Sanajeh: a Python DSL for GPGPU
In order to obtain high performance GPGPU, we usually have to code in a low-level programming language such as CUDA/C with careful memory allocation and access. We propose Sanajeh, a Python DSL that compiles object-oriented programs into GPGPU code. It is based on the single-method multiple-objects (SMMO) model, and allows dynamic object allocation inside of the GPU.
Related Projects
News
- 2 presentations at JSSST 2023 Conference
- Keynote Talk on High-Level Programming Abstractions for GPGPU at CTHPC2022
- Fathul, Arai and Nose Presents Master’s Theses
- Talk on Nested Object Support for GPGPU at ARRAY 2021(@PLDI)
- Master’s theses defense by Ogushi, Niimi, Chenxin and Luthfan
- Masuhara’s Keynote Talk on Object Support for GPGPU at APLAS’20
- Paper on Python DSL for GPGPU at JSSST 2020 Conference (Along With 5 Posters)
- Poster & Demo Presentation at PPL2020