Processing Unit Design - Fundamentals of Computer Organization and Architecture - page 100

implemented using a set of micro-operations. Associated with each micro-operation
is a set of control lines that must be activated to carry out the corresponding micro-
operation. The idea of microprogrammed control is to store the control signals
associated with the implementation of a certain instruction as a microprogram in
a special memory called a control memory (CM). A microprogram consists of a
sequence of microinstructions. A microinstruction is a vector of bits, where each
bit is a control signal, condition code, or the address of the next microinstruction.
Microinstructions are fetched from CM the same way program instructions are
fetched from main memory (Fig. 5.12).
When an instruction is fetched from memory, the op-code field of the instruc-
tion will determine which microprogram is to be executed. In other words, the
op-code is mapped to a microinstruction address in the control memory. The
microinstruction processor uses that address to fetch the first microinstruction in
the microprogram. After fetching each microinstruction, the appropriate control
lines will be enabled. Every control line that corresponds to a “1” bit should be
turned on. Every control line that corresponds to a “0” bit should be left off.
After completing the execution of one microinstruction, a new microinstruction
will be fetched and executed. If the condition code bits indicate that a branch
must be taken, the next microinstruction is specified in the address bits of the cur-
rent microinstruction. Otherwise, the next microinstruction in the sequence will be
fetched and executed.
The length of a microinstruction is determined based on the number of micro-
operations specified in the microinstructions, the way the control bits will be
interpreted, and the way the address of the next microinstruction is obtained. A
microinstruction may specify one or more micro-operations that will be activated
simultaneously. The length of the microinstruction will increase as the number of
parallel micro-operations per microinstruction increases. Furthermore, when each
control bit in the microinstruction corresponds to exactly one control line, the
length of microinstruction could get bigger. The length of a microinstruction
could be reduced if control lines are coded in specific fields in the microinstruction.
Decoders will be needed to map each field into the individual control lines. Clearly,
using the decoders will reduce the number of control lines that can be activated sim-
ultaneously. There is a tradeoff between the length of the microinstructions and the
amount of parallelism. It is important that we reduce the length of microinstructions
to reduce the cost and access time of the control memory. It may also be desirable
that more micro-operations be performed in parallel and more control lines can be
activated simultaneously.

Control

Control data
Register

Control
Memory

External
input

Control Address

Sequencer

Figure 5.12 Fetching microinstructions (control words)