DESCRIPTION

A CUDA benchmark for the Fast force-coupling method

COMPILE

Modify the /src/config.hpp file to select available options. To compile, under the home directory of this project, run

make clean x

where x is one of the given options in the makefile. The default path of the generated executables are under the repository /bin/ .

USAGE

For Imperial College Maths Department users, on the nvidia4 machine, run

nvidia-smi

to check node status, and then type

export CUDA_VISIBLE_DEVICES=x

to select the available node.

Run with

./bin/x

where x is the name of the executable binary.

PYTHON SCRIPT

A Python script is provided to automatically run sequential simulations using a single binary file. This is achieved by replacing the text in a config file which is then read by the binary file.

⚠️ The Python script is very custom written and does not run out of the box: Be very careful here!

To use that, first change the path in 'settings.py' to match your fast fcm directory path. Create the required directory for data saving.

You will need to use the random generator by compiling ''' make RANDOM_GENERATOR '''

To use the script, modify the parameters in file script.py. The member function start_loop can be modified to sweep the simulation parameters. Data generation and data reading/processing are separate process, and can be controled by system arguments passed in the terminal.

GENERATING DATA

Run simulations with

python3 script.py run