Simple adapters are totally dependent on support from the main processor or another master. This approach to adapter design was common in single task systems, when the processor was dedicated to one adapter function at a time. In such cases, the most economic approach is to use the minimum of logic on the adapter card, and depend on software support. Adapters, which use these techniques, are:
With polled adapters, the processor must periodically request information, or poll, the adapter registers for status of the attached device. This can consume much of the processor's power.
With these adapters, the processor is free to perform other duties until an interrupt is received from the adapter. The overhead of servicing the interrupt is several hundred instructions to save and restore the environment.
Devices with high data transfer rates (for example, fast communication adapters) cause high interrupt rates and the overhead could then use up much of the available processor power. This is even more of an issue in a multitasking environment where the situation is more complex and the overhead for interrupt processing is typically higher.
This type of adapter shares a segment of memory (on the adapter) and if combined with the use of interrupts, can be much more efficient than the interrupt per character type. However, the adapter is still dependent on the processor to move data between the adapter's storage and a dynamically allocated buffer in system memory. With the arbitration overhead and large amounts of data (for example, bitmaps for a graphics adapter), this may be a burden on the processor.
The biggest disadvantage with this type of adapter is that there is a limited
number of fixed assignments for shared I/O memory. This limits the number
of possible configurations and designs.