Please read the following link...
posted on
Feb 02, 2009 04:47PM
http://agoracom.com/ir/patriot/forum...
and then the actual European application. Please note that a problem is described then, more importantly, the "inventive/novel step" is given, eg
Computer instructions are usually broken down into sequential pieces, for example: fetch, decode, register read, execute, and store. Each piece will require a single machine cycle. In most Reduced Instruction Set Computer (RISC) chips, instruction require from three to six cycles.
RISC instructions are very parallel. For example, each of 70 different instructions in the SPARC (SUN Computer's RISC chip) has five cycles. Using a technique called "pipelining", the different phases of consecutive instructions can be overlapped.
To understand pipelining, think of building five residential homes. Each home will require in sequence, a foundation, framing, plumbing and wiring, roofing, and interior finish. Assume that each activity takes one week. To build one house will take five weeks.
But what if you want to build an entire subdivision? You have only one of each work crew, but when the foundation men finish on the first house, you immediately start them on the second one, and so on. At the end of five weeks, the first home is complete, but you also have five foundations. If you have kept the framing, plumbing, roofing, and interior guys all busy, from five weeks on, a new house will be completed each week.
This is the way a RISC chip like SPARC appears to execute an instruction in a single machine cycle. In reality, a RISC chip is executing one fifth of five instructions each machine cycle. And if five instructions stay in sequence, an instruction will be completed each machine cycle.
The problems with a pipeline are keeping the pipe full with instructions. Each time on out of sequence instruction such as a BRANCH or CALL occurs, the pipe must be refilled with the next sequence. The resulting dead time to refill the pipeline can become substantial when many IF/THEN/ELSE statements or subroutines are encountered.
The microprocessor 50 has no pipeline as such. The approach of this microprocessor to speed is to overlap instruction fetching with execution of the previously fetched instruction(s). Beyond that, over half the instructions (the most common ones) execute entirely in a single machine cycle of 20 nsec. This is possible because:
There are many others.
Be well