accelerator command

Syntax:

accelerator style args 

• style = off or on or cuda
• args = 0 or more args specific to the style

    off args = none
    on args = none
    cuda args = to be determined 
  

Examples:

accelerator off
accelerator on
accelerator cuda blah 

Description:

Alter settings for use of accelerated versions of various styles. LAMMPS can be built with optional packages which provide accelerated versions of specific atom, pair, fix, compute, and integrate styles.

These are the relevant packages:

• OPT = a handful of pair styles, cache-optimized for faster CPU performance
• GPU = a handful of pair styles and the PPPM kspace_style, optimized to run on one or more GPUs or multicore CPU/GPU nodes
• USER-CUDA = a collection of atom, pair, fix, compute, and intergrate styles, optimized to run on one or more NVIDIA GPUs

See this section of the manual for instructions on how to build LAMMPS with any of these packages.

These styles can be specified explicitly in your input script, e.g. pair_style lj/cut/gpu. If the -accelerator command-line switch is used, you do not need to modify your input script, as discussed in this section of the manual. The command-line suffix (opt,gpu,cuda) is automatically appended when the style is created for atom, pair, fix, compute, and integrate styles. If an accelerated version does not exist, the standard version is created.

If the -accelerator command-line switch is used, you may wish to disable it for one or more input script commands, so that the standard version of the style is used instead of the accelerated one. This can be useful for performance testing or debugging.

The off style allows you to do this. The effect of the -accelerator command-line switch is effectively turned off until another accelerator command is used with the on style.

The cuda style invokes options associated with the use of the USER-CUDA package. These will be described when the USER-CUDA package is released with LAMMPS.

Restrictions:

This cuda style can only be invoked if LAMMPS was built with the USER-CUDA package. See the Making LAMMPS section for more info.

Obviously, you must have GPU hardware and associated software to build and use LAMMPS with either the GPU or USER_CUDA packages.

Related commands:

fix gpu

Default: none