Code Improvements

[jturney]

Improvements

The following are improvements to Ambit that were brought up during the Tensor discussion at Emory with Psi4 developers:

• [ ] Batching blocks/slices to/from disk
• [ ] Memory manager
• [ ] Improved algorithm for contraction chain
• [ ] Permutational symmetry
• [ ] Point group symmetry
• [ ] Finish DiskTensor and ensure compatibility with CoreTensor.

Tests to conduct

• [ ] With current distributed infrastructure, how does CTF perform. We fully acknowledge that we do not utilize CTF fully.
• [ ] I have a GlobalArrays backend for Ambit that needs to be tested and benchmarked.

[dgasmith]

As a note id really like to separate symmetry and contraction chains from the underlying tensor class. Derived classes can be as complicated as needed, but there should be a base class that is as bare bones as possible.

[jturney]

That is already the design of the base class. Rob and I wanted the base class to be as minimal as possible. The contraction chains are handled by expression templates. Symmetry will likely follow the BlockedTensor class style.

[kannon92]

Hey @jturney, I am working at PNNL for the next few months. I will be planning on implementing DSRG with global arrays. I would be more than willing to test GlobalArrays for Ambit. Can you push those changes to Ambit? Or send me the code?

[jturney]

@kannon92 Sure thing! I'll locate it and send it your way.