-
Notifications
You must be signed in to change notification settings - Fork 45
Open
Description
We have logic for tensoring prime transforms to prime-power transforms, and for tensoring prime-power transforms, in both Tensor.hs and RTCommon.hs.
In a personal communication, cpeikert also noted:
All the reindexing stuff (extIndices* and baseIndices*) between Zm^* and [n] has duplication. I toyed with the idea of having a special DSL for binary or ternary (0, +-1) matrices that have a single nonzero entry per row or column, which I think is general enough to capture all the two-index transformations without killing efficiency. We could even preserve the ext/baseIndices vectors if desired, but compute them in a nicer way.
Metadata
Metadata
Assignees
Labels
No labels