|
Uni20
Asynchronous tensor and runtime library.
|
Directories | |
| async | |
| backends | |
| blas | |
| cpu | |
| cublas | |
| ops | |
Files | |
| async.hpp | |
| Opt-in asynchronous Tensor wrappers for dense linear algebra operations. | |
| backend_selector.hpp | |
| Ordered backend selector values shared by tensor storage and linalg dispatch. | |
| contraction_axes.hpp | |
| Normalized axis metadata for fixed-rank pairwise tensor contractions. | |
| contraction_strides.hpp | |
| Stride grouping and direct-GEMM projections for tensor contraction. | |
| decomposition_precision.hpp | |
| dispatch.hpp | |
| Operation-tag backend dispatch over tensor-view operands. | |
| dispatch_diagnostics.cpp | |
| dispatch_diagnostics.hpp | |
| Opt-in structured diagnostics for kernel-dispatch walks. | |
| dispatch_error.hpp | |
| Structured kernel-dispatch failure data. | |
| dispatch_error_presentation.hpp | |
| Presentation adapter for structured kernel-dispatch failures. | |
| elementwise_functions.hpp | |
| Named backend-neutral function objects for elementwise transforms. | |
| kernel_attempt.hpp | |
| Allocation-free runtime results for kernel backend attempts. | |
| least_squares_info.hpp | |
| linalg.hpp | |
| lu_info.hpp | |
| matrix_product_shape.hpp | |
| Backend-independent dense matrix-product shape validation. | |
| operation_scalar.hpp | |
| Numeric coefficient sources shared by synchronous and asynchronous operations. | |
| operation_tags.hpp | |
| Backend-independent operation values used with normalized kernel operands. | |
| precision.hpp | |
| reduction_axes.hpp | |
| Normalized input and surviving axes for fixed-rank tensor reductions. | |
| refinement_info.hpp | |
| solve_info.hpp | |
| svd_least_squares_info.hpp | |