|
Uni20
Asynchronous tensor and runtime library.
|
Files | |
| async.hpp | |
| The Async<T> container: coroutine‐safe asynchronous read/write. | |
| async_errors.hpp | |
| Exception hierarchy for async/dataflow operations. | |
| async_impl.hpp | |
Inline implementation of Async<T> helpers and task await transforms. | |
| async_node.hpp | |
| Debug metadata nodes used for async DAG visualization. | |
| async_ops.hpp | |
| Concepts and helpers for composing Async expressions. | |
| async_task.hpp | |
| Promise-neutral coroutine task ownership and the public AsyncTask type. | |
| async_task_impl.hpp | |
| Inline implementation of promise-neutral task ownership and nesting. | |
| async_task_promise.hpp | |
| Defines shared and concrete task coroutine promises. | |
| async_toys.hpp | |
| Small async helper operations used in tests and examples. | |
| async_traits.hpp | |
| Payload classification and write-through customization for asynchronous values. | |
| awaiters.hpp | |
| Utilities for awaiting on Async<T>: co_await overload, all(), try_await(). | |
| buffers.hpp | |
Awaitable capabilities for shared reads and exclusive mutable access to Async<T>. | |
| cuda_task.hpp | |
| Defines the CUDA task admission type and scheduler interface. | |
| debug_cuda_scheduler.hpp | |
| Deterministic scheduler for ordinary and CUDA task domains. | |
| debug_scheduler.hpp | |
| Coroutine scheduler for AsyncTask. | |
| epoch_context.hpp | |
| Manages one “generation” of write/read ordering in an Async<T> | |
| epoch_queue.hpp | |
| FIFO queue enforcing writer→readers→next-writer ordering. | |
| future_value.hpp | |
Write-once async wrappers (FutureValue and Defer) built on Async<T>. | |
| reverse_value.hpp | |
| Reverse-mode gradient accumulation channel built on reverse epoch queues. | |
| scheduler.hpp | |
| Defines internal task routing and the host-oriented scheduler interface. | |
| shared_storage.hpp | |
| Reference-counted optional in-place storage used by async buffers. | |
| storage_buffer.hpp | |
| Defines an in-place storage helper used by async shared storage. | |
| task_registry.hpp | |
| Selects the task-registry implementation for the current build mode. | |
| task_registry_debug.cpp | |
| task_registry_debug.hpp | |
| Declares the debug task registry used for async coroutine diagnostics. | |
| task_registry_dummy.hpp | |
| Provides a no-op task registry used when async debug tracking is disabled. | |
| task_registry_presentation.hpp | |
| Presentation-layer reports for async task-registry snapshots. | |
| task_registry_snapshot.hpp | |
| Structured async task-registry graph snapshots and diagnostics. | |
| tbb_cuda_scheduler.hpp | |
| Unified oneTBB scheduler for host and multi-device CUDA tasks. | |
| tbb_numa_scheduler.hpp | |
| Scheduler that dispatches work across NUMA-aware TbbScheduler arenas. | |
| tbb_scheduler.hpp | |
| Scheduler implementation using oneAPI oneTBB task_arena + task_group. | |
| tbb_task_submission.hpp | |
| Internal non-blocking oneTBB arena admission helper. | |
| var.hpp | |
| var_toys.hpp | |
Demonstration reverse-mode helpers built on Var<T>. | |