Repository files navigation

Scalable Agent-based GPU Enabled Simulator

• Python 3.11+
• NVIDIA GPU with CUDA drivers or AMD GPU with ROCm 5.7.1+

The most straightforward way is to use conda to install SAGESim and all dependencies:

Add to your current env:

conda env update --name myenv --file sagesim_all_deps.yml --prune

or, create a new env:

conda env create --name sagesimenv --file sagesim_all_deps.yml

However, your system might require specific steps to installingmpi4py and/orcupy depending on your hardware.In that case use your systems recommended instructions to installmpi4py andcupy and execute:

pip install sagesim

• git clone https://code.ornl.gov/sagesim/sagesim
• cd /path/to/clone_repo/examples/sir
• mpiexec -n 4 python run.py

• nan checked by inequality to self. Unfortunate limitation of cupyx.
• Dicts and objects are unsupported.
• *args and **kwargs are unsupported.
• nested functions are unsupported.
• Be sure to usecupy data types and array routines in favor ofnumpy: [https://docs.cupy.dev/en/stable/reference/routines.html]
• for loops must use range iterator only. No 'for each' style loops.
• return does not seem to work well either
• break andcontinue are unsupported!
• Cannot reassign variables withinif orfor statements. Must be assigned at top level of function or new variable declared under subscope.
• -1 indexing does not necessarily work as expected, as it will access the last element of the memory block of the array instead of the logical array. Uselen(my_array) - 1 instead