# Designing Parallel Algorithms: Part 1

### Partitioning

The partitioning stage of a design is intended to expose opportunities for parallel execution. Hence, the focus is on defining a large number of small tasks in order to yield what is termed a fine-grained decomposition of a problem. Just as fine sand is more easily poured than a pile of bricks, a fine-grained decomposition provides the greatest flexibility in terms of potential parallel algorithms. In later design stages, evaluation of communication requirements, the target architecture, or software engineering issues may lead us to forego opportunities for parallel execution identified at this stage. We then revisit the original partition and agglomerate tasks to increase their size, or granularity. However, in this first stage we wish to avoid prejudging alternative partitioning strategies.

A good partition divides into small pieces both the computation associated with a problem and the data on which this computation operates. When designing a partition, programmers most commonly first focus on the data associated with a problem, then determine an appropriate partition for the data, and finally work out how to associate computation with data. This partitioning technique is termed domain decomposition. The alternative approach -- first decomposing the computation to be performed and then dealing with the data -- is termed functional decomposition. These are complementary techniques which may be applied to different components of a single problem or even applied to the same problem to obtain alternative parallel algorithms.

In this first stage of a design, we seek to avoid replicating computation and data; that is, we seek to define tasks that partition both computation and data into disjoint sets. Like granularity, this is an aspect of the design that we may revisit later. It can be worthwhile replicating either computation or data if doing so allows us to reduce communication requirements.

Domain Decomposition

In the domain decomposition approach to problem partitioning, we seek first to decompose the data associated with a problem. If possible, we divide these data into small pieces of approximately equal size. Next, we partition the computation that is to be performed, typically by associating each operation with the data on which it operates. This partitioning yields a number of tasks, each comprising some data and a set of operations on that data. An operation may require data from several tasks. In this case, communication is required to move data between tasks. This requirement is addressed in the next phase of the design process.

The data that are decomposed may be the input to the program, the output computed by the program, or intermediate values maintained by the program. Different partitions may be possible, based on different data structures. Good rules of thumb are to focus first on the largest data structure or on the data structure that is accessed most frequently. Different phases of the computation may operate on different data structures or demand different decompositions for the same data structures. In this case, we treat each phase separately and then determine how the decompositions and parallel algorithms developed for each phase fit together.

Figure 2 illustrates domain decomposition in a simple problem involving a three-dimensional grid. (This grid could represent the state of the atmosphere in a weather model, or a three-dimensional space in an image-processing problem.) Computation is performed repeatedly on each grid point. Decompositions in the x, y, and/or z dimensions are possible. In the early stages of a design, we favor the most aggressive decomposition possible, which in this case defines one task for each grid point. Each task maintains as its state the various values associated with its grid point and is responsible for the computation required to update that state.

Figure 2: Domain decompositions for a problem involving a three-dimensional grid. One-, two-, and three-dimensional decompositions are possible; in each case, data associated with a single task are shaded. A three-dimensional decomposition offers the greatest flexibility and is adopted in the early stages of a design.

Functional Decomposition

Functional decomposition represents a different and complementary way of thinking about problems. In this approach, the initial focus is on the computation that is to be performed rather than on the data manipulated by the computation. If we are successful in dividing this computation into disjoint tasks, we proceed to examine the data requirements of these tasks. These data requirements may be disjoint, in which case the partition is complete. Alternatively, they may overlap significantly, in which case considerable communication will be required to avoid replication of data. This is often a sign that a domain decomposition approach should be considered instead.

While domain decomposition forms the foundation for most parallel algorithms, functional decomposition is valuable as a different way of thinking about problems. For this reason alone, it should be considered when exploring possible parallel algorithms. A focus on the computations that are to be performed can sometimes reveal structure in a problem, and hence opportunities for optimization, that would not be obvious from a study of data alone.

Figure 3: Functional decomposition in a computer model of climate. Each model component can be thought of as a separate task, to be parallelized by domain decomposition. Arrows represent exchanges of data between components during computation: the atmosphere model generates wind velocity data that are used by the ocean model, the ocean model generates sea surface temperature data that are used by the atmosphere model, and so on.

Functional decomposition also has an important role to play as a program structuring technique. A functional decomposition that partitions not only the computation that is to be performed but also the code that performs that computation is likely to reduce the complexity of the overall design. This is often the case in computer models of complex systems, which may be structured as collections of simpler models connected via interfaces. For example, a simulation of the earth's climate may comprise components representing the atmosphere, ocean, hydrology, ice, carbon dioxide sources, and so on. While each component may be most naturally parallelized using domain decomposition techniques, the parallel algorithm as a whole is simpler if the system is first decomposed using functional decomposition techniques, even though this process does not yield a large number of tasks (Figure 3).

Partitioning Design Checklist

The partitioning phase of a design should produce one or more possible decompositions of a problem. Before proceeding to evaluate communication requirements, we use the following checklist to ensure that the design has no obvious flaws. Generally, all these questions should be answered in the affirmative.

• Does your partition define at least an order of magnitude more tasks than there are processors in your target computer? If not, you have little flexibility in subsequent design stages.
• Does your partition avoid redundant computation and storage requirements? If not, the resulting algorithm may not be scalable to deal with large problems.
• Are tasks of comparable size? If not, it may be hard to allocate each processor equal amounts of work.
• Does the number of tasks scale with problem size? Ideally, an increase in problem size should increase the number of tasks rather than the size of individual tasks. If this is not the case, your parallel algorithm may not be able to solve larger problems when more processors are available.
• Have you identified several alternative partitions? You can maximize flexibility in subsequent design stages by considering alternatives now. Remember to investigate both domain and functional decompositions.

Answers to these questions may suggest that, despite careful thought in this and subsequent design stages, we have a "bad'' design. In this situation it is risky simply to push ahead with implementation.

### Coming Up Next

In the next installment of this multi-part article, I examine information flow specified in the communication phase of a design.

### More Insights

 To upload an avatar photo, first complete your Disqus profile. | View the list of supported HTML tags you can use to style comments. | Please read our commenting policy.

# First C Compiler Now on Github

The earliest known C compiler by the legendary Dennis Ritchie has been published on the repository.

# HTML5 Mobile Development: Seven Good Ideas (and Three Bad Ones)

HTML5 Mobile Development: Seven Good Ideas (and Three Bad Ones)

# Building Bare Metal ARM Systems with GNU

All you need to know to get up and running... and programming on ARM

# Amazon's Vogels Challenges IT: Rethink App Dev

Amazon Web Services CTO says promised land of cloud computing requires a new generation of applications that follow different principles.

# How to Select a PaaS Partner

Eventually, the vast majority of Web applications will run on a platform-as-a-service, or PaaS, vendor's infrastructure. To help sort out the options, we sent out a matrix with more than 70 decision points to a variety of PaaS providers.

More "Best of the Web" >>