Expand description
CUDA provider namespace.
Modules§
- cubecl
- Public tenferro-wide CubeCL session (issue #1597).
- raw
- Type-safe raw CUDA extension session (issue #1597).
Structs§
- Cuda
Backend - Cuda
Compute Capability - CUDA compute capability
major.minor(e.g. 9.0 for Hopper-class parts). - Cuda
Device Id - Provider-qualified identity of a CUDA device ordinal.
- Cuda
Device Info - Immutable metadata describing one CUDA device.
- Cuda
Device Uuid - Stable physical or MIG device identity as a 16-byte UUID.
- Cuda
Exec Session - Borrowed CUDA execution capability.
- Cuda
Runtime - CubeCL CUDA runtime wrapper.
- Cuda
Runtime Identity - Opaque identity of one exact CUDA runtime instance.
Enums§
- Cuda
Device Error - Structured failures from CUDA device discovery and initialization.
- GpuExtension
Capability - Capability a GPU extension may query before using a provider-specific path.
Functions§
- cuda_
capabilities - Return the CUDA backend operation capability descriptor table.
- cuda_
devices - Discover CUDA devices visible to the current process.
- cuda_
runtime_ engine_ registration - Build a runtime engine registration for a
CudaBackendand caller-selectedEngineId. - cuda_
runtime_ hardware_ class - Return the canonical CUDA runtime hardware class.
- download_
tensor - Download a CubeCL-managed GPU tensor back to host memory.
- gpu_
available - Returns
trueif a CUDA device can initialize a CubeCL runtime. - upload_
tensor - Upload a host tensor into a CubeCL-managed GPU allocation.
- with_
cuda_ exec_ session - Visit a CUDA execution session through the erased backend-session surface.