The transfer of data using buffers and interrupts
Introduction - speed mismatch
When two devices working at different speeds try to communicate, they have to do so at the speed of the slowest device. This is not good because the CPU gets tied up managing the transfer of a constant stream of data to a printer, for example. That means it can't work on other tasks that may be more urgent than mere printing! However, by using a 'buffer', the problem of working at the speed of the slowest device and slowing the CPU down can largely be overcome. A buffer is simply some memory that improves the efficiency of data transfer between two devices working at different speeds by allowing big blocks of data to be collected together and then sent at once rather than as a stream of data that needs constant CPU management time. A buffer can also be an external piece of memory designed specifically for the purpose of collecting data from a CPU and then taking over the management of the transfer between itself and the device it is connected to.
An example of the use of a buffer
All devices such as printers come with their own buffers and you can usually upgrade the amount of buffer memory you have in the printer. The bigger the buffer in a printer, the more data it can accept before it tells the computer that is sending the data to stop for a minute because it is full and needs to empty itself (by printing). Large buffers are particularly important on networked printers such as those commonly found in schools and colleges. A lot of people will potentially be sending a lot of work at the same time. This could result in the system grinding to a halt, as the server tries to manage all of the jobs being sent to the printer. If, however, the jobs could be sent to the printer and stored there, the server would be free to get on with other things and the printer could print at its leisure. Consider sending a file to a printer. The file is held in memory (RAM). We know that RAM works very quickly and we know that all data in and out of the CPU goes via the ALU - that is one of the functions of the ALU. We also know that the printer churns out pages much slower than the speed a CPU can work at!
Assume there isn't a buffer for a moment.
The CPU will take charge of the data transfer. The data goes from the RAM, via the ALU and out to the printer. The printer works very slowly compared to the CPU, however. If you send data constantly, byte by byte, the CPU will have to slow down how fast it sends the data to the printer so that the printer can receive the data. This means, however, that the CPU is not being used as efficiently as possible.
If you provide a buffer, however, data transfer can be improved.
The transfer of data with a buffer.
When the CPU wants to send a file, all of the data can be transferred to the buffer in one go. The buffer works at the speed of the CPU so data transfer between the CPU and the buffer is very fast. The CPU is now free to do other tasks while the buffer can transfer data to the printer. This immediately improves the transfer of data because the CPU is not waiting around, trying to constantly send data. It can send data in one, high-speed burst and then get on with something else!
Buffers and interrupts
A complication can arise, however, if you need to send a file to a printer but the whole file won't fit into the buffer because the file is too big for it! To deal with this situation, we need to use interrupts. These are simply signals sent to the CPU. They tell the CPU to stop what it is doing and give some time to whoever sent the interrupt. Remember, the CPU, whilst very fast, can only actually carry out one job at a time!
The CPU, via the ALU, fills the buffer with as much of the file as it can. When the buffer is full, it sends an interrupt to the CPU to say 'Stop'. The buffer then sends the data to the printer. The CPU, meanwhile, is free to get on with another job. When the buffer is nearly empty, it sends an interrupt signal to the CPU, to say 'More data, please'. When the CPU gets this interrupt, it checks its priority. The CPU has to decide if the job it is currently doing is more important than filling up the buffer! If it is, then it carries on with what it is doing until it is finished, and then it services the buffer interrupt. If it isn't more important, then it stops doing what it is doing and 'pushes' the contents of its registers onto the stack, thereby saving where it was before the interrupt happened! The CPU then sends the buffer some more of the file until it gets the 'Stop' signal again. The buffer resumes sending data to the printer whilst the CPU 'pops' the contents of the stack back into its registers and continues from where it left off before the interruption. This process is repeated until the whole file has been sent!
The transfer of data between two devices that work at different speeds can be made far more efficient by using a buffer because we can send data in large blocks rather than in lots of small quantities. We have looked in some detail at how you would transfer a file from memory to a printer using a buffer designed for the purpose, but we could equally have used the example of transferring data from memory to a secondary storage device such as a hard disk, for example.
