https://github.com/prg-titech/sanajeh

A DSL for GPGPU programming with Python objects
https://github.com/prg-titech/sanajeh

dynasoar gpgpu parallel python smmo

Last synced: about 2 months ago
JSON representation

A DSL for GPGPU programming with Python objects

• Host: GitHub
• URL: https://github.com/prg-titech/sanajeh
• Owner: prg-titech
• Created: 2019-10-10T06:36:34.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2023-01-24T23:43:39.000Z (almost 2 years ago)
• Last Synced: 2023-03-06T12:57:17.977Z (almost 2 years ago)
• Topics: dynasoar, gpgpu, parallel, python, smmo
• Language: Python
• Homepage:
• Size: 45.6 MB
• Stars: 6
• Watchers: 5
• Forks: 2
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Sanajeh

A DSL for GPGPU programming with Python objects.

# System Requirements

* DynaSOAr's system requirements

* Python 3.7 or later

# Design

![](img/structure.PNG)

Sanajeh allows the usage of SMMO paradigm on Python. Sanajeh analyzes the class definitions and uses an analyzer to separate *.cu* and *.h* definitions. From there, these are compiled into *.so* files. The separate *main* function runs the SMMO program by calling the *.so* files through FFI.

# Compilation

The examples are defined in `/examples`. Three working examples are `nbody`, `collision`, and `wator`. 

Compilation is executed as follows. Omitting the compile option will run the program.

````

src/interface.py path/to/file.py

````

Then run the compiled program with the following command. `--render` is the rendering options. `--cpu` if the program needs to be run sequentially on Python.

````

src/inteface.py --run [--render] [--cpu] path/to/file.py

````

The compilation results are automatically put inside the directory `device_code`.