In order to make the new driving method to work, special
improvements should be made to the design of processor
architectures. Firstly, for each instruction executing step,
such as fetching instruction step, instruction decoding step,
fetching operand step, operating step, writing-back step,
and PC+ step, a certain processor’s functional unit should
be constructed to carry out the given operation. Secondly, a
functional unit should be designed in the way by which the
unit can work as a service provider so that a PD can behave
as a service client. Thus, a PD, according to the certain sequence
of the instruction executing steps, ”travels” around
the processor’s functional units to request the services provided
by the units, that is, the PD may ”go” to an instruction
fetching unit to fetch an instruction, and then the PD may
go to a decoding unit to decode the instruction, , and then
the PD may go to a PC+ unit to have its PC renewed, and
then the PD may go back to the instruction fetching unit to
fetch the next instruction, and so on. The basic architecture
of a CMPP is as flows.
Figure 1. The structure of CMPP with dynamic
PED.