3 ms·
If julia could become a native vector/gpu programming language that is usable by mere mortals this could be a niche that might eventually grow into mainstream.
by streamofdigits 5y ago
If julia could become a native vector/gpu programming language that is usable by mere mortals this could be a niche that might eventually grow into mainstream. But I can't help but notice that e.g. nvidia's github mentions only: Python, C++, C, Go, Cuda. (not sure the order matters).
- eigenspace 5y agoJulia has excellent CUDA support[1], I have no idea why nvidia doesn't promote it more. It's fast, flexible and very featureful. There was a recent thread about it here: https://news.ycombinator.com/item?id=27496679 https://news.ycombinator.com/item?id=27496679 The AMD [2] and Intel [3] support is younger, but developing quickly. There's also [4] for a unified API that works across different GPU vendors to avoid lockin [1] https://cuda.juliagpu.org/dev/ https://cuda.juliagpu.org/dev/ [2] https://github.com/JuliaGPU/AMDGPU.jl https://github.com/JuliaGPU/AMDGPU.jl [3] https://github.com/JuliaGPU/oneAPI.jl https://github.com/JuliaGPU/oneAPI.jl [4] https://github.com/JuliaGPU/KernelAbstractions.jl https://github.com/JuliaGPU/KernelAbstractions.jl