Subject Name: Advanced Computer Architecture

Subject Code: CS-6001

Semester: 6th
Downloaded from be.rgpvnotes.in

Department of Computer Science Engineering

Subject Name: Advance Computer Architecture Subject Code: CS 6001

Subject Notes
Unit-V
Parallel Programming Models
The model of a parallel algorithm is developed by considering a strategy for dividing the data and processing
method and applying a suitable strategy to reduce interactions. Parallel programming models are specifically
designed for multiprocessor, multicomputer or vector / SIMD computers. Parallel computing is the
simultaneous use of multiple compute resources to solve a computational problem:
 A problem is broken into discrete parts that can be solved concurrently
 Each part is further broken down to a series of instructions
 Instructions from each part execute simultaneously on different processors
 An overall control/coordination mechanism is employed.
In this chapter, we will discuss some parallel models.

Shared-Variable Model
A program is collection of processes. Parallelism depends on how inter-process communication (IPC) is
implemented. Fundamental issues in parallel programming are centered on the specification, creation,
suspension, reactivation, migration, termination and synchronization of concurrent processes residing in the
same or different processors.
By limiting the scope and access rights, the process address space may be shared or restricted. To
ensure orderly IPC, a mutual exclusion property requires the exclusive access of a shared object by one
process at a time.

Shared Variable Communication-

Multiprocessor programming is based on the use of shared variables in a common memory for IPC. As shown
in figure 5.1 shared variable IPC demands the use of shared memory and mutual exclusion among multiple
processes accessing the same set of variables.

Figure 5.1: IPC using shared variable

 Critical section.
Issues with shared variable model-

 Memory consistency.
 Atomicity with memory operation.
 Fast synchronization.
 Shared data structure.

Page no: 1 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

Message-Passing Model
Two processes A and B communicate with each other by passing message through a direct network. The
messages may be instructions, data, synchronization, or interrupt signals, etc. Delay caused by message
passing is much longer than shared variable model in a same memory. Two message passing programming
models are introduced here.

Figure 5.2: IPC using message passing

 It synchronizes the sender and receiver process with time and space just like telephone call.
Synchronous message passing-:

 No shared memory.
 No need of mutual exclusion.
 No buffer is used in communication channel.
 It can be blocked by channel being busy or error.
 One message is allowed to be transmitted via a channel at a time.
 Sender and receiver must be coupled in both time and space synchronously.
 Also called blocking communication scheme.

 Does not need to synchronize the sender and receiver in time and space.
Asynchronous message passing-:

 Non-blocking can be achieved.

 Buffers are used to hold the message along the path of connecting channel.
 Critical issue in programming this model is how to distribute or duplicate the program codes and data

 Message passing programming is gradually changing, once the virtual memories from all nodes are
sets over the processing nodes.

combined.

Figure 5.3: (a) Synchronous message passing (b) Asynchronous message passing
Data-Parallel Model
The data parallel code is easier to write and to debug because parallelism is explicitly handled by hardware
synchronization and flow control. Data parallel languages are modified directly from standard serial
programming languages.

Page no: 2 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

Data parallel programming emphasizes local computations and data routing operations such as permutation,
replication, reduction, and parallel prefix. It is applied to fine grain problems using regular grids, stencils, and
multidimensional signal / image data sets.
In data parallel model, tasks are assigned to processes and each task performs similar types of operations on
different data. Data parallelism is a consequence of single operations that is being applied on multiple data
items.
Data-parallel model can be applied on shared-address spaces and message-passing paradigms. In data-parallel
model, interaction overheads can be reduced by selecting a locality preserving decomposition, by using
optimized collective interaction routines, or by overlapping computation and interaction.
The primary characteristic of data-parallel model problems is that the intensity of data parallelism increases
with the size of the problem, which in turn makes it possible to use more processes to solve larger problems.
Example − De se atrix ultiplicatio

Figure 5.4: Data parallel model

Object-Oriented Model
In this model, objects are dynamically created and manipulated. Processing is performed by sending and
receiving messages among objects. Concurrent programming models are built up from low level objects such
as process, queues, and semaphores into high level objects like monitors and program modules.
Program abstraction leads to program modularity and software reusability as is often found in OOP.
Other areas have encouraged the growth of OOP include the development of CAD tools and word processors
with graphics capability. The development of concurrent object oriented programming (COOP) provides an
alternative model for concurrent computing on multiprocessors or on multicomputer. Various object models
differ in the internal behavior of objects and in how they interact with each other.
The idea behind parallel object-oriented programming is to provide suitable abstractions and software
engineering methods for structured application design. As in the traditional object model, objects are defined
as abstract data types, which encapsulate their internal state through well-designed interfaces and thus
represent passive data containers. If we treat this model as a collection of shared objects, we can find an
interesting resemblance with the shared data model.
The object-oriented programming model is by now well established as the state-of-the-art software
engineering methodology for sequential programming, and recent developments are also emerging to
establish object-orientation in the area of parallel programming. The current lack of acceptance of this model
among the scientific community can be explained by the fact that computational scientists still prefer to write
their programs using traditional languages. This is the main difficulty that has been faced by the object-
oriented environments, though it is considered as a promising technique for parallel programming.

 Object are created and manipulated dynamically.

In short we can say in object oriented model-

 Processing is performed using object.

 Concurrent programming model are built up from low level object such as processes, queue and
semaphore.

Page no: 3 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

Functional and Logic Models

Two language oriented programming models for parallel processing are described in this section. The first
model is based on using functional programming languages such as pure, Lisp, SISAL, and Strand 88. The
second model is based on logic programming languages such as Concurrent space Prolog and Parlog.
A functional programming language emphasizes the functionality of a program and should not produce side
effects after execution. There is no concept of storage, assignment, and branching in functional programs. In
other words, the history of any computation performed prior to the evaluation of a functional expression
should be irreverent to the meaning of the expression. The majority of parallel computers designed to support
the functional model were oriented toward Lisp, such as the Multilisp developed at MIT. Other data flow
computers have been used to execute functional programs, including SISAL used in the Manchester dataflow
machine.
Based on predicate logic, logic programming is suitable for knowledge processing dealing with large databases.
This model adopts an implicit search strategy and supports parallelism in the logic inference process. Both
functional and logic programming models have been used in artificial intelligence applications where parallel
processing is very much in demand.

 Two language-oriented programming for parallel processing are purposed.

So we can say in functional and logic models-

 Functional programming model such as LISP, SISAL, Strand 88.

 Logic programming model as prolog.
 Based on predicate logic, logic programming is suitable for solving large database queries.
Parallel Languages and Compilers
The environment for parallel computers is much more demanding than that for sequential computers.
Software for driving parallel computers is still in the early developmental stage. Users are still forced to spend
a lot of time programming hardware details instead of concentrating on program parallelism using high level
abstraction. To break this hardware / software barrier, we need a parallel software environment which
provides better tools for users to implement parallelism and to debug programs. Most of the recently
developed software tools are still in the research and testing stage, and a few have become commercially
available. A parallel language is able to express programs that are executable on more than one processor. In
this we use high level language in source code as it becomes a necessity in modern computer.

Figure 5.5: Compilation phases in parallel code generation

Page no: 4 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

Role of Compiler
The role of compiler is to remove the burden of program optimization and code generation. A parallelizing
compiler consists of the three major phases.

 Flow Analysis
o Program flow pattern in order to determine data and control dependence in the source code.
o Flow analysis is conducted at different execution levels on different parallel computers.
o Instruction level parallelism is exploited in super scalar or VLSI processors, loop level in SIMD,
vector, Task level in multiprocessors, multicomputer, or a network workstation.

 Optimization
o The transformation of user programs in order to explore the hardware capabilities as much as
possible.
o Transformation can be conducted at the loop level, locality level, or prefetching level.
o The ultimate goal of PO is to maximize the speed of code execution.
o It involves minimization of code length and of memory accesses and the exploitation.
o Sometimes should be conducted at the algorithmic level and must involve the programmer.

 Code Generation
o Code generation usually involves transformation from one representation to another, called an
intermediate form.
o Even more demanding because parallel constructs must be included.
o Code generation closely tied to instruction scheduling policies used.
o Optimized to encourage a high degree of parallelism.
o Parallel code generation is very different for different computer classes they are software and
hardware scheduled.

Language Features for Parallelism

 Optimization features.
Language features are classified into six categories-

o This features converting sequentially coded programs into parallel forms.

o The purpose is to match the software with the hardware Parallelism in the target machine.
o Automated parallelizer
o Semi-automated parallelizer (programmers interaction)
o Interactive restructure support (static analyzer, run time static, data flow graph,)

 Availability features.
o Feature enhances user friendliness. Make the language portable to a larger class of parallel
computer
o Expand the applicability of software libraries
o Scalability – scalable to the number of processors available and independent of hardware
topology
o Compatibility-compatible with establishment sequential
o Portability –portable to shared memory multiprocessors, message passing multicomputer, or
both

 Synchronization /communication features.

o Single assignment languages

Page no: 5 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

o Remote producer call

o Data flow languages such as ID
o Send /receive for message passing
o Barriers ,mailbox , semaphores , monitors

 Control of parallelism
o Coarse ,medium, or fine grains
o Explicit versus implicit parallelism
o Global parallelism in the entire program
o Take spilt parallelism
o Shared task queue

 Data parallelism features

o Used to specify how data are accessed and distributed in either SIMD and MIMD computers
o Run- time automatic decomposition
o Mapping specification
o Virtual processor support
o Direct access to shared data
o SPMD(single program multiple data)

 Process management features

o Needed to support the efficient creation of parallel processes, implementation of
multithreading or multitasking.
o Dynamic process creation at run time
o Light weight processes(threads)- compare to UNIX(heavyweight)processes
o Replicated work
o Partitioned networks
o Automatic load balancing

Parallel Programming Environment

An environment for parallel programming consists of hardware platforms, languages supported, OS and
software tools, and application packages. The hardware platforms vary from shared memory, message
passing, vector processing, and SIMD to data flow computers.
To implement a parallel algorithm you need to construct a parallel program. The environment within which
parallel programs are constructed is called the parallel programming environment. There are hundreds of
parallel programming environments. To understand them and organize them in a meaningful way, we need to
sort them with regard to a classification scheme. Currently, various kinds of high performance machines based
on the integration of many processors are available.

 To find the concurrency in the given problem.

Key parallel programming steps:

 To structure the algorithm so that concurrency can be exploited.

 To implement the algorithm in a suitable programming environment.
 To execute and tune the performance of the code on a parallel system.

 Tools support individual process tasks such as checking the consistency of a design, compiling a
Software Tools and Environments

program, comparing test results, etc. Tools may be general-purpose, stand-alone tools (e.g. a word-
processor) or may be grouped into workbenches.

Page no: 6 Follow us on facebook to get real-time updates from RGPV

Downloaded from be.rgpvnotes.in

 Workbenches support process phases or activities such as specification, design, etc. They normally

 Environments support all or at least a substantial part of the software process. They normally include
consist of a set of tools with some greater or lesser degree of integration.

several different integrated workbenches.

The diagram below illustrates this classification and shows some examples of these different classes of CASE
support. Many types of tool and workbench have been left out of this diagram.

Figure 5.6: Tools, workbenches and environments

The term 'workbenches' is not now much used and the term 'environments' has been extended to cover sets
of tools focused on a specific process phase e.g. programming environment or requirements engineering
environment.
In the above diagram environments are classified as integrated environments or process-centered
environments. Integrated environments provide infrastructure support for integrating different tools but are
not concerned with how these tools are used. Process-centered environments are more general. They include
software process knowledge and a process engine which uses this process model to advise engineers on what
tools or workbenches to apply and when they should be used.
In practice, the boundaries between these different classes are blurred. Tools may be sold as a single product
but may embed support for different activities. For example, most word processors now provide a built-in
diagram editor. CASE workbenches for design usually support programming and testing so they are more akin
to environments than specialized workbenches.

Page no: 7 Follow us on facebook to get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]