Work Graph

Implement data dependency graph on top of Async and Machines.

Pipelines should be able to analyze and schedule concurrency at compile time. Syntax can be simple.

Data fusion should also be possible, as well as cache and data locality, CPU topology, etc.