Scribd
Upload
48 views

Parallel Computer Structures

The document discusses three types of parallel computer structures: pipelined computers which use temporal parallelism through overlapping computations, array processors which use spatial parallelism through multiple synchronized arithmetic units, and multiprocessor systems which use asynchronous parallelism through a set of interactive processors sharing resources. It also covers classifications of parallel computers including Flynn's taxonomy based on instruction and data streams, Feng's classification using degree of parallelism, and Handler's classification considering parallelism at subsystem levels.

Uploaded by

Tushar Gupta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
48 views

Parallel Computer Structures

The document discusses three types of parallel computer structures: pipelined computers which use temporal parallelism through overlapping computations, array processors which use spatial parallelism through multiple synchronized arithmetic units, and multiprocessor systems which use asynchronous parallelism through a set of interactive processors sharing resources. It also covers classifications of parallel computers including Flynn's taxonomy based on instruction and data streams, Feng's classification using degree of parallelism, and Handler's classification considering parallelism at subsystem levels.

Uploaded by

Tushar Gupta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 23

Parallel Computer

Structures
Parallel Computer Structures
 Pipelined Computers :
 overlapped computations
 temporal parallelism.
 Array Processors :
 multiple synchronized arithmetic logic units
 spatial parallelism.
 Multiprocessor Systems :
 asynchronous parallelism
 set of interactive processors with shared resources.
2
Pipeline Computers
 Normally, four major steps to execute an
instruction:
 Instruction Fetch (IF)
 Instruction Decoding (ID)
 Operand Fetch (OF)
 Execution (EX)

3
Space-time diagram for a nonpipelined processor

4
Space-time diagram for a pipelined processor

5
Array Computers
 A synchronous parallel computer
 Multiple arithmetic logic units
 processing elements (PE)
 operate in parallel.
 PEs are synchronized
 perform the same function at the same time.
 Appropriate data routing mechanism must be
established among the PEs.
6
Functional structure of
array computer

7
Multiprocessor Systems
 A single computer that includes multiple
processors (computer modules).
 All processors share memory modules, I/O
channels and peripheral devices.
 Controlled by one operating system
 provides interaction between processors and their
programs.
 Local memory and private devices.

8
 Processor communication
 sending messages
 sharing a common memory or through an
interrupt network.
 Interconnections
 Time-shared common bus
 Crossbar switch network
 Multiport memories

9
10
Performance of Parallel Computers
 Speedup of a parallel computer with n identical
processors
 n times faster than single processor.
 Speedup is much less(in practice)
 some processors are idle.
 Actual speedup
 lower-bound log2n (Minsky’s conjecture)
 upper bound n/ln n
 So Commercial processor system consists of
only 2 or 4 processors. 11
12
 T1=1
 fi – probability of assigning the same problem to
i processors.
 Average load = di = 1/i per processor
 fi =1/n.
 Average time required to solve the problemn
on
1
an n- processor system is n i
Tn   f i .d i  i 1

i 1 n
T1 n n
 Average speedup S  n

Tn 1 ln n

i 1 i
13
Architectural Classification
Schemes
1. Flynn’s Classification
 Multiplicity of data and instruction streams.
 Instruction stream
 a sequence of instructions as executed by the
machine.
 Data stream
 a sequence of data including input, partial, or
temporary results, called for by the instruction
stream.
 Flynn’s four machine organizations :
 SISD, SIMD, MISD, MIMD.
15
SISD
 Single Instruction stream-Single Data
stream
 Instructions are executed sequentially but
may be overlapped in their execution
stages (pipelining).

16
SIMD
 Single Instruction stream-Multiple Data stream
 There are multiple PEs supervised by the same
control unit.

17
MISD
 Multiple Instruction stream-Single Data stream

18
MIMD
 Multiple Instruction stream-Multiple Data stream
 Multiprocessor systems and Multi- computer
systems.

19
2. Feng’s classification
 Use the degree of parallelism
 Maximum degree of parallelism
 max: no: of bits that can be processed within a
unit time (P). T

 Average parallelism degree P i

P a
 i 1

T
 Utilization Rate T

 P a
 P
i 1
i

P T .P
20
 Max: parallelism degree, P(C ) of a computer
system C is P(C )= n.m
 Four type of processing methods
1. Word-serial and bit-serial (WSBS)
 n=m=1
 Bit serial processing as one bit is processed
at a time.
2. Word-parallel and bit-serial (WPBS)
 n=1, m>1
 Bis(Bit-slice) processing as an m-bit-slice is
processed at a time.
21
3. Word-serial and bit-parallel (WSBP)
 n>1, m=1
 Word-slice processing as it processed one
word of n-bits at a time.
4. Word-parallel and bit-parallel (WPBP)
 n>1, m>1
 Parallel processing as an array of (n x m)
bits is processed at a time.

22
3. Handler’s Classification
 Degree of parallelism and pipelining built in to
the hardware structures of a computer system.

 Parallel pipeline processing is considered at


three subsystem levels:
 PCU, ALU, BLC

 T(X) = < K x K’ , D x D’ , W x W’ >

 Ex:
TI - ASC <1, 4, 64 x 8>
23

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy