View Source Evision.IPCHandle.CUDA (Evision v0.2.9)

Represents a CUDA IPC handle for sharing data allocated on CUDA device between different OS processes.

• :handle, a binary-encoded cudaIpcMemHandle_t that can be loaded at the native level via memcpy or similar functions.

• :step, the number of bytes between two consecutive rows.

  When there are no gaps between successive rows, the value of step is equal to the number of columns times the size of the data type.

• :rows, the number of rows.

• :cols, the number of columns.

• :channels, the number of channels.

• :type, the data type.

• :device_id is the CUDA device ID that the memory is allocated on.