CUDA Fortran code to solve Lattice Boltzmann equations of a
multi-component flow interacting with rigid spherical particles.
Fabio Bonaccorso IIT-CLNS,Rome Italy
Marco Lauricella IAC-CNR, Rome Italy
Andrea Montessori IAC-CNR, Rome Italy
Adriano Tiribocchi IAC-CNR, Rome Italy
with contributions from:
Giorgio Amati CINECA-CED, Rome Italy
Massimo Bernaschi IAC-CNR, Rome Italy
Sauro Succi IAC-CNR, Rome Italy
The software development process has received funding from the
European Research Council under the Horizon 2020 Programme
Grant Agreement n. 739964 (COPMAT). Marco Lauricella acknowledges
the project “3D-Phys" (PRIN 2017PHRM8X) for support from MIUR.
This is an experimental code. The authors accept no responsibility
for the performance of the code or for the correctness of the results.
The code is licensed under the 3-Clause BSD License (BSD-3-Clause).
Structure
LBcuda is supplied as a main UNIX directory with 1 subdirectory.
All the source code files are contained in the main directory.
The 'Tests' sub-directory contains different test cases that can
help the user to edit new input files. Further the 'Tests'
contains the input files to run the test cases reported in Table 2 and
sub-directory 3 of the main article 'LBcuda: a high-performance CUDA
port of LBsoft for simulation of colloidal systems'.
Compiling LBcuda
The main directory stores a UNIX makefile that assembles the
executable versions of the code both in GPU and multi-GPUs version.
Note the makefile could be eventually modified for several common
workstations and NVIDIA devices into the main directory,
where the code is compiled and linked. Finally, the binary
executable file can be run in the main directory or copy in,
an external location which is intended to be the working directory
from which jobs are submitted for execution and the data files
manipulated.
On Windows system we advice the user to compile LBcuda under the
Windows native Ubuntu shell. Note that the NVIDIA HPC Software
Development Kit (SDK) is necessary to compile and run LBcuda on
the GPU and multi-GPUs parallel mode.
Executing LBcuda
To run LBcuda, it is necessary first to ensure that the program is
compiled and that the file input.dat and eventually the input.xyz
are present in the main directory.
All output data files will be returned to the main directory.
Remember that the input file HAS TO BE NAMED 'input.dat' and put in
the main directory. Few example input files are contained in
the 'Tests' sub-directory which can be used as test cases.
Example command to run LBcuda in serial mode:
mpirun -np 1 ./LBCUDA
Example command to run LBcuda in parallel mode on 8 GPUs:
mpirun -np 8 ./LBCUDA