Package-level declarations
Types
The base for all top-level SpatialML handles.
Global tensor means they do not depend on certain Pipeline, so they can be created and deleted outside the lifecycle of any Pipelines. Instead, they are bound to the framework session. Additionally, they cannot be directly accessed from Pipelines. A Pipeline tries to access a GlobalTensor must first create a PipelineTensorPlaceholder with the same initialization config inside itself, and during each pipeline submission, map the PipelineTensorPlaceholder to the GlobalTensor.
SpatialML pipeline.
A PipelineTensor is a tensor that is bound to a Pipeline. It has the same lifecycle as the Pipeline it is subordinate to. Depending on whether it has local memory, it can categorized into two subclasses: PipelineTensorLocal or PipelineTensorPlaceholder. The former has local memory allocated to it, while the latter has no underlying memory, and works as a run-time reference.
Pipeline tensor with local memory.
The pipeline tensor with no local storage, but only a placeholder to refer to some global tensor. When a pipeline is submitted for execution, the PipelineTensorPlaceholder inside it will be null unless it is mapped to a GlobalTensor of exactly the same config. In that case, any reads from the placeholder will become reads from the mapped GlobalTensor, and writes to it will thus become writes to the mapped GlobalTensor. We have an internal synchronization mechanism to schedule the pipeline executions, so that if a GlobalTensor is being written to, there will no other running pipelines trying to access (read/write) the same GlobalTensor.
A wrapper of the pipeline tensor, with a slice onto this tensor. The slice must be a tensor of SLICE_ARRAY usage.
Specifying the component and the field to be updated in Pipeline.updateSceneGraphProperty.
The exception that SpatialML could throw.
The interface of SpatialML instance. An SpatialMLInstance must be created before any SpatialML APIs or commands are executed.
A handle to each session of SpatialML framework usage. For each session, a dedicated SpatialML container for securely rendering mixed-reality components will be allocated.
Root class of all tensors (global and pipeline).