Conceptual
Login

Polling, Interrupts, and Direct Memory Access

The core principle governing computer architecture I/O operations is the abstraction hierarchy where high-level semantic statements (e.g., print/scan) map to low-level hardware interactions managed by system software layers including drivers, file systems, and kernel interrupt handlers. The domain addresses how peripheral devices are logically addressed via memory-mapped or separate address spaces using port registers, governed by timing characteristics such as latency for seek operations and throughput rates varying from input/output bound (e.g., keyboard) to bandwidth-constrained bulk transfers (e.g., disk). This mechanism resolves the speed mismatch between high-frequency processors and slower peripherals through three primary synchronization models: polling, synchronous interrupt-driven I/O with vectored/non-vectored priorities, and asynchronous Direct Memory Access (DMA), each balancing CPU overhead against data transfer efficiency while addressing cache coherence constraints.

Questions this Concept answers

  • Why does polling waste processor time compared with an interrupt-driven transfer?