Module device_runtime 

Stream-ordered device memory runtime.

crate::CudaProviderBuilder constructs one XlogDeviceRuntime and one crate::GpuMemoryManager that share the provider’s exact CUDA device. The runtime owns the stream pool and composes the production resource stack:

XlogDeviceRuntime
  -> StreamPool of non-blocking streams
  -> LoggingResource (optional)
  -> GlobalDeviceBudget
  -> AsyncCudaResource

GlobalDeviceBudget is the sole byte-admission authority. Optional LoggingResource wraps it so successful operations and typed admission failures are both observable exactly once. DirectCudaResource remains available for crate tests that need a synchronous allocator; canonical providers always use AsyncCudaResource.

Stream-ordered contract: every alloc / dealloc names a stream; reuse across streams requires explicit event/sync. No reliance on the CUDA null/default stream.

Re-exports

pub use async_resource::AsyncCudaResource;

pub use budget::GlobalDeviceBudget;

pub use direct::DirectCudaResource;

pub use logging::InMemorySink;

pub use logging::LogAction;

pub use logging::LogRecord;

pub use logging::LogResult;

pub use logging::LoggingResource;

pub use logging::LoggingSink;

pub use logging::NullSink;

pub use logging::SinkError;

pub use resource::Access;

pub use resource::AllocTag;

pub use resource::AllocationAccounting;

pub use resource::AllocationRequest;

pub use resource::BlockId;

pub use resource::BlockState;

pub use resource::DeviceBlock;

pub use resource::DeviceMemoryResource;

pub use resource::Generation;

pub use resource::ResourceBudgetSnapshot;

pub use resource::ResourceError;

pub use resource::ResourceResult;

pub use resource::StreamId;

pub use runtime::ConditionalGraphStats;

pub use runtime::EventLifecycleStats;

pub use runtime::ResidentCompletionEvent;

pub use runtime::ResidentGraphHandleLifecycleStats;

pub use runtime::XlogDeviceRuntime;

pub use stream_pool::StreamPool;

pub use stream_pool::StreamPoolError;

pub use stream_pool::DEFAULT_MAX_STREAMS;

Modules

async_resource

Stream-bound allocation through the shared raw storage owner.

budget

Per-runtime budget over the canonical memory resource.

direct

Default-device-stream allocation through the shared raw storage owner.

logging

LoggingResource — telemetry decorator for any DeviceMemoryResource.

resource

Core DeviceMemoryResource trait and supporting types.

runtime

XlogDeviceRuntime hosts one provider-owned CUDA device, stream pool, and decorated memory-resource stack.

stream_pool

StreamPool — owned non-blocking CUDA streams indexed by StreamId.