Embedded Systems Design, 2nd Edition

54 Embedded systems design

the results. To prevent pipeline stalling, an internal Harvard

architecture is used to allow simultaneous instruction and oper-

and fetches. It has been optimised for many instructions and

addressing modes so that single-cycle execution can be achieved.

The early pipeline stages are effectively duplicated to allow both

paths of a branch instruction to be processed until the path

decision is taken. This removes pipeline stalls and the subsequent

performance degradation. While integer instructions are being

executed, the floating point unit is free to execute floating point

instructions.

Data

registers

Address

registers

PC

CCR

A7’/ ...