Hi Vishal,
The exact input sequence will be that your data gets latched on the input and transferred to the onboard FIFO (32MB). The MITE, which is the chip that handles communication with the PCI bus, packs the data from the FIFO (32MB) and gets it ready for transport on the PCI bus. The MITE uses algorithms based on current amount of data in the FIFO, the sampling rate etc. to pack the data efficiently for transport. This algorithm is not documented and is proprietary. The data is then read into the PC Memory from the PCI bus using DMA (or interrupts). This memory is your PC RAM and is limited to the size of the buffer you declared in your program. Once again, you won't have access to how or how quickly data is transferred from the MITE to the PC Memory (as it is proprietary).
When you perform a Read in your application environment (LabVIEW/C++...) it directly reads the data from the buffer. Once it reads the data from the buffer, those locations are available for the DMA to copy data to (circular buffer in continuous mode).
You can change whether your device uses DMA or interrupts using Set DAQ Device Info functions. This function can be read in the NI-DAQ C Function Reference (or LabVIEW help).
For output operations, if the buffer size is greater than the onboard FIFO (32MB) or if the output is set for continuous, you would use a circular buffer in PC Memory (RAM). Data would be transferred by the MITE on the 6534 from PC Memory to the onboard FIFO (32MB). As data is transferred from the PC Memory, those memory locations are now available for LabVIEW or your other programming software to write more data to the PC memory. Similarly, as data is output from the onboard FIFO of the 6534, the MITE can use those memory locations to transfer more data from the PC memory.
Errors, -10845 and -10846 describe what happens when you overrun any of the buffers (PC Memory or onboard FIFO). You can get their description in the NI-DAQ C Reference Manual as well.
Ron