Python is in such a weird spot right now, because with numpy and pytorch it’s fucking incredible for big number crunching problems. There is basically no reason to ever write CUDA because there are ten different python dispatched BLAS libraries out there which will skull fuck the most carefully written SIMD algorithms with like six lines of code.
But then it shits itself the moment you need any kind of threaded branching logic, and you realize it is completely cursed for anything production grade.
Python is in such a weird spot right now, because with numpy and pytorch it’s fucking incredible for big number crunching problems. There is basically no reason to ever write CUDA because there are ten different python dispatched BLAS libraries out there which will skull fuck the most carefully written SIMD algorithms with like six lines of code.
But then it shits itself the moment you need any kind of threaded branching logic, and you realize it is completely cursed for anything production grade.
That’s true, I use it for infosec scripting and puzzle challenges like code golf
If I need something more heavy duty I’d put my balls to the grindstone and finally learn rust and go! and do my best to get it right in the first go