NvSciStream Blanket Statements

Generic statements applicable for NvSciStream interfaces.

NvSciStream blanket statements

Input parameters

• NvSciStreamPacket passed as input parameter to NvSciStreamPoolPacketInsertBuffer() or NvSciStreamPoolPacketDelete() API is valid if it is returned from a successful call to NvSciStreamPoolPacketCreate() API and has not been deleted using NvSciStreamPoolPacketDelete().
• NvSciStreamPacket passed as input parameter to NvSciStreamBlockPacketAccept() or NvSciStreamBlockElementAccept() or NvSciStreamProducerPacketPresent() or NvSciStreamConsumerPacketRelease() API is valid if it was received earlier through a NvSciStreamEventType_PacketCreate event and NvSciStreamEventType_PacketDelete event for the same NvSciStreamPacket is not yet received.
• NvSciStreamBlock passed as input parameter to an API is valid if it is returned from a successful call to any one of the block create APIs and has not yet been destroyed using NvSciStreamBlockDelete() unless otherwise stated explicitly.
• An array of NvSciSyncFence(s) passed as an input parameter to an API is valid if it is not NULL and each entry in the array holds a valid NvSciSyncFence. The NvSciSyncFence(s) in the array should be in the order according to the indices specified for the NvSciSyncObj(s) during NvSciStreamBlockSyncObject() API. That is, if the NvSciSyncObj is set with index == 1, its NvSciSyncFence should be the 2nd element in the array. The array should be empty if the synchronousOnly flag received through NvSciStreamEventType_SyncAttr from other endpoint is true.
• NvSciIpcEndpoint passed as input parameter to an API is valid if it is obtained from successful call to NvSciIpcOpenEndpoint() and has not yet been freed using NvSciIpcCloseEndpoint().
• Pointer to NvSciEventService passed as input parameter to an API is valid if the NvSciEventService instance is obtained from successful call to NvSciEventLoopServiceCreate() and has not yet been freed using NvSciEventService::Delete().
• The input parameter for which the valid value is not stated in the interface specification and not covered by any of the blanket statements, it is considered that the entire range of the parameter type is valid.

Output parameters

• Output parameters are passed by reference through pointers. Also, since a null pointer cannot be used to convey an output parameter, API functions return an error code if a null pointer is supplied for a required output parameter unless otherwise stated explicitly. Output parameter is valid only if error code returned by an API function is NvSciError_Success unless otherwise stated explicitly.

Return values

• Any initialization API that allocates memory (for example, Block creation APIs) to store information provided by the caller, will return NvSciError_InsufficientMemory if the allocation fails unless otherwise stated explicitly.
• Any API function which takes a NvSciStreamBlock as input will panic in safety build or return NvSciError_BadParameter in standard build if the NvSciStreamBlock is invalid.
• Any API function which takes a NvSciStreamPacket as input will panic in safety build or return NvSciError_BadParameter in standard build if the NvSciStreamPacket is invalid.
• Any element level interface which operates on a block other than the interfaces for block creation, NvSciStreamBlockConnect(), NvSciStreamBlockEventQuery(), and NvSciStreamBlockEventServiceSetup() will return NvSciError_StreamNotConnected error code if the producer block in the stream is not connected to every consumer block in the stream or the NvSciStreamEventType_Connected event from the block is not yet queried by the application by calling NvSciStreamBlockEventQuery() interface.
• If an API function is invoked with parameters such that two or more failure modes are possible, it is guaranteed that one of those failure modes will occur, but it is not guaranteed which.

Concurrency

• Unless otherwise stated explicitly in the interface specifications, any two or more NvSciStream functions can be called concurrently without any side effects. The effect will be as if the functions were executed sequentially. The order of execution is not guaranteed.