Skip to content
- how can I lean towards octopodes as described by an hpc co-design initiative (Preparing for the future — Rethinking proxy apps)
- abstract definition of algorithms using flexible and parametrized building blocks (the so called “Octopodes”)
- example of octopode : matrix multiplication
- contacted jens domke, one of the authors
- reinforcement learning for optimizing implementation? (ex: google chips ai design)
- i could try direct approach (no building blocks) to compare to building block methods like
- train an model like monocoder to use deal II to autogenerate sim codes