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:

• [x] 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).