v5.0.0

TensorOperations v5.0.0

Diff since v4.1.1

• Support for cuTENSOR v2 and with that, more recent version of CUDA.jl
• Further improvement and extension of the backend and allocator support, with in particular, a package extension for allocating temporaries via Bumper.jl, and a new associated @butensor macro.
• Improved AD support
• Fallback implementations for AbstractArray objects which are not strided.
• Breaking changes in the interface: argument order and argument type of the conjugation flags.
• Restricted to Julia 1.8 and higher

Merged pull requests:

• Update to cuTENSOR 2.0 (#160) (@github-actions[bot])
• Fix premature freeing of temporary tensors during reverse mode AD. (#170) (@lkdvos)
• Add naive Base julia AbstractArray implementation (#171) (@lkdvos)
• Bump julia-actions/cache from 1 to 2 (#172) (@dependabot[bot])
• Switch up argument order (#173) (@lkdvos)
• Change conj flags from Symbol to Bool (#176) (@lkdvos)
• add backend types (#181) (@Jutho)
• Alternative Allocators (#182) (@lkdvos)

Closed issues:

• Manual allocation strategy (#143)
• Freed reference problem when combining cuTENSOR and Zygote (#169)
• TensorOperationscuTENSORExt fails to compile (#174)
• Costchecks are not using GlobalRef (#178)
• Optdata cannot use Int as labeltype (#179)
• install newest version? (#180)