You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
slicing into chunks so we can do active storage on slices distributed across chunks.
The latter problem is encapsulated by imagining that an array X(10,10,10) is distributed across arbitrary hypercubes. How would we identify extract just the first row (ie X(0,,))? To do that the storage has to be given the shape (10,10,10) and the chunk selection (0,**).
This requires implementation in storage of either a buffer before application of reduction, or a reduction method that can handle streaming. The low hung fruit here would be to use a buffer.
An issue for the C-coders, is what will a Python description of shape and chunk selection look like by the time they see them? To answer that we have another issue.
The text was updated successfully, but these errors were encountered:
We need to be able to support
The latter problem is encapsulated by imagining that an array X(10,10,10) is distributed across arbitrary hypercubes. How would we identify extract just the first row (ie X(0,,))? To do that the storage has to be given the shape (10,10,10) and the chunk selection (0,**).
This requires implementation in storage of either a buffer before application of reduction, or a reduction method that can handle streaming. The low hung fruit here would be to use a buffer.
An issue for the C-coders, is what will a Python description of shape and chunk selection look like by the time they see them? To answer that we have another issue.
The text was updated successfully, but these errors were encountered: