Pipelining     Last updated on          

A basic Time-oriented instruction-fetching pipelining-method is shown; but, there are many other ways of methods such as Pipe-oriented, Frequency-oriented, Material-oriented, and ...; Pipelining enhances more throughput gain, and faster processing speed gain; Notice the 2 dimensional Time/Pipe pipelining. SiO2 based technology limits 2 dimensions only. Optical computing in the future will compute more than 2 dimensions. In traditional SiO2 based technology, even though numbers of pipes can be increased to achieve more throughput gain, but more numbers of pipes make very difficult human un-understandable complexity to code. Also see: Driver-to-load ratio;

                   Time1    Time2    Time3    Time4    Time5    Time6    Time7    Time8    TimeN-1    TimeN

Pipe #1:    Task1    Task1   

Pipe #1:    Task0    Task0    Task2    Task2    Task2    Task4    Task4

Pipe #2:    Task2    Task2    Task2    Task2    Task2    Task2    Task2

Pipe #3:    Task3    Task3    Task3

Pipe #3:    Task0    Task0    Task0    Task4    Task4    Task4    Task4

Pipe #4:    Task4    Task4    Task4    Task4    Task4    Task4    Task4    Task4

Pipe #5:    Task5    Task5    Task5    Task5   

Pipe #5:    Task0    Task0    Task0    Task0    Task6    Task6

Pipe #6:    Task6    Task6    Task6    Task6    Task6    Task6

                    TaskN-1

                    TaskN

 

For systems experts only: before developing optical computing, calculate through put gain in Time/Pipe while multitasking; where through put gain = more numbers of executed tasks ;

 

 

Up