Classification of computer Architectures
Flynn’s Classification (1966): is based on the multiplicity of instruction stream
and data stream in a computer system.
Feng’s Classification (1972): is based on serial versus parallel processing.
Handler’s Classification (1977): is determined by the degree of parallelism
and pipelining in various subsystem levels.
2
Feng’s Classification
●Feng (1972) has proposed the use of degree of parallelism to classify
various computer architectures.
●The maximum number of binary digits(bits) that can be processed within a
unit time by a computer system is called maximum degree of parallelism P.
●Let us Pi be the number of bits that can be processed within the ith
processor cycle. Consider processor cycles indexed by i=1,2,3…,T. The
average degree of parallelism, Pa is given by
●In general, Pi≤P. We define the utilization rate µ of a computer system
within T cycles is
●If the computing power is fully utilized (or the parallelism is fully exploited),
we have Pi=P for all I and µ=1 for 100 percent utilization. The utilization
rate depends on the application program being executed.
3
Classification based on Word/Bit Count
Classification based on sequential and parallel operation at bit and word levels :
1) Word Serial and Bit Serial (WSBS): also called as Bit Serial Processing (1 bit is
processed at time); slow
2) Word Parallel and Bit Serial (WPBS): also called as Slice processing (m bit slice at
a time)
3) Word Serial and Bit Parallel (WSBP): also called as word slice processing (1 word
of n bit at time)
4) Word Parallel and Bit Parallel (WPBP): also known as fully parallel processing;
fastest
The maximum degree of parallelism P is given by the product of the number of bits in
a word(n) and number of words in parallel(m). P= m.n
4
Example
5 word length (n)
1) Word Serial and Bit Serial (WSBS)
(1, 1)
2) Word Parallel and Bit Serial (WPBS)
(*, 1)
3) Word Serial and Bit Parallel (WSBP)
(1, *)
4) Word Parallel and Bit Parallel (WPBP)
(*, *)
* ---> greater than 1
Example...
Degree of Parallelism
-> area of rectangle
6
Limitation
1.Fails to project the concurrency in pipeline processors as degree of
parallelism doesn’t account for concurrency handle by pipelined design
Handler’s schema allow pipeline to be specific...
7