Query modes enable the purposeful selection of data subsets for data localities. More...
#include <idistributed_data_locality.h>
Inherits mi::base::Interface_declare< 0x841224fe, ... >.
Public Types | |
enum | Query_mode_flags { } |
Query methods apply to a distributed dataset that is referred to by an unique identifier. More... | |
Public Member Functions | |
virtual const mi::neuraylib::Tag_struct & | get_distributed_data_tag () const =0 |
Query methods apply to a distributed dataset that is refered to by an unique identifier. More... | |
virtual mi::Uint32 | get_query_mode_flags () const =0 |
Should return a combination of Query_mode_flags to indicate required behavior to NVIDIA IndeX related to, for example, how to handle currently unavailable data subsets. More... | |
Query modes enable the purposeful selection of data subsets for data localities.
NVIDIA IndeX manages the assignments of data subsets bound inside a 3D bounding box to hosts and GPUs in the cluster environments. A distributed data job (see IDistributed_data_job
) runs against data subsets (see IData_subset
). The job may be invoked for the entire set of data subsets that belong to a distributed dataset (see IDistributed_data
) or invoked for a subset of data subsets. The interface class IDistributed_data_locality_query_mode
enables an application to specify the set of data subset that shall be considered for distributed processing and analysis. Derived classes of the interface class implement different query methods. For instance, a the data subset selection for job execution commonly relies on the spatial selection, i.e., a region of interest, in 3D space.
When implementing an IDistributed_data_job
, the method IDistributed_data_job::get_scheduling_mode()
has to expose a query method so that the internal NVIDIA IndeX infrastructure can schedule the job to the user-specified subset data selection.
|
pure virtual |
Query methods apply to a distributed dataset that is refered to by an unique identifier.
Each query method applies to a single instance of a distributed dataset. The method get_distributed_data_tag is expected to identify the distributed dataset by means of the unique identifier mi::neuraylib::Tag_struct
. Each derived class is expected to provide the identifier respectively.
mi::neuraylib::Tag_struct
that uniquely identifiers a distribute dataset in the scene description or DiCE's distributed data store.
|
pure virtual |
Should return a combination of Query_mode_flags
to indicate required behavior to NVIDIA IndeX related to, for example, how to handle currently unavailable data subsets.
Query_mode_flags
.