|
Uni20
Asynchronous tensor and runtime library.
|
cuBLAS handle slots and dynamically paired stream execution leases. More...
#include <uni20/backend/cuda/resource_pool.hpp>#include <uni20/backend/cuda/runtime.hpp>#include <cublas_v2.h>#include <cstddef>#include <optional>Go to the source code of this file.
Classes | |
| class | uni20::cublas::ExecutionLease |
| Move-only dynamic pairing of one cuBLAS handle slot and one idle stream. More... | |
| class | uni20::cublas::ExecutionPool |
| Device-local cuBLAS handle pool combined with a shared idle-stream pool. More... | |
| class | uni20::cublas::HandleSlot |
| One expensive reusable cuBLAS handle and its handle-local state. More... | |
Namespaces | |
| namespace | uni20 |
| namespace | uni20::cublas |
| namespace | uni20::cuda |
Functions | |
| ExecutionPool & | uni20::cublas::execution_pool (cuda::DeviceResources &resources) |
| Return the lazily constructed cuBLAS execution pool for one CUDA device. | |
| ExecutionPool & | uni20::cublas::execution_pool (cuda::DeviceResources &resources, std::size_t handle_count) |
| Configure or return the canonical cuBLAS execution pool for one device. | |
cuBLAS handle slots and dynamically paired stream execution leases.