Low-level Tensor Interface I

[devinamatthews]

This issue collects the status of the various discussion topics for defining a tensor interface that is:

• low-level
• dense
• shared-memory
• synchronous
• preferably CPU/GPU portable

As sub-issues are resolved, summaries will be added here.

Sub-issues:

• Terminology (Tensor terminology #1)
• Tensor structure
  • Stride requirements/restrictions
  • Standard layouts
  • Alignment
  • Scalars as tensors
• Naming
  • Arguments
  • Functions
• Types
  • Integral arguments
  • Scalars
• Operations
  • Classification
  • Definition
  • Naming
  • Batching
• Threading
• Algorithm selection

[devinamatthews]

FYI I plan to use the checkboxes to record progress of the discussion, i.e. check off a sub-issue when it is resolved.

Many of these probably merit actual Github issues (e.g. #1), but perhaps some do not and can be briefly discussed here.

[devinamatthews]

I suggest that people add proposals, in the spirit of Paul's proposal, for a low-level (C) tensor contraction interface. From there we can use these concrete examples to address the issues listed above (and others as they arise).