This repo for multi devices sibernetic engine implementation. This implementation was testen on Novosibirsk State University cluster on configuration with several GPU unit from 1 to 8. GPU - NVIDIA Tesla V100. Results was reported here (in Russian), and here (in Russian) this old report.
For building you have to run command make -f makefile
it will create bin folder and create executable file
To clean all building arifacts you have to run command make clean
Usage: ./bin/sibernetic [OPTION]
--multi_dev Run with all available devices.
but default it will run only one.
-f <config_file> Path to configuration file.
-help, -h, -?, --help Print this information.
--p_sort turn on parallel sorting.
-iter_limit=<number> limit for number of iteration simulation will preferme before stop.
-time_limit=0.5 limit of time simulation in seconds (will run 0.5 seconds of simulation).
Full documentation at: <https://github.com/openworm/multi-dev-sibernetic/>
Please report any bugs/issues to: <https://github.com/openworm/multi-dev-sibernetic/issues>
Add it soon ...
Main algorithm idea was published here