v0.9.6

New features and improvements:

• Improved support for complex-valued calculations (TSQR algorithm)
• Julia binding for basic data types and for the TSQR algorithm
• detect CPU cache sizes using likwid
• TT-AMEn+ALS variant for solving linear systems in TT format
• More micro benchmarks and benchmarking results
• improved performance of various building blocks

v0.9.5

New features and improvements:

• Tensor-Train operators (aka a matrix where a normal tensor-train is interpreted as a vector)
• Solvers for linear systems (TT-GMRES, TT-MALS, TT-AMEn)
• ITensor interface
• support for transforming to the QTT (quantized tensor-train format)
• more Multivector operations
• improved performance of various building blocks
  • Q-less tall-skinny QR
  • optimized tensor-train addition
  • avoid copying data through view interfaces and C++ move variants for folding/unfolding 2d into a 3d tensors and vice versa
  • optimized contraction sequences for tensor-train-operator times dense tensor

v0.9.1

Bug-fixes and minor performance improvements and lots of updated benchmarking results/scripts...

v0.9.0

Initial Release

Features:

• C++ header-only classes for tensor-train arithmetic (for multi-core CPUs using OpenMP)
• Fast hybrid-parallel TT-SVD algorithm based on a Q-less TSQR (for multi-core CPU clusters using OpenMP+MPI)
• Simple quantum computer simulator for generic two-qubit quantum gates
• Python bindings compatible with numpy
• Examples for tensor-train solvers for linear system (GMRES) and eigenvalue problems (Jacobi Davidson)