A bus master is any device that can initiate a transfer across the bus. Allowing devices other than the CPU to be bus masters frees CPU cycles to perform other tasks. It also allows a form of parallelism in which peripheral to peripheral data transfers can occur simultaneously with instruction execution.

3. What characteristics of the CPU and of the system bus should be balanced to obtain maximum system performance?

The width of the data bus should equal or exceed CPU word size. The bus clock rate should match the CPU clock rate, though this is difficult or impossible to achieve.

4. What is an interrupt? How is an interrupt generated? How is it processed?

Interrupts are numeric codes and can be generated by peripheral devices, an explicit software instruction, or the CPU itself. Peripheral device interrupts are sent over the system bus, detected by the CPU, and placed in an interrupt register. Software and CPU-generated interrupts are placed directly in the interrupt register by the CPU. The CPU detects an interrupt by checking the contents of the interrupt register after each execution cycle. If an interrupt is present the CPU suspends the current program by performing a PUSH operation. It then branches to the supervisor, which looks up the interrupt code in a table and calls the interrupt handler address it finds there. When the interrupt handler returns to the supervisor a POP operation is performed, which activates the program that was executing when the interrupt was detected.

A physical access describes the storage location to be read or written in terms of the physical organization of storage locations (e.g., track, sector, and head for a disk drive). A logical access assumes that the location to be read or written is contained within a linear address space. Thus, the location is described by a single unsigned integer.

8. What functions does a device controller perform?

A device controller translates logical accesses into physical accesses and translates messages between the bus protocol and the protocol(s) used to control attached devices. A device controller may also perform multiplexing (i.e., allowing multiple devices to share a single bus port).

Differences in control language are resolved by a device controller as it translates bus messages into device messages and vice versa. Differences in physical organization (i.e., deviations from the assumption of a linear address space) are resolved by a device controller as it translates logical accesses into physical accesses. Differences in communication speed and/or unit of data transfer are resolved by using a buffer located within the device or device controller.

A buffer is an area of RAM implemented within a device controller, I/O device, or storage device. It is used to resolve differences in I/O unit size or speed of data production and consumption.

A cache controller attempts to guess what data will be requested next and prefetch that data into the cache.

Homework #4 Solution (p 243)

The most important differences are speed and cost. Static RAM is five to ten times faster and approximately ten times more expensive. Static RAM is composed entirely of transistors. Dynamic RAM is composed of transistors and capacitors. The capacitors require frequent recharging which slows access time.

8. How is data stored and retrieved on a magnetic mass storage device?

For writing, current is routed through a read/write head which generates a magnetic charge at its tip. A bit area of a magnetic recording surface is placed in close proximity to the read/write head and a charge is induced in the surface material. The direction of current flow through the read/write head determines the polarity of the stored charge.

For reading, a bit area of the recording surface is placed in close proximity to the read/write head. The stored charge induces current to flow through the read write/head. A switch connected to the read/write head detects the direction of current flow to interpret it as a bit value.

Disk average access time is a combination of rotational delay, track-to-track seek time (head movement), and head-to-head switching time. Rotational delay is reduced if rotational speed is increased.

11. Why is the recording density of optical disks higher than the recording density of magnetic disks? What factor(s) limit this recording density?

A laser can be tightly focused to a narrow beam. Magnetic charge cannot be easily focused and an attempt to charge a very small area overwrites surrounding areas. Optical bit size is limited by the wavelength of the read laser and the amount of reflected light required to activate the read mechanism.

1. What access time is required of primary storage for the CPU to execute with zero wait states?

One divided by 400,000,000 divided by 2 equals 1.25 nanoseconds.

15/2.5 = 6 wait states

60/2.5 = 24 wait states