What does CPI in Pipelining actually mean?

Question

Require detailed explanation along with formulas and which formula to use depending upon each scenario

Comments

CPI = cycles per instruction

CPI for ideal pipeline = 1.1

rest u have to calculate based on question....it varies

By solving many questions ul get an idea...or post any specific questions where different formulas have been used...may be then u could be helped to know the difference

This is vague :)

---

Correct me if I am wrong, but until now whatever i have analyzed is that CPI has 2 variants

Let's take for Ex 5 stage instruction
IF = 2ns
ID = 1.5ns
OF = 2ns
EX = 3ns
WB = 1ns

while there are buffers/latches/registers having time latencies of let's say 1ns.
 
1.) CPI for non pipelined processors = which is basically summation of all the time latency faced during each instruction while ignoring delays due to other buffers/latches/registers,
CPI = 2+1.5+2+3+1 = 9.5

2.) CPI for pipelined shall be the maximum time latency relative to all the stages in the pipeline
ie = EX  + latency due to interstage latches
(3+1 = 4 )

Any Modification/Editing that can enhance this question is more than welcome
:)

---

CPI is cycles per instruction,ie CPU clock cycles needed to execute an instruction,there us no unit associated with it...when you use something like '1.5ns' that is the total cycle time or execution time etc.

---

@namit right :)