Adaptive Independent Component Analysis on GPU in pure Julia.
- Single CPU performance is slightly faster than the fortran implementation
- GPU is fastest
- 64 core multi-threading CPU is faster for Fortran, this difference appears at utilizing ~8 cores in our testing
We checked our implementation against the Fortran implementation from Jason Palmer. This check is also implemented as a continuous integration check for future versions.
- Float32 did not impact performance in our three tested datasets
In theory, GPU support should work on Nvidia, AMD, Apple-Metal, and Intel as well. In practice, we only could test Nvidia and Apple-Metal so far. The support hinges on KernelAbstractions.jl to support more backends. Please raise an issue if you run into problems, we can surely figure it out!
If you want to make contributions of any kind, please first that a look into our contributing guide directly on GitHub or the contributing page on the website
 Valentin Morlock 💻 📖 🐛 🚇 |
 Benedikt Ehinger 💻 🚇 |
 AlexLulkin 💻 🚇 |