Interrupt Driven I/O

The main disadvantage of programmed I/O as can be noticed is that the CPU is busy waiting for an I/O opportunity and as a result remain tied up for that I/O operation. This disadvantage can be overcome by means of interrupt driven I/O. In Programmed I/O CPU itself checks for an I/O opportunity but in case of interrupt driven I/O the I/O controller interrupts the execution of CPU when ever and I/O operation is required for the computation of the required I/O operation. This way the CPU can perform other computation and interrupted to perform and interrupt service routine only when an I/O operation is required, which is quite an optimal technique.

DMA driven I/O

In case data is needed to transferred from main memory to I/O port this can be done using CPU which will consume 2 bus cycles for a single word, one bus cycle from memory to CPU and other from CPU to I/O port in case of output and the vice versa in case of input. In case no computation on data is required CPU can be bypassed and another device DMA (direct memory access) controller can be used. Its possible to transfer a data word directly from memory to CPU and vice versa in a single bus cycle using the DMA, this technique is definitely faster.

We shall start our discussion with the study of interrupt and the techniques used to program them. We will discuss other methods of I/O as required.
What are interrupts?
Literally to interrupt means to break the continuity of some on going task. When we talk of computer interrupt we mean exactly the same in terms of the processor. When an interrupt occurs the continuity of the processor is broken and the execution branches to an interrupt service routine. This interrupt service routine is a set of instruction carried out by the CPU to perform or initiate an I/O operation generally. When the routine is over the execution of the CPU returns to the point of interruption and continues with the on going process.

Interrupts can be of two types

1. Hardware interrupts
2. Software interrupts

Only difference between them is the method by which they are invoked. Software interrupts are invoked by means of some software instruction or statement and hardware interrupt is invoked by means of some hardware controller generally.

Interrupt Mechanism

Interrupts are quite similar to procedures or function because it is also another form temporary execution transfer, but there some differences as well. Note that when procedures are invoked by there names which represents their addresses is specified whereas in case of interrupts their number is specified. This number can be any 8 bit value which certainly is not its address. So the first question is what is the significance of this number? Another thing should also be noticed that procedures are part of the program but the interrupts invoked in the program are no where declared in the program. So the next question is where do these interrupts reside in memory and if they reside in memory then what would be the address of the interrupt?
Firstly lets see where do interrupts reside. Interrupts certainly reside somewhere in memory, the interrupts supported by the operating system resides in kernel which you already know is the core part of the operating system. In case of DOS the kernel is io.sys which loads in memory at boot time and in case of windows the kernel is kernel32.dll or kernel.dll. these files contain most of the I/O routines and are loaded as required. The interrupts supported by the ROM BIOS are loaded in ROM part of the main memory which usually starts at the address F000:0000H. Moreover it is possible that some device drivers have been installed these device drivers may provide some I/O routines so when the system boots these I/O routines get memory resident at interrupt service routines. So these are the three possibilities.
Secondly a program at compile time does not know the exact address where the interrupt service routine will be residing in memory so the loader cannot assign addresses for interrupt invocations. When a device driver loads in memory it places the address of the services provided by itself in the interrupt vector table. Interrupt Vector Table (IVT) in short is a 1024 bytes sized table which can hold 256 far addresses as each far address occupies 4 bytes. So its possible to store the addresses of 256 interrupts hence there are a maximum of 256 interrupt in a standard PC. The interrupt number is used as an index into the table to get the address of the interrupt service routine.