Hi all, I'm working on a really nice project in collaboration with Yu Feng. https://github.com/rainwoodman/vast In the short term, this will provide a multi-dimensional dense array implementation comparable to NumPy minus the Python runtime overhead. There's already a couple of view routines implemented and we plan to generate code, lot's of it! I have a separate project which provide additional numerical types to be used in the dense array (or anywhere else for that matter): https://github.com/arteymix/numeric-glib Then we'll address computation graph with the goal of reducing the time needed to evaluate and compute gradients when processing data with high dimensionality compared to other existing tools. This should be heavily based on GObject Introspection to conveniently inject existing code into the graph. There's some future plan to implement a Jupyter/IPython kernel for Vala and Genie. I have some drafts of using libvala to rewrite snippets AST into chained blocks and evaluate them using forked processes. It should not be too difficult considering that we already have ZeroMQ bindings and an excellent library for the language. While all these are pet projects, I am pretty optimist that we will have a nice general purpose multi-dimensional array implementation for crunching some serious numbers out of all that. Best regards, -- Guillaume Poirier-Morency <guillaumepoiriermorency gmail com> Étudiant au baccalauréat en Informatique à l'Université de Montréal Chercheur boursier à l'IRIC Mon blog: https://arteymix.github.io/ Clé PGP: B1AD6EA5
Attachment:
signature.asc
Description: This is a digitally signed message part