baracuda-cutensor-sys
Raw FFI bindings + dynamic loader for NVIDIA cuTENSOR — high-performance tensor primitives (contraction, reduction, elementwise, permutation) with arbitrary index permutations.
Symbols resolve lazily via libloading;
no link-time dependency on libcutensor.so / cutensor64_*.dll.
Most users want baracuda-cutensor — that crate exposes typed
Handle, TensorDescriptor, OperationDescriptor, ComputeDescriptor,
PlanPreference, and Plan builders, plus the full op catalog
(Contraction, Reduction, ElementwiseBinary, ElementwiseTrinary,
Permutation, BlockSparseContraction, TrinaryContraction).
What's exposed
- All
cutensor*Descriptor_ttypes. - Plan creation + execution.
- Op functions:
cutensorContract,cutensorReduce,cutensorElementwiseBinaryExecute,cutensorElementwiseTrinaryExecute,cutensorPermute. - Plan-cache I/O.
Part of the baracuda workspace.
License
Dual MIT / Apache-2.0.