Abstract
This white paper presents an efficient “N buffer scheme” to manage I/O in a real-time system like an audio decoder. In general, this N buffer scheme (N > 2) can be applied to any program or a process whose average CPU load is much smaller than its peak CPU load. The 2-buffer ‘pingpong’ approach, which is a special case of “N buffer scheme”, is applicable to general programming tasks, independent of the CPU load profile. Towards the end of the paper, it is shown that the “N buffer scheme” is memory efficient as compared to a single/dual buffer solution, without a performance penalty (i.e., without affecting the CPU load). Alternatively, for a given memory size, the “N buffer scheme” puts lesser load on CPU as compared to a single/ dual buffer solution.

Author Rupesh W Kumbhare

Download this paper