Julian Samaroo
Julian Samaroo
To eliminate code like https://github.com/SciML/DiffEqGPU.jl/blob/6c1886bf4e62a6293490d6c333a1302045a744d0/src/DiffEqGPU.jl#L369-L370, it would be nice if we could select the backend for the user automatically based on the input arguments provided. I propose an `Auto()` backend...
With ROCKernels being merged into KernelAbstractions, we should soon be able to integrate AMDGPU support into this package! @ChrisRackauckas if this is desired, how would you like me to do...
Replaces #129 TODO: - [x] Wire up chunked reading to `loadtable` - [x] Split blocks across multiple workers - [x] Don't scale block size by file size - [x] Write...
Closes #151
We could provide an option to let users define an error handler function which will handle all kinds of task failures (which may necessitate propagating more richly-annotated information on error...
We should propagate the RNG state (just like Base tasks do right now) across tasks. We should also implement a scheduler option to set an RNG that we'll serialize and...
@DrChainsaw @shashi we should add FileTrees reverse CI to Dagger to prevent me from breaking things :smile:
We currently ignore the costs of transferring non-`Chunk` task inputs, but that's silly; they have a real transfer cost, but it's incurred immediately at task transfer to the worker. We...
This PR allows the scheduler to execute itself on all workers in the cluster. We first expand the notion of "thunk ID" to be per-worker, so that we can locally...