drstojanovic / WatershedAnalysis

Watershed analysis algorithm implementation in CUDA and OpenACC

WatershedAnalysis

Watershed analysis algorithm implementation in CUDA and OpenACC

Build Docker image

nvidia-docker build --tag=cudawatershed .

Run NVIDIA Docker image

nvidia-docker run -it --privileged cudawatershed

Remove watch_dog flagto enable reading CPU status

echo '0' > /proc/sys/kernel/nmi_watchdog

Run sequential (single core) OpenACC implementation

./acc/acc_singlecore.exe ./TestData/dem1.tif dem1_acc_result.tif

Run OpenACC implementation on multi-core CPU

./acc/acc_multicore.exe ./TestData/dem1.tif dem1_acc_result.tif

Run OpenACC implementation on many-core GPU

./acc/acc_gpu_35.exe ./TestData/dem1.tif dem1_acc_result.tif

Run sequential (single core) implementation

./cuda_10.exe ./TestData/dem1.tif dem1_cuda_result.tif cpu_st ftm 0.01

Run OpenMP implementation on multi-core CPU

./cuda_10.exe ./TestData/dem1.tif dem1_cuda_result.tif cpu_mt ftm 0.01

Run CUDA implementation on many-core GPU

./cuda_10.exe ./TestData/dem1.tif dem1_cuda_result.tif gpu ftm 0.01

Stop and remove Docker image

nvidia-docker rmi -f cudawatershed