ME/CS/ECE759 High Performance Computing: Final Project

Environment:

OS: Ubuntu 20.04
C++ compiler: gcc 9.4 or newer
CUDA complier: nvcc 10.1 or newer

Compile the code:

Clone the repo and go to root directory. Build the program as following:

cd FinalProject759
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make

After building the codes, you shall have executable file "CS759_project" as well as three share libraries built in the folder "libcfd.so", "libcfd_omp.so", and "libcfd_cuda.so". Here are the explanations for these files.

CS759_project: main entrance of the program
libcfd.so: subroutine for sequential computation
libcfd_omp.so: subroutine for OpenMP
libcfd_cuda.so: subroutine for CUDA

Argument for the program

Runing the program by following argument

./CS759_project arg1 arg2 arg3 arg4 arg5

where:
arg1: size of fluid field
arg2: time step in second
arg3: number of threads for OpenMP
arg4: size of threads block for CUDA
arg5: running mode (0 - all, 1 - no parallel, 2 - OpenMp, 3 - CUDA)

Example

# running non-parallel, OpenMP (10 threads), and CUDA (16*16) threads block
# for a 32*32 1-meter grid at dt=0.01s simulation
./CS759_project 32 0.01 10 16 0