Julian Samaroo
Julian Samaroo
The scheduler is quite close to being able to run multithreaded, with many parts running concurrently already (synchronized via the global scheduler lock). We'll want to make certain "non-critical" `ComputeState`...
We're already pushing some extra work onto the worker nodes (mainly argument fetching and processor selection/load balancing), and it would be beneficial for large DAGs to move more work onto...
We should implement a `select` operator, which returns the result (and probably ID) of the first task to complete out of a set of tasks. This operator should have the...
In order to make it easier to test the correctness of the scheduler, we want to log scheduler decisions, including any relevant context that the scheduler may have used to...
It's not actively tested or maintained, so we have to decide if we want to remove it, or if we want to modernize it and start testing that it works...
So that users don't have to manually do `Dagger.tochunk(x, ...)` for things like GPU arrays, FFMPEG handles, etc.
We don't appear to evict `Chunk`s from workers which don't have another task depending on their value, which is often a good thing because we keep around values that are...
While debugging #258, I found myself wanting to get a full stacktrace back to the task invoking `Dagger.spawn` when the eager listener fails to add a thunk. On error, we...