verlet lists. Not fit for very asymmetric binary mixtures. bin_verlet are verlet lists that works particularly well (both in terms of performances and memory usage) for very asymmetric binary mixtures.

[threads_per_block=2*warpSize] used if backend == CUDA. number of threads per block used internally by CUDA kernel launches. warpSize (for arch <= 20) is always 32.

[debug=0] 0|1 1 if you want verbose logs, 0 otherwise.

SIMULATION OPTIONS The options listed here specify the behaviour of the simulation.

steps number of steps to be performed.

[restart_step_counter=0] 0|1 0 means that the step counter will start from the value read in the configuration file, 1 otherwise.

[CUDA_sort_every=0] used if backend == CUDA. the particles will be reordered (to speed up memory accesses) every CUDA_sort_every list updates.

[seed=time(NULL)] seed for the random number generator. On Unix systesm, it will use by default a number from /dev/urandom + time(NULL)

T temperature of the simulation. It can be expressed in simulation units or kelvin (append a k or K after the value) or celsius (append a c or C after the value). Examples: Value -> Simulation Units 0.1 -> 0.1 300 K -> 0.1 300k -> 0.1 26.85c -> 0.1 26.85 C -> 0.1

verlet_skin if a particle moves more than verlet_skin then the lists will be updated. Its name is misleading: the actual verlet skin is 2*verlet_skin. I'm sorry for that :)

[use_average_seq=1] 0|1 specifies whether to use the default hard-coded average parameters for base-pairing and stacking interaction strengths or not. If sequence dependence is to be used, set this to 0 and specify seq_dep_file.

[seq_dep_file] specifies the file from which the sequence dependent parameters should be read. Mandatory if use_average_seq=no, ignored otherwise. A sample file is provided (sequence_dependent_parameters.txt).

[external_forces=0] 0|1 specifies whether there are external forces acting on the nucleotides or not. If it is set to 1, then a file which specifies the external forces' configuration has to be provided (see external_forces_file).

[external_forces_file] specifies the file containing all the external forces' configurations. Currently there are six supported force types (see EXAMPLES/TRAPS for some examples): - string - twist - trap - repulsion_plane - repulsion_plane_moving - mutual_trap Only 'string' forces are supported in CUDA simulations.

THESE OPTIONS ARE FOR MD SIMULATIONS ONLY:

dt time step of the integration.

thermostat no|refresh|john if backend == CPU no|john if backend == CUDA no means no thermostat will be used. refresh will refresh all the particle's velocities from a maxwellian every newtonian_steps steps. john is an Anderson-like thermostat (see pt).

newtonian_steps required if thermostat != no. number of steps after which a procedure of thermalization will be performed.

pt used if thermostat == john. It's the probability that a particle's velocity will be refreshed during a thermalization procedure.

diff_coeff required if pt is not specified. used internally to automatically compute the pt that would be needed if we wanted such a self diffusion coefficient. Not used if pt is set.

END MD

THESE OPTIONS ARE FOR MC (AND CMMC) SIMULATIONS ONLY

[check_energy_every=10] this number times print_energy_every gives the number of steps after which the energy will be computed from scratch and checked against the actual value computed adding energy differences.

[check_energy_threshold=1e-4] if abs((old_energy - new_energy)/old_energy) > check_energy_threshold then the program will die and warn the user.

ensemble NVT ensemble of the simulation. More ensembles could be added in future versions.

delta_translation maximum displacement (per dimension) for translational moves in simulation units.

delta_translation maximum displacement for rotational moves in simulation units.

END MC


INPUT/OUTPUT The options listed here are used to manage the I/O (read and write configurations, energies and so on)

conf_file initial configuration file.

topology file containing the system's topology.

trajectory_file the main output of the program. All the configurations will be appended to this file as they are printed.

[confs_to_skip=0] valid only if conf_file is a trajectory. Skip the first confs_to_skip configurations and then load in memory the (confs_to_skip+1)th.

[lastconf_file=last_conf.dat] This is the file where the last configuration is saved (when the program finishes or is killed). Set to last_conf.dat by default

[refresh_vel=0] 0|1 if 1 the initial velocities will be refreshed from a maxwellian.

energy_file energy output file.

[print_energy_every=1000] this will make the program print the energies every print_energy_every steps.

[no_stdout_energy=0] 0|1 if 1 the energy will be printed just to the energy_file.

[time_scale=linear] linear|log_lin with linear the configurations will be saved every print_conf_interval. with log_lin the configurations will be saved logarithmically for print_conf_ppc times. After that the logarithmic sequence will restart.

print_conf_interval linear interval if time_scale == linear. First step of the logarithmic scale if time_scale == log_lin.

print_conf_ppc used if time_scale == log_lin. points per logarithmic cycle.

[print_reduced_conf_every=0] every print_reduced_conf_every steps the program will print out the reduced configurations (i.e. confs containing only the centers of mass of strands).

reduced_conf_output_dir used if print_reduced_conf_every > 0. output directory for reduced_conf files.

[log_file=stderr] file where generic and debug informations will be logged. If not specified then stderr will be used.

[print_timings=0] 0|1 if 1 the MD step timing have be printed to a file.

timings_filename used if print_timings == 1. output file where the MD step timing will be appended to.