|
Uni20
Asynchronous tensor and runtime library.
|
#include <uni20/backend/cublas/execution.hpp>#include <uni20/backend/cublas/cublas_error.hpp>#include <uni20/backend/cuda/buffer.hpp>#include <uni20/backend/cuda/cuda_error.hpp>#include <uni20/common/trace.hpp>#include <cuda_runtime_api.h>#include <memory>#include <new>#include <utility>#include <vector>Namespaces | |
| namespace | uni20 |
| namespace | uni20::cublas |
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. | |
| cuda::ResourceLease<HandleSlot> handle |