mGPUKDE_tiled

Multi-GPU-parallel and tile-based kernel density estimation for large-scale spatial point pattern analysis

Citation

Zhang, G., & Xu, J. (2023). Multi-GPU-parallel and tile-based kernel density estimation for large-scale spatial point pattern analysis. ISPRS International Journal of Geo-Information, 12(2), 31. https://doi.org/10.3390/ijgi12020031

Dependencies

1. CUDA 11.7: https://developer.nvidia.com/cuda-11-7-0-download-archive.
2. GDAL: https://gdal.org/index.html.

IDE

Visual Studio 2022.

How to configure the project before building in Visual Studio 2022 (Community Version)?

Step 1. Open the solution by opening the .sln file

Step 2. Edit project properties as below:

Project --> Properties --> Configuration Properties

--> VC++ Directories: edit "Include Directories" and "Library Directories" to point to the GDAL include and lib folders, respectively.

--> CUDA C/C++: configure "CUDA Toolkit Custom Dir"

How to run the program?

A complied mGPUKDE_tiled.exe file is provided in the x64\Debug folder. Below are two example commands to run the program:

1. Run in mode 0 (data are generated on the fly):

mGPUKDE_tiled.exe 0 100 1 0.1 0 1 0 1 0 12 2 density_cpu.tif density_gpu.tif 0

Arguments explanation: mGPUKDE_tiled.exe 0 #points cellsize_edge_correction cellsize_density h_option enable_edge_corection enable_sample_weight skip_omp_parallel skip_gpu_parallel num_cpu_threads num_gpu denfn_seq, denfn_cuda serialized_mode

2. Run in mode 1 (data are read from files):

mGPUKDE_tiled.exe 1 pntsRedwood.csv redwood_edgecorrection.tif redwood_densityestimation.tif 2 1 0 1 0 12 2 density_cpu.tif density_gpu.tif 0

Arguments explanation: mGPUKDE_tiled.exe 1 points_file mask_file_edge_correction mask_file_density h_option enable_edge_corection enable_sample_weight skip_omp_parallel skip_gpu_parallel num_cpu_threads num_gpu denfn_seq, denfn_cuda serialized_mode

Notes: serialized_mode should always be 0 (false)