Expand description
This library is designed to perform tensor network contractions, using partitioning of the network as parallelization strategy. To this end, it ships multiple methods to partition a tensor network for lowest contraction cost, for example, based on simulated annealing. The partitionings can then be contracted in parallel on a distributed-memory system, as common in high-performance computing. Local contractions on a single system are also possible.
Modules§
- builders
- Methods for creating tensor networks.
- contractionpath
- Find and evaluate the cost of contraction paths. Also contains functionality for rebalancing of partitioned tensor networks.
- gates
- Implementation of various quantum gates to be used in tensor networks.
- io
- Methods for importing and exporting tensors and tensor networks.
- mpi
- Methods for MPI interaction, i.e., communication and synchronization between multiple compute nodes.
- tensornetwork
- Code concerning the tensor datastructure itself.
Macros§
- path
- Macro to create (nested) contraction paths, assuming the left tensor is replaced in each contraction.