alt Hacker Newsyou can but through a python interpreter in the mojo process so you get the same numpy speed with mojo<->ptyhon interop overhead. NuMojo is native and also is now starting to support features that numpy can't really do like native GPU execution.
Replies
vovavili • last Tuesday at 8:31 PM
I wonder if the Python-to-Mojo overhead is significant enough where language-native bindings make more sense. I'd be interested in seeing some benchmarks.
insane_dreamer • yesterday at 7:15 PM
> features that numpy can't really do like native GPU execution
we use numpy + jax for that; works well
> numpy can't really do like native GPU execution
I'd be interested to see where GPU code beats NUMPY's SIMD implementation, which is really