marinbenc
dynomightI don't like NumPy https://dynomight.net/numpy/
Pierre ~/
@dynomight I just stumbled upon your article from python's weekly newsletter (issue 700) and I must say it was a pleasing read with meaningful points which I had not even considered up until now. Thanks for the nice read 👍
Redish Lab@dynomight Thank you for this!
In my experience, it is also nearly impossible to make numpy efficient over variables (because it keeps having to shift the variables back to Python, so it is always creating large intermediate variables), particularly when trying to do multi-stage operations.
Benedikt Ehinger@BenediktEhinger My impression is that Julia is much better for the "run loops on CPU" type scenarios. But is it really easier when using a GPU? I know there are a million experimental julia packages that try to do incredible things, but it's very hard to know how smoothly things really work in practice.