Skip to content

Releases: Jutho/TensorOperations.jl

v1.0.2

16 Jan 15:01
Compare
Choose a tag to compare

improve support for 32-bit windows systems

v1.0.1

28 Dec 22:53
Compare
Choose a tag to compare
  • added some precompilation hints
  • improve support for non bitstype element type
  • more robust detection of whether a copied reshape is necessary

v1.0.0

15 Dec 01:38
Compare
Choose a tag to compare

Major overhaul with significant changes (but unchanged API):

  • More robust support for general strided arrays and custom element types
  • Package wide cache for storing temporaries and removing pressure from Julia's GC
  • Contracting with Diagonal objects
  • Expanded and improved documentation
  • Various fixes

v0.7.1

29 Sep 23:27
Compare
Choose a tag to compare

small bugfix when using ncon-style indexing

v0.7.0

04 Sep 14:08
Compare
Choose a tag to compare

Dropped support for Julia 0.7, updated and tested for compatibility with Julia 0.7 and 1.0

v0.6.1

01 Aug 13:06
Compare
Choose a tag to compare

Latest release to support julia 0.6.

Contains a rewritten implementation of the @tensor macro, which immediately generates the appropriate function calls without creating intermediate julia objects. Also various small fixes.

@tensoropt and NCon-style indexing

18 Oct 18:21
Compare
Choose a tag to compare
  • Addition of a @tensoropt macro that will optimize the contraction order of any product of tensors A[...]*B[...]*C[...]*... (see below).
  • The @tensor macro will reorganize the contraction order of the so-called NCON style of specifying indices is respected, i.e. all contracted indices are labelled by positive integers and all uncontracted indices are specified by negative integers. In that case, tensors will be contracted in such an order that indices with smaller integer label will
    be contracted first
  • Better overall type stability, both in the @tensor(opt) environment and with the function based approach. Even the simple function based syntax can now be made type stable by specifing the indices using tuples.
  • Fully compatible with Julia v0.6 and v0.7. (v0.5 no longer supported)