Uni20
Asynchronous tensor and runtime library.
Loading...
Searching...
No Matches
Files
CUDA backend integration

CUDA runtime ownership, completion, and provider integration. More...

Collaboration diagram for CUDA backend integration:

Files

file  buffer.hpp
 Typed CUDA device buffers and scoped stream access guards.
 
file  cuda_error.hpp
 Structured CUDA runtime failures and checked-call helpers.
 
file  cuda_error_presentation.hpp
 Presentation adapter for structured CUDA runtime failures.
 
file  device.hpp
 CUDA device discovery and immutable hardware capability snapshots.
 
file  initialization.hpp
 Stream-ordered diagnostic initialization of CUDA scalar storage.
 
file  resource_pool.hpp
 Fixed-capacity pools and move-only leases for CUDA provider resources.
 
file  runtime.hpp
 CUDA device, stream, resource, and scoped runtime primitives.
 
file  task_awaiters.hpp
 CUDA-specific awaiters for device and runtime resource scheduling.
 

Detailed Description

CUDA runtime ownership, completion, and provider integration.