jiankangren / pyclipsm

PyCLiPSM - Harnessing Heterogeneous Computing Resources on CPUs and GPUs for Accelerated Digital Soil Mapping

PyCLiPSM

Set up computing environment:

1. Install GPU drivers. If you are using NVIDIA GPUs, OpenCL support is included in the driver (https://developer.nvidia.com/opencl). If you are using AMD GPU/CPU, install appropriate OpenCL drivers from ADM (This link provides helpful pointers microsoft/LightGBM#1342). If you are using Intel GPU/CPU, install appropriate OpenCL drivers from Intel (e.g., https://software.intel.com/en-us/articles/opencl-drivers).
2. It's assumed that you already have Python (version 2.7) installed. Anaconda is recommended for installing Python https://www.anaconda.com/distribution/.
3. Once OpenCL drivers and Python are properly installed, you can use pip to install the PyOpenCL package (https://pypi.org/project/pyopencl/).
4. Run pyopencl_test.py to test if PyOpenCL is working properly: python pyopencl_test.py.

Run PyCLiPSM with sample data:

1. Run pyopencl_test.py to see a list of available OpenCL computing platforms/devices on your computer: python pyopencl_test.py
2. Change configurations in the OPENCL_CONFIG variable in utility/config.py accordingly, as well as the OPENCL_CONFIG variable in PyCLiPSM_main.py.
3. Run PyCLiPSM_main.py to get a sense of how to use PyCLiPSM (using example data provided): python PyCLiPSM_main.py

Use PyCLiPSM for your own application:

1. Prepare soil sample data and environmental covariate data following the example data in the "data" directory
2. Change parameters (e.g., data directory, data file names, etc.) in PyCLiPSM_main.py accordingly
3. Run PyCLiPSM_main.py: python PyCLiPSM_main.py

Contact

[email protected]