Description
Dynamically loaded single-device CUDA adapter.
Structs§
- Cuda
Compute Capability - CUDA compute capability reported by the selected physical device.
- Cuda
CsrAction Evidence - Auditable evidence for one CUDA CSR action accepted against a host reference vector.
- Cuda
CsrAction Result - CUDA values plus complete action evidence.
- Cuda
CsrTransfer Evidence - Explicit matrix, input, and output transfer evidence for one CSR action.
- Cuda
Library Versions - CUDA driver, bindings, and sparse-library versions used by one action.
- Cuda
Linear Solve Evidence - Auditable device evidence surrounding the backend-neutral solve report.
- Cuda
Linear Solve Result - Independently accepted solution and CUDA-specific execution evidence.
- Cuda
Linear Solver - One CUDA device selected for CSR Krylov execution.
- Cuda
Linear Transfer Evidence - Explicit matrix, problem-vector, preconditioner, and result transfers for one CUDA linear solve.
- Cuda
Runtime - Dynamically loaded cudarc runtime adapter.
Constants§
- CUDARC_
VERSION - Exact dynamically loaded driver-binding dependency version.
- CUDA_
ADAPTER_ VERSION - Exact Eqiora CUDA adapter package version compiled into this binary.
- CUDA_
BINDING_ TOOLKIT - CUDA toolkit ABI selected for the generated driver bindings.
- CUDA_
LINEAR_ EXECUTION - Stable CUDA execution identity, separate from the solver algorithm.
- CUDA_
LINEAR_ EXECUTION_ PROVIDER - Exact declared CUDA execution release and binding versions.
- CUDA_
LINEAR_ SOLVER_ BACKEND - Stable CUDA Krylov backend identity.
- CUDA_
LINEAR_ SOLVER_ PROVIDER - Exact declared CUDA Krylov solver release.
- CUDA_
RUNTIME_ ID - Stable runtime identity even when the optional adapter is not compiled.
Functions§
- verify_
csr_ action - Execute one finalized
f64CSR action on CUDA and compare it with the independent host operator under an explicit tolerance. - verify_
csr_ action_ against - Execute one finalized
f64CSR action on CUDA and compare it with an explicitly supplied independent reference vector.