From 00828876d1bc1b996d8bbabc9fe0567225ac522e Mon Sep 17 00:00:00 2001 From: Matt Fishman Date: Thu, 17 Dec 2020 13:32:53 -0500 Subject: [PATCH] Update NEWS.md for v0.1.31 --- NEWS.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/NEWS.md b/NEWS.md index f21c2392e2..99013469ca 100644 --- a/NEWS.md +++ b/NEWS.md @@ -1,3 +1,10 @@ +ITensors v0.1.31 Release Notes +============================== +* Update to NDTensors v0.1.20, which includes some more general block sparse slicing operations as well as optimizations for contracting scalar-like (length 1) tensors (NDTensors PR #57). +* Add flux of IndexVal functionality which returns the QN multiplied by the direction of the Index. Make `qn` consistently return the bare QN. Might be breaking for people who were calling `qn(::IndexVal)` and related functions, since now it consistently returns the QN not modified by the Index direction (PR #543). +* Introduce `splitblocks` function for Index, ITensor and MPS/MPO. This splits the QNs of the specified indices into blocks of size 1 and drops nonzero blocks, which can make certain tensors more sparse and improve efficiency. This is particularly useful for Hamiltonian MPOs. Thanks to Johannes Hauschild for pointing out this strategy (PR #540). +* Add Ising YY and ZZ gates to qubits examples (PR #539). + ITensors v0.1.30 Release Notes ============================== * Update to NDTensors v0.1.19, which includes various block sparse optimizations. The primary change is switching the block-offset storage from a sorted vector to a dictionary for O(1) lookup of the offsets. Note this may be a slightly breaking change for users that were doing block operations of block sparse tensors since now blocks have a special type Block that stores a tuple of the block location and the hash (NDTensors PR #54 and ITensors PR #538).