|
Uni20
Asynchronous tensor and runtime library.
|
Files | |
| access.hpp | |
| RAII leases that resolve device descriptors and tensor views to mdspans. | |
| async.hpp | |
| Async aliases for lazy tensor-level views. | |
| basic_tensor.hpp | |
| Owning tensor implementation and configurable extents alias. | |
| concepts.hpp | |
| Concepts for tensor-level objects that resolve mdspan kernel operands. | |
| conjugate.hpp | |
| Read-only lazy conjugation views for tensor-level objects. | |
| conjugate_inplace.hpp | |
| Backend-dispatched in-place conjugation of tensor elements. | |
| copy.hpp | |
| Backend-dispatched tensor copies and inferred owning materialization. | |
| copy_into.hpp | |
| Backend-dispatched copies into existing mdspan and tensor outputs. | |
| cuda_access.hpp | |
| CUDA data-handle acquisition for buffer descriptors and tensor views. | |
| generated.hpp | |
| Compact read-only tensors whose values are generated from their indices. | |
| layout.hpp | |
| mdspec_tensor_view.hpp | |
| Concrete tensor-level views over owned mdspan or mdspec metadata. | |
| output.hpp | |
| Shape preparation contracts for tensor operation outputs. | |
| precision.hpp | |
| reductions.hpp | |
| Tensor sums, inner products, and Euclidean norms with explicit result residency. | |
| reshape.hpp | |
| Explicit aliasing, in-place, and owning tensor reshape operations. | |
| shape.hpp | |
| Checked helpers for constructing runtime tensor extents. | |
| tensor.hpp | |
| General-purpose owning tensors with runtime extents. | |
| transform.hpp | |
| Backend-dispatched elementwise overwrite and update operations. | |